Unnamed: 0
int64 0
10k
| input
stringlengths 9.18k
112k
| output
stringlengths 136
194k
| instruction
stringclasses 1
value |
|---|---|---|---|
9,735 | "<ns0:abstract xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:p>Surface air temperature (T a ) required for real-time environmental modelling applications should be spatially quantified to capture the nuances of local-scale climates. This study created near real-time air temperature maps at a high spatial resolution across Australia. This mapping is achieved using the thin plate spline (TPS) interpolation in concert with a digital elevation model and 'live' recordings garnered from 534 telemetered Australian Bureau of Meteorology (BoM) automatic weather station (AWS) sites. The interpolation was assessed using cross-validation analysis in a 1-year period using 30-minute interval observation. This was then applied to a fully automated mapping system -based in the R programming language -to produce near real-time maps at sub-hourly intervals. The cross-validation analysis revealed broad similarities across the seasons with meanabsolute error ranging from 1.2°C (autumn and summer) to 1.3°C (winter and spring), and corresponding root-mean-square error in the range 1.6°C to 1.7°C. The R 2 and concordance correlation coefficient (P c ) values were also above 0.8 in each season indicating predictions were strongly correlated to the validation data. On an hourly basis, errors tended to be highest during the late afternoons in spring and summer from 3 pm to 6 pm (AEST), particularly for the coastal areas of Western Australia. The mapping system was trialed over a 21-day period from 1 June 2020 to 21 June 2020 with majority of maps completed within 28-minutes of AWS site observations being recorded. All outputs were displayed in a web mapping application to exemplify a real-time application of the outputs.</ns0:p><ns0:p>This study found that the methods employed would be highly suited for similar applications requiring real-time processing and delivery of climate data at high spatiotemporal resolutions across a considerably large land mass.</ns0:p></ns0:div>
</ns0:abstract>
<ns0:body xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:head>Introduction</ns0:head><ns0:p>A timely and accurate source of air temperature (T a ) data is essential for a wide variety of environmental modelling applications requiring real-time monitoring of environmental change <ns0:ref type='bibr' target='#b14'>(Lazzarini et al. 2014)</ns0:ref>. This is often gleaned from a network of in-situ telemetered meteorological weather stations that are streamed over the internet <ns0:ref type='bibr' target='#b32'>(Williams et al. 2011</ns0:ref>). However, datasets of this nature tend to be relevant for a single geographic location that fail to accurately account for the spatial variability between sites that can vary markedly over short distances <ns0:ref type='bibr' target='#b30'>(Webb et al. 2016)</ns0:ref>. For applications that rely on location-specific data, observations are often harvested from stations situated kilometers away from their location of interest, resulting in that data not being truly representative of the desired location <ns0:ref type='bibr' target='#b11'>(Jeffrey et al. 2001;</ns0:ref><ns0:ref type='bibr' target='#b19'>Liu et al. 2018b)</ns0:ref>. Thus, T a can vary considerably over space and time, often attributed to the effects of topographic, coastal and latitudinal factors <ns0:ref type='bibr' target='#b7'>(Hutchinson 1991;</ns0:ref><ns0:ref type='bibr' target='#b9'>Jarvis & Stuart 2001a;</ns0:ref><ns0:ref type='bibr' target='#b29'>Wang et al. 2011)</ns0:ref>, cloud cover <ns0:ref type='bibr' target='#b34'>(Xue et al. 2019)</ns0:ref>, radiative effects from aerosols <ns0:ref type='bibr' target='#b15'>(Li et al. 2017;</ns0:ref><ns0:ref type='bibr' target='#b21'>Mitchell et al. 1995)</ns0:ref> and diurnal variation <ns0:ref type='bibr' target='#b18'>(Liu et al. 2018a</ns0:ref>). As such, T a for the purpose of input to real-time modelling applications need to be spatially quantified to dynamically account for these interactions but also at an appropriate spatial and temporal resolution to account for the subtle nuances of local-scale climates.</ns0:p><ns0:p>There has been a plethora of research aimed at interpolating surface air temperature at various spatiotemporal scales <ns0:ref type='bibr' target='#b7'>(Hutchinson 1991;</ns0:ref><ns0:ref type='bibr' target='#b10'>Jarvis & Stuart 2001b;</ns0:ref><ns0:ref type='bibr' target='#b11'>Jeffrey et al. 2001;</ns0:ref><ns0:ref type='bibr' target='#b13'>Jones et al. 2009;</ns0:ref><ns0:ref type='bibr' target='#b33'>Xu et al. 2018)</ns0:ref>. This is in addition to surface temperature estimated from satellite data <ns0:ref type='bibr' target='#b20'>(Mao et al. 2017;</ns0:ref><ns0:ref type='bibr' target='#b26'>Sobrino et al. 2020)</ns0:ref>. Or from regional reanalysis of global circulation models at high spatiotemporal resolutions <ns0:ref type='bibr' target='#b1'>(Bollmeyer et al. 2015;</ns0:ref><ns0:ref type='bibr' target='#b27'>Su et al. 2019)</ns0:ref>. Despite this, their application to real-time monitoring of climate have been limited, or insufficient for local-scale monitoring purposes. For example, a modelling system based on remote sensing data coupled with in-situ meteorological recordings was able to produce air temperature maps in near realtime across the United Arab Emirates <ns0:ref type='bibr' target='#b14'>(Lazzarini et al. 2014)</ns0:ref>. However, the spatial resolution of ~3km was limited in accounting for lapse rates in highly variable topography, despite the system capable of delivering outputs at very high temporal resolution (every 15-minutes). Similarly, a near real-time drought monitoring tool developed for South Asia <ns0:ref type='bibr' target='#b0'>(Aadhar & Mishra 2017)</ns0:ref>, capable of producing daily minimum and maximum temperatures at a spatial resolution of 0.05° (~5km), would also require further adaptation for high resolution monitoring. This is in addition to a similar system currently used in Australia, where daily minimum (Tmin) and maximum (Tmax) temperatures are produced from Australian Bureau of Meteorology (BoM) weather station sites using thin plate smoothing spline (TPS) interpolation to deliver daily products at 0.05° (~5km) grid resolution <ns0:ref type='bibr' target='#b11'>(Jeffrey et al. 2001;</ns0:ref><ns0:ref type='bibr' target='#b13'>Jones et al. 2009)</ns0:ref>. While both datasets are useful for broad-scale analysis requiring up-to-date daily records, they still lacked the resolution for sub-daily real-time monitoring at the local-scale.</ns0:p><ns0:p>Recently, a near-real time mapping system was developed to produce air temperature maps at high spatiotemporal resolutions across the state of Tasmania, Australia <ns0:ref type='bibr' target='#b31'>(Webb et al. 2020)</ns0:ref>. This used a combination of regression trees (RT) and TPS interpolation and capable of consistently producing maps at a spatial resolution of 80m at 1-hourly time steps. Evaluation of the system showed that the TPS method was highly suited to real-time application due to the speed and relative accuracy of the outputs produced. For example, assessment of the TPS interpolation showed root mean square errors were consistently under 1.5°C, in addition to only requiring 2minutes processing time to produce each map product. In this context, the application would be suited to the estimation of T a across a much larger geographic space at a similar spatiotemporal resolution. As such, there is also an opportunity to apply this approach on a digital platform for real-time access for end-users.</ns0:p><ns0:p>The objective of this study was to apply and extend the methods in <ns0:ref type='bibr' target='#b31'>Webb et al. (2020)</ns0:ref> for production of T a maps across continental Australia. TPS interpolation is used to produce T a maps at sub-hourly intervals (every 30-minutes) based on recordings garnered directly from BoM automatic weather station (AWS) sites. The resulting maps are presented digitally at a spatial resolution of 286m, appropriate for local-scale monitoring purposes. The methods for prediction accuracy are evaluated using historic hourly T a data captured over a 1-year period, in addition to assessing the efficacy of the system for real-time application and subsequent display of outputs in a purpose-built web mapping application.</ns0:p></ns0:div>
<ns0:div><ns0:head>Materials & Methods</ns0:head></ns0:div>
<ns0:div><ns0:head>Approach</ns0:head><ns0:p>The present study consisted of 2 parts. Firstly, evaluation of the TPS methodology using crossvalidation; and secondly, application of the methodology for operational real-time mapping of T a (Fig. <ns0:ref type='figure' target='#fig_0'>1</ns0:ref>). For the evaluation purpose of the study, a historical dataset of 30-minute interval T a recordings was garnered from BoM automatic weather station (AWS) sites for the 1-year period 1 March 2019 to 29 February 2020. This data was used in a leave-one-out cross-validation exercise to assess the prediction performance of the TPS interpolation method. For the application of the methodology for operational real-time mapping, this was tested over a 21-day period from 1 June 2020 to 21 June 2020. For this purpose, a fully automated mapping system was developed using R programming language (R Development Core Team 2015). Processing performance of this mapping system was evaluated for computational efficiency by analyzing each subsequent spatial output (i.e. the time to taken to produce each T a map) and therefore assessed for real-time application. Maps produced from the interpolation process are immediately displayed in a web map application.</ns0:p></ns0:div>
<ns0:div><ns0:head>Air temperature (T a ) data</ns0:head><ns0:p>Air temperature (T a ) data recorded by automatic weather stations (AWS) from the Bureau of Meteorology (BoM) and capable of providing real-time access at 30-minute intervals were considered for primary use in this study (Fig. <ns0:ref type='figure' target='#fig_1'>2</ns0:ref>). For evaluating the accuracy of the model, a requirement was set, where each station used for the real-time application should have historic recordings for the previous year, specifically from 1 March 2019 to 29 February 2020. These historical data were used for cross-validation analysis. It should be noted that not all AWS sites had data available for the entire evaluation period. Thus, AWS sites that had least 15 days of recordings -consisting of 30-minute interval recordings in each season -was considered for the evaluation process. AWS sites that did not meet this criterion were discarded from the analysis (192 in total). Thus, the screening process resulted in 534 AWS sites corresponding to a possible 17567 recording observations in the evaluation period and relevant to each AWS. It should be noted that AWS air temperature observations are recorded using a resistance temperature detector placed within a Stevenson weather screen at 1.2 m above ground (Bureau of Meteorology 2018). All AWS recordings are telemetered into the BoM climate database and publicly accessible via URL (http://www.bom.gov.au/tas/observations/). These are typically displayed at 30-minute intervals. However, due to telemetry and processing delays, readings tend to lag the true observation time of approximately 10-to 20-minutes.</ns0:p></ns0:div>
<ns0:div><ns0:head>Interpolating T a using thin plate smoothing splines (TPS)</ns0:head><ns0:p>T a values garnered from the BoM AWS sites were interpolated on a 30-minute interval basis using thin plate smoothing splines. This was performed to form TPS predictions in the evaluation period (1 March 2019 to 29 February 2020) as well as for application to real-time mapping. Its application involves a trivariate approach whereby latitude, longitude, and elevation variables are used as independent variables, as per <ns0:ref type='bibr' target='#b11'>Jeffrey et al. (2001)</ns0:ref>. The independent variables of latitude and longitude are used for the partial spline component to account for spatial variation, whereas elevation is combined to account for the temperature lapse rates. The spline component of the algorithm is optimised by minimising the generalized cross validation error from the residual sum of squares <ns0:ref type='bibr' target='#b7'>(Hutchinson 1991)</ns0:ref>. In this study, the Fields statistical package <ns0:ref type='bibr' target='#b22'>(Nychka et al. 2017</ns0:ref>) was used to implement the TPS algorithm in R software (R Development Core Team 2015). To guide the mapping of T a , the 9-second Digital Elevation Model (DEM) was used <ns0:ref type='bibr' target='#b8'>(Hutchinson et al. 2008)</ns0:ref>. This was reprojected to Geocentric Datum of Australia 94, Geoscience Australia Lambert projection; and resampled to a spatial resolution of 286 m (roughly equivalent to the spatial resolution of original 9-second DEM). The geographical coordinates of the AWS site locations were then spatially intersected with the newly resampled DEM. This operation provided a consistent template to routinely form TPS models using the AWS observations as data points to the algorithm (on a 30-minute basis). Thus, T a predictions generated by each TPS model were spatially interpolated using the DEM as the z variable, along with the coordinate parameters of the inherent cell properties of the DEM acting as the latitude (x) and longitude (y) variables. This allowed the spline smoothing parameter to be applied continuously across the geographic feature space of the DEM, resulting in a final mapped prediction; saved as GeoTIFF rasters.</ns0:p></ns0:div>
<ns0:div><ns0:head>Evaluating TPS interpolation</ns0:head><ns0:p>The performance of the TPS algorithm was evaluated in the period from 1 March 2019 to 29 February 2020. A leave-one-out cross-validation procedure was employed for each AWS site, similar to the method employed in <ns0:ref type='bibr' target='#b31'>Webb et al. 2020</ns0:ref>. Specifically, the training dataset was split into i parts such that i is equal to the number of AWS sites, i.e. 534. For each AWS in i, the i th AWS site was kept for validation (i.e. using actual recordings from the evaluation period), while the remaining dataset, comprising of the remaining BoM recordings was used for TPS modelling to predict T a at the i th AWS site. This was performed for each 30-minute interval (h) in the evaluation period to produce a set of modelled TPS estimates versus actual AWS recordings at each site. This equated to 17,567 modelled TPS predictions where observed T a -recorded from each corresponding AWS site -could then be compared. Validation metrics used to assess the modelling accuracy against the T a recordings, as per <ns0:ref type='bibr' target='#b31'>Webb et al. 2020</ns0:ref>, included the mean absolute error (MAE), root-mean-square error (RMSE), coefficient of determination (R 2 ) and the concordance coefficient. The concordance coefficient (P c ) was used to assess agreement between TPS predictions ; and actual recordings ; that fall on the 45° line through the origin, as defined 𝑥 𝑦</ns0:p><ns0:p>by <ns0:ref type='bibr' target='#b16'>Lin (1989)</ns0:ref>:</ns0:p><ns0:formula xml:id='formula_0'>𝑝 𝑐 = 2𝑝𝜎 𝑥 𝜎 𝑦 𝜎 2 𝑥 + 𝜎 2 𝑦 + (𝜇 𝑥 -𝜇 𝑥 ) 2</ns0:formula><ns0:p>where for and represent the means for and , respectively, and represent the</ns0:p><ns0:formula xml:id='formula_1'>𝜇 𝑥 𝜇 𝑥 𝑥 𝑦 𝜎 2 𝑥 𝜎 2 𝑦</ns0:formula><ns0:p>corresponding variances, and is the correlation coefficient between and . A concordance 𝑝 𝑥 𝑦 rating close to one indicates strong agreement between predicted and actual T a pairings that fall on the 45° line through the origin.</ns0:p></ns0:div>
<ns0:div><ns0:head>Application to real-time monitoring of T a</ns0:head><ns0:p>The proposed methodology, as advocated by <ns0:ref type='bibr' target='#b31'>Webb et al. 2020</ns0:ref>, was adopted in this study for operational real-time monitoring of T a across continental Australia. However, since the formation of BoM grids and calibration equations were not required in this study, the methodology was retrofitted to consist of two major components. Firstly, the import of 'live' T a data via the internet from the BoM website, and secondly, the mapping of the observations using TPS interpolation. This was trialled over a 21-day period from 1 June 2020 to 21 June 2020, using real-time BoM observations to drive the system which was fully automated using software R (R Development Core Team 2015). Because new BoM observations are typically available every 30-minutes, individual AWS site observations were downloaded at this frequency from the BoM observations portal as comma delimited text files (e.g. http://www.bom.gov.au/fwo/IDT60801/IDT60801.<stationIDnumber>.axf). Thus, the mapping system was programmed to query and import recordings every 30-minutes (bi-hourly) that corresponded to the nearest half-hour at 0-and 30 minutes (past the hour). Because of telemetry and associated processing delays (observation updates varied from station to station), the system was programmed to make queries at 5, 10, 15, 20, 25 and 30 minutes within their 30-minute processing window. In addition, a threshold was set where at least 480 out of the 534 BoM stations (i.e. 90% of total available AWS sites that were used in the evaluation analysis) have available observations before the mapping was allowed to commence in their respective processing window. This served to limit the number of missing observations that could otherwise produce significant inaccuracies in the subsequent mapped product. However, if this threshold was not met during the allocated query times, the mapping was still permitted to commence at the 30-minute mark regardless of the number of observations available (this was subsequently recorded). To provide context of the proposed system, the same procedure by <ns0:ref type='bibr' target='#b31'>Webb et al. (2020)</ns0:ref> found that most observations tended to be imported at the 15-minute mark (from the nearest observation hour) with corresponding TPS maps completed thereafter at the 17-minute mark.</ns0:p><ns0:p>The rationale for this study assumes a similar time frame, albeit at bi-hourly intervals, where observations are imported every 30-minutes (with an import time lag of ~15-minues from the nearest half-hour), followed by T a mapping thereafter. Note that all AWS recording times in this study were standardised to Australian Eastern Standard Time (AEST).</ns0:p><ns0:p>To interpolate the TPS predictions, the processing schema described in <ns0:ref type='bibr' target='#b31'>Webb et al. 2020</ns0:ref> was used. This consisted the Raster package <ns0:ref type='bibr' target='#b6'>(Hijmans & van Etten 2012)</ns0:ref> in combination with the Fields statistical package <ns0:ref type='bibr' target='#b22'>(Nychka et al. 2017</ns0:ref>) using software R (R Development Core Team 2015), to map and subsequently visualize the predictions in a continuous manner across Australia. To improve processing speed, the clusterR function within the Raster package was parameterised to host the TPS algorithm, thereby enabling mapping to occur using multi-core processors. In this manner, the mapping system was hosted on a high-end cloud computing Linux platform (Ubuntu 18.04 LTS (Bionic)) constituting 16 virtual CPU cores and 64GB RAM; made available courtesy of the Australian National eResearch Collaboration Tools and Resources project (NeCTAR). Spatial outputs were saved as individual GeoTIFF raster format at a grid cell resolution of 286m, i.e. equivalent to the spatial resolution of the resampled DEM.</ns0:p></ns0:div>
<ns0:div><ns0:head>Results</ns0:head></ns0:div>
<ns0:div><ns0:head>Assessment of the TPS interpolation procedure</ns0:head><ns0:p>Each of the AWS sites underwent the leave-one-out cross-validation analysis to assess TPS prediction accuracy for T a in the evaluation period: 1 March 2019 to 29 February 2020. This analysis revealed broad similarities across the seasons with MAE values ranging from 1.2°C (autumn and summer) to 1.3°C (winter and spring), and similarly RMSE ranging from 1.6°C to 1.7°C (Table <ns0:ref type='table' target='#tab_0'>1</ns0:ref>). The R 2 and P c values were above 0.8 indicating that the TPS predictions were strongly correlated to the validation data in addition to being highly associated with the 45° line through the origin <ns0:ref type='bibr' target='#b16'>(Lin 1989)</ns0:ref>. This assessment also implied that predictions were relatively consistent across the evaluation period and did not vary substantially on a seasonal basis. Moreover, it implied that the TPS interpolation was more suited to predicting T a in autumn which tended to exhibit superior statistics across all validation measures when compared to the other seasons. This was particularly evident regarding R 2 and P c , which registered the highest values of 0.91 and 0.94, respectively. However, TPS predictions tended to be least accurate in spring and winter which had MAE and RMSE values greater by 0.1°C, when compared to the corresponding MAE and RMSE values in autumn. Interestingly, although spring exhibited comparatively inferior MAE and RMSE values, the R 2 statistics were similar, both registering 0.91. This suggests that while errors were comparatively larger in spring, they were still very highly correlated to the validation data. However, it should be noted that the coefficient of determination may have been unrealistically overestimated for spring since the seasonal data signal was not removed prior to analysis, as advocated in <ns0:ref type='bibr' target='#b11'>Jeffrey et al. (2001)</ns0:ref>.</ns0:p><ns0:p>When looking at the histogram distribution of the MAE and RMSE it was apparent that spring and winter had a notable proportion of AWS sites that exhibited values above 2°C (Fig. <ns0:ref type='figure'>3</ns0:ref>). This contributed to the inflated error values shown in <ns0:ref type='table' target='#tab_0'>1</ns0:ref>).</ns0:p><ns0:p>When viewing the errors spatially, it was clear that most of the larger interpolation errors transpired in regions where there was a lack of neighbouring AWS sites (Fig. <ns0:ref type='figure' target='#fig_3'>4</ns0:ref>). Specifically, the central and western interior parts of Australia tended to exhibit MAE and RMSE values above 2°C, compared to the eastern half where temperatures were consistently predicted within 2°C of the actual T a . Of note was the predominately high errors encountered for the coastal areas of Western Australia (between Geraldton and Port Hedland) during summer and spring where the MAE and RMSE prediction errors regularly exceeded 2.5°C. For example, the Learmonth Airport AWS site (Fig. <ns0:ref type='figure' target='#fig_1'>2</ns0:ref> In terms of the R 2 and P c , low values tended to emanate along the coastal regions, particularly for Western Australia, Northern Territory and North Queensland coastal regions and neighbouring islands (Fig. <ns0:ref type='figure' target='#fig_3'>4</ns0:ref>). For example, the lowest values were encountered for sites Pirlangimpi Airport, Browse Island and Coconut Island in spring with R 2 of 0.01, 0.1 and 0.1, respectively, and P c of 0.09, 0.25 and 0.23, respectively. Summer also encountered low R 2 of 0.05, 0.05 and 0.31, respectively, with corresponding P c of 0.18, 0.2 and 0.52. The same sites in winter also had the lowest R 2 of 0.41, 0.01 and 0.14, respectively, along with corresponding P c of 0.48, 0.08 and 0.27.</ns0:p><ns0:p>When observing MAE (Fig. <ns0:ref type='figure' target='#fig_4'>5</ns0:ref>) and RMSE (Fig. <ns0:ref type='figure'>6</ns0:ref>) over a 24-hour period, it was clear that the high values encountered for the coastal areas of Western Australia in summer and spring tended to occur during afternoons. Specifically, these had MAE and RMSE ranging between 4-6°C for times 3 pm to 6 pm, i.e. 1 pm to 4 pm, Australian Western Standard Time (AWST). Of note was the Learmonth Airport AWS site registering MAE of 6.9°C and RMSE of 7.6°C, peaking at 5 pm (3 pm, AWST) in summer (Fig. <ns0:ref type='figure'>7</ns0:ref>). Similarly, very high error values were encountered for the south-eastern area of Western Australia in spring, notably for the Forrest AWS site at 6 pm, which registered 6.1°C and 6.3°C for MAE and RMSE, respectively. This was in addition to 5.3°C and 7°C, respectively, for the same site in summer, along with the Ceduna AWS site (South Australia) at 6 pm, registering high MAE and RMSE values of 5.4°C and 6°C, respectively. During winter the trend for high MAE and RMSE emanating from central Australia and the coastal fringes of Northern Territory and Western Australia tended to occur during early mornings from 3 am to 9 am (1 am to 7 am, AWST), with MAE and RMSE ranging 3-5°C. The AWS sites with the greatest error in these parts were Adele Island and Yampi Sound which both registered a MAE of 6.4°C and RMSE of 6.4°C and 6.7°C, respectively (Fig. <ns0:ref type='figure' target='#fig_4'>5 & 6</ns0:ref>). Both sites are located in the northern coastal region of Western Australia (NB: the locations of all aforementioned AWS sites are depicted in Fig. <ns0:ref type='figure' target='#fig_1'>2</ns0:ref>).</ns0:p><ns0:p>Regarding R 2 and P c , it was revealed that low values <0.5 were mostly evident during late nights and early mornings, particularly during winter and summer (Fig. <ns0:ref type='figure'>8 & 9</ns0:ref>). This tended to be prevalent throughout the central western interior and the coastal fringes of Western Australia, Northern Territory and North Queensland. Specifically, R 2 and P c <0.5 tended to occur from midnight through to 6 am, suggesting that the TPS predictions were not highly correlated with the validation data during these times. This occurred despite the RMSE and MAE registering values below 2°C for some sites. For example, the AWS site at Browse Island during summer at 12 am registered R 2 and P c values of 0.04 and 0.2, respectively, while MAE and RMSE was 0.9°C and 1.1°C, respectively (Fig. <ns0:ref type='figure'>7</ns0:ref>). This was also encountered for the Coconut Island AWS during winter mornings, e.g. at 9 am where the R 2 and P c values registered 0.0 and 0.1, respectively, while the MAE and RMSE was 1.4°C and 1.7°C, respectively. This suggested that while predictions were reasonably accurate, there was little to no correlation with the validation data.</ns0:p></ns0:div>
<ns0:div><ns0:head>Assessment of mapping T a in near real-time</ns0:head><ns0:p>The TPS methodology was applied to mapping T a in real-time at 30-minute intervals over a 21day period from 1 June 2020 to 21 June 2020. This exercise resulted in 1007 maps being produced which aligned to the total number of 30-minute processing intervals in the trial period; confirming all possible maps were successfully processed. On analysing the map completion times, the majority of the maps were completed at 28-minutes (Fig. <ns0:ref type='figure' target='#fig_0'>10</ns0:ref>). Specifically, 410 and 414 maps were produced for their respective 0-and 30-minute processing intervals. This corresponded directly to the AWS import times (Fig. <ns0:ref type='figure' target='#fig_0'>11</ns0:ref>), with the same proportion of AWS observations reaching the 480-observation threshold import limit at the 15-minute mark; thereby permitting T a mapping to commence. Thus, import times that occurred at 15-minutes, equated to resulting maps being completed at 28-minutes from the AWS observation time. From this, it can be deduced that on all occasions the map processing time was 13-minutes, regardless of the interval being processed. It should be noted that on 35 occasions the 480-observation threshold limit was not reached, resulting in maps -that did not meet this criterion -being produced at the 30-minute mark. This equated to 14 and 21 maps produced at the 0-and 30-minute processing intervals, respectively.</ns0:p></ns0:div>
<ns0:div><ns0:head>Discussion</ns0:head></ns0:div>
<ns0:div><ns0:head>Appraisal of the TPS interpolation procedure</ns0:head><ns0:p>On the whole, the TPS interpolation method was a reliable predictor of T a across Australia with an RMSE of 1.65°C, i.e. when averaged across the seasons (Table <ns0:ref type='table' target='#tab_0'>1</ns0:ref>). When compared to previous studies, this error was similar to <ns0:ref type='bibr' target='#b11'>Jeffrey et al. (2001)</ns0:ref> with RMSE of 1.5°C and 1.9°C for daily maximum and minimum temperatures, respectively; and <ns0:ref type='bibr' target='#b13'>Jones et al. (2009)</ns0:ref> with corresponding RMSE of 1.2°C and 1.7°C. On a seasonal basis the TPS predictions tended to be least accurate in spring and winter which had MAE and RMSE values larger by 0.1°C compared to the same measures in autumn. When viewing these errors spatially, it was clear that the majority of the larger interpolation errors transpired in the central and western interior parts of Australia. This is unsurprising given the station density in these parts are relatively sparse in addition to large temperature variances which tend to produce inflated errors <ns0:ref type='bibr' target='#b11'>(Jeffrey et al. 2001;</ns0:ref><ns0:ref type='bibr' target='#b12'>Jones & Trewin 2000)</ns0:ref>. Of note was the predominately high errors encountered for the coastal areas of Western Australia (between Geraldton and Port Hedland) during summer and spring afternoons where prediction errors were regularly above 2.5°C. This was in addition to high MAE values for individual AWS sites located at Forrest in Western Australia and Ceduna in South Australia. Collectively, these regions tend to experience very strong temperature gradients, particularly concerning maximum temperatures, since their proximity between the coast and inland deserts result in local climate regimes being invariably affected by the relatively cool ocean to the west and hot desert interior to the east <ns0:ref type='bibr' target='#b13'>(Jones et al. 2009)</ns0:ref>. These are increasingly difficult to model with a sparse network of observation sites since these errors tended to be amplified during mid to late afternoons in late spring and summer when the temperature gradients were at their peak. Also, temperatures in these areas vary considerably over short periods leading to a tendency for larger errors <ns0:ref type='bibr' target='#b12'>(Jones & Trewin 2000)</ns0:ref>.</ns0:p><ns0:p>Concerning winter, the trend for high MAE and RMSE in central Australia and coastal fringes of Northern Territory and Western Australia tended to occur during early mornings from 3 am to 9 am (1 am to 7 am, AWST). As acknowledged previously, the accuracy of the mapping was limited in these regions due to an insufficient network of AWS sites. Also, AWS sites in the coastal fringes tend to have tight climate gradients as a result of local maritime effects <ns0:ref type='bibr' target='#b13'>(Jones et al. 2009)</ns0:ref>. This possibly contributed to the low R 2 and P c values encountered in Fig. <ns0:ref type='figure' target='#fig_3'>4</ns0:ref>, 8 & 9despite some AWS sites exhibiting relatively small MAE and RMSE values, e.g. Browse Island and Coconut Island AWS sites (Fig. <ns0:ref type='figure'>7</ns0:ref>). Thus, the predictions tended to be highly variable over a narrower prediction range due to the tighter temperature gradient in these climates. Combined with a sparse network of AWS sites, the TPS method was unable to account for this on a subhourly timescale. Moreover, the spread of AWS sites in remote coastal locations -e.g. Adele Island, Yampi Sound and Pirlangimpi Airport AWS sites -tend to have considerably larger errors as a result of unique and often complex microclimates, thereby compounding the variability <ns0:ref type='bibr' target='#b13'>(Jones et al. 2009)</ns0:ref>. It should also be noted that the larger errors for the central interior parts of Australia may also be due to the weaker link between altitude and temperature -for which the TPS algorithm is reliant <ns0:ref type='bibr' target='#b7'>(Hutchinson 1991)</ns0:ref>. This is because minimum temperatures have a highly variable and complex relationship with topography for which elevation and its association with lapse rates are only one part <ns0:ref type='bibr' target='#b25'>(Rolland 2003;</ns0:ref><ns0:ref type='bibr' target='#b28'>Trewin 2005)</ns0:ref>. Considering minimum temperatures tend to transpire during early mornings -as encountered for AWS sites in winter (Fig. <ns0:ref type='figure' target='#fig_4'>5</ns0:ref> & 6) -a multivariate approach to modelling might be more appropriate along with a denser network of AWS sites. This approach was conducted by <ns0:ref type='bibr' target='#b31'>Webb et al. (2020)</ns0:ref>, and showed errors improved during winter when using regression tree interpolation (along with multiple terrain and satellite covariate datasets). However, the substantially longer processing times may not be appropriate for real-time application, negating its ability to produce outputs in a timely manner as required for this study. Similar experiments contrasting TPS, ordinary kriging and inverse distance weighting interpolation found that TPS was more accurate and required few guiding covariates <ns0:ref type='bibr' target='#b10'>(Jarvis & Stuart 2001b)</ns0:ref>. This justified the selection of the TPS method in the current study, even though kriging can be an equally effective method <ns0:ref type='bibr' target='#b7'>(Hutchinson 1991</ns0:ref>). However, kriging requires considerable computational overhead <ns0:ref type='bibr' target='#b10'>(Jarvis & Stuart 2001b</ns0:ref>) and therefore, in the context of this study, not ideal for real-time application.</ns0:p><ns0:p>It should be commented that the cross-validation analysis adopted in this study would likely overestimate the error since predictions were made at locations that have actual data observations. This would be less of a concern for regions where the number of observation points is numerous, such as for the majority of land areas in south-east Australia -which tended to have more accurate T a predictions compared to the western interior. Nevertheless, this exemplifies that the sparse network of AWS sites in central and western coastal areas of Australia was a notable factor contributing to larger interpolation errors. It should be further commented that while the cross-validation analysis was valid using a static dataset, in reality and as exemplified during real-time application, interpolation could only commence when the predefined threshold of 480 AWS observations was met. Thus, on most occasions' predictions were based on the minimum allowable number of AWS sites and therefore prone to produce less accurate predictions compared to using an entire dataset. To evaluate this scenario, a K-fold cross validation was implemented <ns0:ref type='bibr' target='#b5'>(Hastie et al. 2009)</ns0:ref>. Specifically, the training dataset -represented by AWS observations in each 30-minute interval (h) within the evaluation period -was split into K equal parts using random sampling. Where the Kth part was kept for validation and the remaining K-1 part were combined for TPS modelling in each fold. In this manner, the predictions produced by the modelling were assessed against the held back validation subset and was repeated K times, such that each K validation subset was used once to assess the K-1 model. In this study, K = 10 was specified, representing 90% of the dataset for TPS modelling and 10% for validation in each fold, i.e. equivalent to 480 and 54 AWS sites, respectively.</ns0:p><ns0:p>The result of the K-fold analysis (Table <ns0:ref type='table' target='#tab_3'>2</ns0:ref>) revealed broad similarities with Table <ns0:ref type='table' target='#tab_0'>1</ns0:ref>. For example, the MAE of 1.2°C, 1.2°C, 1.3°C and 1.3°C, respectively for summer, autumn, winter and spring, and corresponding RMSE of 1.7°C, 1.6°C, 1.7°C and 1.8°C, respectively, deviated by no more than 0.1°C when compared to the equivalent measures in Table <ns0:ref type='table' target='#tab_0'>1</ns0:ref>. This indicated that overall the prediction accuracy did not deteriorate greatly when interpolation was based on the minimum allowable number of AWS sites. It also justified that the 480 AWS threshold adopted in this study was an acceptable limit for the real-time application.</ns0:p></ns0:div>
<ns0:div><ns0:head>Appraisal of mapping T a in near real-time and application to digital mapping</ns0:head><ns0:p>The TPS interpolation applied in real-time was capable of producing sub-hourly T a maps typically within 28-minutes of the observation being recorded by the available AWS sites (Fig. <ns0:ref type='figure' target='#fig_0'>10</ns0:ref>). Specifically, import times were generally reached for the predefined threshold of 480 observations at the 15-minute mark (Fig. <ns0:ref type='figure' target='#fig_0'>11</ns0:ref>) which was followed by a 13-minute processing lag. In this regard, maps were consistently available within their 30-minute processing window and had a high degree of temporal reliability -with all possible maps produced in the 21-day trial period. The resulting maps were presented on a digital web mapping platform to allow real-time access and interrogation ability of each output. An example of this application can be accessed at URL http://austemperature.live/ (Fig. <ns0:ref type='figure' target='#fig_1'>12</ns0:ref>). A GeoServer backend was used to host current outputs to allow geospatial representation and sharing of outputs via a Wep Map Service (Open Source Geospatial Foundation 2019). The maps can be spatially queried to reveal temperatures for the current hour and for the previous 3-hrs (at 30-minute intervals). This is enabled via web application packages shiny and leaflet <ns0:ref type='bibr' target='#b3'>(Chang et al. 2019;</ns0:ref><ns0:ref type='bibr' target='#b4'>Cheng et al. 2019</ns0:ref>) within the R programming environment (R Development Core Team 2015). In this fashion, maps can be spatially interrogated via an on-the-fly 'data drilling' for any geographical location in Australia (via mouse click). A facility to view the cross-validation statistics of each map output is also provided as well as the ability to download each newly created map for use in GIS applications.</ns0:p><ns0:p>A potential new feature is to provide an error map for each subsequent map produced (similar to <ns0:ref type='bibr'>Fig. 5 & 6)</ns0:ref>. This would provide an approximate error measure for regions with limited AWS sites which tended to be high, as encountered in this study. Note that rainfall mapping outputs are also presented in the application, although this should be used with caution due to the preliminary nature of this work.</ns0:p></ns0:div>
<ns0:div><ns0:head>Conclusions</ns0:head><ns0:p>The methods described in this study were successful for operational real-time spatial mapping of T a at high spatiotemporal across Australia. The TPS interpolation method was best suited for mapping T a during autumn and was comparatively less accurate during winter and spring. In particular, areas, where there was a lack of AWS sites, tended to underperform. These areas included the central and western interior regions of Australia, as well for the north-west coastal areas of Western Australia and parts of the Northern Territory coastline. On a temporal basis, the errors were amplified during the afternoons, particularly around the coastal regions of Western Australia, during spring and summer. In winter, errors tended to be higher in central Australia and the coastal fringes of Northern Territory and Western Australia, from 3 am to 9 am. In terms of applying the TPS method to real-time operational mapping, the mapping system was able to regularly provide spatial outputs within 28-minutes of AWS site observations being recorded. In addition, it also had a high degree of temporal reliability with all maps produced in the 21-day trial period. Outputs were sequentially displayed on purpose-built web mapping application to exemplify real-time application of the outputs. In this regard, the methodology employed in this study would be highly suited for similar applications requiring real-time processing and delivery of climate data at high spatiotemporal resolutions across a large landmass, suitably complimented with a relatively dense network of observation sites. Manuscript to be reviewed Manuscript to be reviewed Manuscript to be reviewed Manuscript to be reviewed </ns0:p></ns0:div><ns0:figure xml:id='fig_0'><ns0:head>Figure 1</ns0:head><ns0:label>1</ns0:label><ns0:figDesc>Figure 1</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_1'><ns0:head>Figure 2</ns0:head><ns0:label>2</ns0:label><ns0:figDesc>Figure 2</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_2'><ns0:head /><ns0:label /><ns0:figDesc>Figure 3</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_3'><ns0:head>Figure 4</ns0:head><ns0:label>4</ns0:label><ns0:figDesc>Figure 4</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_4'><ns0:head>Figure 5 Interpolated</ns0:head><ns0:label>5</ns0:label><ns0:figDesc>Figure 5</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_5'><ns0:head /><ns0:label /><ns0:figDesc>Figure 8</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_6'><ns0:head>(</ns0:head><ns0:label /><ns0:figDesc>Figure 10</ns0:figDesc></ns0:figure>
<ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='22,42.52,377.62,525.00,214.50' type='bitmap' /></ns0:figure>
<ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='23,42.52,334.79,525.00,395.25' type='bitmap' /></ns0:figure>
<ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='26,42.52,224.62,525.00,296.25' type='bitmap' /></ns0:figure>
<ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='27,42.52,250.12,525.00,325.50' type='bitmap' /></ns0:figure>
<ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='28,42.52,227.32,525.00,322.50' type='bitmap' /></ns0:figure>
<ns0:figure type='table' xml:id='tab_0'><ns0:head>Table 1 .</ns0:head><ns0:label>1</ns0:label><ns0:figDesc>Specifically, spring and winter both had a total of 42 and 46 AWS sites that registered MAE above 2°C, compared to 22 and 16 AWS sites for summer and autumn, respectively. Similarly, spring and winter also had a large proportion of RMSE values above 2°C with 147 and 134 AWS sites, respectively, compared to 95 and 91 AWS sites for summer and autumn, respectively. In regard to R 2 and P c , winter had a greater proportion of AWS sites that exhibited moderate to weak correlation (≤0.7) with 45 and 32 sites, respectively; compared to 37 and 17 in summer, 11 and 10 in spring, and 13 and 8 in autumn. The high proportion of low R 2 values in winter contributed to the lowest R 2 value of 0.86, compared to 0.89, 0.91 and 0.91 for summer, autumn and spring, respectively (Table</ns0:figDesc><ns0:table /></ns0:figure>
<ns0:figure type='table' xml:id='tab_1'><ns0:head /><ns0:label /><ns0:figDesc>) in spring had MAE and RMSE of 3.4°C and 4.3°C, respectively, in addition to summer with corresponding MAE and RMSE of 3.2°C and 4.2°C, respectively. Outside of this cluster, there were also high MAE and RMSE values for individual AWS sites located at Pirlangimpi Airport (Tiwi Islands, Northern Territory) in spring with 3.6°C and 4.3°C, respectively; Forrest in Western Australia during summer with corresponding MAE and RMSE of 2.9°C and 3.8°C, respectively; and Yampi Sound in the Northern Territory during winter with MAE and RMSE of 3.3°C and 4.3°C, respectively. Furthermore, in winter there was a notable cluster of high MAE values emanating from central Australia through to the coastal fringes of Northern Territory and Western Australia (i.e. Darwin through to Broome) with MAE and RMSE consistently above 2°C.</ns0:figDesc><ns0:table /></ns0:figure>
<ns0:figure type='table' xml:id='tab_2'><ns0:head>Table 1 (on next page)</ns0:head><ns0:label>1</ns0:label><ns0:figDesc>Validation statistics for the TPS interpolation procedure showing R 2 , P c , MAE (°C) and RMSE (°C) values -averaged for each AWS site according to the season. sd standard deviation, min minimum, max maximum PeerJ reviewing PDF | (2020:07:50683:1:1:NEW 13 Sep 2020)Manuscript to be reviewed</ns0:figDesc><ns0:table><ns0:row><ns0:cell /><ns0:cell cols='2'>Summer Autumn</ns0:cell><ns0:cell>Winter</ns0:cell><ns0:cell>Spring</ns0:cell></ns0:row><ns0:row><ns0:cell>R 2</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>mean</ns0:cell><ns0:cell>0.89</ns0:cell><ns0:cell>0.91</ns0:cell><ns0:cell>0.86</ns0:cell><ns0:cell>0.91</ns0:cell></ns0:row><ns0:row><ns0:cell>min</ns0:cell><ns0:cell>0.05</ns0:cell><ns0:cell>0.02</ns0:cell><ns0:cell>0.01</ns0:cell><ns0:cell>0.01</ns0:cell></ns0:row><ns0:row><ns0:cell>max</ns0:cell><ns0:cell>0.99</ns0:cell><ns0:cell>0.99</ns0:cell><ns0:cell>0.97</ns0:cell><ns0:cell>0.99</ns0:cell></ns0:row><ns0:row><ns0:cell>sd</ns0:cell><ns0:cell>0.11</ns0:cell><ns0:cell>0.09</ns0:cell><ns0:cell>0.11</ns0:cell><ns0:cell>0.09</ns0:cell></ns0:row><ns0:row><ns0:cell>P c</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>mean</ns0:cell><ns0:cell>0.92</ns0:cell><ns0:cell>0.94</ns0:cell><ns0:cell>0.92</ns0:cell><ns0:cell>0.93</ns0:cell></ns0:row><ns0:row><ns0:cell>min</ns0:cell><ns0:cell>0.18</ns0:cell><ns0:cell>0.14</ns0:cell><ns0:cell>0.18</ns0:cell><ns0:cell>0.09</ns0:cell></ns0:row><ns0:row><ns0:cell>max</ns0:cell><ns0:cell>0.99</ns0:cell><ns0:cell>0.99</ns0:cell><ns0:cell>0.99</ns0:cell><ns0:cell>0.99</ns0:cell></ns0:row><ns0:row><ns0:cell>sd</ns0:cell><ns0:cell>0.09</ns0:cell><ns0:cell>0.08</ns0:cell><ns0:cell>0.09</ns0:cell><ns0:cell>0.08</ns0:cell></ns0:row><ns0:row><ns0:cell>MAE</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>mean</ns0:cell><ns0:cell>1.2</ns0:cell><ns0:cell>1.2</ns0:cell><ns0:cell>1.3</ns0:cell><ns0:cell>1.3</ns0:cell></ns0:row><ns0:row><ns0:cell>min</ns0:cell><ns0:cell>0.5</ns0:cell><ns0:cell>0.5</ns0:cell><ns0:cell>0.5</ns0:cell><ns0:cell>0.6</ns0:cell></ns0:row><ns0:row><ns0:cell>max</ns0:cell><ns0:cell>3.2</ns0:cell><ns0:cell>2.8</ns0:cell><ns0:cell>3.3</ns0:cell><ns0:cell>3.6</ns0:cell></ns0:row><ns0:row><ns0:cell>sd</ns0:cell><ns0:cell>0.4</ns0:cell><ns0:cell>0.4</ns0:cell><ns0:cell>0.5</ns0:cell><ns0:cell>0.5</ns0:cell></ns0:row><ns0:row><ns0:cell>RMSE</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>mean</ns0:cell><ns0:cell>1.6</ns0:cell><ns0:cell>1.6</ns0:cell><ns0:cell>1.7</ns0:cell><ns0:cell>1.7</ns0:cell></ns0:row><ns0:row><ns0:cell>min</ns0:cell><ns0:cell>0.6</ns0:cell><ns0:cell>0.7</ns0:cell><ns0:cell>0.7</ns0:cell><ns0:cell>0.8</ns0:cell></ns0:row><ns0:row><ns0:cell>max</ns0:cell><ns0:cell>4.3</ns0:cell><ns0:cell>3.5</ns0:cell><ns0:cell>4.3</ns0:cell><ns0:cell>4.3</ns0:cell></ns0:row><ns0:row><ns0:cell>sd</ns0:cell><ns0:cell>0.5</ns0:cell><ns0:cell>0.5</ns0:cell><ns0:cell>0.6</ns0:cell><ns0:cell>0.6</ns0:cell></ns0:row></ns0:table><ns0:note>1PeerJ reviewing PDF | (2020:07:50683:1:1:NEW 13 Sep 2020)</ns0:note></ns0:figure>
<ns0:figure type='table' xml:id='tab_3'><ns0:head>Table 2 (on next page)</ns0:head><ns0:label>2</ns0:label><ns0:figDesc>K-fold cross-validation statistics for the TPS interpolation procedure showing R</ns0:figDesc><ns0:table><ns0:row><ns0:cell>2 , P c , MAE</ns0:cell></ns0:row></ns0:table><ns0:note>PeerJ reviewing PDF | (2020:07:50683:1:1:NEW 13 Sep 2020)Manuscript to be reviewed</ns0:note></ns0:figure>
<ns0:note place='foot'>PeerJ reviewing PDF | (2020:07:50683:1:1:NEW 13 Sep 2020)Manuscript to be reviewed</ns0:note>
<ns0:note place='foot' n='3'>pm. (EE) summer, 6 pm. (FF) summer, 9 pm. PeerJ reviewing PDF | (2020:07:50683:1:1:NEW 13 Sep 2020)Manuscript to be reviewed</ns0:note>
<ns0:note place='foot' n='3'>pm. (EE) summer, 6 pm. (FF) summer, 9 pm. PeerJ reviewing PDF | (2020:07:50683:1:1:NEW 13 Sep 2020)</ns0:note>
<ns0:note place='foot' n='3'>pm. (EE) summer, 6 pm. (FF) summer, 9 pm. PeerJ reviewing PDF | (2020:07:50683:1:1:NEW 13 Sep 2020)Manuscript to be reviewed</ns0:note>
<ns0:note place='foot' n='3'>pm. (EE) summer, 6 pm. (FF) summer, 9 pm. PeerJ reviewing PDF | (2020:07:50683:1:1:NEW 13 Sep 2020)Manuscript to be reviewed</ns0:note>
</ns0:body>
" | "Reviewer 1
The criteria for not rejecting an AWS (lines 122-125) suggest that each iteration of the evaluation process the number fo stations actually participating in the TPS procedure would differ. Tamperatures at stations in regions with sparse coverage, could end up being estimated from data coming from quite distant areas, and thus presenting large estimation errors. Could the authors discuss the potential significance of this effect?
Narrowing down the validation period and AWSs used, to those 'overall commonly available' might not prove practical. However, an estimate might be obtained by employing an additional completeness criterion based on the percentage of 'active' nearby stations during each estimation step.
This could also be relevant to the discussion in lines 203-204 and perhaps lead to an additional comment after line 345 on the possible effect of a reduced number of stations available during the TPS interpolation procedure .
Thank you for your comment, but I’m a little unsure on what you’re asking. Are you suggesting that each AWS site have the proportion of active neighbouring AWS quantified in each evaluation step in order to gauge the accuracy of the resulting predictions (especially when there is a lack of neighbouring AWS sites in an evaluation step)? Employing this strategy is impractical to implement in our timeframe and would entail considerable computation and analysis.
Having said that, I agree with your comment and think it is worthwhile having some degree of measure for situations when predictions are based on the minimum allowable number of AWS sites, as stipulated in lines 202-204, (i.e. 480, or 90% of all available AWS recordings is the minimum allowable number sites that must be met before interpolation is allowed). In order to evaluate this, I have performed a K-fold cross validation (Hastie et al. 2009). Specifically, the training dataset (in each evaluation step) was split into K equalized parts by random sampling. For each fold, the Kth part was kept for validation and the remaining parts (K−1) were combined for training (i.e. TPS modelling). The predictions made by the training dataset are assessed against the held back validation data in each evaluation step. This process was repeated K times (folds) where each K subsample was used once to validate each K−1 model. K=10 was specified which is equivalent to 10% of the data held back for validation in each permutation. The results of this are now discussed in lines 393 to 406.
A very interesting article with great potential for application. Proper data handling and analysis, very well writen and presented. Apart from the comment in the 'Validity of the findings' section, just a few more, minor, comments/suggestions:
1. Air temperature is measured with an accuracy of one decimal so using two decimals to report the respective MAE might be considered an exaggeration. In that respect, an estimation of the combined uncertainty -instead of simply the MAE- of the estimations made by the method, might also be of interest.
I agree and have revised table 1 and other relevant figures so error values are rounded to 1 decimal. I have also expanded the result section by incorporating the other validation metrics (RMSE, R2 and Pc) and revised figure 3, 4 and 5 and added additional figures (6,7 & 8) to incorporate the additional validation measures (as requested). I have also accompanied this with additional commentary in the results and discussions sections. Refer lines 243 – 312 & 333 – 385.
2. There are but a few grammatical errors and missing/misused words (i.e. lines 38, 54, 123, 124, 171, 299)
Agreed and rectified accordingly.
3. There is a narration gap in lines 42-43, from generaly speaking about the dependence of (environmental) variables on geographical factors to the specifics of air temperature. Perhaps a bridging sentence could be added.
Agreed and have reworded this sentence to mention air temperature and its associated variability with space and time as well added additional references as recommended by another reviewer. Refer lines 43 – 46.
5. Please indicate how many AWS stations were discarded (line 125)
Now listed in the text (192 in total)
6. Concordance coefficient is indexed as Pc, pc (capital and small p, in the text) and CC in the raw data. Please choose one.
I have retained Pc and changed all instances accordingly in both the manuscript and raw data.
7. The setup of the mapping systems seemed clearly described (lines 196-208), however it was only after lines 288-291 that ts operation was really understood. This is a very important feature of the system. Perhaps the description could be improved and the rationalle behind the overall design, better explained.
I have added some additional text and referenced a previous paper to provide context. Refer lines 200 - 206
8. Line 210: perhaps the authors meant 'visualize'
Yes. Changed accordingly.
9. Line 243: less than 10% is hardly a large proportion
It’s the tail end of the distribution curve, but significant nonetheless given the portrayal of errors in later figures (particularly for RMSE – now included). Nonetheless, I have relabelled as ‘notable’.
10. Line 328: 'minimum' should be removed from here
Agreed and removed accordingly.
11. Raw data provided were easy to identify and use, except for the Excel version of figure 8 (my, older, version of Excel could not display that) and the MAE tiffs (evaluation\data\spatial\mae) that seem to be blank (instead of showing the plates of Figure 5).
Apologies for you not being able to see figure 8. Unfortunately, I am unable to save this as an older version due to the box plot feature being a recent addition to the latest excel release (or it has at least been altered compared to previous releases and therefore not viewable in older versions).
In regard to the MAE tiffs, I have only provided those that are relevant to figure 4. It was not practical for me to supply them for individual tiffs shown in figure 5 (and now 6, 8 & 9) due to the size of each tiff layer (~350Mb) which would have resulted in a total of 44GB required for upload and exhausted my dropbox allocation.
Reviewer 2
The authors applied the thin plate spline method to interpolation T2m that were observed at a sparse in situ meteorological monitoring network across Australia toward the creation of high resolution T2m maps. The paper is generally well written and the analysis sounds meaningful. However, the technique is a bit outdated given the available of more contemporary yet advanced data fusion and assimilation methods.
1.Line 43-44: The variations in temperature in space is also highly regulated by factors such as aerosols via direct radiative effect (Mitchell et al., 1995, doi:10.1175/1520-0442(1995)008<2364:OSTGGA>2.0.CO;2; Li et al. 2017, doi:10.1093/nsr/nwx117 ), clouds (Xue et al., 2019, doi:10.1007/s00382-018-4505-8), synoptic-scale circulation (Redmond and Koch 1991. doi: 10.1029/91WR00690; Domonkos et al. 2003, doi:10.1002/joc.929; Liu et al., 2018, doi: 10.1175/JCLI-D-17-0608.1), which is suggested to be mentioned at least.
Thank you for these interesting and useful references. I have incorporated them as suggested.
2.Line 67: thin plate smoothing spline performed poorly with limited observations, why didn’t have a try on the widely applied Kriging method?
Kriging was tested in the early stages of the analysis but was not adopted for further investigation for several reasons. Firstly, preliminary trials suggested that results were similar, or not much better than TPS. This was also encountered in studies by Hutchison (1991) [https://www.researchgate.net/publication/247765032_The_Application_of_Thin_Plate_Smoothing_Splines_to_Continent-Wide_Data_Assimilation ] and Jarvis and Stuart (2000) [DOI: 10.1175/1520-0450(2001)040<1075:ACASFI>2.0.CO;2]. Secondly, kriging tended to have longer processing times compared to TPS and therefore less suitable for real-time application (the computational overhead associated with kriging was also reported in Jarvis and Stuart (2000)). And thirdly, variogram estimation, in some circumstances, required manual intervention in identifying an appropriate form - which is not ideal in an automated setting as developed for this study (although this could be rectified with further adjustment to the code). Overall, due to the simplicity of TPS algorithm as well as the speed and relatively good accuracy of the resulting maps, it was more adapt for real-time application. I have now explained this reasoning in more detail in lines 380 - 385.
3.Line 128: as per your description, air temperature should be the commonly applied 2m temperature which should be denoted as T2m to be in line with the common usage in meteorology. Please change the usage throughout the paper.
Unfortunately, I must respectively disagree on this front. The reason - all AWS sites are measured at 1.2m above ground (refer: http://www.bom.gov.au/climate/cdo/about/airtemp-measure.shtml). Denoting this as T2M would be misleading to readers, I would think.
4.Methodology: currently, especially at the time where both satellite observations and meteorological reanalysis are widely applied, creating a gridded field of geophysical parameters simply using a 2D interpolation method seems to be out of date and less accurate, though it could be a timely way. A popular method is to fuse/assimilate the in situ observations with a given background field either from satellite observations or model output since they may provide more accurate estimates over remote regions with limited or even no in situ observations.
I wholeheartedly agree that there are more modern methods of incorporating multivariate data into the analysis such as assimilating satellite data and utilising machine learning. The latter of which was investigated in a previous paper (DOI: 10.1007/s00704-020-03259-4). The problem with these methods, as you alluded to, was due to increased processing times. TPS is a comparatively fast method and able to produce a reasonable result in a timelier manner. For example, we tested regression tree modelling in conjunction to assimilating multiple covariate datasets to map air temperature across Australia (similar to TPS used in the current study). We found that processing time increased by 45 minutes (compared to 13 minutes with TPS). For this reason, machine learning methods were overlooked in this paper. However, since you raised this important point, I have mentioned this in the manuscript and provided additional commentary in lines 373 – 380.
5.Line 248-249: As stated by the authors, larger interpolation errors would be introduced in regions with limited neighboring AWS sites. How to deal with this drawback, or just providing estimations with large uncertainty?
Within the example application (URL http://austemperature.live/) we have provided some measure of the uncertainty for each map that is produced (e.g. if you click on cross validation stat link). A possible way of improving this is to provide an error map to accompany each temperature map produced. This can be done in a similar way to what was produced in figures 4, 5 & 6, albeit on an hourly basis. Implementing this feature would provide a spatial measure of the uncertainty in regions with limited AWS sites – although it would increase the processing time to create this measure. Nonetheless, I have mentioned this potential in lines 434– 436.
6.Line 263–264: why large errors were observed in the coastal regions during afternoons in summer and spring? Is this due to instrumental bias or region-specific anomaly? Please try to give some explanation.
This was explained in lines 312-317. However, it probably does deserve some extra attention. As such, I have provided additional commentary in lines 347 – 354.
" | Here is a paper. Please give your review comments after reading it. |
9,738 | "<ns0:abstract xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:p>An understanding of local perceptions of carnivores is important for conservation and management planning. In the central Himalayas, Nepal, we interviewed 428 individuals from 85 settlements using a semi-structured questionnaire to quantitatively assess local perceptions and tolerance of snow leopards and wolves. We used generalized linear mixed effect models to assess influential factors, and found that tolerance of snow leopards was much higher than of wolves. Interestingly, having experienced livestock losses had a minor impact on perceptions of the carnivores. Occupation of the respondents had a strong effect on perceptions of snow leopards but not of wolves. Literacy and age had weak impacts on snow leopard perceptions, but the interaction among these terms showed a marked effect, i.e. being illiterate had a more marked negative impact among older respondents. Among the various factors affecting perceptions of wolves, numbers of livestock owned and gender were the most important predictors. People with larger livestock herds were more negative towards wolves. In terms of gender, males were more positive to wolves than females, but no such pattern was observed for snow leopards.</ns0:p><ns0:p>People's negative perceptions towards wolves were also related to the remoteness of the villages. Factors affecting people's perceptions could not be generalized for the two species, and thus need to be addressed separately. We suggest future conservation projects and programs should prioritize remote settlements.</ns0:p></ns0:div>
</ns0:abstract>
<ns0:body xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:head>INTRODUCTION</ns0:head><ns0:p>Large carnivore co-existence with humans remains a global challenge <ns0:ref type='bibr' target='#b5'>(Athreya et al., 2013)</ns0:ref>, and mitigation of human-carnivore conflicts requires multiple approaches and disciplines <ns0:ref type='bibr' target='#b67'>(Redpath et al., 2013)</ns0:ref>. Among the various aspects of carnivore conflict management, understanding local perceptions is crucial for establishing long term conservation strategies <ns0:ref type='bibr' target='#b6'>(Bagchi & Mishra, 2006;</ns0:ref><ns0:ref type='bibr' target='#b22'>Conforti & de Azevedo, 2003)</ns0:ref>, especially in multi-use landscapes where animal husbandry is the main source of income. An assessment of local perceptions helps in identifying groups of people or villages that are negative towards protection of carnivores, and thus aids conservation authorities to find suitable strategies to improve their tolerance <ns0:ref type='bibr' target='#b70'>(Suryawanshi, 2013;</ns0:ref><ns0:ref type='bibr' target='#b73'>Treves & Karanth, 2003)</ns0:ref>. Further, assessments form a basis for quantifying the effects of conservation management interventions and aid in formulating new strategies if opinions towards conservation change <ns0:ref type='bibr' target='#b29'>(Dressel et al., 2015)</ns0:ref>.</ns0:p><ns0:p>Globally, local perceptions and attitudes towards large carnivores are complex and vary markedly between regions <ns0:ref type='bibr' target='#b69'>(Røskaft et al., 2007)</ns0:ref>. Multiple factors influence local perceptions, including animal behavior, risk of negative encounters, and the length of the period of coexistence <ns0:ref type='bibr' target='#b23'>(Dickman, 2010;</ns0:ref><ns0:ref type='bibr' target='#b29'>Dressel et al., 2015;</ns0:ref><ns0:ref type='bibr' target='#b40'>Kellert et al., 1996;</ns0:ref><ns0:ref type='bibr'>Zimmermann et al., 2001)</ns0:ref>.</ns0:p><ns0:p>Local perceptions also vary among ethnic groups, and are linked to religious-and cultural beliefs <ns0:ref type='bibr' target='#b0'>(Ale et al., 2016;</ns0:ref><ns0:ref type='bibr' target='#b25'>Dickman et al., 2014;</ns0:ref><ns0:ref type='bibr' target='#b40'>Kellert et al., 1996;</ns0:ref><ns0:ref type='bibr' target='#b46'>Li et al., 2013;</ns0:ref><ns0:ref type='bibr' target='#b60'>Mkonyi et al.. 2017)</ns0:ref>.</ns0:p><ns0:p>Socio-demographic variables such as age, sex, income, occupation, literacy, number of livestock owned and loss to predators have all shown to be associated with local perceptions and attitudes of large carnivores <ns0:ref type='bibr' target='#b13'>(Caruso et al., 2020;</ns0:ref><ns0:ref type='bibr' target='#b30'>Fort et al., 2018;</ns0:ref><ns0:ref type='bibr' target='#b39'>Kellert & Berry, 1987;</ns0:ref><ns0:ref type='bibr' target='#b41'>Kideghesho et al., 2007;</ns0:ref><ns0:ref type='bibr' target='#b61'>Naughton-Treves et al., 2003;</ns0:ref><ns0:ref type='bibr' target='#b62'>Oli et al., 1994;</ns0:ref><ns0:ref type='bibr' target='#b69'>Røskaft et al., 2007;</ns0:ref><ns0:ref type='bibr' target='#b72'>Trajçe et al., 2019)</ns0:ref>.</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:04:48442:1:1:NEW 26 Jul 2020)</ns0:p><ns0:p>Manuscript to be reviewed</ns0:p><ns0:p>In the central Himalayas, Snow leopards (Panthera uncia) and Himalayan wolves (Canis lupus chanco) are the two most important predators involved in conflicts with people <ns0:ref type='bibr' target='#b19'>(Chetri et al., 2019a)</ns0:ref>. A recent study from the region revealed that snow leopards were responsible for the majority of predation losses (61.9%); the remaining were from Himalayan wolf (16.8%) and other predators (21.3%) including feral dogs (Canis lupus familiaris), brown bear (Ursus arctos), black bear (Ursus thibetanus), Eurasian lynx (Lynx lynx), golden jackal (Canis aureus) and common leopard (Panthera pardus) <ns0:ref type='bibr' target='#b19'>(Chetri et al., 2019a)</ns0:ref>. The snow leopard is categorized as Vulnerable in the IUCN red list of threatened species <ns0:ref type='bibr' target='#b52'>(McCarthy et al., 2017)</ns0:ref>, whereas wolf is considered as Least Concern <ns0:ref type='bibr' target='#b10'>(Boitani et al., 2018)</ns0:ref>. However, in the national Red Data List of Nepal, wolves are considered as Critically Endangered and snow leopards are considered as Endangered <ns0:ref type='bibr' target='#b37'>(Jnawali et al., 2011)</ns0:ref>. A recent fecal DNA study reported that the snow leopard density within our study area in the central Himalayas was 0.95 (SE 0.19) animals per 100 km 2 <ns0:ref type='bibr' target='#b20'>(Chetri et al., 2019b)</ns0:ref>, but density estimates of wolves from the area are still lacking. The species has received little conservation attention due to its lower conservation status in the IUCN Red list, which has made it difficult to acquire necessary funding for population monitoring.</ns0:p><ns0:p>However, according to <ns0:ref type='bibr' target='#b15'>Chetri (2014)</ns0:ref>, Himalayan wolves are rare in the region and mostly solitary. The wolves that thrive in this landscape are genetically unique to the region as revealed by recent DNA analysis, and they are considered different from the grey wolf lineage <ns0:ref type='bibr' target='#b18'>(Chetri et al., 2016)</ns0:ref>. Both species range widely and often encounter pastoralists.</ns0:p><ns0:p>Although information on livestock depredation by snow leopards and wolves exists from Nepal's Himalaya <ns0:ref type='bibr' target='#b4'>(Aryal et al., 2014;</ns0:ref><ns0:ref type='bibr' target='#b21'>Chetri et al., 2017;</ns0:ref><ns0:ref type='bibr' target='#b62'>Oli et al., 1994;</ns0:ref><ns0:ref type='bibr' target='#b76'>Wegge et al., 2012;</ns0:ref><ns0:ref type='bibr'>Werhahn et</ns0:ref> PeerJ reviewing PDF | (2020:04:48442:1:1:NEW 26 Jul 2020)</ns0:p><ns0:p>Manuscript to be reviewed al., 2019), limited information is available regarding variation in local perceptions and tolerance to these species on a large spatial scale <ns0:ref type='bibr' target='#b35'>(Hanson et al., 2019;</ns0:ref><ns0:ref type='bibr' target='#b44'>Kusi et al., 2019;</ns0:ref><ns0:ref type='bibr' target='#b62'>Oli et al., 1994)</ns0:ref>.</ns0:p><ns0:p>Hence, in our study, we examined local communities' perceptions of snow leopards and wolves in a large area of ~5000 km 2 where livestock depredation has been a main concern in recent decades. The survey covered two protected areas, Annapurna Conservation Area (ACA) and Manaslu Conservation Area (MCA), where ecological studies of snow leopard and wolf had recently been conducted <ns0:ref type='bibr' target='#b16'>(Chetri, 2018)</ns0:ref>. These studies showed that snow leopard density was far lower than previously assumed, and consequently, average annual livestock losses were low (ca.1%) even though livestock constituted large proportions of the diet of both snow leopards and wolves (ca.25%). Despite the low levels of livestock depredation, perceptions of wolves are often negative. Similarly, incidents of mass killings of livestock by snow leopards decreases local tolerance towards their conservation, which in turn may lead to retaliatory killing of carnivores <ns0:ref type='bibr' target='#b36'>(Jackson, 2015;</ns0:ref><ns0:ref type='bibr' target='#b57'>Mishra et al., 2016;</ns0:ref><ns0:ref type='bibr' target='#b59'>Mishra et al., 2004;</ns0:ref><ns0:ref type='bibr' target='#b71'>Suryawanshi et al., 2014;</ns0:ref><ns0:ref type='bibr' target='#b79'>Woodroffe et al., 2005)</ns0:ref>. Surplus killings and injuries of high valued livestock (e.g. horses, milking yaks and cows) not only outrage local communities <ns0:ref type='bibr' target='#b62'>(Oli et al., 1994)</ns0:ref>, but also have negative repercussion that can spread even to distant villages. Integrated conservation and development efforts that were initiated in ACA and MCA in the 1990-ies included conservation awareness campaigns principally targeting snow leopard, but not wolves. Due to the relatively low livestock loss and the considerable conservation efforts in the study area, we expected perceptions of carnivores to be more positive than reported from previous studies from the Himalayan range. Furthermore, we expected perceptions to vary geographically as well as between species due to a bias in the impact of conservation awareness Manuscript to be reviewed campaigns, i.e. tolerance of wolves should be lower, and perceptions could potentially depend on the remoteness of villages. Lastly, we expected perceptions to be affected by socioeconomic and demographic factors, e.g. livestock losses and ownership, gender, age, education and occupation.</ns0:p></ns0:div>
<ns0:div><ns0:head>MATERIAL AND METHODS</ns0:head></ns0:div>
<ns0:div><ns0:head>Ethics statement</ns0:head><ns0:p>Approval and relevant permits required to carry out this research were obtained from the National Trust for Nature Conservation (Ref.no.291), Nepal.</ns0:p></ns0:div>
<ns0:div><ns0:head>Study area</ns0:head><ns0:p>We conducted the study in the Annapurna Manaslu Conservation landscape in the central -Himalayas (N28 29, E83 85; Fig. <ns0:ref type='figure' target='#fig_2'>1</ns0:ref>). Both ACA and MCA are located within this landscape -and are the largest community-based conservation areas in Nepal (ca.9292 km 2 ). It is located in the rain shadow area of the Himalayas. Together with Bhimthang valley, it is the priority landscape for snow leopards conservation in the country (DNPWC 2017). The human population density is 1 per km 2 (CBS, 2012), and agro-pastoralism is the main source of livelihood, although some households are also involved in eco-tourism related entrepreneurs. The overall livestock density in the study area is 35.74 ± 0.10/km 2 <ns0:ref type='bibr' target='#b21'>(Chetri et al., 2017)</ns0:ref>. All accessible areas are used for livestock grazing following the seasonal traditional Tibetan calendar <ns0:ref type='bibr' target='#b19'>(Chetri et al., 2019a)</ns0:ref>.</ns0:p><ns0:p>Grazing areas are designated for seasonal grazing as summer and winter. Areas close to villages are used for year-round grazing (Fig. <ns0:ref type='figure' target='#fig_2'>1</ns0:ref>). Livestock, for example sheep, goats and cows are usually herded and periodically moved among different pastures according to seasons <ns0:ref type='bibr' target='#b21'>(Chetri et al., 2017)</ns0:ref>. Small stocks (sheep and goats) are herded and sometimes accompanied by herding dogs. They are released in the morning and brought back to corrals/pens in the afternoon on a</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:04:48442:1:1:NEW 26 Jul 2020)</ns0:p><ns0:p>Manuscript to be reviewed But in this study we will use VDC as the data was collected prior to this change.</ns0:p><ns0:p>Each VDC has separate designated grazing areas. Among the 21 VDCs in the study area (6621 km 2 ), 2934 km 2 (44.3%) was used for livestock grazing (summer 55.6%, winter 24.6% and 19.8% year-round). The remaining areas (ca.3687 km 2 ) were inaccessible for livestock grazing due to rugged terrain and high altitude (Fig. <ns0:ref type='figure' target='#fig_2'>1</ns0:ref>). Our survey covered 13% of the total number of households within the survey villages <ns0:ref type='bibr'>(CBS, 2012)</ns0:ref>. Due to scattered settlements/villages, vast landscape and remoteness of the area, most of the questionnaires were conducted using locally trained community members, managed through the Unit Conservation Offices (UCOs) of ACA and MCA <ns0:ref type='bibr' target='#b19'>(Chetri et al., 2019a)</ns0:ref>. Before the initiation of the survey, each interviewer was briefed about the purpose of the study and trained in how to conduct the semi-structured questionnaire, and verbal consent was obtained from all subjects. The survey households were selected following the main village trails. We approached the household closest to the main village trail and selected every third household thereafter for interviews. If the inhabitants were absent, we selected the nearest neighbor. For each respondent, we recorded the number of livestock owned, herd composition and livestock loss to snow leopards and wolves during the previous year. We also recorded respondent age, gender, education and occupation. We asked their opinion about</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:04:48442:1:1:NEW 26 Jul 2020)</ns0:p><ns0:p>Manuscript to be reviewed the presence of snow leopards and wolves near their grazing areas and homesteads, and categorized their answers as positive, neutral and negative. We considered questionnaires as invalid when respondents stated that they did not know about the species presence and conflict.</ns0:p><ns0:p>These questionnaires were excluded from the analyses. Hence, although we administered similar questionnaire sets to assess perceptions of snow leopards and wolves, the sample size for wolf perceptions became smaller due to a larger proportion of invalid questionnaires (Table <ns0:ref type='table' target='#tab_1'>1</ns0:ref>). This was mainly because wolves are found only in the northwestern section of ACA and MCA (see Fig. <ns0:ref type='figure' target='#fig_2'>1</ns0:ref>).</ns0:p></ns0:div>
<ns0:div><ns0:head>Data analysis</ns0:head><ns0:p>We used generalized linear mixed effects models (GLMMs) to determine the relationship between response and potential predictor variables with two separate sets of models, one for snow leopards and one for wolves. GLMMs take into account random effects and provide a more flexible approach for analyzing non-normal data <ns0:ref type='bibr' target='#b11'>(Bolker et al., 2009)</ns0:ref>. As a binomial response variable, we categorized opinions of presence of snow leopards and wolves as either positive or negative, as described previously. As explanatory variables, we included factors and covariates that were identified as important predictors of livestock losses in a previous study conducted in the region <ns0:ref type='bibr' target='#b19'>(Chetri et al., 2019a)</ns0:ref>, as well as demographic and socioeconomic variables that have been linked to perceptions of carnivores in previous studies ( see e.g. <ns0:ref type='bibr' target='#b13'>Caruso et al., 2020;</ns0:ref><ns0:ref type='bibr' target='#b39'>Kellert & Berry, 1987;</ns0:ref><ns0:ref type='bibr' target='#b44'>Kusi et al., 2019;</ns0:ref><ns0:ref type='bibr' target='#b62'>Oli et al., 1994;</ns0:ref><ns0:ref type='bibr' target='#b71'>Suryawanshi et al., 2014)</ns0:ref> Manuscript to be reviewed read and write), (vi) gender, (vii) age and (viii) distance to the nearest conservation office (number of walking days required to reach nearest conservation field office -standardized to 8 hrs. walking/day). We standardized all numeric explanatory variables by two standard deviations, following <ns0:ref type='bibr' target='#b31'>Gelman & Hill, (2007)</ns0:ref>. VDC was used as a random effect in all models.</ns0:p><ns0:p>We checked correlation among (continuous) predictors before carrying out the regression analysis, and we did not include collinear variables (rho > 0.6) into the same model. We analyzed the data using R version 3.4.2 (R CoreTeam, 2017).</ns0:p></ns0:div>
<ns0:div><ns0:head>RESULTS</ns0:head><ns0:p>We used only 395 and 327 questionnaires in our analyses regarding snow leopards and wolves, respectively (Table <ns0:ref type='table' target='#tab_1'>1</ns0:ref>), due to the exclusion of respondents that were neutral or unaware of species presence or conflicts. In terms of gender, more than 80% of the respondents were male (Table <ns0:ref type='table' target='#tab_1'>1</ns0:ref>), and approximately 50% of the respondents were illiterate. Among occupations, most respondents belonged to the agro-pastoralist category (Table <ns0:ref type='table' target='#tab_1'>1</ns0:ref>). An analysis of respondents' perceptions in general revealed that local people were more negative towards wolves (n=276, 84.4%) than towards snow leopards (n=209, 52.9%) (Table <ns0:ref type='table' target='#tab_1'>1</ns0:ref>). </ns0:p></ns0:div>
<ns0:div><ns0:head>Local perceptions of snow leopards</ns0:head><ns0:p>We compared 22 candidate models to assess perceptions of snow leopards (Table <ns0:ref type='table' target='#tab_3'>S1</ns0:ref>). The two highest ranking models had a small difference in AICc value (ΔAICc = 0.27) and Akaike</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:04:48442:1:1:NEW 26 Jul 2020)</ns0:p><ns0:p>Manuscript to be reviewed weights (0.45 and 0.39). Both models included the predictor variables occupation, sex and the interaction between age and literacy (Table <ns0:ref type='table' target='#tab_2'>2</ns0:ref>). The top ranking model in terms of AICc also included ownership, but due to the marginal effect of removing this variable, we present here the simpler second ranking model (Fig. <ns0:ref type='figure' target='#fig_3'>2</ns0:ref>). The occupation of the respondents had a strong effect on their perceptions of snow leopards. Among the three categories of occupation, there was only a slight difference in perceptions between agro-pastoralists (income from both agriculture and livestock) and herders (income solely from livestock herding). On the contrary, respondents with other additional sources of income (e.g. tourism) were far more positive towards snow leopards (i.e. OCCOTHER, Fig. <ns0:ref type='figure' target='#fig_3'>2</ns0:ref>). Furthermore, sex was included in the model, but the effect was weak, i.e. men were more positive towards snow leopards than women. The main effects of literacy and age had very weak impacts on perceptions, but the interaction between these terms showed a marked effect. As illustrated in Fig. <ns0:ref type='figure' target='#fig_3'>2</ns0:ref>, being literate had a more marked negative impact among the older respondents.</ns0:p></ns0:div>
<ns0:div><ns0:head>Local perceptions of wolves</ns0:head><ns0:p>We compared 24 candidate models to assess perceptions of wolves (Table <ns0:ref type='table' target='#tab_6'>S2</ns0:ref>). The highest ranking model performed far better than the other candidates (Akaike weight = 0.70, Table <ns0:ref type='table' target='#tab_2'>2</ns0:ref>).</ns0:p><ns0:p>This model included different predictors than the model for snow leopards, with one exception, i.e. sex (Fig. <ns0:ref type='figure' target='#fig_4'>3</ns0:ref>). In this case, male respondents were markedly more positive than females. Other predictor variables were livestock loss (numbers lost to wolves), herd composition, distance to the nearest conservation field office and livestock ownership. The latter predictor had a marked effect on perceptions, i.e. respondents with larger herds were more negative.</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:04:48442:1:1:NEW 26 Jul 2020)</ns0:p><ns0:p>Manuscript to be reviewed </ns0:p></ns0:div>
<ns0:div><ns0:head>DISCUSSION</ns0:head><ns0:p>In our study landscape, a far larger proportion of respondents were negative towards wolves than to snow leopards. This was also observed by <ns0:ref type='bibr' target='#b44'>Kusi et al., (2019)</ns0:ref> in upper Dolpa and Humla areas, located in the western region of Nepal. This perception is common in areas where wolves coexist with other large predators, for example brown bear and lynx <ns0:ref type='bibr' target='#b72'>(Trajçe et al., 2019)</ns0:ref>. A possible cause is related to the difference in behavior of wolves compared to other carnivores. Snow leopards are cryptic, avoid humans and are more nocturnal than wolves <ns0:ref type='bibr' target='#b54'>(McCarthy et al., 2005;</ns0:ref><ns0:ref type='bibr' target='#b55'>Mech & Boitani, 2010)</ns0:ref>. Wolves are more active during the day and attack livestock both during day and night <ns0:ref type='bibr' target='#b80'>(Xu et al., 2015)</ns0:ref>. Furthermore, research has shown that greater visibility and howling behavior of wolves may reinforce negative perceptions <ns0:ref type='bibr' target='#b40'>(Kellert et al., 1996)</ns0:ref>.</ns0:p><ns0:p>Social norms and cultural beliefs also play an important role in perceptions of the two carnivores.</ns0:p><ns0:p>Cultural sentiments, religious belief and folklore associated with snow leopards have a strong positive influence on their conservation <ns0:ref type='bibr' target='#b0'>(Ale et al., 2016;</ns0:ref><ns0:ref type='bibr' target='#b45'>Li et al., 2014)</ns0:ref>. Further, the local practice of non-violence (e.g. Tsum valley of MCA) and protection of forest and landscape in the name of monasteries <ns0:ref type='bibr' target='#b45'>(Li et al., 2014)</ns0:ref> have also played an important role in snow leopard</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:04:48442:1:1:NEW 26 Jul 2020)</ns0:p><ns0:p>Manuscript to be reviewed conservation. The Buddhist communities traditionally do not kill wildlife because it was considered a sin in their religion <ns0:ref type='bibr' target='#b45'>(Li et al., 2014)</ns0:ref>. The snow leopard is often considered as a symbol of the mountains, and the charisma of the species promotes attention both in terms of research and conservation efforts from global and national conservation authorities <ns0:ref type='bibr' target='#b53'>(McCarthy et al., 2016)</ns0:ref>. In contrast, wolves are traditionally depicted as merciless and evil creatures in legends and folklore <ns0:ref type='bibr' target='#b27'>(Dingwall, 2001;</ns0:ref><ns0:ref type='bibr' target='#b51'>Marvin, 2012)</ns0:ref>. A recent study from Spiti, India showed that more than 98% of the survey respondents claimed that wolves were not safe for livestock and their presence was highly disliked by the communities <ns0:ref type='bibr' target='#b48'>(Lyngdoh & Habib, 2019)</ns0:ref>. Similar trends have been observed in parts of Europe <ns0:ref type='bibr' target='#b27'>(Dingwall, 2001;</ns0:ref><ns0:ref type='bibr' target='#b51'>Marvin, 2012)</ns0:ref> and America <ns0:ref type='bibr' target='#b34'>(Grima et al., 2019)</ns0:ref>. This is not surprising as dislike for wolves is common across the globe <ns0:ref type='bibr' target='#b9'>(Bhatia et al., 2016;</ns0:ref><ns0:ref type='bibr' target='#b29'>Dressel et al., 2015;</ns0:ref><ns0:ref type='bibr' target='#b38'>Kansky et al., 2014;</ns0:ref><ns0:ref type='bibr' target='#b44'>Kusi et al., 2019;</ns0:ref><ns0:ref type='bibr' target='#b48'>Lyngdoh & Habib, 2019;</ns0:ref><ns0:ref type='bibr' target='#b71'>Suryawanshi et al., 2014)</ns0:ref>. The negative perceptions of the wolf in our study area thus probably due to fear and cultural bias as reported in many other studies <ns0:ref type='bibr' target='#b47'>(Linnell et al., 2002;</ns0:ref><ns0:ref type='bibr' target='#b64'>Prokop et al., 2011)</ns0:ref>.</ns0:p><ns0:p>In our study area, analysis of livestock depredation revealed higher losses from snow leopards compared to wolves <ns0:ref type='bibr' target='#b19'>(Chetri et al., 2019a)</ns0:ref>, but still the tolerance level of local communities towards snow leopards was higher. Tolerance to snow leopards in ACA has changed to become more positive compared with an earlier study <ns0:ref type='bibr' target='#b62'>(Oli et al., 1994)</ns0:ref>, probably as a result of continued efforts to increase awareness as part of an ongoing conservation program <ns0:ref type='bibr' target='#b28'>(DNPWC, 2017)</ns0:ref>. No such efforts have targeted wolves, or other coexisting carnivores in this area.</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:04:48442:1:1:NEW 26 Jul 2020)</ns0:p><ns0:p>Manuscript to be reviewed</ns0:p><ns0:p>Our model revealed that literacy, age, occupation, number of livestock owned and gender affected perceptions towards snow leopards and wolves. However, the predictors for the two species were different (see Fig. <ns0:ref type='figure' target='#fig_3'>2</ns0:ref> and 3), i.e. the latter two appeared in the best model for wolf perceptions and the former three for snow leopard. Regarding literacy and age, only the interaction between these terms had an influence on perceptions of snow leopards, but not the main effects. Being illiterate was associated with negative perceptions among older respondents.</ns0:p><ns0:p>Possibly, younger people had more exposure to snow leopard conservation campaigns, regardless of literacy. Several earlier studies have shown that older people are more negative towards large predators and usually less supportive of their conservation than the younger generation <ns0:ref type='bibr' target='#b8'>(Bencin et al., 2016;</ns0:ref><ns0:ref type='bibr' target='#b39'>Kellert & Berry, 1987;</ns0:ref><ns0:ref type='bibr' target='#b43'>Kleiven et al., 2004;</ns0:ref><ns0:ref type='bibr' target='#b69'>Røskaft et al., 2007;</ns0:ref><ns0:ref type='bibr' target='#b78'>Williams et al., 2002)</ns0:ref>.</ns0:p><ns0:p>Occupation influenced perceptions of snow leopards, i.e. people with sources of income other than animal husbandry were more positive. Elsewhere, it was also reported that people having smaller landholdings and few economic opportunities other than livestock herding are more negative towards snow leopards and wolves <ns0:ref type='bibr' target='#b6'>(Bagchi & Mishra, 2006;</ns0:ref><ns0:ref type='bibr' target='#b26'>Din et al., 2017)</ns0:ref>. In a study of jaguars (Panthera onca), <ns0:ref type='bibr' target='#b13'>Caruso et al., (2020)</ns0:ref> found a similar pattern; people's perceptions and attitudes were strongly influenced by occupation and economic benefits through ecotourism. In Ladakh, India snow leopard based ecotourism has become popular and provides income generation opportunities to the local communities <ns0:ref type='bibr' target='#b36'>(Jackson, 2015;</ns0:ref><ns0:ref type='bibr' target='#b50'>Maheshwari & Sathyakumar, 2019;</ns0:ref><ns0:ref type='bibr' target='#b74'>Vannelli et al., 2019)</ns0:ref>.</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:04:48442:1:1:NEW 26 Jul 2020)</ns0:p><ns0:p>Manuscript to be reviewed</ns0:p><ns0:p>Regarding perceptions of wolves, males were more positive than females. This pattern was also reported in earlier studies <ns0:ref type='bibr' target='#b39'>(Kellert & Berry, 1987;</ns0:ref><ns0:ref type='bibr' target='#b69'>Røskaft et al., 2007;</ns0:ref><ns0:ref type='bibr' target='#b71'>Suryawanshi et al., 2014)</ns0:ref>, and has been explained by women having less contact with conservation agencies <ns0:ref type='bibr' target='#b32'>(Gillingham & Lee, 1999</ns0:ref>).An other study suggested that the negative attitudes of women might be a result of greater perceived risk or fear <ns0:ref type='bibr' target='#b63'>(Prokop & Fančovičová, 2010)</ns0:ref>. As suggested by <ns0:ref type='bibr' target='#b44'>Kusi et al. (2019)</ns0:ref>, men in the Himalayas often migrate outside of villages for seasonal work and may thus have been more exposed to alternative attitudes to nature and conservation. In addition, men frequently venture into the pasture for livestock grazing activities and presumably had more encounters with wolves, which make them understand their behaviour and threats. High encounter rates with wolves either in the wild or in captivity, may promote more positive perceptions of the animals <ns0:ref type='bibr' target='#b3'>(Arbieu et al., 2020)</ns0:ref>.</ns0:p><ns0:p>People holding large livestock herds were more negative towards wolves, which agrees with a study from western China <ns0:ref type='bibr' target='#b80'>(Xu et al., 2015)</ns0:ref>. A possible explanation is that owners with larger herds have a higher risk of suffering losses in the central Himalayas, especially farmers having a larger herd of goat/sheep <ns0:ref type='bibr' target='#b21'>(Chetri et al., 2017)</ns0:ref>. It is, however, notable that having experienced losses did not affect perceptions of snow leopards, and the effect of perception on wolves was weak. This is in contrast with a recent study from the Nepal Himalayas <ns0:ref type='bibr' target='#b44'>(Kusi et al., 2019)</ns0:ref>, and may be due to the fact that average losses in our study area were quite low (~1% of all livestock holdings).</ns0:p><ns0:p>In our study area, the National Trust for Nature Conservation (NTNC) has been implementing community-based conservation projects and programs since 1992. The overall goal is to</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:04:48442:1:1:NEW 26 Jul 2020)</ns0:p><ns0:p>Manuscript to be reviewed conserve biodiversity of global significance with the active participation of local communities.</ns0:p><ns0:p>Integrated conservation and development efforts have therefore addressed the communities' needs and demands while actively mobilizing local people in conservation efforts. However, even after 2-3 decades of conservation initiatives, local perceptions and tolerance towards carnivores are still rather negative, particularly towards wolves. We therefore recommend a wider perspective of future awareness campaigns to include a broader specter of species and conservation issues. During interviews, we observed that remote settlements had rarely been visited by conservation authorities, and the inhabitants there had limited knowledge of compensation policies for livestock losses and human injury. Local perceptions on wolves tended to be more negative with increasing distances from conservation field offices, and this has been reported in earlier research in parts of ACA <ns0:ref type='bibr' target='#b62'>(Oli et al., 1994)</ns0:ref>. The factors underlying negative perceptions of distant settlements are probably due to a limited local involvement in community conservation programs. In the future, distant and remote settlements require more rigorous conservation outreach and awareness activities.</ns0:p></ns0:div>
<ns0:div><ns0:head>CONCLUSIONS</ns0:head><ns0:p>This study has investigated local villagers' perception of snow leopards and wolves in the central Himalayas, Nepal. In general, the perceptions of locals were more positive towards snow leopards than to wolves. People having larger herds of livestock (goat/sheep) with limited access to conservation programs were more likely to have negative perceptions towards wolves.Our results showed that multiple factors influence local perceptions of the two carnivores and that perception factors cannot be generalized for the two species. Thus, theyneed to be addressed separately. We suggest that future conservation projects and programs prioritize remote settlements. Furthermore, considering the substantial influence of occupation on people's Manuscript to be reviewed </ns0:p><ns0:note type='other'>Figure 2</ns0:note></ns0:div><ns0:figure xml:id='fig_0'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:04:48442:1:1:NEW 26 Jul 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_1'><ns0:head /><ns0:label /><ns0:figDesc>regular basis. Similarly, milking cows and, yaks are brought back to corrals/pens in the afternoon or in the morning for milking. Livestock are kept in corrals/pens for protection against predators corrals are traditionally made of mud walls and stones. Over the last decade there have been considerable changes in the lifestyle of the local people due to the development of roads in ACA. Despite these changes, traditional agro-pastoral lifestyles remain intact, and most importantly, traditional livestock grazing and collective village level decision making and implementation is still functional. In ACA, most farmers prefer to raise goats (Capra hircus) and sheep (Ovis aries) due to abundant pastureland, whereas in MCA farmers prefer cattle-yak hybrids (dzo, Jhopas, Bos spp.) as they are both grazers and browsers. Similarly, in the central part of ACA, farmers prefer yaks (Bos grunniens) due to dominant scrub vegetation. Lulu cows (Bos taurus sp.) and horses (Equus ferus caballus) are common in all areas. Among the main wild ungulates, bharal (Pseudois nayaur) and Himalayan tahr (Hemitragus jemlahicus) are widespread, whereas Tibetan argali (Ovis ammon hogdsoni), kiang (Equus kiang), and Tibetan gazelle (Procapra picticaudata) have overlapping grazing areas with livestock in the north-western parts of ACA. Apart from snow leopards and wolves, other carnivores include golden jackal, red fox (Vulpes vulpes), Himalayan black bear, Tibetan sand fox (Vulpes ferrilata), brown bear, Eurasian lynx and several species of weasel (Mustela spp.), and marten (Martes spp.).</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_2'><ns0:head>Figure 1 :</ns0:head><ns0:label>1</ns0:label><ns0:figDesc>Figure 1: Study area with location of survey villages and grazing areas.</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_3'><ns0:head>Figure 2 :</ns0:head><ns0:label>2</ns0:label><ns0:figDesc>Figure 2: Parameter estimates based on Generalized Linear Mixed-Effects Models of</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_4'><ns0:head>Figure 3 :</ns0:head><ns0:label>3</ns0:label><ns0:figDesc>Figure 3: Parameter estimates based on Generalized Linear Mixed-Effects Models of</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_5'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:04:48442:1:1:NEW 26 Jul 2020) Manuscript to be reviewed perceptions of carnivores, certain parts of the landscape, for example Manang of ACA and Tsum valley of MCA, should be tested for the development of wildlife based ecotourism.</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_6'><ns0:head>Figure 1</ns0:head><ns0:label>1</ns0:label><ns0:figDesc>Figure 1</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_7'><ns0:head>Figure 2</ns0:head><ns0:label>2</ns0:label><ns0:figDesc>Figure 2 Parameter estimates based on Generalized Linear Mixed-Effects Models of factors affecting perceptions of snow leopards.</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_8'><ns0:head>Figure 3 Figure 3</ns0:head><ns0:label>33</ns0:label><ns0:figDesc>Figure 3</ns0:figDesc></ns0:figure>
<ns0:figure type='table' xml:id='tab_1'><ns0:head>Table 1 : Overview of respondent characteristics in the Annapurna-Manaslu landscape, central Himalayas, Nepal.</ns0:head><ns0:label>1</ns0:label><ns0:figDesc /><ns0:table /></ns0:figure>
<ns0:figure type='table' xml:id='tab_2'><ns0:head>Table 2 : Model selection for perception towards the snow leopard and the wolf.</ns0:head><ns0:label>2</ns0:label><ns0:figDesc /><ns0:table /></ns0:figure>
<ns0:figure type='table' xml:id='tab_3'><ns0:head>Table 1 : Overview of respondent characteristics in the Annapurna-Manaslu landscape, 2 central Himalayas, Nepal.</ns0:head><ns0:label>1</ns0:label><ns0:figDesc>Dots and solid lines represent parameter estimates and 95% CI. The estimates are on the logit scale. The strength of the effect of parameters is indicated by the distances between the solid horizontal and the dotted vertical line. VDC (Village Development Committee) is included as a random effect. OWN=Total livestock holding, DIST=Distance from the nearest conservation field office to respondent household, COMP=Proportion of large livestock, Only respondents that responded to perceptions and share their experiences were included.</ns0:figDesc><ns0:table><ns0:row><ns0:cell>LOSS=Total livestock loss, SEXM=Male.</ns0:cell></ns0:row></ns0:table><ns0:note>1 3 Notes. 4 5 Number in parenthesis indicates percentage of individual respondents in each category. PeerJ reviewing PDF | (2020:04:48442:1:1:NEW 26 Jul 2020) Manuscript to be reviewed 6 ACA, Annapurna Conservation Area</ns0:note></ns0:figure>
<ns0:figure type='table' xml:id='tab_4'><ns0:head>Table 2 (on next page)</ns0:head><ns0:label>2</ns0:label><ns0:figDesc /><ns0:table /></ns0:figure>
<ns0:figure type='table' xml:id='tab_5'><ns0:head>Table 2</ns0:head><ns0:label>2</ns0:label><ns0:figDesc>Model selection for perception towards the snow leopard and the wolf. continuous variables were standardized by 2 standard deviations (as per<ns0:ref type='bibr' target='#b31'>Gelman and Hill, 2007)</ns0:ref> and all models included a varying intercept on VDC (Village Development Committee).VDC is included as a random effect. AGE: age of the respondent, COMP: composition of the herd, i.e. proportion of large stock animals, LIT: literacy (yes / no), LOSS: number of domestic animals lost to the carnivore, OCC: respondent's occupation (Herding, Agriculture, Other), OWN: number of domestic animals owned, SEX: gender of the respondent. DIST=Distance from the nearest conservation field office to respondent household. Only the top 10 models are presented for each analysis.</ns0:figDesc><ns0:table><ns0:row><ns0:cell>All</ns0:cell></ns0:row></ns0:table><ns0:note>PeerJ reviewing PDF | (2020:04:48442:1:1:NEW 26 Jul 2020)Manuscript to be reviewed 1</ns0:note></ns0:figure>
<ns0:figure type='table' xml:id='tab_6'><ns0:head>Table 2 : Model selection for perception towards the snow leopard and the wolf. Model df logLik AICc delta weight</ns0:head><ns0:label>2</ns0:label><ns0:figDesc>All continuous variables were standardized by 2 standard deviations (as per<ns0:ref type='bibr' target='#b31'>Gelman and Hill, 2007)</ns0:ref> and all models included a varying intercept on VDC (Village Development Committee).VDC is included as a random effect. AGE: age of the respondent, COMP: composition of the herd, i.e. proportion of large stock animals, LIT: literacy (yes / no), LOSS: number of domestic animals lost to the carnivore, OCC: respondent's occupation (Herding, Agriculture, Other), OWN: number of domestic animals owned, SEX: gender of the respondent. DIST=Distance from the nearest conservation field office to respondent household. Only the top 10 models are presented for each analysis.</ns0:figDesc><ns0:table><ns0:row><ns0:cell>Snow Leopard</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>OWN + OCC + SEX + AGE * LIT</ns0:cell><ns0:cell cols='2'>9 -233.24 484.9 0</ns0:cell><ns0:cell>0.45</ns0:cell></ns0:row><ns0:row><ns0:cell>OCC + SEX + AGE * LIT</ns0:cell><ns0:cell cols='2'>8 -234.42 485.2 0.27</ns0:cell><ns0:cell>0.39</ns0:cell></ns0:row><ns0:row><ns0:cell cols='3'>COMP * LOSS + OWN + OCC + SEX + AGE * LIT 12 -231.64 488.1 3.15</ns0:cell><ns0:cell>0.09</ns0:cell></ns0:row><ns0:row><ns0:cell>COMP * LOSS + OCC + SEX + AGE * LIT</ns0:cell><ns0:cell>11 -233.63 490</ns0:cell><ns0:cell>5.02</ns0:cell><ns0:cell>0.04</ns0:cell></ns0:row><ns0:row><ns0:cell>COMP * LOSS + OWN + OCC + SEX + AGE</ns0:cell><ns0:cell>10 -234.74 490</ns0:cell><ns0:cell>5.11</ns0:cell><ns0:cell>0.04</ns0:cell></ns0:row><ns0:row><ns0:cell>OCC</ns0:cell><ns0:cell cols='3'>4 -245.06 498.2 13.28 0</ns0:cell></ns0:row><ns0:row><ns0:cell>COMP * LOSS + OWN + OCC</ns0:cell><ns0:cell cols='3'>8 -241.52 499.4 14.47 0</ns0:cell></ns0:row><ns0:row><ns0:cell>COMP * LOSS + OWN + OCC + SEX</ns0:cell><ns0:cell cols='3'>9 -240.97 500.4 15.46 0</ns0:cell></ns0:row><ns0:row><ns0:cell>AGE</ns0:cell><ns0:cell cols='3'>3 -247.76 501.6 16.64 0</ns0:cell></ns0:row><ns0:row><ns0:cell>LIT</ns0:cell><ns0:cell>3 -251.48 509</ns0:cell><ns0:cell cols='2'>24.08 0</ns0:cell></ns0:row><ns0:row><ns0:cell>Wolf</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>COMP + DIST + LOSS + SEX + OWN</ns0:cell><ns0:cell cols='2'>7 -104.38 223.1 0</ns0:cell><ns0:cell>0.7</ns0:cell></ns0:row><ns0:row><ns0:cell>COMP * LOSS + OWN + OCC + SEX</ns0:cell><ns0:cell cols='2'>9 -104.88 228.3 5.22</ns0:cell><ns0:cell>0.05</ns0:cell></ns0:row><ns0:row><ns0:cell>SEX</ns0:cell><ns0:cell cols='2'>3 -111.33 228.7 5.62</ns0:cell><ns0:cell>0.04</ns0:cell></ns0:row><ns0:row><ns0:cell>OWN</ns0:cell><ns0:cell cols='2'>3 -111.63 229.3 6.22</ns0:cell><ns0:cell>0.03</ns0:cell></ns0:row><ns0:row><ns0:cell>COMP + LOSS + OWN</ns0:cell><ns0:cell cols='2'>5 -109.59 229.4 6.26</ns0:cell><ns0:cell>0.03</ns0:cell></ns0:row><ns0:row><ns0:cell>COMP + LOSS + OWN + DIST</ns0:cell><ns0:cell cols='2'>6 -108.75 229.8 6.65</ns0:cell><ns0:cell>0.03</ns0:cell></ns0:row></ns0:table><ns0:note>PeerJ reviewing PDF | (2020:04:48442:1:1:NEW 26 Jul 2020) Manuscript to be reviewed COMP * LOSS + OWN + OCC + SEX + AGE PeerJ reviewing PDF | (2020:04:48442:1:1:NEW 26 Jul 2020)</ns0:note></ns0:figure>
<ns0:note place='foot' n='7'>MCA, Manaslu Conservation AreaPeerJ reviewing PDF | (2020:04:48442:1:1:NEW 26 Jul 2020)</ns0:note>
</ns0:body>
" | "July 17th, 2020
Dear Editors,
We thank all the reviewers for their generous comments on the manuscript. The comments are quite useful and we have edited and made changes to address their concerns.
Our responses are in bold letters. We believe that the manuscript is now suitable for publication in PeerJ.
Dr. Madhu Chetri
On behalf of all authors.
________________________________________________
Dear authors
Thanks for your interesting contribution.
Three reviews have been received and reviewers agree your manuscript is interesting but found major and minor issues to solve.
I will be glad to reconsider a resubmission if all comments performed by reviewers are taken into account and a detailed response letter to each point is received.
Best regards!
Comments from the reviewers
Reviewer: María Losada
Basic reporting
The strength of this study is it includes relevant results for further research on the conservation strategies of these carnivore species in the central Himalayas. Moreover, the introduction and background are sufficient in terms of placing the reader in the study context. Regarding the format, the manuscript is correctly structured, fulfills the journal standards, and contains all the sections needed. Overall, the manuscript is clearly written in a professional language. However, several improvements with regard to basic reporting are required.
Major comments:
1. With regard to the Introduction and background of the study, a relevant point to explain here is that this research constitutes a further step in a long-term study (if not, explain why). I recommend including a justification of the novelty of the current study in relation to your previous studies in this matter (Chetri 2018).
Response: We have restructured the introduction: we moved one of the last paragraphs up to the beginning of the introduction (lines 50-60), and we added text in the last 2 paragraphs to improve the justification of the study, and underline the value of conducting the study in an area with good knowledge of carnivore ecology and livestock losses, lines 114-146.
2. Some grammar errors can be found along the manuscript and paragraph structure should be improved by polishing edition, particularly within the Introduction and Methods sections. Moreover, I highly encourage you to improve the paragraph structure in the Introduction section as follows: Lines 49-80 one paragraph is too long. I think that this aspect can be managed by creating two different paragraphs for each species information, focusing on local tolerance and ecological aspects (before Line 67), and then one more paragraph explaining common information (from Line 67 to 80). Furthermore, I suggest putting Lines 92-99 in another paragraph, as a specific statement of the questions formulated for this study. Lines 98-99, the size of the study area fits better in the Study area subsection.
Response: It was a nice suggestion, thank you. We fully agree and made adjustments as suggested.
3. I have some doubts relative to the questionnaire design because of certain questions formulated were excluded from the analyses, such as livestock numbers of each species (instead of the total number of domestic animals owned), season, losses experienced (yes/no) and reported, habitat type, etc. Why did you exclude these data from the analyses and results sections? Please, justify this and include a statement clarifying these omissions.
Response: We included a clarifying statement in the data analyses section of the methods chapter. Generally, we tried to simplify the analyses by reducing the number of explanatory variables without losing relevant information. Otherwise, we believe the analyses would be too complex, and the most important results would be lost/overlooked. We selected questions based on our previous study on predictors of livestock losses (Chetri et al. 2019) and other relevant studies (see added statement in the methods, lines 224-235).
Minor comments:
1. (Line 62) Why the density estimates of wolves are still lacking? Is it because of the lack of recent DNA fecal study of wolves, or because this species is not easy to detect within the study area? How rare are the wolf species in the study area? Please, explain this and include this information if possible.
Response: Thank you for pointing out this. We have added a sentence for clarity, lines 72-80.
2. In general, in-text citations are well performed. Although they should be improved according to the Journal Instructions for Authors (check the Reference format section). A comma is required after the list of all author names and before the publication year, for example, Athreya et al., 2013. In line 41, remove the middle name of the second author (Cesar). In-Line 62, Chetri (2014) is not included in the Reference section. In line 306, add the publication year in the in-text citation. Moreover, the Reference Section’s guidelines recommend choosing an option to give the DOI on each reference. For example: DOI: 10.1016/j.jnc.2016.09.004. Please, apply these modifications along the manuscript.
Response: Thank you so much for pointing out this. We have corrected it as suggested.
3. According to PeerJ Ethic policies, an Ethics statement should be provided in the Materials and Methods section of the manuscript whenever the research was conducted on humans or human tissue; on animals or animal tissue. Please, include this statement related to Human participant's approval documentation obtained from the relevant approval body in the Material & Methods section (Line 147). If not being able to include it, please provide a statement explaining why it is not necessary.
Response: We have added a sentence as suggested. And also included an Ethics statement at the beginning of the methods section (lines, 151-153).
4. In accordance with the PeerJ submission guidelines, all tables and figures titles should be in bold and with a colon after the figure or table name. Please check the format recommended and implement these changes along the manuscript but also in the figure and table files loaded. In more detail, Figure 1 requires more space at the map bottom to avoid cutting the map legend; the font of geographic coordinates of the map grid have to be increased to clear reading. If possible, try to reduce the white space inside the figure file loaded and cite map sources in the legend if you used anyone. In Figure 2, I recommend to define the VDC acronym in the legend but also to increase the font size in both axis titles and axis lines (and the same for Fig. 3). Table 1 can be improved as below: No. or number instead of No; “included or valid questionnaires” instead of “included respondents”; “invalid/excluded questionnaires” instead of “neutral responses”; all the units (i.e. years) should be placed in the row title in parentheses; all the acronyms and abbreviations defined in the footnote (VCD, ACA, MCA); all the footnote markers have to be referenced in the table; “proportions” should be changed by “percentage” or instead of that, specify what means here. Why are not included the answers relative to livestock losses and the loss numbers? Please, provide this information or explain why is not included. In Table 2, VDC would be defined in the legend (as well as in Table S1 and Table S2 in the Supplemental File), and selected models for each species highlighted in bold and this formatting explained in the legend. Moreover, you need to include the distance from the nearest conservation field office to the respondent household (DIST) as a predictor variable in the legend of Table 2 and Table S1, because it is included as a predictor variable in both cases.
Response: Thank you very much for helping out with minor details. We have corrected it as suggested. We have included livestock losses and the loss number in our analysis (please see supplementary data files). These were explanatory variables in our models.
5. Thank you for sharing the raw data and ensuring you follow our Data Sharing policy. However, supplemental files provided (peerj-48442-Raw_data_Snow_leopards.csv; peerj-48442-Raw_data_Himalayan_wolves.csv) require more descriptive column identifiers, including measurement units for each variable and acronyms or abbreviations definitions to easy reading. The file names provided in the “How are you submitting your raw data or code?” section do not correspond with the required ones. Please, correct it in this section.
Response: We have updated accordingly.
Experimental design
The manuscript submitted is within the scope of this journal. The research aims sound rather clear and constitute relevant points for the study. In general, the research performance fulfills the basic technical and ethical standards (with the exception of minor comments made in the basic reporting). The methods section is rather detailed and contains relevant information to replicate the study. However, several improvements relative to the experimental design should be implemented, thus I would give some feedback relative to the Introduction and the Methods sections.
Major comments:
1. It is necessary to be more specific when defining the questions and justifying the need for this study in relation to the lack of knowledge about the problem analyzed (Lines 93-99). In Line 93, remove “in an area”.
Response: We have restructured the introduction: we moved one of the last paragraphs up to the beginning of the introduction, and we added text in the last 2 paragraphs to improve the justification of the study, and underline the value of conducting the study in an area with good knowledge of carnivore ecology and livestock losses. We defined more clearly the questions we addressed, lines 50-60 and lines 114-146.
2. I have found the questionnaire design used for data collection (Appendix S1) could be improved in order to answer the questions formulated. I believe that this issue can be solved if you justify the design provided in the Semi-structured questionnaire survey subsection and explain the reasons for excluding certain questions from the analyses in the data analysis subsection.
Response: We included a clarifying statement in the data analyses section of the methods chapter. Generally, we tried to simplify the analyses by reducing the number of explanatory variables without losing relevant information. Otherwise, we believe the analyses would be too complex, and the most important results would be lost/overlooked. We selected questions based on our previous study on predictors of livestock losses (Chetri et al. 2019) and other relevant studies (see added statement in the methods, lines 224-240).
3. Why did you exclude the neutral respondents from the analyses? I believe that if you do not explain the exclusion reasons resembles a bias. Please, remove the neutral category from the answers because the response variable fits a binomial distribution (Line 155).
Response: Agree, we have removed the line as suggested. We removed neutral respondents because these did not add any information relevant for the analyses.
4. Change these two sentences (Lines 155-156) as follows: “We consider as invalid questionnaires when respondents did not want to answer or stated that they did not know about the species presence and conflict, thus they were excluded from the analyses” (see Appendix S1 comments). In Lines 158-159, change “neutral respondents” by “excluded or invalid questionnaires”.
Response: Thank you for pointing out this, we have updated as suggested, lines 224-229.
Minor comments:
1. In Line 110, I believe that the preferred religion by the majority of the population is not relevant to the study. Please remove it, or instead of, please explain why the religion was not included as a factor in the analyses.
Response: We agree and have removed the sentence.
2. In the stud area subsection, there should be two different paragraphs between Lines 112-123 and Lines 123-128, or instead, try to summarize the information given in these lines.
Response: Agreed, We have separated the paragraphs as suggested.
3. In Lines 126-128, I have not found the link in this sentence to the rest of the paragraph, please remove it or create another paragraph to explain the distribution and ecological aspects of the study species.
Response: Agreed, we have removed the sentence, lines 187-190.
4. In Line 167, are there any differences among the answers of individuals who have chosen the “others” option? It is possible to explain which answers were given as the “others” category and how did you manage this issue in the analyses.
Response: “Others” is one category. These were relatively few and could therefore not be subdivided into several categories. The “others” are people benefitting from tourism, and may thus be positively impacted by the presence of large carnivores. We added a clarifying statement in the data analyses section (lines, 244-245).
Validity of the findings
I consider this study includes relevant findings for further research on the conservation strategies of these carnivore species in the central Himalayas. The statistical analyses seem well performed and the sample size used for the separated models is fairly balanced, and for both species, it allows a large number of factors and covariables to be included in model construction. Another good point is VDC was included as a random factor in both cases. Moreover, all the results were fully reported, and the figure design chosen to show the estimated effects is the most recommended. Most of the ideas argued in the Discussion section show a deep commitment to the labor of your team, and in general, most of the results are broadly supported. The conclusions highlight the most remarkable points of your study, but not all of them. To summarize, I believe the findings provided are rather robust and statistically sound, although certain clarifications in the Data analysis and the Results sections are required, as well as in the Discussion and the Conclusion sections remain open with respect to the background of the study.
Major comments:
1. The impact and the novelty of this research are not justified in the Introduction and Discussion sections. This manuscript will gain relevance if you improve these two aspects. For instance, which aspects of your research make a difference from what Kusi et al. (2019) observed?
Response: See previous comments about justification of the study in the introduction. Kusi et al. 2019 reported that having experienced livestock depredation was a main predictor of the general negative attitude toward wolves but not in the case of snow leopards. In our study, we did not find that experiencing loss was important. This is an important difference that is discussed in lines 398-402. We consider our study and Kusi et al. (2019) to be complementary rather than contradictory. Hence, throughout the discussion, we have also emphasized aspects that are similar in our studies.
2. I assumed that you performed a correlation test among predictor variables before the model construction; please confirm this doubt. In Line 28, how did you evaluate that the answer according to gender is independent of the number of losses? If possible, explain this point in the data analysis subsection. Moreover, I suggest you consider including the VDC as a nested factor of the number of losses variable.
Response: We did check the correlation among (continuous) predictors before carrying out the regression analysis, and we did not include collinear variables (rho > 0.6) into the same mode (lines, 250-252). Both effects of gender (SEX) and of the number lost (LOSS) were included in the best model (for wolf), so they are not completely independent. We corrected the wording in the text and in the abstract (lines, 28-30). VDC was included as a random effect because it had a large number (21) of levels. Variable LOSS is a count rather than a grouping variable, so we cannot really include it as a random effect. However, if by 'VDC as nested factor of LOSS', you mean the 'random-slope' model in which the slope / effect for LOSS varies across VDCs (i.e. model defined by '... LOSS + (1 + LOSS | VDC)', then we considered this model during selection of the random component, but it was not selected. However, if you mean the model with a fixed effect of LOSS plus random variation in intercept among VDC within LOSS (i.e. '... LOSS + (1 | VDC:LOSS)'), then we believe this model makes little biological sense.
3. According to Li et al. (2014), why did you not include religion as a predictive factor in both separated analyses? I suggest thinking about it particularly when you use this fact to give support in the Discussion section.
Response: It was a good question. However, we did not include religion as a predictive factor as all the locals residing in the region are Buddhist, only a few outsiders who are working as labor or teacher are non-Buddhist. So it was not possible to segregate data according to religion.
Minor comments:
1. Several points from the Discussion and the Conclusion sections remain open with respect to the background of the study. We have changed the introduction, see previous comments. Between Lines 245-252, you suggested that all the commented cultural aspects are likely determining how have evolved local perceptions on snow leopards and Himalayan wolves. By contrast, you did not include religion as another factor in the analyses and you did not comment on the relevance of this aspect for the study species conservation in the introduction. We did not include religion since almost all were Buddhists (see comment above Secondly, you suggested that the increasing tolerance for snow leopards could be linked with the continued efforts to increase awareness as part of the ongoing DNPWC conservation program (Line 266). However, you did not take into account for these measures in the data analysis and the introduction sections. We added text regarding conservation efforts in the end of the introduction (lines, 134-141). Thirdly, as Caruso et al (2020) suggest (Lines 285-286), ecotourism development could be indirectly influencing local perceptions towards study species, though it was considered as the “other” category of occupation variable for the analyses. The “other” category included too few respondents to be divided into sub-categories, as described previously. Lastly, the Oli et al. (1994) study reported that negative perceptions on wolves increased with the distances from conservation field offices, even so, you did not compare it when finding a weak effect of this variable (Lines 327-328). We referred to Oli et al. (1994) in this section, so we do not understand the problem here. I encourage you to think about the ideas suggested and if possible, implement them in the new version of the manuscript.
Response: We hope we have addressed the comments properly in the preceding text.
2. The study concludes that the negative perceptions for wolves found in the study area are (Line 337: not is corrected) probably due to fear and cultural bias as previously reported, although I also find interesting that this species suffers a scarcity in conservation awareness policies, and by contrast, the positive perception on snow leopards could be linked with higher conservation efforts. Apart from that, you have not controlled by this cultural bias in your analyses. Thus, I encourage you to implement these concluding remarks in the further version.
Response: We did not include any questions directly addressing this cultural bias in the questionnaires. Hence, it is not possible to include in the analyses. We have emphasized the difference in perceptions of wolves and snow leopards in the introduction and the discussion (as pointed out by the reviewer).
Comments for the author
I really appreciate the study efforts during the data recording and the commitment of your team with the projects developed for both conservation species (ACA and MCA. I encourage you to improve this manuscript to provide the impact deserved of these study findings in order to promote further researches in conservation designing policies within the study area. I recommend you to apply the feedback provided in this review in order to get a new and improved version of the article for the second round of review.
Response: We are grateful for the valuable comments and suggestions from the reviewer. We hope our revisions have addressed the comments properly.
Reviewer 2
Basic reporting
I like to appreciate authors for conducting research in the world’s remotest area. It is a good research conducted focusing perception of people towards snow leopard and wolves. Definitely, the attitude of local people need to be understood for better conservation planning. Study has well speculated the perception of local people using generalized mixed effect model. However, the reasons behind the perception are not well discussed based on this study. I would like to suggest authors to focus rather on introduction and discussion while doing revision. I have pointed my comments line by line as follows.
Experimental design
No comment
Validity of the findings
Authors are suggested to support results with their own observation. I suggest authors to give backup from their own findings. This study should find out the reasons behind negative or positive perception relying on their studied area.
Response: We have tried our best to provide a balanced interpretation of our results, partially referring to our own research within the study area and also including other studies from the same region and elsewhere.
Comments for the author
Line 39 – for better clearance write “Human-carnivore conflict”
Response: Thank you for your suggestion, we have made changes, line 39.
Line 62 – Two sentence are contradicting each other. One sentence is saying density has not been estimated and another is saying rare. In these regions, very less study have been done. That does not mean species is rare.
Response: Thank you for pointing this out, we have improved the sentences, lines 72-80.
Line 68/69 – Depredation mortality is rather low than other causes of death. Then, why livestock depredation is big issue?
Response: Livestock depredation by carnivores is a big issue in the Himalayas simply because local people are highly dependent on livestock farming and agriculture. Even a single livestock loss has a big impact in financial terms. For example, milking cows, riding horses and yaks are very expensive in the region. Also, incidents of mass killings may have a massive impact on economy. These points have been addressed in the introduction, lines 123-132.
Line 70/72 –Better to talk about prey killing behavior of wolves? Author should justify why killing is brutal?
Response: We reworded this section, lines 125-126.
It is suggested to authors to talk about the situation of retaliatory killing and livestock depredation in relation to perception.
Response: We added text, lines 123-132. We also emphasized this point more in a previous paper about factors influencing livestock losses from the same area (Chetri et al. 2019a).
Line 94-why livestock depredation is main concern? author should justify the statement somewhere in introduction.
Response: See comment above
Line 116 – Author is suggested to change “abundant flat plain” to “abundant pasture land”
Response: Thank you for your suggestion, we have made changes, line 178.
Line 135/136 – VDCs are no more administrative unit in Nepal since 2015. VDC are changed to ward in most of cases and new rural municipalities have been established throughout the country. So, authors are suggested to review the study area accordingly.
Response: Thank you for pointing out this, we have clarified now and included sentences in lines, 201-203. Our fieldwork was done before this administrative change.
Study area: Authors are suggested to touch briefly about human population and livestock numbers in study area. This will give readers idea about how human pressure and livestock pressure.
Response: Thank you for pointing out this, we have given human population density in line, 160. Following your suggestion, we have added livestock information in lines, 162-163.
Line 159/160 – not clear. Authors are meant to say northwestern section of ACA and MCA or northwestern section of all studied settlements.
Response: We have corrected, line 230.
Line 162-172. In data analysis section; authors are suggested to justify “why your data fit to generalized mixed effect model” at the beginning and also suggested to include residual plot as supplementary information. As the model is mixed effect, it is good idea to plot by partially pooling. This will help to show relation with each group.
Response: We have restructure the sentence (lines 232-235). We have provided full models and data sets as supplementary information. Given the size of the data set (395 observations) and the fact that VDC has 21 levels, we included VDC as a random effect to avoid overfitting when considering VDC as a fixed effect. The GLMM *does* 'partial pooling', therefore, we thought it's not necessary to *show* partial pooling.
Line 174/175 –These two lines rather fit in methodology.
Response: We removed these lines since they were already in the methods section, lines 255-257.
Discussion
Line 236-243 –Authors are suggested to support results with their own observation. Here in these lines, authors have supported their observation with other research. In discussion section, I found that findings of this research are justified by other research, for example in Line 252 to 260. I suggest authors to give backup from their own study. This study should find out the reasons behind negative or positive perception based on their studied area.
Response: As pointed out previously, we have tried our best to provide a balanced interpretation of our results, partially referring to our own research within the study area and also including other studies from the same region and elsewhere. For instance, we did not have own data on snow leopard and wolf activity patterns, and thus we need to refer to other studies (line 318-326).
Line 247/249 – I think that traditional practice of grazing does not fit here. How are you relating this statement with your research finding?
Response: We removed the sentence, lines 330-332.
Line 251/252 – “Buddhist community did not kill wildlife”. It seems that they do kill wildlife now. Authors are suggested to change the structure of sentence.
Response: We have rephrased this sentence, lines 334-335.
Line 275/276 – As authors have studied perception and age. They are suggested to compare age vs perception to conclude in different perception of young and old people. As you have mentioned about religious belief. Possibly, old people are more religious and should be more positive towards snow leopard. You have also mentioned that the conservation practice in the region is quite old. I think that that program must have been focusing entire population (young people, old people, both male and female) of your study area without any bias
Response: As pointed out previously, we could not include religion as a factor since practically all respondents were Buddhists. We did include age in the models. However, while age alone had little impact on perceptions, the interaction between age and literacy was much stronger. We agree with the reviewer that awareness programs have focused on the entire population, but we do believe that influence on perceptions may vary according to e.g. age and literacy, as shown in our analyses.
Line 289-292 – these sentences are suitable for conclusion or in concluding part of discussion.
Response: We rephrased and moved these sentences to the end of the conclusion part, lines 435-439.
Line 305/306 – Incomplete citation
Response: Thank you for pointing out this, we have updated the citation, line 393.
Conclusion
Line 332 to 342: authors are suggested to include only their findings in conclusion.
Response: Thank you for the suggestion. We have rephrased the conclusion part, lines 423-439.
Reviewer 3
Basic reporting
no comment
Experimental design
The article is framed within the social sciences but with clear applicability in environmental sciences.
Given the scope of the journal I would recommend the authors to make brief recommendations for the conservation (or to foster tolerance) of these two species at the end of the manuscript
Methods are well described in general but I have doubts in the definition of the variable COMP (proportion of large stock). I interpret with this that only certain species (large ones) are considered for this variable, if so, which are they?
Response: It is the proportion of large stock (Yak, horses, cows) relative to all livestock owned, e.g. a value of 0.5 would mean that half of the owned livestock were large. We have specified in line 243 and in text of table 2. We also refer to our previous study (Chetri et al. 2019) in line 243.
Validity of the findings
no comment
Comments for the author
I have enjoyed reading this article which deals with the local perceptions of two large carnivores in the central Himalayas. As the authors said in the abstract ' An understanding of local perceptions of carnivores is important for conservation and management planning' However, I believe that the article may benefit if the authors connect these perceptions with management planning by including a series of recommendations based on these results that may in some way favor co-existence in the area of study. I understand that they include a few sentences during the discussion but I think it would be useful to address this in further detail.
Response: We added text in the conclusion part of the manuscript, lines 423-439.
I don't know where it would fit better (introduction or discussion) but I think it would be good to include a small paragraph about the benefits (apex predator, ecotourism opportunities...) that these carnivores can bring in the area and not only focus on them as livestock predators.
Response: We fully agree. We have included a sentence in lines 372-375 and in lines 435-439.
I'm sorry, possibly I have not understood well, throughout the text you mention two areas ACA and MCA. I understand that they differ in the type of species of livestock (among others characteristics I guess) but then I do not see that the analysis takes into account whether the respondents are from one area or another. Would it not be appropriate to include the regions in the analysis or the type of livestock they had?
Response: This is a very good suggestion, and we had planned to do this. However, there were too few respondents in the Manaslu region for this analyses to make any sense. The area is far more sparsely populated within the altitude range of snow leopard habitat where our sampling was conducted.
Appendix S1 Questionnaire:
-Is there any relationship between having an encounter with the species (question 6 of the questionnaire) and the type of perception (positive/negative)?
-It would be interesting for the reader to have some more descriptive results from the questionnaire (for example questions 3, 4, 6 and 10)
-Questions 8 and 9: You ask why but I find no mention of these questions in the manuscript. Only just a suggestion, I find it interesting if the authors decide to analyze these questions. For example, a content analysis would be a good tool to determine which words could be associated with positive/negative perceptions of species.
Response: We decided to disregard some of the questions because we felt the manuscript would be overloaded with information. This could take focus away from the more important issues.
I have other minor suggestions:
L31 future conservation projectS
Response: Agree, we have corrected, line 33.
L 64 ACA this acronym is not yet defined in the text
Response: Thank you for pointing out this, we have defined now, lines 120-121.
L66 Do you have any knowledge about the distribution ranges of these species that could be projected on the map?
Response: We are planning a separate paper on snow leopard distribution patterns in the area based on scats and sign- Also, we have a density distribution map in Chetri et al., 2009b. Figure one would be too overloaded with information if it included snow leopard distribution (which is all over the area within a given altitude range). Wolf distribution data are too poor to present in a figure, but we have written about it’s approximate distribution in the manuscript text, lines 229-230.
L107 What kind of tourism? (mountaineering, or wildlife watching) Are there eco-tourism options for these species in the area?
Response: Thank you for specifying this, it’s eco-tourism, we have updated now, line 162.
L137 do you know how these VDC (or differences between ACA and MCA) manage the attacks of the carnivores (legal hunting, lethal control, fences, aversive conditions...)
Response: They usually herded small stocks, for example goats and sheep which are of high risk from predators. We have added a paragraph, lines 167-173.
L236 Typo: This perception (without s)
Response: Thank you, we have corrected, line 320.
L300 I think it should be 'may have been'
Response: Agree, we have corrected, line 389.
L 304 some mistake by the reference manager.
Response: Thank you for pointing out this, we have corrected now, line 393.
L309 I think that “actually” it is not necessary here
Response: Agree, we have corrected, line 398.
L312 delete comma
Response: Thank you, we have corrected, line 402.
L315-316 so why did you do not ask it in the questionnaire. I’m wondering if it also could be related with tolerance to snow leopard and wolves.
Response: we do not understand what statement the reviewer referred to?
L327 “ perceptions of distant settlements ARE due to THE limited nature…”
Response: Thank you for pointing out this, we have made corrections, lines 418-419.
L315-216 it would have been a good idea to include in the questionnaire some questions related to people's participation in conservation tasks
Response: Thank you for the suggestion. In the future, we will incorporate this into our questionnaire.
L335 “…perceptions for THE Wolf in our study area ARE probably…”
Response: Thank you for pointing out this, we have made corrections.
" | Here is a paper. Please give your review comments after reading it. |
9,739 | "<ns0:abstract xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:p>An understanding of local perceptions of carnivores is important for conservation and management planning. In the central Himalayas, Nepal, we interviewed 428 individuals from 85 settlements using a semi-structured questionnaire to quantitatively assess local perceptions and tolerance of snow leopards and wolves. We used generalized linear mixed effect models to assess influential factors, and found that tolerance of snow leopards was much higher than of wolves. Interestingly, having experienced livestock losses had a minor impact on perceptions of the carnivores. Occupation of the respondents had a strong effect on perceptions of snow leopards but not of wolves. Literacy and age had weak impacts on snow leopard perceptions, but the interaction among these terms showed a marked effect, i.e. being illiterate had a more marked negative impact among older respondents. Among the various factors affecting perceptions of wolves, numbers of livestock owned and gender were the most important predictors. People with larger livestock herds were more negative towards wolves. In terms of gender, males were more positive to wolves than females, but no such pattern was observed for snow leopards.</ns0:p><ns0:p>People's negative perceptions towards wolves were also related to the remoteness of the villages. Factors affecting people's perceptions could not be generalized for the two species, and thus need to be addressed separately. We suggest future conservation projects and programs should prioritize remote settlements.</ns0:p></ns0:div>
</ns0:abstract>
<ns0:body xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:head>INTRODUCTION</ns0:head><ns0:p>Large carnivore co-existence with humans remains a global challenge <ns0:ref type='bibr' target='#b4'>(Athreya et al., 2013)</ns0:ref>, and mitigation of human-carnivore conflicts requires multiple approaches and disciplines <ns0:ref type='bibr' target='#b63'>(Redpath et al., 2013)</ns0:ref>. Among the various aspects of carnivore conflict management, understanding local perceptions is crucial for establishing long term conservation strategies <ns0:ref type='bibr' target='#b5'>(Bagchi & Mishra, 2006;</ns0:ref><ns0:ref type='bibr' target='#b22'>Conforti & Azevedo, 2003)</ns0:ref>, especially in multi-use landscapes where animal husbandry is the main source of income. An assessment of local perceptions helps in identifying groups of people or villages that are negative towards protection of carnivores, and thus aids conservation authorities to find suitable strategies to improve their tolerance <ns0:ref type='bibr' target='#b66'>(Suryawanshi, 2013;</ns0:ref><ns0:ref type='bibr' target='#b70'>Treves & Karanth, 2003)</ns0:ref>. Further, assessments form a basis for quantifying the effects of conservation management interventions and aid in formulating new strategies if opinions towards conservation change <ns0:ref type='bibr' target='#b29'>(Dressel, Sandström & Ericsson, 2015)</ns0:ref>.</ns0:p><ns0:p>Globally, local perceptions and attitudes towards large carnivores are complex and vary markedly between regions <ns0:ref type='bibr' target='#b65'>(Røskaft et al., 2007)</ns0:ref>. Multiple factors influence local perceptions, including animal behavior, risk of negative encounters, and the length of the period of coexistence <ns0:ref type='bibr' target='#b23'>(Dickman, 2010;</ns0:ref><ns0:ref type='bibr' target='#b29'>Dressel, Sandström & Ericsson, 2015;</ns0:ref><ns0:ref type='bibr' target='#b40'>Kellert et al., 1996;</ns0:ref><ns0:ref type='bibr' target='#b79'>Zimmermann, Wabakken & Dötterer, 2001)</ns0:ref>. Local perceptions also vary among ethnic groups, and are linked to religious-and cultural beliefs <ns0:ref type='bibr' target='#b0'>(Ale, Shah & Jackson, 2016;</ns0:ref><ns0:ref type='bibr' target='#b24'>Dickman et al., 2014;</ns0:ref><ns0:ref type='bibr' target='#b40'>Kellert et al., 1996;</ns0:ref><ns0:ref type='bibr' target='#b45'>Li et al., 2013;</ns0:ref><ns0:ref type='bibr' target='#b56'>Mkonyi et al., 2017)</ns0:ref>. Socio-demographic variables such as age, sex, income, occupation, literacy, number of livestock owned and loss to predators have all shown to be associated with local perceptions and attitudes of large carnivores <ns0:ref type='bibr'>(Caruso</ns0:ref> PeerJ reviewing PDF | (2020:04:48442:2:0:NEW 10 Sep 2020)</ns0:p><ns0:p>Manuscript to be reviewed <ns0:ref type='bibr'>et al., 2020;</ns0:ref><ns0:ref type='bibr' target='#b30'>Fort et al., 2018;</ns0:ref><ns0:ref type='bibr' target='#b39'>Kellert & Berry, 1987;</ns0:ref><ns0:ref type='bibr' target='#b41'>Kideghesho et al., 2007;</ns0:ref><ns0:ref type='bibr' target='#b57'>Naughton-Treves, Grossberg & Treves, 2003;</ns0:ref><ns0:ref type='bibr' target='#b59'>Oli, Taylor & Rogers, 1994;</ns0:ref><ns0:ref type='bibr' target='#b65'>Røskaft et al., 2007;</ns0:ref><ns0:ref type='bibr' target='#b69'>Trajçe et al., 2019</ns0:ref>).</ns0:p><ns0:p>In the central Himalayas, Snow leopards (Panthera uncia) and Himalayan wolves (Canis lupus chanco) are the two most important predators involved in conflicts with people <ns0:ref type='bibr' target='#b19'>(Chetri et al., 2019a)</ns0:ref>. A recent study from the region revealed that snow leopards were responsible for the majority of predation losses (61.9%); the remaining were from Himalayan wolf (16.8%) and other predators (21.3%) including feral dogs (Canis lupus familiaris), brown bear (Ursus arctos), black bear (Ursus thibetanus), Eurasian lynx (Lynx lynx), golden jackal (Canis aureus) and common leopard (Panthera pardus) <ns0:ref type='bibr' target='#b19'>(Chetri et al., 2019a)</ns0:ref>. The snow leopard is categorized as Vulnerable in the IUCN red list of threatened species <ns0:ref type='bibr' target='#b50'>(McCarthy et al., 2017)</ns0:ref>, whereas wolf is considered as Least Concern <ns0:ref type='bibr' target='#b9'>(Boitani, Phillips & Jhala, 2018)</ns0:ref>. However, in the national Red Data List of Nepal, wolves are considered as Critically Endangered and snow leopards are considered as Endangered <ns0:ref type='bibr' target='#b37'>(Jnawali et al., 2011)</ns0:ref>. A recent fecal DNA study reported that the snow leopard density within our study area in the central Himalayas was 0.95 (SE 0.19) animals per 100 km 2 <ns0:ref type='bibr' target='#b20'>(Chetri et al., 2019b)</ns0:ref>, but density estimates of wolves from the area are still lacking <ns0:ref type='bibr' target='#b77'>(WWF, 2015;</ns0:ref><ns0:ref type='bibr' target='#b17'>Chetri et al., 2016;</ns0:ref><ns0:ref type='bibr' target='#b21'>Chetri, Odden & Wegge, 2017)</ns0:ref>. The species has received little conservation attention due to its lower conservation status in the IUCN Red list, which has made it difficult to acquire necessary funding for population monitoring. According to <ns0:ref type='bibr' target='#b14'>Chetri (2014)</ns0:ref>, Himalayan wolves are rare in the region and mostly solitary. The wolves that thrive in this landscape are genetically unique to the region as revealed by recent DNA analysis, and they are considered different from the grey wolf lineage <ns0:ref type='bibr' target='#b17'>(Chetri et al., 2016)</ns0:ref>. Both species range widely and often encounter pastoralists. Manuscript to be reviewed Although information on livestock depredation by snow leopards and wolves exists from Nepal's Himalaya <ns0:ref type='bibr' target='#b3'>(Aryal et al., 2014;</ns0:ref><ns0:ref type='bibr' target='#b21'>Chetri, Odden & Wegge, 2017 ;</ns0:ref><ns0:ref type='bibr' target='#b19'>Chetri et al., 2019a;</ns0:ref><ns0:ref type='bibr' target='#b59'>Oli, Taylor & Rogers, 1994;</ns0:ref><ns0:ref type='bibr' target='#b72'>Wegge, Shrestha & Flagstad, 2012;</ns0:ref><ns0:ref type='bibr' target='#b73'>Werhahn et al., 2019)</ns0:ref>, limited information is available regarding variation in local perceptions and tolerance to these species on a large spatial scale <ns0:ref type='bibr' target='#b34'>(Hanson et al., 2019;</ns0:ref><ns0:ref type='bibr' target='#b43'>Kusi et al., 2019;</ns0:ref><ns0:ref type='bibr' target='#b59'>Oli, Taylor & Rogers 1994)</ns0:ref>. Hence, in our study, we examined local communities' perceptions of snow leopards and wolves in a large area of ~5000 km 2 where livestock depredation has been a main concern in recent decades. The survey covered two protected areas, Annapurna Conservation Area (ACA, hereafter) and Manaslu Conservation Area <ns0:ref type='bibr'>(MCA, hereafter)</ns0:ref>, where ecological studies of snow leopard and wolf had recently been conducted <ns0:ref type='bibr' target='#b15'>(Chetri, 2018)</ns0:ref>. These studies showed that snow leopard density was far lower than previously assumed, and consequently, average annual livestock losses were low (ca.1%) even though livestock constituted large proportions of the diet of both snow leopards and wolves (ca.25%). Despite the low levels of livestock depredation, perceptions of wolves are often negative. Similarly, incidents of mass killings of livestock by snow leopards decreases local tolerance towards their conservation, which in turn may lead to retaliatory killing of carnivores <ns0:ref type='bibr' target='#b35'>(Jackson, 2015;</ns0:ref><ns0:ref type='bibr'>Mishra, Redpath & Suryawanshi, 2016;</ns0:ref><ns0:ref type='bibr' target='#b55'>Mishra et al., 2004;</ns0:ref><ns0:ref type='bibr' target='#b68'>Suryawanshi et al., 2014;</ns0:ref><ns0:ref type='bibr' target='#b76'>Woodroffe et al., 2005)</ns0:ref>. Surplus killings and injuries of high valued livestock (e.g. horses, milking yaks and cows) not only outrage local communities <ns0:ref type='bibr' target='#b59'>(Oli, Taylor & Rogers, 1994)</ns0:ref>, but also have negative repercussion that can spread even to distant villages. Manuscript to be reviewed wolves. Due to the relatively low livestock losses and the considerable conservation efforts in the study area, we expected perceptions of carnivores to be more positive than reported from previous studies from the Himalayan range. Furthermore, we expected perceptions to vary geographically as well as between species due to a bias in the impact of conservation awareness campaigns, i.e. tolerance of wolves should be lower, and perceptions could potentially depend on the remoteness of villages. Lastly, we expected perceptions to be affected by socioeconomic and demographic factors, e.g. livestock losses and ownership, gender, age, education and occupation.</ns0:p></ns0:div>
<ns0:div><ns0:head>MATERIAL AND METHODS</ns0:head></ns0:div>
<ns0:div><ns0:head>Ethics statement</ns0:head><ns0:p>Approval and relevant permits required to carry out this research were obtained from the National Trust for Nature Conservation (Ref.no.291), Nepal.</ns0:p></ns0:div>
<ns0:div><ns0:head>Study area</ns0:head><ns0:p>We conducted the study in the Annapurna Manaslu Conservation landscape in the central -Himalayas (N28 29, E83 85; Fig. <ns0:ref type='figure' target='#fig_3'>1</ns0:ref>). Both ACA and MCA are located within this landscape -and are the largest community-based conservation areas in Nepal (ca.9292 km 2 ). It is located in the rain shadow area of the Himalayas. Together with Bhimthang valley, it is the priority landscape for snow leopards conservation in the country <ns0:ref type='bibr' target='#b28'>(DNPWC, 2017)</ns0:ref>. The human population density is 1 per km 2 (CBS, 2012), and agro-pastoralism is the main source of livelihood, although some households are also involved in eco-tourism related entrepreneurs.</ns0:p><ns0:p>The overall livestock density in the study area is 35.74 ± 0.10/km 2 <ns0:ref type='bibr' target='#b21'>(Chetri, Odden & Wegge, 2017)</ns0:ref>. All accessible areas are used for livestock grazing following the seasonal traditional Tibetan calendar <ns0:ref type='bibr' target='#b19'>(Chetri et al., 2019a)</ns0:ref>. Grazing areas are designated for seasonal grazing as But in this study we will use VDC as the data was collected prior to this change.</ns0:p><ns0:p>Each VDC has separate designated grazing areas. Among the 21 VDCs in the study area (6621 km 2 ), 2934 km 2 (44.3%) was used for livestock grazing (summer 55.6%, winter 24.6% and 19.8% year-round). The remaining areas (ca.3687 km 2 ) were inaccessible for livestock grazing due to rugged terrain and high altitude (Fig. <ns0:ref type='figure' target='#fig_3'>1</ns0:ref>). Our survey covered 13% of the total number of households within the survey villages <ns0:ref type='bibr'>(CBS, 2012)</ns0:ref>. Due to scattered settlements/villages, vast landscape and remoteness of the area, most of the questionnaires were conducted using locally trained community members, managed through the Unit Conservation Offices (UCOs) of ACA and MCA <ns0:ref type='bibr' target='#b19'>(Chetri et al., 2019a)</ns0:ref>. Before the initiation of the survey, each interviewer was briefed about the purpose of the study and trained in how to conduct the semi-structured questionnaire, Manuscript to be reviewed and verbal consent was obtained from all subjects. The survey households were selected following the main village trails. We approached the household closest to the main village trail and selected every third household thereafter for interviews. If the inhabitants were absent, we selected the nearest neighbor. For each respondent, we recorded the number of livestock owned, herd composition and livestock loss to snow leopards and wolves during the previous year. We also recorded respondent age, gender, education and occupation. We asked their opinion about the presence of snow leopards and wolves near their grazing areas and homesteads, and categorized their answers as positive, neutral and negative. We considered questionnaires as invalid when respondents stated that they did not know about the species presence and conflict (i.e., neutral responses). These questionnaires were excluded from the analyses. Hence, although we administered similar questionnaire sets to assess perceptions of snow leopards and wolves, the sample size for wolf perceptions became smaller due to a larger proportion of invalid questionnaires (Table <ns0:ref type='table' target='#tab_1'>1</ns0:ref>). This was mainly because wolves are found only in the northwestern section of ACA and MCA (see Fig. <ns0:ref type='figure' target='#fig_3'>1</ns0:ref>).</ns0:p></ns0:div>
<ns0:div><ns0:head>Data analysis</ns0:head><ns0:p>We used generalized linear mixed effects models (GLMMs) to determine the relationship between response and potential predictor variables with two separate sets of models, one for snow leopards and one for wolves. GLMMs take into account random effects and provide a more flexible approach for analyzing non-normal data <ns0:ref type='bibr' target='#b10'>(Bolker et al., 2009)</ns0:ref>. As a binomial response variable, we categorized opinions of presence of snow leopards and wolves as either positive or negative, as described previously. As explanatory variables, we included factors and covariates that were identified as important predictors of livestock losses in a previous study conducted in the region <ns0:ref type='bibr' target='#b19'>(Chetri et al., 2019a)</ns0:ref>, as well as demographic and socioeconomic variables that have been linked to perceptions of carnivores in previous studies (see e.g. <ns0:ref type='bibr' target='#b12'>Caruso et al., 2020;</ns0:ref><ns0:ref type='bibr' target='#b39'>Kellert & Berry, 1987;</ns0:ref><ns0:ref type='bibr' target='#b43'>Kusi et al., 2019;</ns0:ref><ns0:ref type='bibr' target='#b59'>Oli, Taylor & Rogers 1994;</ns0:ref><ns0:ref type='bibr' target='#b68'>Suryawanshi et al., 2014)</ns0:ref> hrs. walking/day). We did not include religion as a predictive factor as all the locals residing in the region are Buddhist, only a few outsiders who are working as labor or teacher are non-Buddhist. We standardized all numeric explanatory variables by two standard deviations, following <ns0:ref type='bibr' target='#b31'>Gelman & Hill (2007)</ns0:ref>. VDC was used as a random effect in all models. We checked correlation among (continuous) predictors before carrying out the regression analysis, and we did not include collinear variables (rho > 0.6) into the same model. We analyzed the data using R version 3.4.2 (R CoreTeam, 2017).</ns0:p></ns0:div>
<ns0:div><ns0:head>RESULTS</ns0:head><ns0:p>We used only 395 and 327 questionnaires in our analyses regarding snow leopards and wolves, respectively (Table <ns0:ref type='table' target='#tab_1'>1</ns0:ref>), due to the exclusion of respondents that were neutral or unaware of species presence or conflicts. In terms of gender, more than 80% of the respondents were male (Table <ns0:ref type='table' target='#tab_1'>1</ns0:ref>), and approximately 50% of the respondents were illiterate. Among occupations, most respondents belonged to the agro-pastoralist category (Table <ns0:ref type='table' target='#tab_1'>1</ns0:ref>). An analysis of respondents'</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:04:48442:2:0:NEW 10 Sep 2020)</ns0:p><ns0:p>Manuscript to be reviewed perceptions in general revealed that local people were more negative towards wolves (n=276, 84.4%) than towards snow leopards (n=209, 52.9%) (Table <ns0:ref type='table' target='#tab_1'>1</ns0:ref>). </ns0:p></ns0:div>
<ns0:div><ns0:head>Local perceptions of snow leopards</ns0:head><ns0:p>We compared 22 candidate models to assess perceptions of snow leopards (Table <ns0:ref type='table' target='#tab_5'>S1</ns0:ref>). The two highest ranking models had a small difference in AICc value (ΔAICc = 0.27) and Akaike weights (0.45 and 0.39). Both models included the predictor variables occupation, sex and the interaction between age and literacy (Table <ns0:ref type='table' target='#tab_2'>2</ns0:ref>). The top ranking model in terms of AICc also included ownership, but due to the marginal effect of removing this variable, we present here the simpler second ranking model (Fig. <ns0:ref type='figure' target='#fig_5'>2</ns0:ref>). The occupation of the respondents had a strong effect on their perceptions of snow leopards. Among the three categories of occupation, there was only a slight difference in perceptions between agro-pastoralists (income from both agriculture and livestock) and herders (income solely from livestock herding). On the contrary, respondents with other additional sources of income (e.g. tourism) were far more positive towards snow leopards (i.e. OCCOTHER, Fig. <ns0:ref type='figure' target='#fig_5'>2</ns0:ref>). Furthermore, sex was included in the model, but the effect was weak, i.e. men were more positive towards snow leopards than women. The main effects of literacy and age had very weak impacts on perceptions, but the interaction between these terms showed a marked effect. As illustrated in Fig. <ns0:ref type='figure' target='#fig_5'>2</ns0:ref>, being illiterate had a more marked negative impact among the older respondents.</ns0:p></ns0:div>
<ns0:div><ns0:head>Local perceptions of wolves</ns0:head><ns0:p>PeerJ reviewing PDF | (2020:04:48442:2:0:NEW 10 Sep 2020)</ns0:p><ns0:p>Manuscript to be reviewed</ns0:p><ns0:p>We compared 24 candidate models to assess perceptions of wolves (Table <ns0:ref type='table' target='#tab_8'>S2</ns0:ref>). The highest ranking model performed far better than the other candidates (Akaike weight = 0.70, Table <ns0:ref type='table' target='#tab_2'>2</ns0:ref>).</ns0:p><ns0:p>This model included different predictors than the model for snow leopards, i.e. sex and ownership (Fig. <ns0:ref type='figure' target='#fig_6'>3</ns0:ref>). In this case, male respondents were markedly more positive than females.</ns0:p><ns0:p>Other predictor variables were livestock loss (numbers lost to wolves), herd composition, distance to the nearest conservation field office and livestock ownership. The latter predictor had a marked effect on perceptions, i.e. respondents with larger herds were more negative. </ns0:p></ns0:div>
<ns0:div><ns0:head>DISCUSSION</ns0:head><ns0:p>In our study landscape, a far larger proportion of respondents were negative towards wolves than to snow leopards. This was also observed by <ns0:ref type='bibr' target='#b43'>Kusi et al. (2019)</ns0:ref> in upper Dolpa and Humla areas, located in the western region of Nepal. This perception is common in areas where wolves coexist with other large predators, for example brown bear and lynx <ns0:ref type='bibr' target='#b69'>(Trajçe et al., 2019)</ns0:ref>. Manuscript to be reviewed livestock both during day and night <ns0:ref type='bibr' target='#b78'>(Xu, Yang & Dou, 2015)</ns0:ref>. Furthermore, research has shown that greater visibility and howling behavior of wolves may reinforce negative perceptions <ns0:ref type='bibr' target='#b40'>(Kellert et al., 1996)</ns0:ref>.</ns0:p><ns0:p>Social norms and cultural beliefs also play an important role in perceptions of the two carnivores.</ns0:p><ns0:p>Cultural sentiments, religious belief and folklore associated with snow leopards have a strong positive influence on their conservation <ns0:ref type='bibr' target='#b0'>(Ale, Shah & Jackson, 2016;</ns0:ref><ns0:ref type='bibr' target='#b44'>Li et al., 2014)</ns0:ref>. Further, the local practice of non-violence (e.g. Tsum valley of MCA) and protection of forest and landscape in the name of monasteries <ns0:ref type='bibr' target='#b44'>(Li et al., 2014)</ns0:ref> have also played an important role in snow leopard conservation. The Buddhist communities to which most of our respondents belong traditionally do not kill wildlife because it was considered a sin in their religion <ns0:ref type='bibr' target='#b44'>(Li et al., 2014)</ns0:ref>. The snow leopard is often considered as a symbol of the mountains, and the charisma of the species promotes attention both in terms of research and conservation efforts from global and national conservation authorities <ns0:ref type='bibr' target='#b51'>(McCarthy et al., 2016)</ns0:ref>. In contrast, wolves are traditionally depicted as merciless and evil creatures in legends and folklore <ns0:ref type='bibr' target='#b27'>(Dingwall, 2001;</ns0:ref><ns0:ref type='bibr' target='#b49'>Marvin, 2012)</ns0:ref>. A recent study from Spiti, India showed that more than 98% of the survey respondents claimed that wolves were not safe for livestock and their presence was highly disliked by the communities <ns0:ref type='bibr' target='#b47'>(Lyngdoh & Habib, 2019)</ns0:ref>. Similar trends have been observed in parts of Europe <ns0:ref type='bibr' target='#b27'>(Dingwall, 2001;</ns0:ref><ns0:ref type='bibr' target='#b49'>Marvin, 2012)</ns0:ref> and America <ns0:ref type='bibr' target='#b33'>(Grima, Brainard & Fisher, 2019)</ns0:ref>. This is not surprising as dislike to wolves is common across the globe <ns0:ref type='bibr' target='#b7'>(Bhatia et al., 2016;</ns0:ref><ns0:ref type='bibr' target='#b29'>Dressel, Sandström & Ericsson, 2015;</ns0:ref><ns0:ref type='bibr' target='#b38'>Kansky, Kidd & Knight, 2014;</ns0:ref><ns0:ref type='bibr' target='#b43'>Kusi et al., 2019;</ns0:ref><ns0:ref type='bibr' target='#b47'>Lyngdoh & Habib, 2019;</ns0:ref><ns0:ref type='bibr' target='#b68'>Suryawanshi et al., 2014)</ns0:ref>. The negative perceptions of the wolf in our study area thus probably</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:04:48442:2:0:NEW 10 Sep 2020)</ns0:p><ns0:p>Manuscript to be reviewed due to fear and cultural bias as reported in many other studies <ns0:ref type='bibr' target='#b46'>(Linnell et al., 2002;</ns0:ref><ns0:ref type='bibr' target='#b61'>Prokop, Usak & Erdogan, 2011)</ns0:ref>, an issue to be considered in future conservation plans in ACA and MCA.</ns0:p><ns0:p>In our study area, analysis of livestock depredation revealed higher losses from snow leopards compared to wolves <ns0:ref type='bibr' target='#b19'>(Chetri et al., 2019a)</ns0:ref>, but still the tolerance level of local communities towards snow leopards was higher. Tolerance to snow leopards in ACA has changed to become more positive compared with an earlier study <ns0:ref type='bibr' target='#b59'>(Oli, Taylor & Rogers, 1994)</ns0:ref>, probably as a result of continued efforts to increase awareness as part of an ongoing conservation program <ns0:ref type='bibr' target='#b28'>(DNPWC, 2017)</ns0:ref>. No such efforts have targeted wolves, or other coexisting carnivores in this area.</ns0:p><ns0:p>Our model revealed that literacy, age, occupation, number of livestock owned and gender affected perceptions towards snow leopards and wolves. However, the predictors for the two species were different (see Fig. <ns0:ref type='figure' target='#fig_5'>2</ns0:ref> and 3), i.e. the latter two appeared in the best model for wolf perceptions and the former three for snow leopard. Regarding literacy and age, only the interaction between these terms had an influence on perceptions of snow leopards, but not the main effects. Being illiterate was associated with negative perceptions among older respondents.</ns0:p><ns0:p>Possibly, younger people had more exposure to snow leopard conservation campaigns, regardless of literacy. Several earlier studies have shown that older people are more negative towards large predators and usually less supportive of their conservation than the younger generation <ns0:ref type='bibr' target='#b6'>(Bencin, Kioko & Kiffner, 2016;</ns0:ref><ns0:ref type='bibr' target='#b39'>Kellert & Berry, 1987;</ns0:ref><ns0:ref type='bibr' target='#b42'>Kleiven, Bjerke & Kaltenborn, 2004;</ns0:ref><ns0:ref type='bibr' target='#b65'>Røskaft et al., 2007;</ns0:ref><ns0:ref type='bibr' target='#b74'>Williams, Ericsson & Heberlein, 2002)</ns0:ref>.</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:04:48442:2:0:NEW 10 Sep 2020)</ns0:p><ns0:p>Manuscript to be reviewed</ns0:p><ns0:p>Occupation influenced perceptions of snow leopards, i.e. people with sources of income other than animal husbandry were more positive. Elsewhere, it was also reported that people having smaller landholdings and few economic opportunities other than livestock herding are more negative towards snow leopards and wolves <ns0:ref type='bibr' target='#b5'>(Bagchi & Mishra, 2006;</ns0:ref><ns0:ref type='bibr' target='#b25'>Din et al., 2017)</ns0:ref>. In a study of jaguars (Panthera onca), <ns0:ref type='bibr' target='#b12'>Caruso et al. (2020)</ns0:ref> found a similar pattern; people's perceptions and attitudes were strongly influenced by occupation and economic benefits through ecotourism. In Ladakh, India snow leopard based ecotourism has become popular and provides income generation opportunities to the local communities <ns0:ref type='bibr' target='#b35'>(Jackson, 2015;</ns0:ref><ns0:ref type='bibr' target='#b48'>Maheshwari & Sathyakumar, 2019;</ns0:ref><ns0:ref type='bibr' target='#b71'>Vannelli et al., 2019)</ns0:ref>.</ns0:p><ns0:p>Regarding perceptions of wolves, males were more positive than females. This pattern was also reported in earlier studies <ns0:ref type='bibr' target='#b39'>(Kellert & Berry, 1987;</ns0:ref><ns0:ref type='bibr' target='#b65'>Røskaft et al., 2007;</ns0:ref><ns0:ref type='bibr' target='#b68'>Suryawanshi et al., 2014)</ns0:ref>, and has been explained by women having less contact with conservation agencies <ns0:ref type='bibr' target='#b32'>(Gillingham & Lee, 1999)</ns0:ref>. Another study suggested that the negative attitudes of women might be a result of greater perception of risk or fear <ns0:ref type='bibr' target='#b60'>(Prokop & Fančovičová, 2010)</ns0:ref>. As suggested by <ns0:ref type='bibr' target='#b43'>Kusi et al. (2019)</ns0:ref>, men in the Himalayas often migrate outside of villages for seasonal work and may thus have been more exposed to alternative attitudes to nature and conservation. In addition, men frequently venture into the pasture for livestock grazing activities and presumably had more encounters with wolves, which make them understand their behaviour and threats. High encounter rates with wolves either in the wild or in captivity, may promote more positive perceptions of the animals <ns0:ref type='bibr' target='#b2'>(Arbieu et al., 2020)</ns0:ref>.</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:04:48442:2:0:NEW 10 Sep 2020)</ns0:p><ns0:p>Manuscript to be reviewed</ns0:p><ns0:p>People holding large livestock herds were more negative towards wolves, which agrees with a study from western China <ns0:ref type='bibr'>(Xu, Yang & Dong, 2015)</ns0:ref>. A possible explanation is that owners with larger herds have a higher risk of suffering losses in the central Himalayas, especially farmers having a larger herd of goat/sheep <ns0:ref type='bibr' target='#b21'>(Chetri, Odden & Wegge, 2017)</ns0:ref>. It is, however, notable that having experienced losses did not affect perceptions of snow leopards, and the effect of perception on wolves was weak. This is in contrast with a recent study from the Nepal Himalayas where livestock depredation by wolves is the main predictor of the negative attitude towards wolves <ns0:ref type='bibr' target='#b43'>(Kusi et al., 2019)</ns0:ref>. However, such a pattern was not recorded in our study area and may be due to the fact that average losses in our study area were quite low (~1% of all livestock holdings).</ns0:p><ns0:p>In our study area, the National Trust for Nature Conservation (NTNC) has been implementing community-based conservation projects and programs since 1992. The overall goal is to conserve biodiversity of global significance with the active participation of local communities.</ns0:p><ns0:p>Integrated conservation and development efforts have therefore addressed the communities' needs and demands while actively mobilizing local people in conservation efforts. However, even after 2-3 decades of conservation initiatives, local perceptions and tolerance towards carnivores are still rather negative, particularly towards wolves. We therefore recommend a wider perspective of future awareness campaigns to include a broader specter of species and conservation issues with particular efforts to the Himalayan wolf. During interviews, we observed that remote settlements had rarely been visited by conservation authorities, and the inhabitants there had limited knowledge of compensation policies for livestock losses and human injury. Local perceptions on wolves tended to be more negative with increasing distances from conservation field offices, and this has been reported in earlier research in parts of ACA <ns0:ref type='bibr' target='#b59'>(Oli, Taylor & Rogers, 1994)</ns0:ref>. The factors underlying negative perceptions of distant settlements are probably due to a limited local involvement in community conservation programs. In the future, distant and remote settlements require more rigorous conservation outreach and awareness activities.</ns0:p></ns0:div>
<ns0:div><ns0:head>CONCLUSIONS</ns0:head><ns0:p>This study has investigated local villagers' perception of snow leopards and wolves in the central Himalayas, Nepal. In general, the perceptions of locals were more positive towards snow leopards than to wolves. People having larger herds of livestock (goat/sheep) with limited access to conservation programs were more likely to have negative perceptions towards wolves. Our results showed that multiple factors influence local perceptions of the two carnivores and that perception factors cannot be generalized for the two species. Thus, they need to be addressed separately. We suggest that future conservation projects and programs prioritize remote settlements. Furthermore, considering the substantial influence of occupation on people's perceptions of carnivores, certain parts of the landscape, for example Manang of ACA and Tsum valley of MCA, should be tested for the development of wildlife based ecotourism.</ns0:p><ns0:note type='other'>Figure 1</ns0:note><ns0:p>Study area with location of survey villages and grazing areas.</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:04:48442:2:0:NEW 10 Sep 2020)</ns0:p><ns0:p>Manuscript to be reviewed 4 Only respondents that responded to perceptions and share their experiences were included.</ns0:p><ns0:note type='other'>Figure 2</ns0:note><ns0:p>5 Number in parenthesis indicates percentage of individual respondents in each category.</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:04:48442:2:0:NEW 10 Sep 2020)</ns0:p><ns0:p>Manuscript to be reviewed 6 ACA (hereafter), Annapurna Conservation Area.</ns0:p><ns0:p>7 MCA (hereafter), Manaslu Conservation Area.</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:04:48442:2:0:NEW 10 Sep 2020)</ns0:p><ns0:p>Manuscript to be reviewed </ns0:p></ns0:div><ns0:figure xml:id='fig_0'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:04:48442:2:0:NEW 10 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_1'><ns0:head /><ns0:label /><ns0:figDesc>Integrated conservation and development efforts that were initiated in ACA and MCA in the 1990-ies included conservation awareness campaigns principally targeting snow leopard, but not PeerJ reviewing PDF | (2020:04:48442:2:0:NEW 10 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_2'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:04:48442:2:0:NEW 10 Sep 2020)Manuscript to be reviewed summer and winter. Areas close to villages are used for year-round grazing (Fig.1). Livestock, for example sheep, goats and cows are usually herded and periodically moved among different pastures according to seasons<ns0:ref type='bibr' target='#b21'>(Chetri, Odden & Wegge, 2017)</ns0:ref>. Small stocks (sheep and goats) are herded and sometimes accompanied by herding dogs. They are released in the morning and brought back to corrals/pens in the afternoon on a regular basis. Similarly, milking cows and, yaks are brought back to corrals/pens in the afternoon or in the morning for milking. Livestock are kept in corrals/pens for protection against predators corrals are traditionally made of mud walls and stones.Over the last decade there have been considerable changes in the lifestyle of the local people due to the development of roads in ACA. Despite these changes, traditional agro-pastoral lifestyles remain intact, and most importantly, traditional livestock grazing and collective village level decision making and implementation is still functional. In ACA, most farmers prefer to raise goats (Capra hircus) and sheep (Ovis aries) due to abundant pastureland, whereas in MCA farmers prefer cattle-yak hybrids (dzo, Jhopas, Bos spp.) as they are both grazers and browsers.Similarly, in the central part of ACA, farmers prefer yaks (Bos grunniens) due to dominant scrub vegetation. Lulu cows (Bos taurus sp.) and horses (Equus ferus caballus) are common in all areas. Among the main wild ungulates, bharal (Pseudois nayaur) and Himalayan tahr(Hemitragus jemlahicus) are widespread, whereas Tibetan argali (Ovis ammon hogdsoni), kiang (Equus kiang), and Tibetan gazelle (Procapra picticaudata) have overlapping grazing areas with livestock in the north-western parts of ACA. PeerJ reviewing PDF | (2020:04:48442:2:0:NEW 10 Sep 2020) Manuscript to be reviewed Apart from snow leopards and wolves, other carnivores include golden jackal, red fox (Vulpes vulpes), Himalayan black bear, Tibetan sand fox (Vulpes ferrilata), brown bear, Eurasian lynx and several species of weasel (Mustela spp.), and marten (Martes spp.).</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_3'><ns0:head>Figure 1 :</ns0:head><ns0:label>1</ns0:label><ns0:figDesc>Figure 1: Study area with location of survey villages and grazing areas.</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_4'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:04:48442:2:0:NEW 10 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_5'><ns0:head>Figure 2 :</ns0:head><ns0:label>2</ns0:label><ns0:figDesc>Figure 2: Parameter estimates based on Generalized Linear Mixed-Effects Models of</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_6'><ns0:head>Figure 3 :</ns0:head><ns0:label>3</ns0:label><ns0:figDesc>Figure 3: Parameter estimates based on Generalized Linear Mixed-Effects Models of</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_7'><ns0:head /><ns0:label /><ns0:figDesc>A possible cause is related to the difference in behavior of wolves compared to other carnivores. Snow leopards are cryptic, avoid humans and are more nocturnal than wolves (McCarthy, Fuller & Munkhtsog, 2005; Mech & Boitani, 2010). Wolves are more active during the day and attack PeerJ reviewing PDF | (2020:04:48442:2:0:NEW 10 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_8'><ns0:head>Figure 2</ns0:head><ns0:label>2</ns0:label><ns0:figDesc>Figure 2 Parameter estimates based on Generalized Linear Mixed-Effects Models of factors affecting perceptions of snow leopards.</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_9'><ns0:head>Figure 3 Figure 3</ns0:head><ns0:label>33</ns0:label><ns0:figDesc>Figure 3</ns0:figDesc></ns0:figure>
<ns0:figure type='table' xml:id='tab_1'><ns0:head>Table 1 : Overview of respondent characteristics in the Annapurna-Manaslu landscape, central Himalayas, Nepal.</ns0:head><ns0:label>1</ns0:label><ns0:figDesc /><ns0:table /></ns0:figure>
<ns0:figure type='table' xml:id='tab_2'><ns0:head>Table 2 : Model selection for perception towards the snow leopard and the wolf.</ns0:head><ns0:label>2</ns0:label><ns0:figDesc /><ns0:table /></ns0:figure>
<ns0:figure type='table' xml:id='tab_3'><ns0:head>Table 1 (on next page)</ns0:head><ns0:label>1</ns0:label><ns0:figDesc /><ns0:table /></ns0:figure>
<ns0:figure type='table' xml:id='tab_4'><ns0:head>Table 1</ns0:head><ns0:label>1</ns0:label><ns0:figDesc>Overview of respondent characteristics in the Annapurna-Manaslu landscape, central Himalayas, Nepal.</ns0:figDesc><ns0:table><ns0:row><ns0:cell>Only respondents that responded to perceptions and share their experiences were included.</ns0:cell></ns0:row><ns0:row><ns0:cell>Number in parenthesis indicates percentage of individual respondents in each category. ACA,</ns0:cell></ns0:row><ns0:row><ns0:cell>Annapurna Conservation Area; MCA, Manaslu Conservation Area.</ns0:cell></ns0:row></ns0:table><ns0:note>PeerJ reviewing PDF | (2020:04:48442:2:0:NEW 10 Sep 2020)Manuscript to be reviewed 1</ns0:note></ns0:figure>
<ns0:figure type='table' xml:id='tab_5'><ns0:head>Table 1 : Overview of respondent characteristics in the Annapurna-Manaslu landscape, 2 central Himalayas, Nepal.</ns0:head><ns0:label>1</ns0:label><ns0:figDesc /><ns0:table><ns0:row><ns0:cell>Snow leopards</ns0:cell><ns0:cell>Wolves</ns0:cell></ns0:row></ns0:table><ns0:note>3 Notes.</ns0:note></ns0:figure>
<ns0:figure type='table' xml:id='tab_6'><ns0:head>Table 2 (on next page)</ns0:head><ns0:label>2</ns0:label><ns0:figDesc /><ns0:table /></ns0:figure>
<ns0:figure type='table' xml:id='tab_7'><ns0:head>Table 2</ns0:head><ns0:label>2</ns0:label><ns0:figDesc>Model selection for perception towards the snow leopard and the wolf. continuous variables were standardized by 2 standard deviations (as per<ns0:ref type='bibr' target='#b31'>Gelman and Hill, 2007)</ns0:ref> and all models included a varying intercept on VDC (Village Development Committee).VDC is included as a random effect. AGE: age of the respondent, COMP: composition of the herd, i.e. proportion of large stock animals, LIT: literacy (yes / no), LOSS: number of domestic animals lost to the carnivore, OCC: respondent's occupation (Herding, Agriculture, Other), OWN: number of domestic animals owned, SEX: gender of the respondent. DIST=Distance from the nearest conservation field office to respondent household. Only the top 10 models are presented for each analysis.</ns0:figDesc><ns0:table><ns0:row><ns0:cell>All</ns0:cell></ns0:row></ns0:table><ns0:note>PeerJ reviewing PDF | (2020:04:48442:2:0:NEW 10 Sep 2020)Manuscript to be reviewed 1</ns0:note></ns0:figure>
<ns0:figure type='table' xml:id='tab_8'><ns0:head>Table 2 : Model selection for perception towards the snow leopard and the wolf. Model df logLik AICc delta weight</ns0:head><ns0:label>2</ns0:label><ns0:figDesc>All continuous variables were standardized by 2 standard deviations (as per<ns0:ref type='bibr' target='#b31'>Gelman and Hill, 2007)</ns0:ref> and all models included a varying intercept on VDC (Village Development Committee).VDC is included as a random effect. AGE: age of the respondent, COMP: composition of the herd, i.e. proportion of large stock animals, LIT: literacy (yes / no), LOSS: number of domestic animals lost to the carnivore, OCC: respondent's occupation (Herding, Agriculture, Other), OWN: number of domestic animals owned, SEX: gender of the respondent. DIST=Distance from the nearest conservation field office to respondent household. Only the top 10 models are presented for each analysis.</ns0:figDesc><ns0:table><ns0:row><ns0:cell>Snow Leopard</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>OWN + OCC + SEX + AGE * LIT</ns0:cell><ns0:cell cols='2'>9 -233.24 484.9 0</ns0:cell><ns0:cell>0.45</ns0:cell></ns0:row><ns0:row><ns0:cell>OCC + SEX + AGE * LIT</ns0:cell><ns0:cell cols='2'>8 -234.42 485.2 0.27</ns0:cell><ns0:cell>0.39</ns0:cell></ns0:row><ns0:row><ns0:cell cols='3'>COMP * LOSS + OWN + OCC + SEX + AGE * LIT 12 -231.64 488.1 3.15</ns0:cell><ns0:cell>0.09</ns0:cell></ns0:row><ns0:row><ns0:cell>COMP * LOSS + OCC + SEX + AGE * LIT</ns0:cell><ns0:cell>11 -233.63 490</ns0:cell><ns0:cell>5.02</ns0:cell><ns0:cell>0.04</ns0:cell></ns0:row><ns0:row><ns0:cell>COMP * LOSS + OWN + OCC + SEX + AGE</ns0:cell><ns0:cell>10 -234.74 490</ns0:cell><ns0:cell>5.11</ns0:cell><ns0:cell>0.04</ns0:cell></ns0:row><ns0:row><ns0:cell>OCC</ns0:cell><ns0:cell cols='3'>4 -245.06 498.2 13.28 0</ns0:cell></ns0:row><ns0:row><ns0:cell>COMP * LOSS + OWN + OCC</ns0:cell><ns0:cell cols='3'>8 -241.52 499.4 14.47 0</ns0:cell></ns0:row><ns0:row><ns0:cell>COMP * LOSS + OWN + OCC + SEX</ns0:cell><ns0:cell cols='3'>9 -240.97 500.4 15.46 0</ns0:cell></ns0:row><ns0:row><ns0:cell>AGE</ns0:cell><ns0:cell cols='3'>3 -247.76 501.6 16.64 0</ns0:cell></ns0:row><ns0:row><ns0:cell>LIT</ns0:cell><ns0:cell>3 -251.48 509</ns0:cell><ns0:cell cols='2'>24.08 0</ns0:cell></ns0:row><ns0:row><ns0:cell>Wolf</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>COMP + DIST + LOSS + SEX + OWN</ns0:cell><ns0:cell cols='2'>7 -104.38 223.1 0</ns0:cell><ns0:cell>0.7</ns0:cell></ns0:row><ns0:row><ns0:cell>COMP * LOSS + OWN + OCC + SEX</ns0:cell><ns0:cell cols='2'>9 -104.88 228.3 5.22</ns0:cell><ns0:cell>0.05</ns0:cell></ns0:row><ns0:row><ns0:cell>SEX</ns0:cell><ns0:cell cols='2'>3 -111.33 228.7 5.62</ns0:cell><ns0:cell>0.04</ns0:cell></ns0:row><ns0:row><ns0:cell>OWN</ns0:cell><ns0:cell cols='2'>3 -111.63 229.3 6.22</ns0:cell><ns0:cell>0.03</ns0:cell></ns0:row><ns0:row><ns0:cell>COMP + LOSS + OWN</ns0:cell><ns0:cell cols='2'>5 -109.59 229.4 6.26</ns0:cell><ns0:cell>0.03</ns0:cell></ns0:row><ns0:row><ns0:cell>COMP + LOSS + OWN + DIST</ns0:cell><ns0:cell cols='2'>6 -108.75 229.8 6.65</ns0:cell><ns0:cell>0.03</ns0:cell></ns0:row></ns0:table><ns0:note>PeerJ reviewing PDF | (2020:04:48442:2:0:NEW 10 Sep 2020) Manuscript to be reviewed COMP * LOSS + OWN + OCC + SEX + AGE PeerJ reviewing PDF | (2020:04:48442:2:0:NEW 10 Sep 2020)</ns0:note></ns0:figure>
</ns0:body>
" | "September 10th, 2020
Dear Editors,
We thank the reviewers for their great efforts and comments on the manuscript. The comments are quite useful and we have edited and made changes to address their concerns. We will also like to thank the editors for your great efforts in bringing the manuscript into this shape.
Our responses are in bold letters. We believe that the manuscript has improved a lot and is now suitable for publication in PeerJ.
Thank you very much.
Dr. Madhu Chetri
On behalf of all authors.
___________________________________________________________
Comments from the reviewers
Reviewer: María Losada
Basic reporting
The manuscript fulfills the general standards and all improvements suggested were effectively performed. I consider the introduction to have improved notably, and the background and the aims of the study are now sufficient to justify the novelty of the current research. The polishing edition is notable in the new version of the manuscript. However, some minor comments are included below for improving it.
Minor comments:
1. Thank you for including all the necessary corrections regarding the in-text citations and references section. In Line 42, please remove “de” before “Azevedo”. In Line 55, please remove the second dot in the last citation. In Line 82, I would suggest adding the reference citation (Chetri et al., 2019a), as an example of the background of the study. Please notice that parentheses () are not required after “et al.” in Line 194. In Lines 200, 252, and 304, a comma (,) is not required after the author names. In Line 512, Mijiddorj, Alexander and Samelius (2018) reference is not cited in the text. In Line 577, Zimmermann, Wabakken and Dötterer (2001) reference should be in a separate line. According to PeerJ submission guidelines, for three or fewer authors, you need to list all author names (e.g. Zimmermann, Wabakken and Dötterer, 2001). Please check the references section and in-text citations.
Response: Thank you for pointing out this, we have updated the references as well as citation in the texts.
2. I appreciate all the improvements implemented in the introduction section. However, I believe that the question concerning the density of Himalayan wolves has not been resolved yet. I would suggest taking into account the following references “WWF (2015) Non-Invasive Genetic Population Survey of Snow Leopard and Wolf. Final report. WWF, Kathmandu, Nepal”; “Chetri et al. (2016) ZooKeys; and Chetri et al. (2017) PLoS ONE” and include them in Line 72. In Line 75, please consider removing “However” in the beginning, there is no opposition here. Thank you for defining the acronyms in Lines 86-87. I would suggest adding the following specification: (ACA, hereafter) and (MCA, hereafter). In Line 101, I would recommend implementing the plural in “low livestock loss”.
Response: Thank you for pointing out this, we have added the references and followed the suggestion accordingly.
3. I believe that the statement in the data analysis subsection helps to clarify the selection of the most appropriate predictors recorded through the semi-structured questionnaire. Thank you for including all the recommended predictors in the analysis, and all the changes needed in all tables and figures. Please consider including units (years) for “Mean age” and correct “parentheses” in the footnote of Table 1. Moreover, I recommend to add the definition of ownership predictor, and to replace “LIT1” with “LITyes”, and “AGE:LIT1” with “AGE:LITyES” in the legend of Figure 2 (please, check if the figure shown matches the selected model). Please, consider replacing “snow leopard” with “Himalayan wolf” when defining the LOSS predictor in the footnote of Table S2 (Supplemental files).
Response: Thank you for pointing out this, we have upgraded accordingly. We have also corrected Table 1. Also legend of Figure 2 and also predictor in the footnote of Table S2.
Experimental design
The study aims are well defined in the introduction section, and the novelty of the current research is well justified. Moreover, the research complies with technical and ethical requirements, and the methods section shows significant improvements, according to the comments received. Thank you for removing neutral respondents and for including a statement in the data analyses subsection in order to clarify the study design and model selection. However, I will provide some feedback in order to improve the manuscript.
Minor comments:
1. Thank you for improving the paragraph structure in the study area subsection. In Line 137, I recommend changing “Bos spp.” to the italic font.
Response: Thank you for your suggestion, we have corrected, line 159.
2. In Line 176, I understand that the invalid questionnaires belong to respondents who gave neutral answers or ignored the presence of study species or conflicts with them. If that is correct, I would suggest including “(i.e., neutral responses)” after 'conflict' for further clarification.
Response: Thank you for your suggestion, we have updated accordingly, line 200.
3. I agree with the reasons for excluding religion from the analyses. However, I would consider this explanation (or similar) to be included in Line 199: “we did not include religion as a predictive factor as all the locals residing in the region are Buddhist, only a few outsiders who are working as labor or teacher are non-Buddhist.”
Response: Thank you for your suggestion, we have added a sentence, lines 223-225 .
Validity of the findings
The results of this study are widely supported. The findings are quite robust and statistically sound, with a significant application to future conservation plans for these large carnivore species in the Himalayan range. I would highlight the improvements made in the introduction as well as in the discussion and conclusion sections, which are now further strengthened in terms of the aims and background of this study. Thank you for including all the comments provided previously, particularly with reference to statistical analysis. However, I will provide some feedback to improve your manuscript.
Response: Thank you for your feeback and suggestions. We have improved as per your suggestions.
Minor comments:
1. Results are rather strong and robust. However, I would suggest replacing “literate” with “being illiterate” in Line 231. Moreover, I would recommend checking if Figure 2 matches with the selected model. In this case, please consider adding 'and ownership' after “sex” in Line 237.
Response: Thank you, we have corrected, line 266.
2. Thank you for all your answers and changes implemented in the discussion section. In Line 266, I recommend adding “, to which most of our respondents belong,” after “Buddhist communities”. In Line 275, I would suggest replacing “dislike for” with “aversion to”. Please consider replacing “An other” with “another” in Line 313, and “perceived” with “perception of” in Line 314. In Line 353, please leave a space between “Theyneed”.
Response: Thank you, we have updated accordingly.
3. With regard to the impact of your findings, I suggest remarking which aspects are different from the study of Kusi et al. (2019). For example, your research was performed in a different study area. In Line 329, I would recommend comparing your results here in more detail (Fig. 3) in contrast to Kusi et al.
Response: Thank you for your suggestion. We have edited the last paragraph, lines 377-381.
4. Thank you for emphasizing the difference in perceptions of wolves and snow leopards in the introduction and the discussion sections. However, I recommend adding “an issue to be considered in future conservation plans in ACA and MCA” at the end of Line 279. In Line 339, after “conservation issues”, please consider adding “with particular efforts to the Himalayan wolf”.
Response: Thank you, we have followed your suggestions, line 319 and 391 .
Comments for the author
I really appreciate all the improvements implemented in the current version of the manuscript. I would highlight the research efforts and the commitment of your research group with the projects developed for both conservation species in ACA and MCA. I highly encourage you to improve this manuscript and give these results the impact they deserve, with the ultimate goal of getting more funding for monitoring these two carnivore species and promoting future conservation and management strategies in remote settlements, particularly focused on the Himalayan wolf. I recommend you to apply the feedback provided in this review in order to get a more polished version of the article before publication.
Response: Thank you for your efforts in reviewing and providing us very useful suggestions.
" | Here is a paper. Please give your review comments after reading it. |
9,740 | "<ns0:abstract xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:p>Harmonia axyridis is an invasive alien ladybird in North America and Europe. Studies show that multiple natural enemies are using Ha. axyridis as a new host. However, thus far, no research has been undertaken to study the effects of simultaneous infections of multiple natural enemies on Ha. axyridis. We hypothesized that high thallus densities of the ectoparasitic fungus Hesperomyces virescens on a ladybird weaken the host's defenses, thereby making it more susceptible to infection by other natural enemies. We examined mortality of the North American-native Olla v-nigrum and Ha. axyridis co-infected with He.</ns0:p><ns0:p>virescens and an entomopathogenic fungus-either Beauveria bassiana or Metarhizium brunneum. Laboratory assays revealed that He. virescens-infected O. v-nigrum individuals are more susceptible to entomopathogenic fungi, but Ha. axyridis does not suffer the same effects. This is in line with the enemy release hypothesis, which predicts that invasive alien species in new geographic areas experience reduced regulatory effects from natural enemies compared to native species. Considering our results, we can ask how He. virescens affects survival when confronted by other pathogens that previously had little impact on Ha. axyridis.</ns0:p></ns0:div>
</ns0:abstract>
<ns0:body xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:head>Introduction</ns0:head><ns0:p>PeerJ reviewing PDF | (2020:04:48281:1:0:NEW 6 Aug 2020)</ns0:p></ns0:div>
<ns0:div><ns0:head>Manuscript to be reviewed</ns0:head><ns0:p>In nature, hosts may be exploited by more than one natural enemy. These organisms interact with each other and with their hosts <ns0:ref type='bibr' target='#b11'>(Furlong & Pell, 2005)</ns0:ref>. These complex interactions shape the population structure and dynamics of all organisms in the system. Natural enemies also compete with one another, and the impact on the host can be either synergistic, additive, or antagonistic <ns0:ref type='bibr' target='#b45'>(Shapiro-Ilan et al., 2012)</ns0:ref>. These interactions can be manifested in various aspects of host fitness or mortality. For example, biological control of Drosophila suzukii (Diptera, Drosophilidae), an important pest of fruit and berry crops, can be improved by treatments combining multiple natural enemies, which have an additive effect <ns0:ref type='bibr' target='#b30'>(Renkema & Cuthbertson, 2018)</ns0:ref>. At the same time, dual infections (even if causing an increase in host mortality) may be deleterious to one or both pathogens in terms of pathogen growth, fecundity, or other fitness parameters.</ns0:p><ns0:p>Harmonia axyridis (Coleoptera, Coccinellidae), native to eastern Asia, has rapidly increased its global range and is now present on all continents except Antarctica <ns0:ref type='bibr' target='#b20'>(Roy et al., 2016;</ns0:ref><ns0:ref type='bibr' target='#b5'>Camacho-Cervantes et al., 2017;</ns0:ref><ns0:ref type='bibr' target='#b18'>Hiller & Haelewaters, 2019)</ns0:ref>. Even though it has repeatedly been introduced for its beneficial properties as a biological control agent against aphid pests, its negative effects on native ladybird communities in invaded areas <ns0:ref type='bibr' target='#b23'>(Koch et al., 2006;</ns0:ref><ns0:ref type='bibr' target='#b20'>Honěk et al. 2016;</ns0:ref><ns0:ref type='bibr'>Brown & Roy, 2018)</ns0:ref> and on food production <ns0:ref type='bibr' target='#b23'>(Koch et al., 2006)</ns0:ref> have raised serious concerns since the early 2000s <ns0:ref type='bibr' target='#b20'>(Roy et al., 2016)</ns0:ref>. It is now a model organism for studying invasive alien species <ns0:ref type='bibr' target='#b41'>(Roy & Wajnberg, 2008;</ns0:ref><ns0:ref type='bibr'>Brown et al., 2018)</ns0:ref> and it has been listed in Europe as 'one of the worst' invasive species <ns0:ref type='bibr' target='#b28'>(Nentwig et al., 2018)</ns0:ref>. Harmonia axyridis is often reported as a host to several natural enemies. These include parasites (Hesperomyces virescens, Coccipolipus hippodamiae, Parasitylenchus bifurcatus), parasitoids (phorid and tachinid flies, Dinocampus coccinellae, Homalotylus spp., Tetrastichinae spp.), pathogens (bacteria, fungi, nematodes, protozoans), and predators (bugs, lacewings, ladybirds, and spiders) <ns0:ref type='bibr' target='#b12'>(Garcés & Williams, 2004;</ns0:ref><ns0:ref type='bibr' target='#b33'>Riddick et al., 2008;</ns0:ref><ns0:ref type='bibr' target='#b31'>Riddick, 2010;</ns0:ref><ns0:ref type='bibr' target='#b16'>Harding et al., 2011;</ns0:ref><ns0:ref type='bibr' target='#b29'>Raak-van den Berg et al., 2014;</ns0:ref><ns0:ref type='bibr' target='#b15'>Haelewaters et al., 2017;</ns0:ref><ns0:ref type='bibr' target='#b6'>Ceryngier et al. 2018)</ns0:ref>. Independent studies show that natural enemies of native ladybirds have recently employed Ha. axyridis as a new host, sometimes simultaneously <ns0:ref type='bibr' target='#b29'>(Raak-van den Berg et al., 2014;</ns0:ref><ns0:ref type='bibr' target='#b15'>Haelewaters et al., 2017;</ns0:ref><ns0:ref type='bibr' target='#b6'>Ceryngier et al. 2018;</ns0:ref><ns0:ref type='bibr' target='#b22'>Knapp et al., 2019)</ns0:ref>. Review of the effects of parasites, pathogens, and parasitoids of Ha. axyridis shows that they have only limited potential for controlling population densities of their host when acting alone <ns0:ref type='bibr'>(Roy et al., 2008;</ns0:ref><ns0:ref type='bibr' target='#b31'>Riddick, 2010;</ns0:ref><ns0:ref type='bibr' target='#b15'>Haelewaters et al., 2017;</ns0:ref><ns0:ref type='bibr' target='#b6'>Ceryngier et al., 2018)</ns0:ref>. Thus far, no studies have focused on the effects of infections of multiple natural enemy on Ha. axyridis.</ns0:p><ns0:p>Hesperomyces virescens (Ascomycota, Laboulbeniomycetes, Laboulbeniales) is a common obligate ectoparasite of ladybirds <ns0:ref type='bibr' target='#b20'>(Roy et al., 2016;</ns0:ref><ns0:ref type='bibr' target='#b15'>Haelewaters et al., 2017)</ns0:ref>. Contrary to most multicellular fungi, He. virescens as well as other members of the Laboulbeniales order lack hyphae, instead they form 3-dimensional multicellular thalli by determinate growth <ns0:ref type='bibr' target='#b2'>(Blackwell et al., 2020)</ns0:ref>. Laboulbeniales, including He. virescens, cannot be grown in axenic culture and no asexual stages are known, which makes their study challenging <ns0:ref type='bibr' target='#b13'>(Haelewaters et al., 2021)</ns0:ref>. Given locally high prevalence of He. virescens on ladybird hosts <ns0:ref type='bibr' target='#b32'>(Riddick & Cottrell, 2010;</ns0:ref><ns0:ref type='bibr' target='#b15'>Haelewaters et al., 2017)</ns0:ref> and the abundance of entomopathogenic fungal strains in the environment <ns0:ref type='bibr' target='#b39'>(Roy & Cottrell, 2008)</ns0:ref>, we examined mortality of native and invasive He. virescens-infected ladybirds exposed to either Beauveria bassiana or Metarhizium brunneum (Ascomycota, Sordariomycetes, Hypocreales) (sensu <ns0:ref type='bibr' target='#b9'>Cottrell & Shapiro-Ilan, 2003</ns0:ref><ns0:ref type='bibr'>, 2008)</ns0:ref>. Because He. virescens forms a branched, non-septate, rhizoidal haustorium <ns0:ref type='bibr' target='#b49'>(Weir & Beakes, 1996)</ns0:ref> that penetrates the host's exoskeleton and makes contact with the body fluid for nutrient uptake, we hypothesized that high thallus densities with concomitant haustorial formation by He. virescens weaken host defenses, thus increasing the host's susceptibility to infection by other natural enemies. With this experiment, we assess how He. virescens affects ladybird survival when exposed to other natural enemies that alone have little impact on Ha. axyridis and compare results with a North American-native ladybird of similar body size, Olla v-nigrum. If He. virescens-on its own and in combination with other natural enemies-significantly impacts survival of the invasive ladybird but not the native one, then the results of this work could have consequences toward a pest management strategy to control infestations of vineyards and agroecosystems by Ha. axyridis.</ns0:p></ns0:div>
<ns0:div><ns0:head>Materials & Methods</ns0:head></ns0:div>
<ns0:div><ns0:head>Field collections and laboratory colonies</ns0:head><ns0:p>Harmonia axyridis and Olla v-nigrum ladybirds were collected for the purpose of establishing laboratory colonies of Hesperomyces-infected and non-infected ladybirds. Specimens were collected at overwintering sites at the 485-ha USDA-ARS, Southeastern Fruit and Tree Nut Research Laboratory, located in Byron, Georgia, <ns0:ref type='bibr'>USA (32.657792,</ns0:ref>. Sex and age of field-collected specimens were not determined to reduce dispersal of fungal propagules <ns0:ref type='bibr'>(Cottrell & Riddick, 2012)</ns0:ref>. All specimens were brought to the laboratory and housed in individual Petri plates (10 cm diam.) with 1/3 of a piece of a cotton dental wick (Deerpack Products, LLC, Miami, Florida) drenched in water for hydration. Ladybirds were housed in environmental chambers at 25 ± 1 ˚C and photoperiod of 14:10 (L:D) h. Food was provided 3× per week in the form of Ephestia kuehniella eggs (Lepidoptera, Pyralidae) and an artificial meat-based diet (Beneficial Insectary, Redding, California). Olla v-nigrum and Ha. axyridis ladybirds were maintained within the Petri plates for 14d and 21d32, respectively, at which time ladybirds were visually examined for presence of Hesperomyces using a dissecting microscope at 50× magnification. Eggs were harvested from ovipositing ladybirds and used to establish clean (free from fungal growth) laboratory-reared colonies of ladybirds with known age.</ns0:p></ns0:div>
<ns0:div><ns0:head>Laboratory rearing of ladybirds</ns0:head><ns0:p>During examination for presence/absence of Hesperomyces, ladybirds were divided into two groups, infected and non-infected. Both groups of ladybirds were placed into plastic rearing containers of 19 × 13.5 × 9 cm (Pioneer Plastics, North Dixon, Kentucky), which were modified with two 3-cm diameter circular openings, one that was covered by 1 × 1 mm mesh to allow for air flow; and the second that was covered with a removable #7 rubber stopper to allow for feeding routinely as well as adding newly emerged laboratory-reared ladybirds. Routine maintenance included transferring ladybirds into fresh rearing containers at the end of each 7d period, which included nutrient supplementations of laboratory-reared yellow pecan aphids, Monelliopsis pecanis (Hemiptera, Aphididae).</ns0:p><ns0:p>The first laboratory generation of adults emerged about one month after placement in rearing containers. Emerging adults were placed into fresh rearing containers and stored into a separate incubator (25 ± 1 ˚C, 14:10 (L:D) h) for 7 days. Similar to field-captured O. v-nigrum and Ha. axyridis, M. pecanis aphids were used as a diet augmentation. As the study progressed, we also incorporated black pecan aphids, Melanocallis caryaefoliae (Hemiptera, Aphididae), in the ladybird diet (3× per week).</ns0:p></ns0:div>
<ns0:div><ns0:head>Artificial transmissions of Hesperomyces</ns0:head><ns0:p>Not only did we need the ladybirds for our experiments to be of the same age, we also needed to artificially infect a subset of these 'clean' laboratory-grown, adult ladybirds with Hesperomyces virescens. Exposure to Hesperomyces was conducted via tumbling of the field-captured 'source' ladybirds (infected with Hesperomyces) with randomly selected laboratory-reared 'target' ladybirds <ns0:ref type='bibr' target='#b10'>(Cottrell & Shapiro-Ilan, 2008)</ns0:ref>. A total of 25 target ladybirds were mixed with 5 Hesperomyces-infected source ladybirds in a 1.6 × 5.8 cm glass tube, which was placed on a hotdog roller (Nostalgia Electrics, Green Bay, Wisconsin) for 5 min. This procedure was repeated for at least 160 target ladybirds of both species. We only performed intra-specific artificial transmissions of Hesperomyces, meaning from source Ha. axyridis to target Ha. axyridis and from source O. v-nigrum to target O. v-nigrum. Both Hesperomyces-exposed target ladybirds and clean (unexposed) ladybirds were fed a diet of M. pecanis aphids for 24h. We did a second tumbling experiment using randomly selected emerged adults from the second cohort of laboratory-reared colonies. More tumbling experiments were performed to increase quantities of Hesperomyces-infected ladybirds, but source/target numbers were changed to 100/40.</ns0:p><ns0:p>To reduce competition for food, ladybirds from all laboratory colonies were transferred from the plastic rearing containers to 14-cm diameter Petri plates. Ladybirds were provided with water ad libitum, E. kuehniella eggs, and artificial meat-based diet. Finally, for assay preparation, the ladybirds were transferred back to clean 19 × 13.5 × 9 cm plastic rearing containers by species.</ns0:p></ns0:div>
<ns0:div><ns0:head>Dual fungal infections assay</ns0:head><ns0:p>Within 24 hours preceding the assay, 160 non-infected and 160 Hesperomyces-infected ladybirds of each species (Ha. axyridis and O. v-nigrum) were each placed into sterile test tubes, one individual per test tube. Test tubes were then closed with a sterile foam stopper to prevent ladybirds from escaping while allowing for air flow. Infected ladybirds were divided into categories according to numbers of thalli per specimen. Because the assay would assess potential interactions between fungal infections, we aimed at selecting heavily Hesperomyces-infected ladybirds; as a baseline, we only used specimens in our bioassays with 14 or more thalli each.</ns0:p><ns0:p>The assay started by pipetting a 1 mL of 2.5 × 105 conidia/mL suspension to each test tube <ns0:ref type='bibr' target='#b9'>(Cottrell & Shapiro-Ilan, 2003</ns0:ref><ns0:ref type='bibr'>, 2008)</ns0:ref>. Treatments included native B. bassiana (native Bb), a commercial B. bassiana strain (GHA Bb; Mycotrol ES, Mycotech, Butte, Montana), M. brunneum strain F52 (Mb, isolated from a tortricid moth, Austria 1971; Novozymes, Franklinton, North Carolina), and double-distilled water (ddH 2 O) as a control treatment. Ladybirds were submerged and swirled for 5 s, after which the suspension was removed again using a pipette and each ladybird was placed into a 6 cm-diameter Petri plate. Any remaining droplets of excess suspension was removed by touching only the droplet with a Kimwipe tissue (Kimtech Science Brand, Kimberly-Clark Worldwide, Roswell, Georgia). Petri plates with treated ladybirds were placed into an incubator (25 ± 1 ˚C, 14:10 (L:D) h). Food and cotton rolls drenched in water were provided ad libitum, and Petri plates were replaced as needed in all treatments and replications simultaneously. Ladybirds were observed for mortality and entomopathogeninduced mycosis at day 14. During assay #1, we made daily observations for ladybird mortality and mycosis. Upon death of a given ladybird, ample water was added to the cotton roll to provide moisture for entomopathogen growth and Parafilm was applied around the Petri plate to prevent spreading of the fungus. Deaths of ladybirds and visual confirmations of mycosis were recorded.</ns0:p><ns0:p>We performed 8 different treatments for each ladybird species: 1) He. virescens-positive + native Bb, 2) He. virescens-positive + GHA Bb, 3) He. virescens-positive + Mb, 4) He. virescenspositive + ddH 2 O (control), 5) He. virescens-negative + native Bb, 6) He. virescens-negative + GHA Bb, 7) He. virescens-negative + Mb, and 8) He. virescens-negative + ddH 2 O (double control). In a single assay, we replicated every treatment 3 or 4 times. We performed the entire assay with all treatments and replicated 3 times, using 6-10 ladybirds for each treatment. Note that M. brunneum treatments were used only in assay #3 (Table <ns0:ref type='table'>S1</ns0:ref>). Over all assays done during this study, we used 1,289 specimens of ladybirds (667 O. v-nigrum and 622 Ha. axyridis) (Table <ns0:ref type='table'>S2</ns0:ref>).</ns0:p></ns0:div>
<ns0:div><ns0:head>Statistical analyses</ns0:head><ns0:p>All statistical analyses were performed in the R language and open-access environment for statistical computing v.3.5.0. We used generalized linear mixed models (function glmer(), Rpackage lme4; <ns0:ref type='bibr' target='#b1'>Bates et al., 2015)</ns0:ref> to analyze the effect of the different treatments (GHA Bb, native Bb, Mb) on the survival of Ha. axyridis and O. v-nigrum in relation to prior infection with Hesperomyces. We modeled the binary response variable survival (alive/dead) of each ladybird individual for both host species separately, and used Hesperomyces infection status as well as the interaction of Hesperomyces infection status with treatment as explaining variables. Further, to correct for variation within replicates and assays, we included the random effect of treatment nested in replicate nested in assay. We compared our candidate models to a respective Nullmodel using likelihood ratio tests and, furthermore, calculated pseudo R 2 -values (function r2(), R package sjstats; <ns0:ref type='bibr' target='#b26'>Lüdecke, 2018)</ns0:ref> to evaluate model fit. To visualize the modeling results and obtained model estimates as forest plots, we used the function plot_model() implemented in the R package sjstats <ns0:ref type='bibr' target='#b26'>(Lüdecke, 2018)</ns0:ref>. For assay #1, we further fitted Kaplan-Meier curves to daily mortality data and tested for significant differences in mortality between ladybird species using the function survfit() of the R package survival <ns0:ref type='bibr' target='#b47'>(Therneau & Lumley, 2019)</ns0:ref>.</ns0:p></ns0:div>
<ns0:div><ns0:head>Results</ns0:head><ns0:p>Our candidate models for both host species Ha. axyridis and O. v-nigrum were significantly better at explaining survival relative to chance variation (Chi-squared test, χ 2 = 156.7, P < 0.001; χ 2 = 153.0, P < 0.001, respectively). The overall model fit was high for both candidate models (Ha. axyridis: Nagelkerke's R 2 = 0.40; O. v-nigrum: Nagelkerke's R 2 = 0.53) suggesting the variance is well described by our applied models.</ns0:p><ns0:p>We found a significant negative effect on ladybird survival of the M. brunneum treatment on He. virescens-negative Ha. axyridis (Fig. <ns0:ref type='figure' target='#fig_0'>1</ns0:ref>, Table <ns0:ref type='table'>1</ns0:ref>), whereas B. bassiana treatments did not affect the survival of He. virescens-negative individuals. Infection with He. virescens significantly affected Ha. axyridis survival over all treatments (Fig. <ns0:ref type='figure' target='#fig_0'>1</ns0:ref>). However, there was no additional effect detectable among co-infection treatments for He. virescens-positive ladybirds (Table <ns0:ref type='table'>1</ns0:ref>). Each treatment applied to O. v-nigrum had a significantly negative effect on the survival for both He. virescens-negative and -positive ladybirds (Fig. <ns0:ref type='figure' target='#fig_0'>1</ns0:ref>, Table <ns0:ref type='table'>2</ns0:ref>). Finally, we found an additional negative effect of all co-infection treatments on the survival of He. virescens-positive O. vnigrum (Fig. <ns0:ref type='figure' target='#fig_0'>1</ns0:ref>, Table <ns0:ref type='table'>2</ns0:ref>). These results suggest that there is no effect of dual infections on Ha. axyridis, whereas O. v-nigrum is highly affected by simultaneous exposure to both He. virescens and an entomopathogenic fungus. Percentages of ladybird mortality by treatment are also presented in tabulated form in Table <ns0:ref type='table'>S3</ns0:ref>.</ns0:p><ns0:p>When comparing the daily survival of Ha. axyridis and O. v-nigrum, no significant differences were found in He. virescens-positive only treatments. However, when co-infected O. v-nigrum showed a significantly lower survival compared to Ha. axyridis for native and GHA B. bassiana strains (log rank test, P = 0.0014 and P < 0.001, respectively; Fig. <ns0:ref type='figure'>2</ns0:ref>).</ns0:p></ns0:div>
<ns0:div><ns0:head>Discussion</ns0:head><ns0:p>Research on the additive effects of multiple natural enemies on a given host is rare, likely because of the complexity involved in designing robust bioassays that include all partners of the system. Combining the natural enemies Orius insidiosus (Hemiptera, Anthocoridae) and Heterorhabditis bacteriophora (Rhabditida, Heterorhabditidae) resulted in the largest decline in larvae of Drosophila suzukii <ns0:ref type='bibr' target='#b30'>(Renkema & Cuthbertson, 2018)</ns0:ref>, which causes major economic losses to fruit crops in its invasive range, spanning North and South America and Europe <ns0:ref type='bibr' target='#b25'>(Lee et al., 2011)</ns0:ref>. The addition of O. insidiosus resulted in 50% fewer D. suzukii larvae compared to treatment with only H. bacteriophora. Plutella xylostella (Lepidoptera, Plutellidae), an important cosmopolitan pest of brassicaceous crops, offers another example. This organism shows resistance to almost all chemical insecticides <ns0:ref type='bibr' target='#b43'>(Sarfraz et al., 2005)</ns0:ref>. Pandora blunckii and Zoophthora radicans (Zoopagomycota, Entomophthoromycetes, Entomophthorales) both infect P. xylostella in the field. In co-inoculation studies with Pa. blunckii and Z. radicans in P. xylostella larvae, larval cadavers (three days post mortality) were most frequently found with conidia of a single entomopathogen, usually the one that had been inoculated first (prior 'residency')-meaning that the other species was excluded <ns0:ref type='bibr' target='#b42'>(Sandoval-Aguilar et al., 2015)</ns0:ref>. In general, the presence of competing species in the same host resulted in a decreased proportion of P. xylostella larvae that were infected compared to single inoculations.</ns0:p><ns0:p>Regarding Ha. axyridis, the following co-infections of natural enemies have been observed in nature: He. virescens + Coccipolipus hippodamiae mites (Acarina, Podapolipidae) in the USA, Austria, and the Netherlands <ns0:ref type='bibr' target='#b7'>(Christian, 2001;</ns0:ref><ns0:ref type='bibr' target='#b31'>Riddick, 2010;</ns0:ref><ns0:ref type='bibr' target='#b29'>Raak-van den Berg et al., 2014)</ns0:ref> and He. virescens + Parasitylenchus bifurcatus nematodes (Nematoda, Allantonematidae) in the Czech Republic, Germany, and the Netherlands (Raak-van den <ns0:ref type='bibr' target='#b29'>Berg et al., 2014;</ns0:ref><ns0:ref type='bibr' target='#b15'>Haelewaters et al., 2017;</ns0:ref><ns0:ref type='bibr'>Herz & Kleespiel, 2012)</ns0:ref>. Given the status of Ha. axyridis as an invasive alien species, these findings demand a better understanding of interactions among the different natural enemies and their potential role in limiting populations of Ha. axyridis. To date, bioassays to determine mortality of ladybirds induced by infection by one or more natural enemies have not yet been performed. Likewise, bioassays including Laboulbeniales have only been carried out in one study <ns0:ref type='bibr' target='#b24'>(Konrad et al., 2015)</ns0:ref>.</ns0:p><ns0:p>When we started this study, He. virescens was considered a single species with multiple ladybird hosts, potentially with multiple strains that infect only a single species, or one closely related <ns0:ref type='bibr'>(Cottrell & Riddick, 2012</ns0:ref>). Yet, it was recently shown that He. virescens is a complex of multiple species, each with its own ladybird host <ns0:ref type='bibr'>(Haelewaters et al., 2018)</ns0:ref>. Isolates of He. virescens from Ha. axyridis and O. v-nigrum in fact represent two different species of Hesperomyces. In other words, the experiments in the current study allow us to make comparisons between two host species, each with their own specific fungal parasite. To further disentangle the interactions in future experiments, we must infect Ha. axyridis and O. v-nigrum ladybirds with the species of Hesperomyces specific to Olla and Harmonia, respectively, perform bioassays, and compare mortality rates under different treatments with our current results. Analyzing interactions among natural enemies only make sense when the taxa considered represent single biological species.</ns0:p><ns0:p>We found a significant negative effect of Hesperomyces-only infection on the survival of both ladybird hosts (Fig. <ns0:ref type='figure' target='#fig_1'>1a</ns0:ref>, Table <ns0:ref type='table'>S3</ns0:ref>). Previous work has shown Hesperomyces infections to result in decreased mating frequency of female ladybirds, lower (male) survival rates in winter, and impeded sensing ability and flexibility of legs in heavily infected ladybirds <ns0:ref type='bibr' target='#b27'>(Nalepa & Weir, 2007;</ns0:ref><ns0:ref type='bibr' target='#b31'>Riddick, 2010;</ns0:ref><ns0:ref type='bibr' target='#b15'>Haelewaters et al., 2017)</ns0:ref>. One study implicated parasitism by He. virescens as the cause of late summer mortality of Chilocorus bipustulatus ladybirds <ns0:ref type='bibr'>(Kamburov et al., 1967)</ns0:ref> but this was later disputed based on controlled laboratory experiments <ns0:ref type='bibr' target='#b0'>(Applebaum et al., 1971)</ns0:ref>. Our research is the first to explicitly link Hesperomyces infection with increased ladybird mortality.</ns0:p><ns0:p>Our findings on the effects of Hesperomyces on ladybird survival provided a unique opportunity for setting up dual infection assays-the first such experiments to be conducted on ladybirds. When first infected with He. virescens and then exposed to either B. bassiana or M. brunneum, Ha. axyridis mortality was not increased. This result was unexpected. We had hypothesized that Ha. axyridis with high thallus densities of He. virescens would have lowered host defenses against other pathogens. In contrast, the mechanism fostering low susceptibility of Ha. axyridis to entomopathogenic fungi <ns0:ref type='bibr' target='#b9'>(Cottrell & Shapiro-Ilan, 2003;</ns0:ref><ns0:ref type='bibr' target='#b22'>Knapp et al., 2019)</ns0:ref> is not compromised by infection with He. virescens. Similarly, infection of O. v-nigrum by He. virescens-only increased mortality but-in contrast to Ha. axyridis-there was significantly higher mortality when co-infected by entomopathogenic fungi. Differential susceptibility to entomopathogenic fungi was reported by <ns0:ref type='bibr' target='#b9'>Cottrell & Shapiro-Ilan (2003)</ns0:ref>, who showed that native B. bassiana was pathogenic to O. v-nigrum but not to Ha. axyridis. We confirm these results regarding the native strain but we also found the same differential pattern for the GHA strain of B. bassiana, whereas in the earlier study this strain was reported to be pathogenic to neither ladybird species <ns0:ref type='bibr' target='#b9'>(Cottrell & Shapiro-Ilan, 2003)</ns0:ref>. It is perhaps surprising that we detect the GHA strain to be pathogenic to native ladybirds in contrast to the previous results, but ladybird populations may become more susceptible over time for various reasons and natural enemies also become better adapted <ns0:ref type='bibr' target='#b22'>(Knapp et al., 2019)</ns0:ref>. We note that differential susceptibility has also been reported for entomopathogenic nematodes-again, Ha. axyridis was less susceptible compared to O. v-nigrum <ns0:ref type='bibr' target='#b46'>(Shapiro-Ilan & Cottrell, 2005)</ns0:ref>.</ns0:p><ns0:p>In addition, our data are the first account of differential susceptibility to M. brunneum between the invasive Ha. axyridis and the native O. v-nigrum. Whereas infection with M. brunneum had a significantly negative effect on the survival of He. virescens-negative Ha. axyridis, this effect was not visible in the dual infection treatment. The infection with Laboulbeniales probably decreased the susceptibility of Ha. axyridis to infection by M. brunneum, similar to the findings of <ns0:ref type='bibr' target='#b24'>Konrad et al. (2015)</ns0:ref>. These authors found that Laboulbenia-infected Lasius neglectus ants (Hymenoptera, Formicidae) showed a decreased susceptibility to Metarhizium brunneum. This protection against Metarhizium was positively correlated with parasite load. Information on the parasite load of He. virescens on ladybirds in nature is nonexistent. In our bioassays, we selected ladybirds bearing 14 or more fungal thalli as He. virescens-positive specimens. Previous work from a long-term ATBI project in the Netherlands (van Wielink, 2017) points at an average of 19.8 ± 4.9 thalli and a maximum of 129 thalli per Ha. axyridis specimen (n = 270). No such data are available for O. v-nigrum. In other words, based on the available information, the artificial parasite load in our bioassays seems to closely mimic the natural conditions.</ns0:p><ns0:p>Our results provide direct support for the enemy release hypothesis <ns0:ref type='bibr' target='#b21'>(Jeffries & Lawton, 1984)</ns0:ref>. This hypothesis is illustrative for the success of invasive alien species and stipulates that an invasive species in new geographic regions will experience reduced regulatory effects from natural enemies compared to native species, resulting in increased population growth of the invasive species <ns0:ref type='bibr' target='#b8'>Colautti et al., 2004;</ns0:ref><ns0:ref type='bibr' target='#b40'>Roy et al., 2011)</ns0:ref>. However, invasions are dynamic <ns0:ref type='bibr' target='#b44'>(Schultheis et al., 2015;</ns0:ref><ns0:ref type='bibr' target='#b15'>Haelewaters et al., 2017)</ns0:ref> and this escape-from-enemies could be lost as invasive species acquire new enemies over time <ns0:ref type='bibr' target='#b19'>(Hokkanen & Pimentel, 1989)</ns0:ref>. Support for enemy release explaining the success of Ha. axyridis has come from two studies that reported decreased susceptibility of Ha. axyridis to entomopathogenic fungi <ns0:ref type='bibr' target='#b9'>(Cottrell & Shapiro-Ilan, 2003)</ns0:ref> and entomopathogenic nematodes <ns0:ref type='bibr' target='#b46'>(Shapiro-Ilan & Cottrell, 2005)</ns0:ref> compared to the native American ladybird species. Our work adds another level of complexity by the addition of a second natural enemy to the interactions. Again, we find differential susceptibility between the invasive and native ladybird species-with a reduced regulatory effect of the tested natural enemies on Ha. axyridis.</ns0:p></ns0:div>
<ns0:div><ns0:head>Conclusions</ns0:head><ns0:p>In this paper, we show a negative effect of infection by Hesperomyces virescens on the survival of both Harmonia axyridis and Olla v-nigrum ladybirds (Fig. <ns0:ref type='figure' target='#fig_1'>1a</ns0:ref>). This is the first study to unequivocally link Hesperomyces infection with increased host mortality and only the second to perform bioassays with hosts co-infected with Laboulbeniales and a second entomopathogenic fungus <ns0:ref type='bibr' target='#b24'>(Konrad et al., 2015)</ns0:ref>. However, the susceptibility to a secondary entomopathogenic fungus was only elevated in the native American ladybird species (O. v-nigrum), whereas the globally invasive Ha. axyridis showed no significant increase in mortality when co-infected with either Beauveria bassiana or Metarhizium brunneum (Figs. <ns0:ref type='figure' target='#fig_0'>1, 2</ns0:ref>). These findings are consistent with the enemy release hypothesis <ns0:ref type='bibr' target='#b21'>(Jeffries & Lawton, 1984)</ns0:ref> and highlight the difficulty in controlling this invasive alien species. Future studies are needed to elaborate population-specific effects on native and commercial strains of entomopathogenic fungi used in pest control. </ns0:p></ns0:div><ns0:figure xml:id='fig_0'><ns0:head>Figure 1</ns0:head><ns0:label>1</ns0:label><ns0:figDesc>Figure 1</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_1'><ns0:head>( A )</ns0:head><ns0:label>A</ns0:label><ns0:figDesc>Percentages of ladybird mortality by treatment (left Harmonia axyridis, right Olla vnigrum). (B-C) Forest plots illustrating the results of our modelling approach showing the treatment effects on survival of ladybirds (negative effect in red (odds ratio < 1), positive effect in blue (odds ratio > 1); ** P < 0.01, *** P < 0.001). (B) Olla v-nigrum. (C) Harmonia axyridis. Photo credits: Olla v-nigrum, Roberto Güller (Flickr); Harmonia axyridis, Andreas Trepte (Wikimedia Commons). Drawings of Hesperomyces thalli by André De Kesel.</ns0:figDesc></ns0:figure>
<ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='17,42.52,306.37,525.00,208.50' type='bitmap' /></ns0:figure>
<ns0:note place='foot'>PeerJ reviewing PDF | (2020:04:48281:1:0:NEW 6 Aug 2020)</ns0:note>
<ns0:note place='foot'>PeerJ reviewing PDF | (2020:04:48281:1:0:NEW 6 Aug 2020)Manuscript to be reviewed</ns0:note>
</ns0:body>
" | "Dear editor,
Firstly, I am sorry about the delay in submitting this revised manuscript. A lot of other manuscripts (some
invited), book chapters, job applications, and student projects had absolute priority. I appreciate your
summary of the comments and I was happy to hear from you as well as the reviewers that the study is
robust and in need of minor changes only. Below, I will respond to each of the points made by the
reviewers in blue for clarity.
I will note that several of the comments were mere suggestions, mentions of alternative approaches, or
questions—not necessarily in need of action items within the manuscript. Although I responded in detail
to the comments below, the manuscript has not been revised in a major way—I assume this is not a
problem, given the minor revisions decision. During the process of revision, I did make some other,
mostly minor edits throughout the manuscript to improve the quality of what will hopefully be the final
paper.
Thank you again for the chance of considering our work for publication in PeerJ.
Very best wishes,
For all co-authors,
Danny Haelewaters
EDITOR
Reviewers have now revised the manuscript. Reviewers think the manuscript is robust, and the findings
are interesting. I definitively agree with that. However, the material and methods section requires
incorporating some observed aspects (pay special attention to reviewer 2). These points are all apparently
easily approachable. In general, the manuscript involves minor changes. Please respond to each of the
observations (point by point) made by the reviewers.
[# PeerJ Staff Note: Reviewer 3 declared a potential Conflict of Interest, and the Editor was aware of this
when making their decision #]
REVIEWER 1
Basic reporting
The English text was very clear and unambiguous throughout the manuscript. The literature references
and background in the field was provided. The structure of the manuscript including figures and tables are
professional. The raw data was shared in supplementary files. A hypothesis was presented and tested.
Experimental design
The research question was well-defined and meaningful. The research clearly fills a knowledge gap. The
experimentation was suitably rigorous. The materials and methods were described in sufficient detail.
Validity of the findings
All data have been provided and they are adequately robust and appear statistically sound. Conclusions
are well stated, linked to the hypothesis statement and supported by the results.
Comments for the author
Dear Authors:
This is an interesting article seeking to experimentally manipulate pathogen densities to better understand
why the invasive ladybird Harmonia axyridis has a competitive advantage over native ladybirds such as
Olla v-nigrum in North America.
Response: Thank for this and the above comments.
A few suggestions to improve the text:
1. Please add a few sentences in the Introduction or Materials and Methods to tell the readers that He.
virescens is likely two separate stains in this study. One for Ha. axyridis, and one for O. v-nigrum. I see
that you mentioned the two strains in the Discussion section.
Response: The inclusion of this information in the Introduction would take the attention away from the
aim of the paper, in my opinion. We really set it up as described in the Introduction (rationale and
hypothesis) and Materials & Methods, and then in the Discussion we explain the results in the light of
recent findings (including Hesperomyces virescens being a species complex)—I think that is an
appropriate way of structuring the paper. If the editor agrees, it would be my preference to keep this
unchanged.
2. In Figure 1, please change the x-axis label, maybe to Mean Percentage (+/- SE) of Mortality.
Response: The reviewer is absolutely correct. Somehow, an old version of Figure 1 was uploaded to the
system—so while the caption was updated to reflect the correct axis, the figure itself showed another axis
(with the confusing “manifold” of mortality). This has now been changed with the accurate Figure 1 that
is part of the revised paper. Apologies.
3. In Figure 2, I am hesitant to believe that the authors can honestly compare the infection rates between
ladybird species, if He. virescens strains differ between ladybird species. Do the authors have any
unpublished data to show that one strain is not more virulent than the other?
Response: Figure 2 does indicate that there is no difference in virulence of the He. virescens strains to
their respective ladybird hosts. Moreover, the primary comparison is not focused the relative virulence of
He. virescens strains but rather on the relative virulence of B. bassiana and M. brunneum when applied to
ladybirds that had already been infected with He. virescens. The same strains of B. bassiana and M.
brunneum were applied to each ladybird species. This approach simulates natural conditions where dual
infections would occur.
4. Line 426: Add italics to the sci. name.
Response: Done
5. Line 461, 469: BioControl
Response: Correct, twice.
REVIEWER 2
Basic reporting
The theoretical framework must be more in line with the objective of the study. The information on lines
40-50 gives contradictory messages and it is unclear whether this study was designed from a fundamental
ecological perspective, that is, to understand the impact of biotic interactions on the structure and
dynamics of populations or if it was designed in a more applied perspective to contribute to finding the
best solutions to control populations of Ha. axyridis. I wonder, is this period really necessary?
Response: This first paragraph of the introduction was deliberately written to showcase how much we do
not know about dual infections—and, as the reviewer correctly points out, that results are contradictory.
Rather than a criticism, I see this as a strong justification for more work in this area of research, and that
is exactly why we set out this experiment. To respond to the reviewer’s specific question: the project was
indeed designed from an applied perspective (potential biocontrol of Harmonia axyridis). I made this
explicit in the revised manuscript, by adding the following sentence at the end of the Introduction:
(lines 91–95) “If He. virescens—on its own and in combination with other natural enemies—significantly
impacts survival of the invasive ladybird but not the native one, then the results of this work could have
consequences toward a pest management strategy to control infestations of vineyards and agroecosystems
by Ha. axyridis.”
Literature, article structure, figures, tables and raw data are adequate.
Manuscript use a clear, unambiguous and correct text. We found nice profissional standards of courtesy
and expression.
Experimental design
My major comments goes to experimental design.
The study adresses a timely topic of research and is of interest to ecologists, both those working in areas
of biological invasion as well as those in the field of applied ecology. PeerJ is a suitable venue for its
publication.
It is not clear why the procedure described in the lines 136-138 was carried out. Can I assume that this
was only done to obtain more Hesperomyces infected ladybirds? Again, it is not clear why this procedure
presented in lines 138-139 was carried out. Can I assume that this was done to obtain diferent
Hesperomyces infectations levels?
Response: I added the following sentence for clarity:
(lines 135–137) “Not only did we need the ladybirds for our experiments to be of the same age, we also
needed to artificially infect a subset of these “clean” laboratory-grown, adult ladybirds with
Hesperomyces virescens.”
I recomend that authors explain to what extent “14 or more thalli” exert a significant imumitarian
response on ladybirds and that this response does not change differently to any value above 14 thalli and
thus, does not alter susceptibility of hosts to aditional infections (see lines 151-153). This fact may
explain the high variance in the percentages of ladybird mortality.
Response: I am not 100% sure if I understand what the reviewer means with this comment. The 14
thallus-cut off was decided based on the 648 infected ladybirds that were available for the experiment.
There is not more to it than this. Second, this cut off “might” explain the variance, however the variance
is comparable over all treatments! Finally, let us not forget that this study is the first one attempting to use
Laboulbeniales fungi in an experimental setting like this one. No study has done any sort of analysis of
the effect of number of thalli on a given host. In this study (Discussion), we pointed out that, on average,
there are 19.8 ± 4.9 thalli on Harmonia axyridis ladybirds. Our cut off of 14 closely mimics the natural
average...
Authors should explain why they tested 1 mL of 2.5 × 105 conidia/mL suspension and not another
concentration? Moreover, despite considering acceptable submerging ladybirds in a suspension with
conidia, wouldn't it have been possible to use Potter's tower to spray ladybirds?
Response: It is true that there are probably multiple ways to apply conidia to the ladybirds, but we used
the method that had presented by Cottrell & Shapiro-Ilan (2008), both co-authors of this manuscript. We
note that the Potter’s tower has not been used widely in recent years—a search on Wed of Science shows
2 mentions in 2011, 2 in 2012, 2 in 2013–2015, 1 in 2018, and none in 2017, 2017, and 2019.
Why statistical comparisons of daily mortality between H. axyridis and O. nigr are performed without
data of control tratments?
Response: Figure 2A represents the both species with Hesperomyces-only, which can be regarded as the
baseline control, whereas Figures 2B & 2C show the results of dual infections. The main purpose of
Figure 2 is to visualize the effect of dual infections versus single infections, which is the subject of the
paper.
Validity of the findings
The study intends to test the effect effects of infections of multiple natural enemy on Ha. axyridis. Two
merits/novelties can be highlighted from this study; i) the setting up of dual infection assays conducted on
ladybirds and ii) the first account of differential susceptibility to M. anisopliae between the invasive Ha.
axyridis and the native O. v-nigrum. The study is of interest to ecologists, both those working in areas of
biological invasion as well as those in the field of applied ecology.
Comments for the author
The study intends to test the effect effects of infections of multiple natural enemy on Ha. axyridis. Two
merits/novelties can be highlighted from this study; i) the setting up of dual infection assays conducted on
ladybirds and ii) the first account of differential susceptibility to M. anisopliae between the invasive Ha.
axyridis and the native O. v-nigrum.
The experiments in the current study, contrary to what the authors advocated, does not fully allow to
make comparisons of infections, and dual infections, between two host species. I believe that the study
only allows the characterization of the effect of their two specific fungal parasite. This is clearly
recognized by the authors when they refer to the need for additional experiments to disentangle the
interactions.
Response: The fact that additional studies are needed to elucidate host-pathogen relationships further
does not negate the validity of this study’s findings. We clearly show that there is no effect in mortality
for both ladybird species when infected with Hesperomyces. However, there are significant differences
when co-infected. As a result, we very well can—and I think we should—discuss these findings.
Nevertheless, further experiments are necessary to adapt the experimental procedures to the very recent
knowledge about the status of Hesperomyces virescens as a species complex, and this is discussed in
proper detail for clarity.
(lines 263 – 274): “When we started this study, He. virescens was considered a single species with
multiple ladybird hosts, potentially with multiple strains that infect only a single species, or one closely
related (Cottrell & Riddick, 2012). Yet, it was recently shown that He. virescens is a complex of multiple
species, each with its own ladybird host (Haelewaters et al., 2018). Isolates of He. virescens from Ha.
axyridis and O. v-nigrum in fact represent two different species of Hesperomyces. In other words, the
experiments in the current study allow us to make comparisons between two host species, each with their
own specific fungal parasite. To further disentangle the interactions in future experiments, we must infect
Ha. axyridis and O. v-nigrum ladybirds with the species of Hesperomyces specific to Olla and Harmonia,
respectively, perform bioassays, and compare mortality rates under different treatments with our current
results. Analyzing interactions among natural enemies only make sense when the taxa considered
represent single biological species.”
In view of this, it does not make sense to compare the daily mortality rate as shown in figure 2.
Response: Quite the opposite; in our opinion it does. The main purpose of Figure 2 is to visualize the
effect of dual infections versus single infection, by ladybird species. This is the subject of the paper.
Minor comments:
Line 40: I recommend to replace “In nature and in agricultural ecosystems…” by “In nature…”.
Response: Done as suggested by the reviewer.
Line 130-134: I recommend mentioning here the fact that the adult states have been mixed.
Response: Agreed. I stated that we infected adult ladybirds in the added sentence—as follows:
(lines 135–137) “Not only did we need the ladybirds for our experiments to be of the same age, we also
needed to artificially infect a subset of these “clean” laboratory-grown, adult ladybirds with
Hesperomyces virescens.”
In the Introduction, we had already stated that Laboulbeniales only infect adult stages so the fact that only
adult ladybirds were used to artificially infect other adult ladybirds was implicit. With the abovementioned addition, there is no more question to it.
REVIEWER 3
Basic reporting
This manuscript aims to assess how mortality of two ladybird species, one native and one invasive, is
mediated by the co-infection by entomopathogenic fungi. The question raised by the authors is very
interesting and novel, dealing with the interactions among different parasitic fungi on the survival of two
ladybird beetles. And the results are very interesting and meaningful. It is well written, with appropriate
literature and good structure.
Nevertheless, I think that the introduction (background/context) should be modified in order to be more
clear about 1) the previous information regarding these types of interactions; 2) the interaction between H.
virescens and other fungi; 3) the fungi species selected and their origins; and 4) how the origin (of both
ladybird beetles and fungi) could mediate the outcomes of these interactions (how co-evolutionary
processes may mediate the interactions among these fungi with the ladybird beetle host species). Some of
this information is provided very late in the manuscript, in the Discussion.
Response: 1) We did this both in the Introduction and in the Discussion). 2) Nothing is known thus far
(all appropriate references are cited and in none of them any interaction with other fungi is discussed). 3)
I think we clearly state this in the Introduction (lines 79–81) and the Materials & Methods (details about
strains provided). 4) Also provided, in the Discussion. The way we structured the paper is as follows:
provide the necessary background information to the reader in the Introduction, leading to the rationale
and main hypothesis; and then discuss what we know to provide context to our results in the Discussion.
Adding parts of this—in my opinion—would disrupt the flow of the article (because there would be parts
of a discussion in the Introduction, but some other parts also in the Discussion …).
Very importantly, the hypothesis should include an explanation for why native and invasive ladybird
species should respond differently to native or non-native fungi.
Response: Our hypothesis is not about comparing native and invasive ladybirds, but about Hesperomyces
affecting its ladybird hosts (lines 87–89). We only included Olla v-nigrum in the experiment to ensure we
would not propose using Hesperomyces virescens as a biological control while it might kill off native
ladybirds. In order to know how Harmonia axyridis and Olla v-nigrum might react to infection by an
under-understood fungus like Hesperomyces virescens, we need more data. Putting forward a hypothesis
would be a mere guess. For example, is Hesperomyces an American-native fungus or did Harmonia
axyridis bring it from Asia? See, getting into this discussion is an entirely different paper on its own. I
absolutely appreciate this comment and I think it is, indeed, very important to gain a better understanding
about native and invasive ladybird species responding to native or non-native parasitic fungi, but as long
as do not know the native/non-native/invasive status of the fungi, it’s just guessing and I would not want
to publish a hypothesis that is the result of a guess.
Experimental design
In general, the methods and statistical analyses are appropriate, and described with sufficient detail. A few
issues that are not completely clear in the text are clarified when looking at the supplementary Tables.
Experiments include the establishment of laboratory–reared colonies of ladybirds, thus controlling for the
initial parasitism state (absence of parasitic fungi), and age. It would be necessary to include a better
justification of why using native B- bassiana and a commercial B. bassiana strain There´s nothing in the
hypothesis dealing with fungi origin (native, exotic, natural or commercial strains). The authors should
revise the hypothesis.
Response: We actually had already provided background on why we used native and commercial
Beauveria bassiana—with citations. Strains were used based on previous work by Cottrell & Shapiro
(2003, 2008). In the revised manuscript, however, I cite these references in the Introduction so it is clearer
where our decision comes from to use these fungi Beauveria bassiana and Metarhizium brunneum. The
rationale & hypothesis are edited as follows:
(lines 80–90) “Given locally high prevalence of He. virescens on ladybird hosts (Riddick & Cottrell,
2010; Haelewaters et al., 2017) and the abundance of entomopathogenic fungal strains in the environment
(Roy & Cottrell, 2008), we examined mortality of native and invasive He. virescens-infected ladybirds
exposed to either Beauveria bassiana or Metarhizium brunneum (Ascomycota, Sordariomycetes,
Hypocreales) (sensu Cottrell & Shapiro-Ilan, 2003, 2008). Because He. virescens forms a branched, nonseptate, rhizoidal haustorium (Weir & Beakes, 1996) that penetrates the host’s exoskeleton and makes
contact with the body fluid for nutrient uptake, we hypothesized that high thallus densities with
concomitant haustorial formation by He. virescens weaken host defenses, thus increasing the host’s
susceptibility to infection by other natural enemies.”
Validity of the findings
What it is not so clear to me is the Result section (see specific comments). Specially, I would like to see
in the figures the results for the control treatments.
Response: As previously mentioned, Figure 1 was incorrect. And Figure 2 was created to illustrated dual
infections versus single infections. Results for control treatments (Hesperomyces-negative) can be found
in Figure 1 and Tables S2 and S3.
Hesperomyces virescens is a complex of multiple species, which is only reported in the Discussion.
Because of this, in the Materials and methods I think it is necessary to be clearer about the origin of the
Hesperomyces used in the trials. Did they come from H. axyridis? O. v-nigrum, both?
Response: This is a great comment. I edited the text as follows:
(lines 140 – 142) “We only performed intra-specific artificial transmissions of Hesperomyces, meaning
from source Ha. axyridis to target Ha. axyridis and from source O. v-nigrum to target O. v-nigrum.”
Specific comments:
L. 138-139: why the change in source/target numbers?
Response: For efficacy reasons. Putting 2 versus 5 source ladybirds in a tube with 5 target ladybirds will
not affect the resulting numbers of thalli on targets. This tumbling procedure (and evaluation of its
success) was presented by Cottrell & Shapiro-Ilan (2008), which is cited in the manuscript.
144: number of ladybirds per container? I suppose H. axiridis and O v-nigrum, infected and non-infected
with H. virescens were reared in different containers. Yes?
Response: This was maybe done for half an hour time just to facilitate the experimental set-up, so a very
minor detail in the methodology. I edited this sentence as follows:
(lines 152–153) “Finally, for assay preparation, the ladybirds were transferred back to clean 19 × 13.5 × 9
cm plastic rearing containers by species.”
L. 148: individually placed in the tubes?
Response: Yes, indeed. I made this unambiguous in the revision:
(lines 156–159) “Within 24 hours preceding the assay, 160 non-infected and 160 Hesperomyces-infected
ladybirds of each species (Ha. axyridis and O. v-nigrum) were each placed into sterile test tubes, one
individual per test tube. Test tubes were then closed with a sterile foam stopper to prevent ladybirds from
escaping while allowing for air flow.”
L. 186-187: the controls are also treatments. How are they included in the analyses (in Table 1, the Pvalues are between each treatment and the corresponding control?
Response: Control treatments are set as baseline for comparison of further treatments, meaning GHA Bb,
native Bb, and Mb were compared to control (once control Hvir-negative for single infections including
Hvir and once control Hvir-positive for dual infections).
There are no comparisons among all different treatments)?
Response: No, why should there be some? The aim of this study was to study the effect of dual infections
on host survival. Because there are significant differences between Hesperomyces-negative and
Hesperomyces-positive beetles, this needs to be taken into consideration when comparing the survival for
the co-infections, leading to the comparison of co-infections with the respective Hesperomyces-positive
control.
L. 191-192: trials = assays? Pease, be consistent
Response: The reviewer is absolutely correct. All corrected to assays (not trials).
L. 210. It is not clear to me how Table 1 confirm this result. It seems to me that it would be more
appropriate citing Table 1 at the end of sentence in Line 211 (after “no additional effect detectable among
co-infection treatments…”.
Response: Agreed and change made.
L. 330-331. I don´t see this result in Fig. 1 because, again, controls are not shown.
In addition, I would like to see the results of all eight treatments in Figure 1 (as in Table S3). Why it is not
reported the Controls without H. virescens for H. axyridis and O. v-nigrum? (i.e., only 7 bars –and not 8shown for each ladybird species). Is it that all results are the difference of these treatments with the
corresponding control? Sorry I am not familiar with the concept “manifold of mortality”. Please, clarify.
Response: An older version of Figure 1 had been uploaded but this has now been corrected.
Table 1 and 2 should be merged.
Response: We do not agree with this suggestion; these are two independent analyses/models so why
combine the results?
Fig. 2: An interesting result is that apparently the survival of O. v-nigrum (in presence of H. virescens) is
more affected by the infection of native Bb than GHA Bb. Again, why the controls (survival of H.
axyridis and O. v-nigrum without H. virescens) are not shown here?
Response: The main purpose of Figure 2 is to visualize the effect of dual infections versus single
infections, which is the subject of the paper.
Comments for the author
My general comments are already stated in the previous sections
" | Here is a paper. Please give your review comments after reading it. |
9,741 | "<ns0:abstract xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:p>Harmonia axyridis is an invasive alien ladybird in North America and Europe. Studies show that multiple natural enemies are using Ha. axyridis as a new host. However, thus far, no research has been undertaken to study the effects of simultaneous infections of multiple natural enemies on Ha. axyridis. We hypothesized that high thallus densities of the ectoparasitic fungus Hesperomyces virescens on a ladybird weaken the host's defenses, thereby making it more susceptible to infection by other natural enemies. We examined mortality of the North American-native Olla v-nigrum and Ha. axyridis co-infected with He.</ns0:p><ns0:p>virescens and an entomopathogenic fungus-either Beauveria bassiana or Metarhizium brunneum. Laboratory assays revealed that He. virescens-infected O. v-nigrum individuals are more susceptible to entomopathogenic fungi, but Ha. axyridis does not suffer the same effects. This is in line with the enemy release hypothesis, which predicts that invasive alien species in new geographic areas experience reduced regulatory effects from natural enemies compared to native species. Considering our results, we can ask how He. virescens affects survival when confronted by other pathogens that previously had little impact on Ha. axyridis.</ns0:p></ns0:div>
</ns0:abstract>
<ns0:body xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:head>Introduction</ns0:head><ns0:p>PeerJ reviewing PDF | (2020:04:48281:2:0:CHECK 7 Sep 2020)</ns0:p></ns0:div>
<ns0:div><ns0:head>Manuscript to be reviewed</ns0:head><ns0:p>In nature, hosts may be exploited by more than one natural enemy. These organisms interact with each other and with their hosts <ns0:ref type='bibr' target='#b11'>(Furlong & Pell, 2005)</ns0:ref>. These complex interactions shape the population structure and dynamics of all organisms in the system. Natural enemies also compete with one another, and the impact on the host can be either synergistic, additive, or antagonistic <ns0:ref type='bibr' target='#b47'>(Shapiro-Ilan et al., 2012)</ns0:ref>. These interactions can be manifested in various aspects of host fitness or mortality. For example, biological control of Drosophila suzukii (Diptera, Drosophilidae), an important pest of fruit and berry crops, can be improved by treatments combining multiple natural enemies, which have an additive effect <ns0:ref type='bibr' target='#b32'>(Renkema & Cuthbertson, 2018)</ns0:ref>. At the same time, dual infections (even if causing an increase in host mortality) may be deleterious to one or both pathogens in terms of pathogen growth, fecundity, or other fitness parameters.</ns0:p><ns0:p>Harmonia axyridis (Coleoptera, Coccinellidae), native to eastern Asia, has rapidly increased its global range and is now present on all continents except Antarctica <ns0:ref type='bibr' target='#b22'>(Roy et al., 2016;</ns0:ref><ns0:ref type='bibr' target='#b5'>Camacho-Cervantes et al., 2017;</ns0:ref><ns0:ref type='bibr' target='#b20'>Hiller & Haelewaters, 2019)</ns0:ref>. Even though it has repeatedly been introduced for its beneficial properties as a biological control agent against aphid pests, its negative effects on native ladybird communities in invaded areas <ns0:ref type='bibr' target='#b25'>(Koch et al., 2006;</ns0:ref><ns0:ref type='bibr' target='#b22'>Honěk et al. 2016;</ns0:ref><ns0:ref type='bibr'>Brown & Roy, 2018)</ns0:ref> and on food production <ns0:ref type='bibr' target='#b25'>(Koch et al., 2006)</ns0:ref> have raised serious concerns since the early 2000s <ns0:ref type='bibr' target='#b22'>(Roy et al., 2016)</ns0:ref>. It is now a model organism for studying invasive alien species <ns0:ref type='bibr' target='#b43'>(Roy & Wajnberg, 2008;</ns0:ref><ns0:ref type='bibr'>Brown et al., 2018)</ns0:ref> and it has been listed in Europe as 'one of the worst' invasive species <ns0:ref type='bibr' target='#b30'>(Nentwig et al., 2018)</ns0:ref>. Harmonia axyridis is often reported as a host to several natural enemies. These include parasites (Hesperomyces virescens, Coccipolipus hippodamiae, Parasitylenchus bifurcatus), parasitoids (phorid and tachinid flies, Dinocampus coccinellae, Homalotylus spp., Tetrastichinae spp.), pathogens (bacteria, fungi, nematodes, protozoans), and predators (bugs, lacewings, ladybirds, and spiders) <ns0:ref type='bibr' target='#b13'>(Garcés & Williams, 2004;</ns0:ref><ns0:ref type='bibr' target='#b35'>Riddick et al., 2008;</ns0:ref><ns0:ref type='bibr' target='#b33'>Riddick, 2010;</ns0:ref><ns0:ref type='bibr' target='#b18'>Harding et al., 2011;</ns0:ref><ns0:ref type='bibr' target='#b31'>Raak-van den Berg et al., 2014;</ns0:ref><ns0:ref type='bibr'>Haelewaters et al., 2017;</ns0:ref><ns0:ref type='bibr' target='#b6'>Ceryngier et al. 2018)</ns0:ref>. Independent studies show that natural enemies of native ladybirds have recently employed Ha. axyridis as a new host, sometimes simultaneously <ns0:ref type='bibr' target='#b31'>(Raak-van den Berg et al., 2014;</ns0:ref><ns0:ref type='bibr'>Haelewaters et al., 2017;</ns0:ref><ns0:ref type='bibr' target='#b6'>Ceryngier et al. 2018;</ns0:ref><ns0:ref type='bibr' target='#b24'>Knapp et al., 2019)</ns0:ref>. Review of the effects of parasites, pathogens, and parasitoids of Ha. axyridis shows that they have only limited potential for controlling population densities of their host when acting alone <ns0:ref type='bibr'>(Roy et al., 2008;</ns0:ref><ns0:ref type='bibr' target='#b33'>Riddick, 2010;</ns0:ref><ns0:ref type='bibr'>Haelewaters et al., 2017;</ns0:ref><ns0:ref type='bibr' target='#b6'>Ceryngier et al., 2018)</ns0:ref>. Thus far, no studies have focused on the effects of infections of multiple natural enemy on Ha. axyridis.</ns0:p><ns0:p>Hesperomyces virescens (Ascomycota, Laboulbeniomycetes, Laboulbeniales) is a common obligate ectoparasite of ladybirds <ns0:ref type='bibr' target='#b22'>(Roy et al., 2016;</ns0:ref><ns0:ref type='bibr'>Haelewaters et al., 2017)</ns0:ref>. Although known since 1891, it was shown only recently that He. virescens is in reality a complex of multiple hostspecific species <ns0:ref type='bibr'>(Haelewaters et al., 2018)</ns0:ref>. Contrary to most multicellular fungi, He. virescens as well as other members of the Laboulbeniales order lack hyphae, instead they form 3-dimensional multicellular thalli by determinate growth <ns0:ref type='bibr' target='#b2'>(Blackwell et al., 2020)</ns0:ref>. Laboulbeniales, including He. virescens, cannot be grown in axenic culture and no asexual stages are known, which makes their study challenging <ns0:ref type='bibr' target='#b14'>(Haelewaters et al., 2021)</ns0:ref>. Given locally high prevalence of He. virescens on ladybird hosts <ns0:ref type='bibr' target='#b34'>(Riddick & Cottrell, 2010;</ns0:ref><ns0:ref type='bibr'>Haelewaters et al., 2017)</ns0:ref> and the abundance of entomopathogenic fungal strains in the environment <ns0:ref type='bibr' target='#b41'>(Roy & Cottrell, 2008)</ns0:ref>, we examined mortality of native and invasive He. virescens-infected ladybirds exposed to either Beauveria bassiana or Metarhizium brunneum (Ascomycota, Sordariomycetes, Hypocreales) (sensu <ns0:ref type='bibr' target='#b9'>Cottrell & Shapiro-Ilan, 2003</ns0:ref><ns0:ref type='bibr'>, 2008)</ns0:ref>. Because He. virescens forms a branched, non-septate, rhizoidal haustorium <ns0:ref type='bibr' target='#b51'>(Weir & Beakes, 1996)</ns0:ref> that penetrates the host's exoskeleton and makes contact with the body fluid for nutrient uptake, we hypothesized that high thallus densities with concomitant haustorial formation by He. virescens weaken host defenses, thus increasing the host's susceptibility to infection by other natural enemies. With this experiment, we assess how He. virescens affects ladybird survival when exposed to other natural enemies that alone have little impact on Ha. axyridis and compare results with a North American-native ladybird of similar body size, Olla v-nigrum. If He. virescens-on its own and in combination with other natural enemies-significantly impacts survival of the invasive ladybird but not the native one, then the results of this work could have consequences toward a pest management strategy to control infestations of vineyards and agroecosystems by Ha. axyridis.</ns0:p></ns0:div>
<ns0:div><ns0:head>Materials & Methods</ns0:head></ns0:div>
<ns0:div><ns0:head>Field collections and laboratory colonies</ns0:head><ns0:p>Harmonia axyridis and Olla v-nigrum ladybirds were collected for the purpose of establishing laboratory colonies of Hesperomyces-infected and non-infected ladybirds. Specimens were collected at overwintering sites at the 485-ha USDA-ARS, Southeastern Fruit and Tree Nut Research Laboratory, located in Byron, Georgia, <ns0:ref type='bibr'>USA (32.657792,</ns0:ref>. Sex and age of field-collected specimens were not determined to reduce dispersal of fungal propagules <ns0:ref type='bibr'>(Cottrell & Riddick, 2012)</ns0:ref>. All specimens were brought to the laboratory and housed in individual Petri plates (10 cm diam.) with 1/3 of a piece of a cotton dental wick (Deerpack Products, LLC, Miami, Florida) drenched in water for hydration. Ladybirds were housed in environmental chambers at 25 ± 1 ˚C and photoperiod of 14:10 (L:D) h. Food was provided 3× per week in the form of Ephestia kuehniella eggs (Lepidoptera, Pyralidae) and an artificial meat-based diet (Beneficial Insectary, Redding, California). Olla v-nigrum and Ha. axyridis ladybirds were maintained within the Petri plates for 14d and 21d32, respectively, at which time ladybirds were visually examined for presence of Hesperomyces using a dissecting microscope at 50× magnification. Eggs were harvested from ovipositing ladybirds and used to establish clean (free from fungal growth) laboratory-reared colonies of ladybirds with known age.</ns0:p></ns0:div>
<ns0:div><ns0:head>Laboratory rearing of ladybirds</ns0:head><ns0:p>During examination for presence/absence of Hesperomyces, ladybirds were divided into two groups, infected and non-infected. Both groups of ladybirds were placed into plastic rearing containers of 19 × 13.5 × 9 cm (Pioneer Plastics, North Dixon, Kentucky), which were modified with two 3-cm diameter circular openings, one that was covered by 1 × 1 mm mesh to allow for air flow; and the second that was covered with a removable #7 rubber stopper to allow for feeding routinely as well as adding newly emerged laboratory-reared ladybirds. Routine maintenance included transferring ladybirds into fresh rearing containers at the end of each 7d period, which included nutrient supplementations of laboratory-reared yellow pecan aphids, Monelliopsis pecanis (Hemiptera, Aphididae).</ns0:p><ns0:p>The first laboratory generation of adults emerged about one month after placement in rearing containers. Emerging adults were placed into fresh rearing containers and stored into a separate incubator (25 ± 1 ˚C, 14:10 (L:D) h) for 7 days. Similar to field-captured O. v-nigrum and Ha. axyridis, M. pecanis aphids were used as a diet augmentation. As the study progressed, we also incorporated black pecan aphids, Melanocallis caryaefoliae (Hemiptera, Aphididae), in the ladybird diet (3× per week).</ns0:p></ns0:div>
<ns0:div><ns0:head>Artificial transmissions of Hesperomyces</ns0:head><ns0:p>Not only did we need the ladybirds for our experiments to be of the same age, we also needed to artificially infect a subset of these 'clean' laboratory-grown, adult ladybirds with Hesperomyces virescens. Exposure to Hesperomyces was conducted via tumbling of the field-captured 'source' ladybirds (infected with Hesperomyces) with randomly selected laboratory-reared 'target' ladybirds <ns0:ref type='bibr' target='#b10'>(Cottrell & Shapiro-Ilan, 2008)</ns0:ref>. A total of 25 target ladybirds were mixed with 5 Hesperomyces-infected source ladybirds in a 1.6 × 5.8 cm glass tube, which was placed on a hotdog roller (Nostalgia Electrics, Green Bay, Wisconsin) for 5 min. This procedure was repeated for at least 160 target ladybirds of both species. We only performed intra-specific artificial transmissions of Hesperomyces, meaning from source Ha. axyridis to target Ha. axyridis and from source O. v-nigrum to target O. v-nigrum. Both Hesperomyces-exposed target ladybirds and clean (unexposed) ladybirds were fed a diet of M. pecanis aphids for 24h. We did a second tumbling experiment using randomly selected emerged adults from the second cohort of laboratory-reared colonies. More tumbling experiments were performed to increase quantities of Hesperomyces-infected ladybirds, but source/target numbers were changed to 100/40.</ns0:p><ns0:p>To reduce competition for food, ladybirds from all laboratory colonies were transferred from the plastic rearing containers to 14-cm diameter Petri plates. Ladybirds were provided with water ad libitum, E. kuehniella eggs, and artificial meat-based diet. Finally, for assay preparation, the ladybirds were transferred back to clean 19 × 13.5 × 9 cm plastic rearing containers by species.</ns0:p></ns0:div>
<ns0:div><ns0:head>Dual fungal infections assay</ns0:head><ns0:p>Within 24 hours preceding the assay, 160 non-infected and 160 Hesperomyces-infected ladybirds of each species (Ha. axyridis and O. v-nigrum) were each placed into sterile test tubes, one individual per test tube. Test tubes were then closed with a sterile foam stopper to prevent ladybirds from escaping while allowing for air flow. Infected ladybirds were divided into categories according to numbers of thalli per specimen. Because the assay would assess potential interactions between fungal infections, we aimed at selecting heavily Hesperomyces-infected ladybirds; as a baseline, we only used specimens in our bioassays with 14 or more thalli each.</ns0:p><ns0:p>The assay started by pipetting a 1 mL of 2.5 × 10 5 conidia/mL suspension to each test tube <ns0:ref type='bibr' target='#b9'>(Cottrell & Shapiro-Ilan, 2003</ns0:ref><ns0:ref type='bibr'>, 2008)</ns0:ref>. Treatments included native B. bassiana (native Bb), a commercial B. bassiana strain (GHA Bb; Mycotrol ES, Mycotech, Butte, Montana), M. brunneum strain F52 (Mb, isolated from a tortricid moth, Austria 1971; Novozymes, Franklinton, North Carolina), and double-distilled water (ddH 2 O) as a control treatment. Ladybirds were submerged and swirled for 5 s, after which the suspension was removed again using a pipette and each ladybird was placed into a 6 cm-diameter Petri plate. Any remaining droplets of excess suspension was removed by touching only the droplet with a Kimwipe tissue (Kimtech Science Brand, Kimberly-Clark Worldwide, Roswell, Georgia). Petri plates with treated ladybirds were placed into an incubator (25 ± 1 ˚C, 14:10 (L:D) h). Food and cotton rolls drenched in water were provided ad libitum, and Petri plates were replaced as needed in all treatments and replications simultaneously. Ladybirds were observed for mortality and entomopathogeninduced mycosis at day 14. During assay #1, we made daily observations for ladybird mortality and mycosis. Upon death of a given ladybird, ample water was added to the cotton roll to provide moisture for entomopathogen growth and Parafilm was applied around the Petri plate to prevent spreading of the fungus. Deaths of ladybirds and visual confirmations of mycosis were recorded.</ns0:p><ns0:p>We performed 8 different treatments for each ladybird species: 1) He. virescens-positive + native Bb, 2) He. virescens-positive + GHA Bb, 3) He. virescens-positive + Mb, 4) He. virescenspositive + ddH 2 O (control), 5) He. virescens-negative + native Bb, 6) He. virescens-negative + GHA Bb, 7) He. virescens-negative + Mb, and 8) He. virescens-negative + ddH 2 O (double control). In a single assay, we replicated every treatment 3 or 4 times. We performed the entire assay with all treatments and replicated 3 times, using 6-10 ladybirds for each treatment. Note that M. brunneum treatments were used only in assay #3 (Table <ns0:ref type='table'>S1</ns0:ref>). Over all assays done during this study, we used 1,289 specimens of ladybirds (667 O. v-nigrum and 622 Ha. axyridis) (Table <ns0:ref type='table'>S2</ns0:ref>).</ns0:p></ns0:div>
<ns0:div><ns0:head>Statistical analyses</ns0:head><ns0:p>All statistical analyses were performed in the R language and open-access environment for statistical computing v.3.5.0. We used generalized linear mixed models (function glmer(), Rpackage lme4; <ns0:ref type='bibr' target='#b1'>Bates et al., 2015)</ns0:ref> to analyze the effect of the different treatments (GHA Bb, native Bb, Mb) on the survival of Ha. axyridis and O. v-nigrum in relation to prior infection with Hesperomyces. We modeled the binary response variable survival (alive/dead) of each ladybird individual for both host species separately, and used Hesperomyces infection status as well as the interaction of Hesperomyces infection status with treatment as explaining variables. Further, to correct for variation within replicates and assays, we included the random effect of treatment nested in replicate nested in assay. We compared our candidate models to a respective Nullmodel using likelihood ratio tests and, furthermore, calculated pseudo R 2 -values (function r2(), R package sjstats; <ns0:ref type='bibr' target='#b28'>Lüdecke, 2018)</ns0:ref> to evaluate model fit. To visualize the modeling results and obtained model estimates as forest plots, we used the function plot_model() implemented in the R package sjstats <ns0:ref type='bibr' target='#b28'>(Lüdecke, 2018)</ns0:ref>. For assay #1, we further fitted Kaplan-Meier curves to daily mortality data and tested for significant differences in mortality between ladybird species using the function survfit() of the R package survival <ns0:ref type='bibr' target='#b49'>(Therneau & Lumley, 2019)</ns0:ref>.</ns0:p></ns0:div>
<ns0:div><ns0:head>Results</ns0:head><ns0:p>Our candidate models for both host species Ha. axyridis and O. v-nigrum were significantly better at explaining survival relative to chance variation (Chi-squared test, χ 2 = 156.7, P < 0.001; χ 2 = 153.0, P < 0.001, respectively). The overall model fit was high for both candidate models (Ha. axyridis: Nagelkerke's R 2 = 0.40; O. v-nigrum: Nagelkerke's R 2 = 0.53) suggesting the variance is well described by our applied models.</ns0:p><ns0:p>We found a significant negative effect on ladybird survival of the M. brunneum treatment on He. virescens-negative Ha. axyridis (Fig. <ns0:ref type='figure' target='#fig_0'>1</ns0:ref>, Table <ns0:ref type='table'>1</ns0:ref>), whereas B. bassiana treatments did not affect the survival of He. virescens-negative individuals. Infection with He. virescens significantly affected Ha. axyridis survival over all treatments (Fig. <ns0:ref type='figure' target='#fig_0'>1</ns0:ref>). However, there was no additional effect detectable among co-infection treatments for He. virescens-positive ladybirds (Table <ns0:ref type='table'>1</ns0:ref>). Each treatment applied to O. v-nigrum had a significantly negative effect on the survival for both He. virescens-negative and -positive ladybirds (Fig. <ns0:ref type='figure' target='#fig_0'>1</ns0:ref>, Table <ns0:ref type='table'>1</ns0:ref>). Finally, we found an additional negative effect of all co-infection treatments on the survival of He. virescens-positive O. vnigrum (Fig. <ns0:ref type='figure' target='#fig_0'>1</ns0:ref>, Table <ns0:ref type='table'>1</ns0:ref>). These results suggest that there is no effect of dual infections on Ha. axyridis, whereas O. v-nigrum is highly affected by simultaneous exposure to both He. virescens and an entomopathogenic fungus. Percentages of ladybird mortality by treatment are also presented in tabulated form in Table <ns0:ref type='table'>S3</ns0:ref>.</ns0:p><ns0:p>When comparing the daily survival of Ha. axyridis and O. v-nigrum, no significant differences were found in Hesperomyces-positive only treatments (log rank test, P = 0.4). However, when co-infected O. v-nigrum showed a significantly lower survival compared to Ha. axyridis for GHA and native B. bassiana strains (log rank test, P = 0.0014 and P < 0.001, respectively). Fig. <ns0:ref type='figure' target='#fig_1'>2</ns0:ref> shows how survival is significantly different between the two ladybird species when coinfected with both Hesperomyces and Beauveria bassiana (GHA and native).</ns0:p></ns0:div>
<ns0:div><ns0:head>Discussion</ns0:head><ns0:p>Research on the additive effects of multiple natural enemies on a given host is rare, likely because of the complexity involved in designing robust bioassays that include all partners of the system. Combining the natural enemies Orius insidiosus (Hemiptera, Anthocoridae) and Heterorhabditis bacteriophora (Rhabditida, Heterorhabditidae) resulted in the largest decline in larvae of Drosophila suzukii <ns0:ref type='bibr' target='#b32'>(Renkema & Cuthbertson, 2018)</ns0:ref>, which causes major economic losses to fruit crops in its invasive range, spanning North and South America and Europe <ns0:ref type='bibr' target='#b27'>(Lee et al., 2011)</ns0:ref>. The addition of O. insidiosus resulted in 50% fewer D. suzukii larvae compared to treatment with only H. bacteriophora. Plutella xylostella (Lepidoptera, Plutellidae), an important cosmopolitan pest of brassicaceous crops, offers another example. This organism shows resistance to almost all chemical insecticides <ns0:ref type='bibr' target='#b45'>(Sarfraz et al., 2005)</ns0:ref>. Pandora blunckii and Zoophthora radicans (Zoopagomycota, Entomophthoromycetes, Entomophthorales) both infect P. xylostella in the field. In co-inoculation studies with Pa. blunckii and Z. radicans in P. xylostella larvae, larval cadavers (three days post mortality) were most frequently found with conidia of a single entomopathogen, usually the one that had been inoculated first (prior 'residency')-meaning that the other species was excluded <ns0:ref type='bibr' target='#b44'>(Sandoval-Aguilar et al., 2015)</ns0:ref>. In general, the presence of competing species in the same host resulted in a decreased proportion of P. xylostella larvae that were infected compared to single inoculations.</ns0:p><ns0:p>Regarding Ha. axyridis, the following co-infections of natural enemies have been observed in nature: He. virescens + Coccipolipus hippodamiae mites (Acarina, Podapolipidae) in the USA, Austria, and the Netherlands <ns0:ref type='bibr' target='#b7'>(Christian, 2001;</ns0:ref><ns0:ref type='bibr' target='#b33'>Riddick, 2010;</ns0:ref><ns0:ref type='bibr' target='#b31'>Raak-van den Berg et al., 2014)</ns0:ref> and He. virescens + Parasitylenchus bifurcatus nematodes (Nematoda, Allantonematidae) in the Czech Republic, Germany, and the Netherlands (Raak-van den <ns0:ref type='bibr' target='#b31'>Berg et al., 2014;</ns0:ref><ns0:ref type='bibr'>Haelewaters et al., 2017;</ns0:ref><ns0:ref type='bibr'>Herz & Kleespiel, 2012)</ns0:ref>. Given the status of Ha. axyridis as an invasive alien species, these findings demand a better understanding of interactions among the different natural enemies and their potential role in limiting populations of Ha. axyridis. To date, bioassays to determine mortality of ladybirds induced by infection by one or more natural enemies have not yet been performed. Likewise, bioassays including Laboulbeniales have only been carried out in one study <ns0:ref type='bibr' target='#b26'>(Konrad et al., 2015)</ns0:ref>.</ns0:p><ns0:p>When we started this study, He. virescens was considered a single species with multiple ladybird hosts, potentially with multiple strains that infect only a single species, or closely related ones <ns0:ref type='bibr'>(Cottrell & Riddick, 2012)</ns0:ref>. It was recently shown that He. virescens is a complex of multiple species, each with its own ladybird host <ns0:ref type='bibr'>(Haelewaters et al., 2018)</ns0:ref>, which calls for caution in reviewing reports from the extensive body of literature on Hesperomyces findings (summarized in <ns0:ref type='bibr'>Haelewaters & De Kesel, 2017)</ns0:ref>. This also means that isolates of He. virescens from Ha. axyridis and O. v-nigrum in fact represent two different species of Hesperomyces. In other words, the experiments in the current study allow us to make comparisons between two host species, each with their own specific fungal parasite. Future experiments are needed to further disentangle these interactions. Even though horizontal transmission of Hesperomyces among ladybird species is rare <ns0:ref type='bibr'>(Cottrell & Riddick, 2012)</ns0:ref>, we should try to infect Ha. axyridis and O. vnigrum ladybirds with the species of Hesperomyces specific to Olla and Harmonia, respectively, perform bioassays, and compare mortality rates under different treatments with our current results. Analyzing interactions among natural enemies only make sense when the taxa considered represent single biological species.</ns0:p><ns0:p>We found a significant negative effect of Hesperomyces-only infection on the survival of both ladybird hosts (Fig. <ns0:ref type='figure' target='#fig_0'>1a</ns0:ref>, Table <ns0:ref type='table'>S3</ns0:ref>). Previous work has shown Hesperomyces infections to result in decreased mating frequency of female ladybirds, lower (male) survival rates in winter, and impeded sensing ability and flexibility of legs in heavily infected ladybirds <ns0:ref type='bibr' target='#b29'>(Nalepa & Weir, 2007;</ns0:ref><ns0:ref type='bibr' target='#b33'>Riddick, 2010;</ns0:ref><ns0:ref type='bibr'>Haelewaters et al., 2017)</ns0:ref>. One study implicated parasitism by He. virescens as the cause of late summer mortality of Chilocorus bipustulatus ladybirds <ns0:ref type='bibr'>(Kamburov et al., 1967)</ns0:ref> but this was later disputed based on controlled laboratory experiments <ns0:ref type='bibr' target='#b0'>(Applebaum et al., 1971)</ns0:ref>. Our research is the first to explicitly link Hesperomyces infection with increased ladybird mortality.</ns0:p><ns0:p>Our findings on the effects of Hesperomyces on ladybird survival provided a unique opportunity for setting up dual infection assays-the first such experiments to be conducted on ladybirds. When first infected with He. virescens and then exposed to either B. bassiana or M. brunneum, Ha. axyridis mortality was not increased. This result was unexpected. We had hypothesized that Ha. axyridis with high thallus densities of He. virescens would have lowered host defenses against other pathogens. In contrast, the mechanism fostering low susceptibility of Ha. axyridis to entomopathogenic fungi <ns0:ref type='bibr' target='#b9'>(Cottrell & Shapiro-Ilan, 2003;</ns0:ref><ns0:ref type='bibr' target='#b24'>Knapp et al., 2019)</ns0:ref> is not compromised by infection with He. virescens. Similarly, infection of O. v-nigrum by He. virescens-only increased mortality but-in contrast to Ha. axyridis-there was significantly higher mortality when co-infected by entomopathogenic fungi. Differential susceptibility to entomopathogenic fungi was reported by <ns0:ref type='bibr' target='#b9'>Cottrell & Shapiro-Ilan (2003)</ns0:ref>, who showed that native B. bassiana was pathogenic to O. v-nigrum but not to Ha. axyridis. We confirm these results regarding the native strain but we also found the same differential pattern for the GHA strain of B. bassiana, whereas in the earlier study this strain was reported to be pathogenic to neither ladybird species <ns0:ref type='bibr' target='#b9'>(Cottrell & Shapiro-Ilan, 2003)</ns0:ref>. It is perhaps surprising that we detect the GHA strain to be pathogenic to native ladybirds in contrast to the previous results, but ladybird populations may become more susceptible over time for various reasons and natural enemies also become better adapted <ns0:ref type='bibr' target='#b24'>(Knapp et al., 2019)</ns0:ref>. We note that differential susceptibility has also been reported for entomopathogenic nematodes-again, Ha. axyridis was less susceptible compared to O. v-nigrum <ns0:ref type='bibr' target='#b48'>(Shapiro-Ilan & Cottrell, 2005)</ns0:ref>.</ns0:p><ns0:p>In addition, our data are the first account of differential susceptibility to M. brunneum between the invasive Ha. axyridis and the native O. v-nigrum. Whereas infection with M. brunneum had a significantly negative effect on the survival of He. virescens-negative Ha. axyridis, this effect was not visible in the dual infection treatment. The infection with Laboulbeniales probably decreased the susceptibility of Ha. axyridis to infection by M. brunneum, similar to the findings of <ns0:ref type='bibr' target='#b26'>Konrad et al. (2015)</ns0:ref>. These authors found that Laboulbenia-infected Lasius neglectus ants (Hymenoptera, Formicidae) showed a decreased susceptibility to Metarhizium brunneum. This protection against Metarhizium was positively correlated with parasite load. Information on the parasite load of He. virescens on ladybirds in nature is nonexistent. In our bioassays, we selected ladybirds bearing 14 or more fungal thalli as He. virescens-positive specimens. Previous work from a long-term ATBI project in the Netherlands (van Wielink, 2017) points at an average of 19.8 ± 4.9 thalli and a maximum of 129 thalli per Ha. axyridis specimen (n = 270). No such data are available for O. v-nigrum. In other words, based on the available information, the artificial parasite load in our bioassays seems to closely mimic the natural conditions.</ns0:p><ns0:p>Our results provide direct support for the enemy release hypothesis <ns0:ref type='bibr' target='#b23'>(Jeffries & Lawton, 1984)</ns0:ref>. This hypothesis is illustrative for the success of invasive alien species and stipulates that an invasive species in new geographic regions will experience reduced regulatory effects from natural enemies compared to native species, resulting in increased population growth of the invasive species <ns0:ref type='bibr' target='#b8'>Colautti et al., 2004;</ns0:ref><ns0:ref type='bibr' target='#b42'>Roy et al., 2011)</ns0:ref>. However, invasions are dynamic <ns0:ref type='bibr' target='#b46'>(Schultheis et al., 2015;</ns0:ref><ns0:ref type='bibr'>Haelewaters et al., 2017)</ns0:ref> and this escape-from-enemies could be lost as invasive species acquire new enemies over time <ns0:ref type='bibr' target='#b21'>(Hokkanen & Pimentel, 1989)</ns0:ref>. Support for enemy release explaining the success of Ha. axyridis has come from two studies that reported decreased susceptibility of Ha. axyridis to entomopathogenic fungi <ns0:ref type='bibr' target='#b9'>(Cottrell & Shapiro-Ilan, 2003)</ns0:ref> and entomopathogenic nematodes <ns0:ref type='bibr' target='#b48'>(Shapiro-Ilan & Cottrell, 2005)</ns0:ref> compared to the native American ladybird species. Our work adds another level of complexity by the addition of a second natural enemy to the interactions. Again, we find differential susceptibility between the invasive and native ladybird species-with a reduced regulatory effect of the tested natural enemies on Ha. axyridis.</ns0:p></ns0:div>
<ns0:div><ns0:head>Conclusions</ns0:head><ns0:p>In this paper, we show a negative effect of infection by Hesperomyces virescens on the survival of both Harmonia axyridis and Olla v-nigrum ladybirds (Fig. <ns0:ref type='figure' target='#fig_0'>1a</ns0:ref>). This is the first study to unequivocally link Hesperomyces infection with increased host mortality and only the second to perform bioassays with hosts co-infected with Laboulbeniales and a second entomopathogenic fungus <ns0:ref type='bibr' target='#b26'>(Konrad et al., 2015)</ns0:ref>. However, the susceptibility to a secondary entomopathogenic fungus was only elevated in the native American ladybird species (O. v-nigrum), whereas the globally invasive Ha. axyridis showed no significant increase in mortality when co-infected with either Beauveria bassiana or Metarhizium brunneum (Figs. <ns0:ref type='figure' target='#fig_1'>1, 2</ns0:ref>). These findings are consistent with the enemy release hypothesis <ns0:ref type='bibr' target='#b23'>(Jeffries & Lawton, 1984)</ns0:ref> and highlight the difficulty in controlling this invasive alien species. Future studies are needed to elaborate population-specific effects on native and commercial strains of entomopathogenic fungi used in pest control. </ns0:p></ns0:div><ns0:figure xml:id='fig_0'><ns0:head>Figure 1</ns0:head><ns0:label>1</ns0:label><ns0:figDesc>Figure 1</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_1'><ns0:head>Figure 2</ns0:head><ns0:label>2</ns0:label><ns0:figDesc>Figure 2</ns0:figDesc></ns0:figure>
<ns0:note place='foot'>PeerJ reviewing PDF | (2020:04:48281:2:0:CHECK 7 Sep 2020)</ns0:note>
<ns0:note place='foot'>PeerJ reviewing PDF | (2020:04:48281:2:0:CHECK 7 Sep 2020)Manuscript to be reviewed</ns0:note>
</ns0:body>
" | "Dear editor,
We added our responses in blue under each of the reviewers’ comments and suggestions.
EDITOR’S DECISION
We now have the second round of reviews from our reviewers. Overall, this second version almost
completely satisfies the suggestions of reviewers, including this editor. I strongly recommend
resubmitting a latest version as soon as possible with the points indicated by the reviewers.
Response: Thank you for the swift process. We made edits based on the input—minor edits, as you
suggested. The title of Figure 2 was changed to be in line with what the purpose was of this figure. From
reviewer #1’s comment, we realized it might have been confusing without proper title. Finally, we note
that we did not agree with reviewer #3 about changing the hypothesis. Our hypothesis is the one we
formulated when we started the experiment; it would be very bad practice changing it based on the
results.
REVIEWER #1
Basic reporting
Same as before (in my previous review)
Experimental design
Same as before (in my previous review)
Validity of the findings
Same as before (as in my previous review)
Comments for the author
In the Intro. or M & M section, please indicate that two strains of the fungus (Hesperomyces virescens)
are likely present in the field. Please add a sentence or two to reveal this important information to your
readers.
Response: We agree that this is indeed very important information. We added a note in lines 75–77 as
per this suggestion:
“Although known since 1891, it was shown only recently that He. virescens is in reality a complex of
multiple host-specific species (Haelewaters et al., 2018).”
As a note: this is also discussed in detail in lines 268–282. (We added some more details in this revision.)
In Figure 2, I believe it is difficult to compare infection rate between lady beetle species, if the fungus (H.
virescens) exists as two distinct strains, one on Harmonia axyridis and the other on Olla v-nigrum. The
Results should be slightly revised to address this problem.
Response: The purpose of Figure 2 is to show how whether the co-infection patterns are different
between ladybird species. These patterns are very different (no effect vs strong effect) so it is crucial to
showcase. It is explained this way in the results (lines 229 – 232) and it is also an important part of our
discussion (lines 288–307). We thank the reviewer for making us realize that this figure or its purpose
might not have been clearly defined. To increase clarity, we changed the title of Figure 2 as follows:
“Figure 2. Effect of Hesperomyces-infection and co-infection with Hesperomyces and Beauveria bassiana
on the survival of ladybirds”.
We also made slight changes to the paragraph in the Results section at lines 229–235.
REVIEWER #3
Basic reporting
Regarding the Introduction, I do not see major changes. Only one final paragraph was added and none
of my suggestions or concerns were considered, and their arguments for not doing so are not
convincing. I think the authors should modified the introduction (theoretical background/context) as
requested before (reviewer 2 also pointed out that theoretical framework must be more in line with the
objective of the study; in this case, the comparison between the invasive H. axyridis and the native Olla
v-nigrum to co-infection). It is not adequate to mention O. v-nigrum just at the end of the Introduction,
without any justification what it is pretended to test with the inclusion of this native species in the
analyses.
Response: Olla v-nigrum is only added at the end of the introduction because of how the experiment
was set up. This is tied to Reviewer #3’s third concern about changing the hypothesis, which we strongly
disagree with. In our initial discussions when designing this project, the hypothesis as written in the
introduction was put forward. The idea was to use a prevalent natural enemy (Hesperomyces virescens)
to regulate the invasive Harmonia axyridis, in combination with the common entomopathogenic fungi,
which we know were common in the envrionment. One major concern was that, in the case we would
find effects, we needed to know whether these were species-specific (i.e. only occurring in H. axyridis)
or more general. This is when we decided to add Olla v-nigrum in the experiment, as a comparison. The
introduction is written in this same logic.
The origins of the fungi should be mentioned in the Introduction. For H. virescens, Roy et al. (2011;
Biocontrol) state that is probably native to North America or at least present there since 1931. See also
Orlova-Bienkowskaja et al. (2018) Plos One. Are the other species native to Northamerica? If not, when
they were introduced to the country? For how long have they been in contact with H. axyridis and O. vnigrum in US? This information is very important when comparing the response of native and invasive
ladybird species to the infection by pathogens.
Response: The reviewer mentions Roy et al. (2011), who treated the fungus as a single species. We now
know that Hesperomyces virescens is a complex of multiple species, so the fungus that was described by
Thaxter in 1891 from Chilocorus stigma is not the same one as the fungus on Harmonia axyridis. Where
Hesperomyces ex Harmonia axyridis is native, we do not know—this topic deserves a manuscript on its
own (which I am working towards, by the way). Currently, this is all mere speculation.
I insist the hypothesis has to be revised, including what was expected when comparing native versus an
exotic ladybird species because the manuscript, since the beginning (title), put special emphasis in how
native versus exotic ladybird species respond to co-infection by ectoparasitic and entomopathogenic
fungi.
Response: We discussed this suggestion a bit among co-authors, and we do not understand why we
would need to revise the hypothesis. It *is* the exact hypothesis that we used to construct this
experiment. It would be very bad practice to change a hypothesis based on the results, or a reviewer’s
opinion, for that matter.
Experimental design
No comment
Validity of the findings
No comment
Comments for the author
Most of my suggestions to the original version of the manuscript were considered in this new version.
Nevertheless, I think that the Introduction has to be modified according to my comments above.
Response: We did include an explicit statement in the introduction about Hesperomyces virescens being
a species complex. This was perhaps “hidden” away in the discussion. Other than this, (i) we refrain from
adding speculation about the origin of the different Hesperomyces species and (ii) the hypothesis has
not changed for reasons discussed above.
I insist Tables 1 and 2 should be merged. They both address the same thing, but for the two species of
coccinellids.
Response: The tables have now been merged as per the reviewer’s suggestion.
Figure 2 is incorrect. All three plots are the same (see Fig. 2 in the original). Also, I insist that controls
must be included in this figure. This point was also raised by reviewer 2 and the authors' response is not
satisfactory. The design includes controls without H. viriscens and these should be included in all
analyzes.
Response: You are correct. Figure 2 is corrected now (plots are different). Thank you for noting this
issue. We only recorded daily mortality data for ladybirds with Hesperomyces-positive only treatments
and co-infection treatments (Hesperomyces + Beauveria bassiana) with the specific purpose to compare
single infection to co-infection. For the uninfected ladybirds, we have the mortality data at day 14. We
explained this in our previous rebuttal. The writing has now improved in the paper (lines 229–235) and
caption of Figure 2.
" | Here is a paper. Please give your review comments after reading it. |
9,742 | "<ns0:abstract xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:p>Background. Type 2 diabetes mellitus (T2DM) is a metabolic disease affecting a huge population worldwide. Teucrium polium L. has been used as a folk medicine for the treatment of T2DM in Anatolia, Turkey. The antihyperglycemic effect of the plant was reported previously. However, there was no detailed study on the underlying molecular mechanisms. In this study, we generated a research plan to clarify the active constituents of the extract and uncover the molecular mechanisms using network pharmacology analysis.</ns0:p></ns0:div>
<ns0:div><ns0:head>Methods.</ns0:head><ns0:p>For this purpose, we composed a dataset of 126 compounds for the phytochemical profile of the aerial parts of T. polium. Drug-likeness of the compounds was evaluated, and 52 compounds were selected for further investigation. 252 T2DM related targets hit by selected compounds were subjected to DAVID database.</ns0:p><ns0:p>Results. The KEGG pathway analysis showed enrichment for TNF signaling pathway, insulin resistance, HIF-1 signaling pathway, apoptosis, PI3K-AKT signaling pathway, FOXO signaling pathway, insulin signaling pathway, and type 2 diabetes mellitus which are related to T2DM . AKT1, IL6, STAT3, TP53, INS, and VEGFA were found to be key targets in protein-protein interaction. Besides these key targets, with this study the role of GSK3β, GLUT4, and PDX1 were also discussed through literature and considered as important targets in the antidiabetic effect of T. polium. Various compounds of T. polium were shown to interact with the key targets activating PI3K-AKT and insulin signaling pathways.</ns0:p></ns0:div>
<ns0:div><ns0:head>Conclusions.</ns0:head><ns0:p>According to these findings, mainly phenolic compounds were identified as the active components and IRS1/PI3K/AKT signaling and insulin resistance were identified as the main pathways regulated by T. polium. This study reveals the relationship of the compounds in T. polium with the targets of T2DM in human. Our findings suggested the use of T. polium as an effective herbal drug in the treatment of T2DM and provides new insights for further research on the antidiabetic effect of T. polium .</ns0:p></ns0:div>
</ns0:abstract>
<ns0:body xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:head>Introduction</ns0:head><ns0:p>Diabetes mellitus is a metabolic disease characterized by high blood glucose levels. According to the reports of WHO, about 422 million people live with diabetes. Diabetes was a direct cause of about 1.6 million deaths only in 2016. Among adults, 90% of the patients have type 2 diabetes mellitus (T2DM) <ns0:ref type='bibr' target='#b30'>(Holman et al. 2015)</ns0:ref>. In T2DM, β-cell dysfunction and/or insulin resistance results with hyperglycemia and high glucose levels in blood. Patients with T2DM are under risk for some complications that diabetes can cause such as cardiovascular disease, renal disease, diabetic retinopathy and neuropathy <ns0:ref type='bibr' target='#b70'>(Zheng et al. 2018)</ns0:ref>. In T2DM, multitarget treatment is used to overcome the defects caused in the organism <ns0:ref type='bibr' target='#b27'>(He et al. 2019;</ns0:ref><ns0:ref type='bibr' target='#b65'>Vuylsteke et al. 2015)</ns0:ref>. The genus Teucrium L., a member of Lamiaceae family (Subfamily Ajugoideaea), has a cosmopolitan distribution and including about 250 species spread worldwide <ns0:ref type='bibr' target='#b59'>(Stevens 2017)</ns0:ref>. In Turkey, Teucrium polium L. is known as 'Acıyavşan' and used as a traditional medicine for the treatment of diabetes <ns0:ref type='bibr' target='#b4'>(Arıtuluk & Ezer 2012)</ns0:ref>. In Algeria and Iran, T. polium is used traditionally for the treatment of diabetes <ns0:ref type='bibr' target='#b15'>(Chinsembu 2019;</ns0:ref><ns0:ref type='bibr' target='#b54'>Rezaei et al. 2015)</ns0:ref>. The infusion or decoction of aerial parts is used frequently for the treatment of diabetes, stomachache, hemorrhoid, common colds, abdominal pains, antipyretics, and sunstroke as internally <ns0:ref type='bibr' target='#b64'>(Tuzlacı 2016)</ns0:ref>. T. polium has been shown to have mainly flavonoids, phenolic acids, phenylethanoid glycosides, and terpenoids mainly diterpenoids <ns0:ref type='bibr' target='#b7'>(Bahramikia & Yazdanparast 2012)</ns0:ref>. Phytotherapeutic effects of T. polium, such as antioxidant, antimutagenic, cytotoxic, anticancer, hepatoprotective, antiinflammatory, hypolipidemic, hypoglycemic, antinociceptive, antispasmodic, antiulcer, antibacterial, antiviral, and antifungal activities have been shown by in vivo or in vitro assays <ns0:ref type='bibr' target='#b6'>(Bahramikia & Yazdanparast 2011;</ns0:ref><ns0:ref type='bibr' target='#b26'>Hasani-Ranjbar et al. 2010)</ns0:ref>. The hypoglycemic effect of T. polium was shown by several reports <ns0:ref type='bibr' target='#b22'>(Esmaeili & Yazdanparast 2004;</ns0:ref><ns0:ref type='bibr' target='#b25'>Gharaibeh et al. 1988;</ns0:ref><ns0:ref type='bibr' target='#b57'>Shahraki et al. 2007;</ns0:ref><ns0:ref type='bibr'>Yazdanparas et al. 2005)</ns0:ref>. The hypoglycemic effect of T. polium was observed by <ns0:ref type='bibr'>Gharaibeh et. al. (1998)</ns0:ref> for the first time. In the research, the decoction of aerial parts was tested through three different administrations (oral, intraperitoneal, and intravenous) in normoglycemic and streptozotocin-induced hyperglycemic rats. In all administration ways of T. polium decoction, it had caused a decrease in blood glucose concentration. The decoction of T. polium had a decrease of 20.5% in blood glucose concentration by oral administration while intraperitoneal and intravenous administration of the decoction had a decrease of 26.5% and 44%, respectively. The study has suggested that the hypoglycemic effect of T. polium was a result of an increase in the peripheral utilization of glucose <ns0:ref type='bibr' target='#b25'>(Gharaibeh et al. 1988)</ns0:ref>. In another study, administration of ethanol-water (7:3) extract of aerial parts of T. polium per six weeks, resulted in a decrease of 64% in blood glucose levels of streptozocin-induced hyperglycemic rats. The result of this study also had proved that T. polium extract reduced blood glucose concentration by increasing pancreatic insulin secretion dosedependently <ns0:ref type='bibr' target='#b22'>(Esmaeili & Yazdanparast 2004)</ns0:ref>. Although there are reports about the hypoglycemic effect of T. polium, there is still a lack of information for underlying mechanisms. In a recent study, to elucidate molecular mechanisms, effects of T. polium extract on pancreatic islets cells regeneration was investigated. It was found that the antidiabetic effect of T. polium was connected with the antioxidant defense system and Pdx1 expression in the JNK pathway <ns0:ref type='bibr' target='#b61'>(Tabatabaie & Yazdanparast 2017)</ns0:ref>. Plants have been used for the treatment of various diseases in folk medicine. These herbal preparations consist of multiple compounds that target multiple proteins in an organism. This suits well with the multicomponent-multitarget paradigm <ns0:ref type='bibr' target='#b69'>(Zhang et al. 2019)</ns0:ref>. Network pharmacology provides information to understand the underlying mechanisms of therapeutic and adverse effects of these multicomponent therapeutics <ns0:ref type='bibr' target='#b31'>(Hopkins 2008;</ns0:ref><ns0:ref type='bibr' target='#b40'>Keith et al. 2005;</ns0:ref><ns0:ref type='bibr' target='#b47'>Li & Zhang 2013)</ns0:ref>. Unlike the trend in drug research studies held in the 20th century which aims single components affecting single targets, nowadays researchers focus on multicomponent therapeutics. Network pharmacology is a rising trend in the 21st century, mainly after 2010 <ns0:ref type='bibr' target='#b49'>(Lu et al. 2019)</ns0:ref>. Network pharmacology studies revealed molecular mechanisms of several Traditional Chinese Medicine (TCM) recipes in the treatment of complex diseases already <ns0:ref type='bibr' target='#b12'>(Chen et al. 2019;</ns0:ref><ns0:ref type='bibr' target='#b13'>Chen et al. 2018;</ns0:ref><ns0:ref type='bibr' target='#b67'>Xiang et al. 2019)</ns0:ref>.</ns0:p><ns0:p>In this study, the underlying mechanisms of T. polium in the treatment of diabetes were aimed to be elucidated. For this purpose, firstly the phytochemical content of the plant was screened through a detailed literature search. Compounds reported from T. polium were selected based on their drug-likeness properties and screened for their potential targets that play a role in the biological processes. Therapeutical targets of T2DM and targets of the compounds were merged for further investigations. Protein-protein interaction (PPI) network of common targets was constructed. Key targets were determined and the role of targets in T2DM pathways was discussed.</ns0:p></ns0:div>
<ns0:div><ns0:head>Materials & Methods</ns0:head><ns0:p>Literature based search for phytochemical content of T. polium Previous phytochemical studies on T. polium were reviewed and compounds that were isolated or determined listed. Literature search was performed using 'Scopus' and 'Web of Science -Clarivate' databases with the keyword 'Teucrium polium' upto June 2020. After the review process, it was found that 126 compounds were reported from T. polium by several reports. Due to their structure, compounds were listed under 3 groups (phenolics, terpenoids and amino acid derivatives). All the compounds were converted to Canonical SMILES format using PubChem (https://pubchem.ncbi.nlm.nih.gov/) or CS Chemdraw Ultra.</ns0:p></ns0:div>
<ns0:div><ns0:head>Evaluation of drug likeness of the compounds</ns0:head><ns0:p>The absorption and permeation abilities of the compounds in the extract play a critical role in the biological activity observed. In this study, Lipinski's rule of five was used to filter compounds which possess good absorption and permeation so that could be a new drug candidate. According to this criteria compounds which have; i) molecular weight (MW) greater than 500, ii) the calculated logP value above 5, iii) more than 5 hydrogen bond donors (HBD) and iv) more than 10 hydrogen bond acceptors (HBA) were filtered <ns0:ref type='bibr' target='#b48'>(Lipinski et al. 2001;</ns0:ref><ns0:ref type='bibr' target='#b63'>Turner & Agatonovic-Kustrin 2007)</ns0:ref>. All 126 compounds were subjected to SWISSADME to obtain data for pharmacokinetics and drug-likeness <ns0:ref type='bibr' target='#b18'>(Daina et al. 2017)</ns0:ref>. 52 compounds meet the 'Lipinski's rule of 5' and further used to construct 'Compound-Target' network. The detailed results for all the dataset of 126 compounds can be found in Table <ns0:ref type='table'>S1</ns0:ref>.</ns0:p></ns0:div>
<ns0:div><ns0:head>Construction of 'Compound-Target' network</ns0:head><ns0:p>In an herbal extract, each compound has an interaction with specific targets. The biological effects of the extract are a result of these interactions. Targets of the selected 52 compounds were searched through TCSMP (Traditional Chinese Medicine Systems Pharmacology Database and Analysis Platform), and SYMMAP databases <ns0:ref type='bibr' target='#b55'>(Ru et al. 2014;</ns0:ref><ns0:ref type='bibr' target='#b66'>Wu et al. 2018)</ns0:ref>. Compounds were also subjected to SwissTargetPrediction for target fishing <ns0:ref type='bibr' target='#b19'>(Daina et al. 2019)</ns0:ref>. During these screening, targets were limited for Homo sapiens. To avoid confusion, targets obtained were screened for UniprotKB ID for their unique identifiers (http://www.uniprot.org/) <ns0:ref type='bibr' target='#b17'>(Consortium 2018)</ns0:ref>. Duplicate targets for the same compounds were removed and 'Compound-Target' network was obtained including 704 targets (Table <ns0:ref type='table'>S2</ns0:ref>). Cytoscape 3.8.0 was used for the visualization of the network <ns0:ref type='bibr' target='#b58'>(Shannon et al. 2003)</ns0:ref>.</ns0:p><ns0:p>Collection of T2DM targets and network construction T2DM related genes were collected from DisGeNET (https://www.disgenet.org/). 'Type 2 diabetes mellitus' was used as a keyword. DisGeNET is a platform with collections of genes associated with diseases <ns0:ref type='bibr' target='#b8'>(Bauer-Mehren et al. 2010)</ns0:ref>. 1513 genes related to T2DM were obtained. The data obtained were transferred to Cytoscape 3.8.0.</ns0:p><ns0:p>'Compound-Target-Disease' network and the 'Protein-Protein Interaction' network construction For further investigation, the intersection of 'Compound-Target' network and T2DM related genes were set as 'Compound-Target-Disease' network. This network consisted of 252 genes (Table <ns0:ref type='table'>S3</ns0:ref>). For an illustration of the roles of selected genes in biological systems, the STRING database (http://string-db.org/, version 11) was used. STRING is a database that helps understanding associations between expressed proteins in a cellular function <ns0:ref type='bibr' target='#b60'>(Szklarczyk et al. 2018)</ns0:ref>. Protein-protein interaction (PPI) map of 252 genes were generated. The confidence score was set as high (> 0.7). The key targets were defined using topological analysis. Topological network parameters cover some properties such as; degree distributions, stress centrality, betweenness centrality, closeness centrality <ns0:ref type='bibr' target='#b20'>(Doncheva et al. 2012)</ns0:ref>. In this study, degree distributions were selected to identify key targets.</ns0:p><ns0:p>Gene enrichment analysis DAVID Bioinformatics Resources 6.8 was used in gene enrichment analysis. DAVID database integrates biological knowledge with analytical tools that provide bioinformatic annotations <ns0:ref type='bibr' target='#b33'>(Huang et al. 2008;</ns0:ref><ns0:ref type='bibr' target='#b34'>Huang et al. 2009)</ns0:ref>. The 252 genes which were common for compounds and disease were uploaded to DAVID (https://david.ncifcrf.gov/). The results were listed based on their p values. Top 20 results with lower p value were selected. The results of gene ontology (GO) function and (Kyoto Encyclopedia of Genes and Genomes) KEGG pathway analysis were evaluated and discussed <ns0:ref type='bibr' target='#b5'>(Ashburner et al. 2000;</ns0:ref><ns0:ref type='bibr' target='#b16'>Consortium 2019;</ns0:ref><ns0:ref type='bibr' target='#b51'>Mi et al. 2019)</ns0:ref>.</ns0:p><ns0:p>Construction of 'Compound-Target-Pathway' network 'Compound-Target-Disease' network and selected KEGG pathways were intersected to give 'Compound-Target-Pathway' network. Cytoscape was used for visualization.</ns0:p></ns0:div>
<ns0:div><ns0:head>Results</ns0:head><ns0:p>Screening of chemical compounds in T. polium and selection for the potential active compounds Through a detailed literature search, 126 compounds were listed in the aerial parts of T. polium (Table <ns0:ref type='table'>S4</ns0:ref>). Mainly phenolic compounds (flavonoids, phenylethanoid glycosides, phenolic acids), terpenoids (secoiridoids, iridoids, sesquiterpenoids, diterpenoids, triterpenoids) and amino acid derivatives (cyanogenic glycosides) were identified. Drug-likeness of these compounds were scanned through Lipinski's rule of 5 (the parameters of the compounds were given in Table <ns0:ref type='table'>S1</ns0:ref>). 80 compounds that met the selected criteria were searched for their potential targets using SYMMAP, TCMSP and Swiss Target Prediction databases. The databases provided information for 52 compounds (Table <ns0:ref type='table'>1</ns0:ref>).</ns0:p><ns0:p>Analysis of compound-target interactions and determination of the common targets of T. polium and T2DM related genes There is a total of 704 genes related to 52 compounds. The 'Compound-Target' network consists of 756 nodes and 4023edges. Quercetin, apigenin and luteolin were found to be in relation with more targets than the other compounds (edge numbers were 254, 184 and 158 respectively). For further investigation, the common targets for T. polium and T2DM were determined. Firstly, 1513 T2DM related genes were imported from DisGeNet. All the targets were converted to Uniprot IDs to avoid confusion. The merge process of compound-target network and diseasetarget network resulted in 252 common targets (Fig. <ns0:ref type='figure'>1</ns0:ref>). These targets were selected for further investigation to understand the mechanisms of T. polium in the treatment of T2DM.</ns0:p></ns0:div>
<ns0:div><ns0:head>Construction of PPI networks and determination of the key targets</ns0:head><ns0:p>To understand the metabolic processes, PPIs play a key role. It comprises a network including direct and indirect interactions between proteins which give researchers new insights in understanding biological phenomena <ns0:ref type='bibr' target='#b36'>(Ijaz et al. 2018;</ns0:ref><ns0:ref type='bibr' target='#b60'>Szklarczyk et al. 2018)</ns0:ref>. To clarify the key targets in the 'Compound-Target-Disease' network, the target genes were subjected to STRING 11.0 using a confidence score of > 0.7 (high) to achieve PPI network. The PPI network had 252 nodes and 1912 edges. According to topological analysis, degree distributions were evaluated. Degree shows the interaction numbers of the targets within the network. The nodes with a higher degree are referred to as a hub. The hub plays a key role in the biological process as it is related with more targets. Two-fold of the mean of the degree was selected as a threshold for the determination of key targets. 37 targets with a higher degree than 30.3 was thought to have a critical role for the mechanism of action (AKT1, INS, VEGFA, IL6, TP53, STAT3, MAPK1, APP, TNF, MAPK8, CXCL8, EGFR, PIK3CA, PIK3R1, SRC, MMP9, IL10, PTGS2, IL1B, CCL2, RELA, HRAS, GAPDH, PTEN, IL2, IL4, MTOR, TLR4, CASP3, JAK2, ICAM1, ESR1, FGF2, CXCL10, PPARG, MMP2, MAPK14). Acacetin (TP4), apigenin (TP5), jaceosidin (TP12), luteolin (TP13), quercetin (TP14), and caffeic acid (TP15) showed higher interactions with the key targets and might have a role in the antidiabetic effects of T. polium (Fig. <ns0:ref type='figure'>2</ns0:ref>). PPI network for key targets had 37 nodes and 425 edges. Interleukin-6 (IL6), signal transducer and activator of transcription 3 (STAT3), mitogen-activated protein kinase 1 (MAPK1), insulin (INS), and vascular endothelial growth factor A (VEGFA) were the proteins with a higher number of interactions (Fig. <ns0:ref type='figure'>3</ns0:ref>).</ns0:p></ns0:div>
<ns0:div><ns0:head>Gene enrichment analysis using DAVID database for GO and KEGG</ns0:head><ns0:p>PeerJ reviewing PDF | (2020:07:50669:1:0:NEW 9 Sep 2020)</ns0:p><ns0:p>Manuscript to be reviewed 252 common targets were subjected to DAVID database for gene enrichments. GO and KEGG gene enrichment results were put in order according to their p values. GO enrichment results were given in three parts; molecular, biological, and cellular. Top 20 results for each analysis were plotted in a graph produced by Graphpad Prism 6 (Fig. <ns0:ref type='figure'>4, 5</ns0:ref>). The results for GO analysis were evaluated through related terms option of DAVID database. According to biological process results; response to drug (GO:0042493), negative regulation of apoptotic process (GO:0043066) and positive regulation of transcription from RNA polymerase II promoter (GO:0045944) showed higher target numbers in count (Fig. <ns0:ref type='figure'>4</ns0:ref>). Negative regulation of apoptotic process, inflammatory response (GO:0006954), positive regulation of cell proliferation (GO:0008284), glucose homeostasis (GO:0042593) and glucose transport (GO:0015758) were found to be related with at least one of the KEGG pathways such as PI3K-Akt signaling pathway (hsa04151), TNF signaling pathway (hsa04668), insulin resistance (hsa04931), insulin signaling pathway (hsa04910), FoxO signaling pathway (hsa04068), adipocytokine signaling pathway (hsa04920), AMPK signaling pathway (hsa04152) and type 2 diabetes mellitus (hsa04930). Molecular function results with higher target numbers were protein binding (GO:0005515), protein homodimerization activity (GO:0042803), and identical protein binding (GO:0042802) (Fig. <ns0:ref type='figure'>4</ns0:ref>). Kinase activity (GO:0016301) and insulin receptor substrate binding (GO:0043560) were found to be related to at least one of the KEGG pathways such as insulin resistance, insulin signaling pathway, FoxO signaling pathway, PI3K-Akt signaling pathway, and type 2 diabetes mellitus. KEGG enrichment results supported these findings. Results of 252 common targets were listed as (related with T2DM); TNF signaling pathway, insulin resistance, apoptosis, HIF-1 signaling pathway, PI3K-Akt signaling pathway, FoxO signaling pathway, insulin signaling pathway and type 2 diabetes mellitus (Fig. <ns0:ref type='figure'>5</ns0:ref>). The top 20 results according to the p values suggested that, compounds reported from T. polium might also leads new insights for the treatment of cancer (Table <ns0:ref type='table'>2</ns0:ref>).</ns0:p></ns0:div>
<ns0:div><ns0:head>Discussion</ns0:head><ns0:p>The potent hypoglycemic effect of T. polium extract has been reported by several reports <ns0:ref type='bibr' target='#b22'>(Esmaeili & Yazdanparast 2004;</ns0:ref><ns0:ref type='bibr' target='#b25'>Gharaibeh et al. 1988;</ns0:ref><ns0:ref type='bibr' target='#b61'>Tabatabaie & Yazdanparast 2017;</ns0:ref><ns0:ref type='bibr'>Yazdanparas et al. 2005)</ns0:ref>. According to the study performed by Tabatabaie and Yazdanparast, T. polium extract lowered fasting blood glucose levels closely to the control group. In addition to the hypoglycemic effect, T. polium treated rats had lower triglyceride and cholesterol levels when compared with diabetic rats <ns0:ref type='bibr' target='#b61'>(Tabatabaie & Yazdanparast 2017</ns0:ref>). These experimental data show the potential of T. polium as a promising herb in the treatment of T2DM. According to the findings of this study, we are considering that T. polium show its antidiabetic effect via enhancing β-cell number and function and exhibiting insulin-like effect through PI3K-AKT pathway. Previous reports performed on the extract and compounds of T. polium support our findings of the network pharmacology assisted analysis. According to the KEGG enrichment results in our study, 18 key targets take part in PI3K-AKT pathway and 11 key targets take part in insulin resistance pathway (Table <ns0:ref type='table'>2</ns0:ref>). For this view, it is essential to understand the roles of these pathways in T2DM. T2DM is appeared owing to two fundamental defects that are insulin resistance and impaired βcell function which are caused by long-term hyperglycemia <ns0:ref type='bibr' target='#b11'>(Cheatham & Kahn 1995)</ns0:ref>. Insulin resistance is characterized as reduced insulin sensitivity in the target tissue (like skeletal muscle, liver, and adipose). It is connected with the pathogenesis of metabolic diseases like obesity, type 2 diabetes, hypertension, cardiovascular diseases, and fatty liver disease <ns0:ref type='bibr' target='#b21'>(Draznin 2020</ns0:ref>). INS, one of the hub genes in the PPI network in our study, regulates glucose metabolism and ensure metabolic homeostasis. It also promotes glycogen synthesis, lipid metabolism, protein synthesis and degradations, gene transcriptions, etc. <ns0:ref type='bibr' target='#b11'>(Cheatham & Kahn 1995)</ns0:ref>. Additionally, it is an important regulator of pancreatic β-cells growth and proliferation through the phosphatidylinositol 3-kinase (PI3K)/AKT (also known as protein kinase B-PKB) pathway <ns0:ref type='bibr' target='#b23'>(Fujimoto & Polonsky 2009)</ns0:ref>. Insulin receptor activation through insulin binding stimulates PI3K-AKT signaling pathway. AKT is determined as one of the key targets in our study and 4ꞌ-O-methyl luteolin (TP2), 6hydroxy luteolin (TP3), apigenin (TP5), isoscutellarein (TP11), jaceosidin (TP12), luteolin (TP13), quercetin (TP14), 20-O-acetyl teucrasiatin (TP24), and capitatin (TP25), 4β,5α-epoxy-7αH-germacr-10( <ns0:ref type='formula'>14</ns0:ref>)-en,1β-hydroperoxyl,6β-ol (TP36), 4α,5β-epoxy-7αH-germacr-10( <ns0:ref type='formula'>14</ns0:ref>)en,1β-hydroperoxyl,6α-ol (TP38), 5,3',4'-trihydroxy-3,7-dimethoxyflavone (TP47), jaranol (TP48) showed interaction with the target AKT1 that could have a role in the antidiabetic effect of T. polium (Fig. <ns0:ref type='figure'>2</ns0:ref>). AKT activation, promotes cell survival, proliferation, and growth by controlling key signaling nodes such as glycogen synthase kinase 3 (GSK3), Forkhead Box O (FoxO) transcriptions factors, tuberous sclerosis complex 2 (TSC2) and mechanistic target of rapamycin (mTOR) complex 1 (mTORC1) <ns0:ref type='bibr' target='#b50'>(Manning & Toker 2017)</ns0:ref>. In a previous study about molecular mechanisms of the effects of T. polium extract on pancreatic β-cells regeneration, it was showed that c-jun N-terminal kinase (JNK) pathway provoked with oxidative stress leads to the inactivation of pancreas/duodenum homeobox protein 1 (PDX1) via FoxO1. In the same study, T. polium extract increased FoxO1 phosphorylation and, promoted the expression of PDX1 <ns0:ref type='bibr' target='#b61'>(Tabatabaie & Yazdanparast 2017)</ns0:ref>. Tyrosol (TP21), a polyphenol reported from T. polium, also was shown to inhibit ER-stress induced β-cell apoptosis by JNK phosphorylation <ns0:ref type='bibr' target='#b45'>(Lee et al. 2016)</ns0:ref>. JNK pathway induces insulin receptor substrate 1 (IRS1) inhibition by causing serine phosphorylation which is an important step in downstream of insulin receptor signaling. This impairs the insulin signaling pathway by inhibiting IRS1-IRS2/PI3K-AKT pathway <ns0:ref type='bibr'>(Kaneto et al. 2005)</ns0:ref>. IRS1-IRS2/PI3K-AKT pathway inactivates FoxO1 which supports pancreatic β-cells proliferation by enhancing PDX1 expression <ns0:ref type='bibr' target='#b41'>(Kitamura et al. 2002)</ns0:ref>. PDX1 has a very important role in pancreatic β-cell function and survival, is regulated through FoxO1 and glycogen synthase kinase 3 beta (GSK3β) <ns0:ref type='bibr' target='#b23'>(Fujimoto & Polonsky 2009)</ns0:ref>. In skeletal muscle, AKT inactivates GSK3β by phosphorylation that results in a reduction in the phosphorylation of a few GSK3 substrates such as Glycogen Synthase (GS) <ns0:ref type='bibr' target='#b29'>(Hermida et al. 2017)</ns0:ref>. In pancreatic β-cells of the islets, GSK3β inhibited by AKT, does not phosphorylate PDX1. PDX1 is a critical regulator of pancreatic development and activates glucose transporter 2 (GLUT2), INS, and glucokinase genes <ns0:ref type='bibr' target='#b35'>(Humphrey et al. 2010</ns0:ref>). Thus, pharmacological inhibition of GSK3β could be substantial in type 2 diabetes treatment <ns0:ref type='bibr' target='#b56'>(Sacco et al. 2019)</ns0:ref>. Even GSK3β has not been found as a key target in our PPI network, a large group of compounds reported in T. polium interact with GSK3β, listed as, 4ꞌ,7-dimethoxy apigenin (TP1), 4ꞌ-O-methyl luteolin (TP2), 6-hydroxy luteolin (TP3), acacetin (TP4), apigenin (TP5), cirsilineol (TP7), cirsiliol (TP8), cirsimaritin (TP9), eupatorin (TP10), isoscutellarein (TP11), jaceosidin (TP12), luteolin (TP13), quercetin (TP14), 20-O-acetyl teucrasiatin (TP24), capitatin (TP25), clerodane-6,7-dione (TP26), (1R,6R,7R,8S,11R)-1,6-dihydroxy-4,11-dimethyl-germacran-4(5), 10( <ns0:ref type='formula'>14</ns0:ref>)-dien-8,12olide (TP31), prunasin (TP33), teucladiol (TP40), 4β,6β-dihydroxy-1α,5β(H)-guai-9-ene (TP41), oplopanone (TP42), ladanein (TP45), salvigenin (TP46), 5,3',4'-trihydroxy-3,7dimethoxyflavone (TP47), jaranol (TP48) (Table <ns0:ref type='table'>S5</ns0:ref>). Both phenolic compounds, terpenoids and a cyanogenic glycoside found to interact with GSK3β. Furthermore, glucose transporter 4 (GLUT4) induction through insulin-stimulated PI3K/AKT has an important role in whole-body glucose homeostasis by glucose intake at adipose tissue, cardiomyocytes and skeletal muscle cells <ns0:ref type='bibr' target='#b42'>(Klip et al. 2019)</ns0:ref>. Previous reports on flavonoids like apigenin (TP5), quercetin (TP14), kaempferol showed enhancement in GLUT4 translocation <ns0:ref type='bibr' target='#b3'>(Alkhalidy et al. 2015;</ns0:ref><ns0:ref type='bibr' target='#b32'>Hossain et al. 2014;</ns0:ref><ns0:ref type='bibr' target='#b37'>Jiang et al. 2019)</ns0:ref>. According to the findings of our study, apigenin (TP5), luteolin (TP13), and quercetin (TP14) showed interaction with the target GLUT4 (Table <ns0:ref type='table'>S5</ns0:ref>). By <ns0:ref type='bibr'>Kadan et al. (2018)</ns0:ref>, it was found that extract of T. polium increased translocation of GLUT4 in L6 muscle cells in the absence and presence of insulin when compared with the control. These findings indicated that insulin-like activity of T. polium is a result of increase in GLUT4 translocation <ns0:ref type='bibr'>(Kadan et al. 2018)</ns0:ref>. Quercetin (TP14) and gallic acid (TP16) increased glucose uptake through IRS1/PI3K/AKT signaling <ns0:ref type='bibr' target='#b24'>(Gandhi et al. 2014;</ns0:ref><ns0:ref type='bibr' target='#b37'>Jiang et al. 2019</ns0:ref>). Furthermore, it was shown that gallic acid and p-coumaric acid (TP17) ameliorate insulin shortage and insulin resistance <ns0:ref type='bibr' target='#b1'>(Abdel-Moneim et al. 2018)</ns0:ref>. A previous study on rats showed that vanillic acid (TP19) upregulates hepatic insulin signaling, insulin receptor, phosphatidylinositol-3 kinase, glucose transporter 2, and phosphorylated acetyl CoA carboxylase expression <ns0:ref type='bibr' target='#b10'>(Chang et al. 2015)</ns0:ref>. Kaempferol treatment also showed to improve β-cell mass in diabetic mice <ns0:ref type='bibr' target='#b3'>(Alkhalidy et al. 2015)</ns0:ref>. Similarly, cirsimaritin (TP9) suppresses apoptosis in βcells <ns0:ref type='bibr' target='#b44'>(Lee et al. 2017</ns0:ref>). IL6, another hub gene in the PPI network in this study, is a proinflammatory cytokine that has a complex role in T2DM. In this study, luteolin (TP13), quercetin (TP14) and teupolin VIII (TP29) showed interaction with the target IL6 (Table <ns0:ref type='table'>S5</ns0:ref>). Several studies showed that chronic inflammation plays a role in T2DM <ns0:ref type='bibr' target='#b46'>(Lehrskov & Christensen 2019)</ns0:ref>. Studies on the patients with T2DM showed an increase in IL6 levels in the plasma <ns0:ref type='bibr' target='#b2'>(Akbari & Hassan-Zadeh 2018)</ns0:ref>. The biological role of IL6 depends on the signaling pathway <ns0:ref type='bibr' target='#b2'>(Akbari & Hassan-Zadeh 2018)</ns0:ref>. Although there are different opinions on the effects of IL6 in T2DM, recent studies showed that the absence of IL6 resulted in hyperglycemia and higher fat levels in people with obesity <ns0:ref type='bibr' target='#b43'>(Kurauti et al. 2017;</ns0:ref><ns0:ref type='bibr' target='#b46'>Lehrskov & Christensen 2019)</ns0:ref>. However, IL6 also increases the expression of insulin degrading enzyme which is important in glucose metabolism. This enzyme also degrades amyloid β. For this view, IL6 is an important cytokine both in two closely related diseases Alzheimer's disease and T2DM <ns0:ref type='bibr' target='#b43'>(Kurauti et al. 2017)</ns0:ref>. STAT3 activation also negatively regulates another common target GSK3β of Alzheimer's disease and T2DM <ns0:ref type='bibr' target='#b53'>(Moh et al. 2008</ns0:ref>). STAT3 has a role in cell differentiation in various systems including immune and endocrine systems. However recent studies showed that STAT3 suppression together with Pdx1 expression increased the number of β-cells <ns0:ref type='bibr' target='#b52'>(Miura et al. 2018)</ns0:ref>. Caffeic acid (TP15), t-ferulic acid (TP18), 4α-[(β-D-glucopyranosyloxy)methyl]-5α-(2-hydroxyethyl)-3-methylcyclopent-2-en-1-one (TP22), 5α-[2-(β-D-glucopyranosyloxy)ethyl]-4α-hydroxymethyl-3-methylcyclopent-2-en-1-one (TP23), and teupolin VIII (TP29) showed interaction with the target STAT3 (Table <ns0:ref type='table'>S5</ns0:ref>). VEGFA is a growth factor that has an important role in angiogenesis, vasculogenesis and endothelial cell growth. Apigenin (TP5), luteolin (TP13), quercetin (TP14) and prunasin (TP33) showed interaction with the target VEGFA (Table <ns0:ref type='table'>S5</ns0:ref>). Hyperglycemic situations result in overexpression of VEGFA which is a critical factor in diabetic complications such as diabetic retinopathy <ns0:ref type='bibr' target='#b9'>(Caldwelll et al. 2003)</ns0:ref>.</ns0:p></ns0:div>
<ns0:div><ns0:head>Conclusions</ns0:head><ns0:p>In the last decade, network pharmacology driven omics methods play an important role to understand the role of the herbal prescriptions used as a folk medicine in various diseases. In this study, molecular mechanisms of T. polium in the treatment of T2DM were evaluated with the help of bioinformatics. Though there were studies on T. polium extract's antihyperglycemic effect via in vitro and in vivo assays, the underlying molecular mechanisms has not totally determined yet. We constructed a comprehensive dataset of compounds reported in T. polium. 126 compounds previously isolated or determined from the aerial parts of the plant were listed through a detailed literature search. In this view, this study serves the most detailed data on the content of the phytochemical profile of T. polium so far. In the present network pharmacological analysis, insulin resistance and PI3K-AKT signaling pathway were shown to take place in the center of the mechanism of action of T. polium. T. polium is an insulin-sensitizing plant. Even though insulin resistance has an important role in the pathophysiology of T2DM, insulin resistance does not result in T2DM in all cases. It turns to T2DM with a loss in β-cell mass in pancreatic islets. The results of the present network pharmacology studies taken together with the previously reported data, suggested that T. polium could be a promising herb for the treatment of T2DM through ameliorating insulin resistance and enhancing β-cell mass.</ns0:p></ns0:div>
<ns0:div><ns0:head>Figure 1</ns0:head><ns0:p>The scheme for intersection of T. polium and type 2 diabetes mellitus (T2DM) targets. PeerJ reviewing PDF | (2020:07:50669:1:0:NEW 9 Sep 2020)</ns0:p><ns0:p>Manuscript to be reviewed Manuscript to be reviewed </ns0:p></ns0:div><ns0:figure xml:id='fig_0'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:07:50669:1:0:NEW 9 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='17,42.52,178.87,525.00,288.75' type='bitmap' /></ns0:figure>
<ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='18,42.52,255.37,525.00,252.00' type='bitmap' /></ns0:figure>
<ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='19,42.52,178.87,525.00,294.75' type='bitmap' /></ns0:figure>
<ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='22,42.52,199.12,525.00,421.50' type='bitmap' /></ns0:figure>
<ns0:note place='foot'>PeerJ reviewing PDF | (2020:07:50669:1:0:NEW 9 Sep 2020)</ns0:note>
</ns0:body>
" | "
Dear Prof. Gwyn Gould,
Thank you for giving us an opportunity to revise the manuscript titled ‘Constitution of a comprehensive phytochemical profile and network pharmacology based investigation to decipher molecular mechanisms of Teucrium polium L. in the treatment of type 2 diabetes mellitus’ (Manuscript ID: 50669). We are grateful to you and reviewers for spending time and dedicating these valuable suggestions.
In revised manuscript, we have carefully considered reviewers’ comments and suggestions. You could find our point by point responses after of each comment.
Due to the comment of Reviewer 3, material methods section has revised by adding the keyword and the databases where data collected. We again checked the databases and noticed a few research articles within our search criteria. I and the co-author Ms. Neziha Yagmur Diker decided to include these studies in our study to avoid inconvenience. We apologize to miss these articles before our first manuscript submission. Even the dataset of our study changed, the findings remained similar. We did not need to rewrite discussion and conclusion section as the selected key targets, gene enrichment analysis results were similar and suits with our discussion and conclusion. All the figures were updated, and the number of the figures were decreased. Table 1 has moved to supplementary section according to your and the reviewers’ comments.
We hope that our edits and the responses we provide below satisfactorily address all the issues and concerns you and the reviewers have noted.
Sincerely,
Vahap Murat Kutluay, PhD
Department of Pharmacognosy, Faculty of Pharmacy
Hacettepe University
Reviewer 1 (Anonymous)
Basic reporting
The MS is well organized, but lack innovation.
Experimental design
The experimental design and method used in this MS is out of date. As many informations from databases were not reliably, so that some of the obtained results will mistake. In a word, integrate databases in quite not enough.
Validity of the findings
As I said previous, I can't judge the fingdings is true or not even without any experimental validation.
Comments for the Author
An increasing phenomenon is appeared that many researchers would like to choose network pharmacology to study herbal medicies. Howerer, if there is any improvement or innovation and just follow the routine which used previous. I don't think it is a good way to promote the development of herbal drugs. At least, experimental validation is necessary.
________________________________________
Reviewer 2 (Ankita Punetha)
Basic reporting
Diabetes is a metabolic disease, with type 2 diabetes being the most common, usually in adults. It occurs either when there is an insufficient insulin production by pancreas or when the produced insulin is not effectively utilized in the body. These disturbances result in hyperglycaemia, or raised blood sugar, which eventually leads to serious damage to many of the body's system over time. Diabetes prevalence has dramatically risen over the years and creates concerns. Amidst this, finding new remediation or new insights to Type 2 diabetes milletus (T2DM) will be a beneficial contribution.
Authors have used bioinformatics approach to characterize the therapeutic potential of the compounds from Teucrium polium L. in the treatment of T2DM. A comprehensive phytochemical profile mined from existing literature is presented along with network-based pharmacology investigation to decipher molecular mechanisms involved.
Introduction highlights the use of the plant extracts in the treatment in various ailments. It sheds light into its bioactive compounds and the phytotherapeutic effects.
The article is well structured, and professionally written. There are some minor issues that needs to be corrected and some sentences require rephrasing for clarity. The details of which are provided along with other comments.
Overall, the article is well within the scope and it will be of interest to the readers of the journal. The article is well organized into sections and sub-sections.
We would like to thank to the reviewer for this detailed review. We believe that the edits we made according to the reviewer’s comments have increased the quality of the manuscript.
Experimental design
Research question is well defined and meaningful. The authors experimental design seems appropriate for network pharmacological analysis.
Validity of the findings
Authors have mentioned, in this study they found GSK3β, GLUT4, and PDX1 to play important role in the antidiabetic effect of T. polium. But looking at the results from this study, it hard to find their role. They are not presented in any figures or table.
Although authors have discussed the compounds that interact with GSK3β and GLUT4 in the discussion section, but it is not clear whether authors found it in this study by their analysis or presented them from previous literature.
For PDX1 the interacting compounds are not mentioned. It will be appropriate to refer to the findings in the study, mentioning whether it is presented (table or a figure).
In conclusions, authors mention insulin resistance and PI3K-AKT signaling pathway as the center of the mechanism of action. Authors can elaborate how they finalized these pathways to be the center. Based on the network pharmacological analysis presented in this study, there are other pathways also that seem to be involved more than insulin resistance (Figure 6 and Table 3). For instance, HIF-1 signaling pathway, TNF signaling pathway, and FoxO signaling pathway.
*The comments in ‘Validity of the findings’ was replied in ‘Comments for Author’ part of Reviewer 2, owing to the fact that there were same comments.
Comments for the Author
Overall, the study is interesting and a useful addition to the area. However, there are certain concerns that needs be addressed to bridge the knowledge gap. A revision is recommended for improving the manuscript.
All the comments of the Reviewer 2 were discussed by authors. All changes and responses were given below.
Page 2, Line 56: Authors have mentioned 'Besides these key targets, with this study GSK3β, GLUT4, and PDX1 were also shown to play important role in the antidiabetic effect of T. polium'. Authors are recommended to refer in which step they identified them.
We apologise for the inconvenience. The addressed sentence could cause some misunderstanding. GSK3β, GLUT4, and PDX1 were included in this study through the findings in the literature.
Page 2, Line 56 (In revised manuscript, Track Changes in Microsoft Word is on): We revised the sentence as ‘Besides these key targets, with this study the role of GSK3β, GLUT4, and PDX1 were also discussed through literature and considered as important targets in the antidiabetic effect of T. polium’.
Later in discussion section (page 8 and 9), authors have mentioned the compounds that interact with GSK3β and GLUT4 but it is not clear whether authors found them in their study or mined them from previous literature. If found in this study, the corresponding figure/ table needs to be presented.
GSK3β and GLUT4 were mined from literature. So, they are not included in the figures. The role of GSK3β and GLUT4 in the antidiabetic activity of T. polium extract and related compounds reported in previous studies. As GSK3β interacts with 25 of 52 compounds in our study, the role of GSK3β in the extract’s antidiabetic activity was discussed together with the findings with our study.
Similarly, for PDX1 the information is missing, which compounds interact and how authors identified them.
In our study PDX1 is not a protein in our ‘compound-target’ network. PDX1 was included according to the literature search. A study by Tabatabaie and Yazdanparast, 2017, reported the role of PDX1 in the antidiabetic activity of the T. polium extract. In this view we integrated the findings of our study with PDX1. Discussed targets in our study such as GSK3β and FOXO1 are found to be in close relationship with PDX1 that could affect the biological role.
‘Tabatabaie PS, and Yazdanparast R. 2017. Teucrium polium extract reverses symptoms of streptozotocin-induced diabetes in rats via rebalancing the Pdx1 and FoxO1 expressions. Biomedicine & Pharmacotherapy 93:1033-1039. DOI:10.1016/j.biopha.2017.06.082’
Page 2, Line 60: Authors mention only 2 pathways to be the main pathways regulated by T. polium. It will appropriate to mention how they arrived at this conclusion. On the other hand, the pathway analysis (Figure 6 and Table 3) show HIF-1 signaling pathway, TNF signaling pathway, FoxO signaling pathway, and insulin signaling pathway as well.
PI3K/AKT and insulin resistance pathway were pointed as main pathways according to the following reasons;
PI3K/AKT pathway was found to be in relation with the highest number of target in gene enrichment analysis (Figure 5).
Insulin resistance pathway was selected as it intersects with other pathways found in the study like TNF signaling pathway, insulin signaling pathway, and PI3K/AKT pathway.
We agreed with the reviewer and HIF1 signaling pathway has added in the following sentences. We mentioned these pathways relation with type 2 diabetes mellitus as;
Page 2, Line 52-54 (In revised manuscript, Track Changes in Microsoft Word is on): ‘The KEGG pathway analysis showed enrichment for TNF signaling pathway, insulin resistance, HIF-1 signaling pathway, apoptosis, PI3K-AKT signaling pathway, FOXO signaling pathway, insulin signaling pathway, and type 2 diabetes mellitus which are related to T2DM.’
Page 2, Line 266-269 (In revised manuscript, Track Changes in Microsoft Word is on): ‘KEGG enrichment results supported these findings. Results of 252 common targets were listed as (related with T2DM); TNF signaling pathway, insulin resistance, apoptosis, HIF-1 signaling pathway, PI3K-Akt signaling pathway, FoxO signaling pathway, insulin signaling pathway and type 2 diabetes mellitus (Fig. 5).’
Figure 3: The legend mentions 32 key targets but only 31 are shown in the figure. 'ESR1' is missing. Check for the discrepancy. (ESR1 is present in PPI network- Figure 4)
Figure 7: The interaction of 29 compounds with 31 key targets is presented. Again, ESR1 is missing.
We thank to the reviewer for this comment. Due to the comments of Reviewer 3, as the content of Figure 3 and Figure 7 were similar, we updated these figures and formed a new figure (Figure 2 in revised manuscript) to replace Figure 3 and 7.
Other minor comments:
Page 2, Line 61-62: It will appropriate to rephrase the following sentence for clarity, 'This study reveals the relationship of the compounds and targets in T. polium'. The current sentence gives an impression that targets are in T. polium whereas targets are being investigated in humans.
Agreed.
Page 2, Line 62-63 (In revised manuscript, Track Changes in Microsoft Word is on): The addressed sentence has changed as “This study reveals the relationship of the compounds in T. polium with the targets of T2DM in human.”
Page 2, Line 68: It seems there is a typing error, '...about 1,6 million deaths...', it should be 1.6 million.
Page 2, Line 70 (In revised manuscript, Track Changes in Microsoft Word is on): The typing error has corrected.
Page 2, Line 74-75: Revise the sentence for clarity to the readers, 'Teucrium L. genus....'.
Page 2, Line 76-77 (In revised manuscript, Track Changes in Microsoft Word is on): The addressed sentence has changed as “The genus Teucrium L., a member of Lamiaceae family (Subfamily Ajugoideaea), has a cosmopolitan distribution and including about 250 species spread worldwide.”
Page 2, Line 76-78: Revise the sentence for clarity to the readers, 'In Turkish folk medicine, treatment....'.
Page 2 and 3, Line 78-82 (In revised manuscript, Track Changes in Microsoft Word is on): The sentence in line 75, “T. polium L. has been used as a traditional ….” have revised together with the sentence addressed by the reviewer and changed as “In Turkey, Teucrium polium L. is known as “Acıyavşan” and used as a traditional medicine for the treatment of diabetes (Arıtuluk & Ezer 2012). In Algeria and Iran, T. polium is used traditionally for the treatment of diabetes.”
A reference (Arıtuluk & Ezer 2012) have added to refer the traditional use of the plant in diabetes in Turkey. The reference list at the end of the manuscript have also updated.
‘Arıtuluk ZC, and Ezer N. 2012. Halk Arasında Diyabete Karşı Kullanılan Bitkiler Türkiye -II Hacettepe Üniversitesi Eczacılık Fakültesi Dergisi 32:179-208. Retrieved from https://dergipark.org.tr/en/pub/hujpharm/issue/49831/639072’
Page 3, Line 76-86: Italicise in vivo and in vitro.
Page 3, Line 89-90 (In revised manuscript, Track Changes in Microsoft Word is on): The typing error has corrected.
Page 3, Line 93-94: Revise the sentence for clarity to the readers, 'In blood glucose concentration, ...'.
Page 3, Line 97-99 (In revised manuscript, Track Changes in Microsoft Word is on): The sentence has revised as ‘The decoction of T. polium had a decrease of 20.5% in blood glucose concentration by oral administration while intraperitoneal and intravenous administration of the decoction had a decrease of 26.5% and 44%, respectively.’
Page 3, Line 115: Acronym TCM can be described.
Page 4, Line 122 (In revised manuscript, Track Changes in Microsoft Word is on): TCM has described as “….Traditional Chinese Medicine (TCM)…” and added to the text.
Page 3, Line 120: In the following sentence it will be appropriate to write 'Compounds reported from T. polium were selected based on their drug-likeness properties...'.
Agreed.
Page 4, Line 126-128 (In revised manuscript, Track Changes in Microsoft Word is on): The sentence changed as the reviewer’s suggestion.
Page 4, Line 136-138: Revise the sentences to bring more for clarity.
Page 4, Line 144 (In revised manuscript, Track Changes in Microsoft Word is on): To increase the clarity the following sentence was deleted, ‘Aqueous extract of T. polium is orally used traditionally in Turkey.’
Page 4, Line 143: It will be appropriate to write, ' ...compounds, were subjected to...'.
Agreed.
Page 4, Line 151 (In revised manuscript, Track Changes in Microsoft Word is on): The sentence changed as the reviewer’s suggestion.
Page 5, Line 164: Remove extra punctuation mark after the sentence.
Page 5, Line 172 (In revised manuscript, Track Changes in Microsoft Word is on): The typing error has corrected.
Page 5, Line 182: It will be appropriate to write, 'The results were listed based on their p-values.', instead of '...due to their p-values.'.
Agreed.
Page 5, Line 190-191 (In revised manuscript, Track Changes in Microsoft Word is on): The sentence has changed as the reviewer’s suggestion.
Page 6, Line 217: Authors can mention STRING 11.0 instead of String.
Agreed.
Page 6, Line 227 (In revised manuscript, Track Changes in Microsoft Word is on): The sentence changed as the reviewer’s suggestion.
Page 6, Line 237: For the consistency throughout the manuscript, it will better if authors use either p-values or p values.
Agreed. All have changed to “p value”.
Page 7, Line 278: It will be appropriate to write, '..cell function which are caused by...:, since authors are referring to two defects that arise due to long-term hyperglycemia or revise the sentence accordingly to bring more clarity.
Agreed.
Page 8, Line 289 (In revised manuscript, Track Changes in Microsoft Word is on): The sentence have changed as “..cell function which are caused by...”.
Page 8, Line 302: Italicise T. polium.
The typing error has corrected.
Page 8, Line 315-136: Authors mention, 'Even GSK3β has not been found as a key target in our PPI network, a large ...'. But authors have mentioned in the results that they found GSK3β as a target in this study. It will appropriate to correct the sentence and provide the reference to figure/ table/ supplementary data where it was found.
GSK3β, was included in this study through the findings in the literature. We integrated the findings of our study with the data in the literature. As mentioned above we also revised the sentence in abstract’s results subheading as ‘Besides these key targets, with this study the role of GSK3β, GLUT4, and PDX1 were also discussed through literature and considered as important targets in the antidiabetic effect of T. polium’.
As recommended supplementary data was added. The Supplementary table (Table S5 in revised manuscript) consisted of 52 compounds and their interactions with 252 genes (GSK3β exists in the table).
Page 8, Line 316-317: Italicise T. polium.
Page 9, Line 330 (In revised manuscript, Track Changes in Microsoft Word is on): The typing error has corrected.
Page 9, Line 326-328: Authors mention they find interaction of the listed compounds with GLUT4 in this study. It is recommended to provide the reference to figure/ table/ supplementary data where it was found.
Page 9, Line 345 (In revised manuscript, Track Changes in Microsoft Word is on): As recommended supplementary data was added. The Supplementary table (Table S5 in revised manuscript) consisted of 52 compounds and their interactions with 252 genes (GLUT4 exists in the table).
Page 9, Line 328 and 330: Italicise T. polium.
Page 9, Line 345 and 347 (In revised manuscript, Track Changes in Microsoft Word is on): The typing error has corrected.
Page 10, Line 370: Italicise in vivo and in vitro.
Page 10, Line 387 (In revised manuscript, Track Changes in Microsoft Word is on): The typing error has corrected.
Page 10, Line 369-371: Revise the following sentence for clarity, 'Though there were studies...'.
Page 10, Line 386-388 (In revised manuscript, Track Changes in Microsoft Word is on): The sentence has revised as ‘Though there were studies on T. polium extract’s antihyperglycemic effect via in vitro and in vivo assays, the underlying molecular mechanisms has not totally determined yet.
References
Page 12, Line 453, 459: DOI is missing.
Page 13, Line 481: DOI is missing.
Page 13, Line 511-512: DOI can be moved to line 511.
Page 13, Line 515: DOI is missing.
Page 15, Line 557, 568: DOI is missing.
Page 15, Line 586: Proper punctuation is required.
Page 15, Line 593: DOI is missing and '-' after punctuation needs to be removed.
Page 16, Line 602: DOI is missing and reference has repeated 2020. Delete the repetition.
Page 16, Line 610: Inconsistent referencing format. Please check.
Page 16, Line 612, 619, 621, 633: DOI is missing.
Typing errors have corrected. We have checked the references again for the DOI number, four of them added but could not able to find DOI number for the references given below. Journal’s website and crossref.org was used for the search.
• Esmaeili MA, and Sadeghi H. 2009. Pancreatic β-cell protective effect of rutin and apigenin isolated from Teucrium polium. Pharmacologyonline 2:341-353.
• Kadan S, Sasson Y, Abu-Reziq R, Saad B, Benvalid S, Linn T, Cohen G, and Zaid H. 2018. Teucrium polium extracts stimulate GLUT4 translocation to the plasma membrane in L6 muscle cells. Advancement in Medicinal Plant Research 6:1-8.
• Shahraki MR, Arab MR, Mirimokaddam E, and Palan MJ. 2007. The effect of Teucrium polium (Calpoureh) on liver function, serum lipids and glucose in diabetic male Rats. Iranian Biomedical Journal 11:65-68
• Venditti A, Frezza C, Zadeh SMM, Foddai S, Serafini M, and Bianco A. 2017b. Secondary metabolites from Teucrium polium L. collected in Southern Iran. Arabian Journal of Medicinal & Aromatic Plants 3:108-123.
• Yazdanparast R, Esmaeili MA, and Ashrafi Helan J. 2005. Teucrium polium extract effects pancreatic function of streptozotocin diabetic rats: a histopathological examination. Iranian Biomedical Journal 9:81-85.
Figures
Figure 1 Legend: Since the color coding of the nodes remains the same in all figures. Authors can mention in Figure 1 that the color coding scheme is consistent across figures instead of repeating the same text in Figure 3 and Figure 7.
Figure 1 was removed due to the comments of Reviewer 3. As the content of Figure 3 and Figure 7 were similar, we updated these figures and formed a new figure (Figure 2 in revised manuscript) to replace Figure 3 and 7. In revised manuscript there is only one figure left from these 3 figures so the repetition has removed.
Figure 7 Legend: The following text is redundant as in Figure 5, 'Bars represent p-values...'. It can be clearly mentioned in Figure 5 that these representations are consistent in the plots instead of duplicating text.
In Figure 5 and 6 legend (Figure 4 and 5 in revised manuscript), as reviewer suggested the sentence 'Bars represent p-values...' have removed. (We think that reviewer has written Figure 7 legend by mistake as that figure does not includes such a term.)
Tables
Table 1 Legend: Authors can simply mention T. polium. In the table heading, it will be better if authors write References instead of Ref.
Table 1 has moved to Supplementary files and numbered as Table S4. The legend of the table is ‘The list of 126 compounds that were reported from Teucrium polium L.’. Ref. has changed as References.
Table 2 Legend: It will be appropriate to write, '...29 compounds that were used in ...'.
Corrected.
Supplementary data
Supplemental_data_S2 and Suplemental_data_S3: The heading can be corrected to 'Gene Symbol' instead of 'Gene Symbole'.
Corrected.
________________________________________
Reviewer 3 (Anonymous)
Basic reporting
The authors proposed an analysis to decipher molecular mechanisms of Teucrium polium L. in the treatment of type 2 diabetes mellitus. Although some vital analyses have been performed, there are some major points that need to be addressed:
English needs to be polished. The manuscript should be formatted better and some spelling and grammar should be checked carefully. For examples, some grammatical errors and typos are as follows:
- ... were put in to DAVID database ...
Page 7, Line 246 (In revised manuscript, Track Changes in Microsoft Word is on): Changed as ‘….were subjected to DAVID database….’
- Our findings suggest its use as an effective herbal drug in the treatment of T2DM and provides new insights ...
Page 2, Line 63-65 (In revised manuscript, Track Changes in Microsoft Word is on): Changed as ‘Our findings suggested the use of T. polium as an effective herbal drug in the treatment of T2DM and provides new insights for further research on the antidiabetic effect of T. polium.’
- The 'Compound-Target' network consist of ...
Page 6, Line 212-213 (In revised manuscript, Track Changes in Microsoft Word is on): Changed as ‘The 'Compound-Target' network consists of ...’
- The top 20 results due to the p-values also suggested that compounds reported from T. polium might also leads new insights for the treatment of cancer.
Page 7, Line 269-271 (In revised manuscript, Track Changes in Microsoft Word is on): Changed as ‘The top 20 results according to the p-values also suggested that compounds reported from T. polium might leads new insights for the treatment of cancer.’
- ...
Experimental design
The authors should show detail on how they performed literature search i.e., which keywords or sources that they used?
Page 4, Line 136-137 (In revised manuscript, Track Changes in Microsoft Word is on): As recommended following sentence has added to Material&Methods part, ‘Literature search was performed using ‘Scopus’ and ‘Web of Science – Clarivate’ databases with the keyword ‘Teucrium polium’ upto June 2020.’
The authors transformed the compounds to SMILE format using PubChem (https://pubchem.ncbi.nlm.nih.gov/) or CS Chemdraw Ultra, so in which cases/compounds they use PubChem or Chemdraw Ultra?
In order to achieve SMILE format of the compounds, firstly we have checked PubChem. In case of the compound is missing in the Pubchem database, we have generated SMILE codes using CS Chemdraw Ultra.
Why did the authors set the confidence score as high (> 0.7).
We have set the confidence score as high to eliminate the interactions between the proteins that are false positives. The paragraph added below also gives a brief information about the confidence score.
“The minimum required interaction score puts a threshold on the confidence score, such that only interaction above this score are included in the predicted network. Lower score mean more interaction, but also more false positives. The confidence score is the approximate probability that a predicted link exists between two enzymes in the same metabolic map in the KEGG database. Confidence limits are as follows:
• low confidence - 0.15 (or better),
• medium confidence - 0.4,
• high confidence - 0.7,
• highest confidence - 0.9.”
This text in italic was taken from http://version10.string-db.org/help/getting_started/
( Accession date 20 August 2020)
GO database or analysis has been used in previously biomedical works such as PMID: 31277574 and PMID: 31921391. Therefore, the authors should refer more works in this description to attract broader readership.
Two articles that recommended by Reviewer 3 have perused by us. For PMID: 31277574, this study focuses on identifying the electron transport proteins with high accuracy by using bioinformatics studies such as novel deep neural network architecture. Additionally, for PMID: 31921391, this study is based on predicting vesicular transportation protein with new strategy that includes gated recurrent units and position-specific scoring matrix profiles. Despite these are very valuable studies in their scope and research field, we believe that these publications will not properly fit with our manuscript. We thank to the reviewer for suggesting these publications, we will consider these in our future works.
Methodology has not been described clearly and it makes hard to reproduce the results. The authors should describe it clearly.
We believe that with the revisions made through the reviewers’ comments increased the clarity of the methodology. The main additions are listed as;
The limitations in literature search were added.
All the disease related targets of the compounds were listed and added as Supplementary material. This table includes 252 genes and the proteins that interact with these targets.
Validity of the findings
Some figures and network analyses are redundant and necessary. For example, Fig. 1, 3, or 4 contained few information and no-sense.
Figure 1 was removed due to the comments of Reviewer 3. As the content of Figure 3 and Figure 7 were similar, we updated these figures and formed a new figure (Figure 2 in revised manuscript) to replace Figure 3 and 7. We prefer to keep Figure 4 (Figure 3 in revised manuscript), as it has a role in visualizing the findings of the study.
Some tables are long and it is better if the authors could put them to supplementary materials.
Agreed. Table 1 has moved to Supplementary files and numbered as Table S4.
The authors should validate the results on an unseen data.
We will consider this comment for our future studies. We plan to expand our studies to perform experimental studies with the guidance of these findings with new projects.
Comments for the Author
No comment.
" | Here is a paper. Please give your review comments after reading it. |
9,743 | "<ns0:abstract xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:p>Background. The coronavirus disease 2019 (COVID-19) pandemic, caused by severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2), is an ongoing global health crisis, directly and indirectly impacting all spheres of human life. Some pharmacological measures have been proposed to prevent COVID-19 or reduce its severity, such as vaccinations. Previous reports indicate that influenza vaccination appears to be negatively correlated with COVID-19-associated mortality, perhaps as a result of heterologous immunity or changes in innate immunity. The understanding of such trends in correlations could prevent deaths from COVID-19 in the future. The aim of this study was therefore to analyze the association between COVID-19 related deaths and influenza vaccination rate (IVR) in elderly people worldwide, and a negative association was expected. Methods. To determine the association between COVID-19 deaths and influenza vaccination, available data sets from countries with more than 0.5 million inhabitants were analyzed (in total 39 countries).To accurately estimate the influence of IVR on COVID-19 deaths and mitigate effects of confounding variables, a sophisticated ranking of the importance of different variables was performed, including as predictor variables IVR and some potentially important geographical and socioeconomic variables as well as variables related to nonpharmaceutical intervention. Results. The results showed a positive association between COVID-19 deaths and IVR of people ≥ 65 years-old. Further exploration is needed to explain these findings, and additional work on this line of research may lead to prevention of deaths associated with COVID-19.</ns0:p></ns0:div>
</ns0:abstract>
<ns0:body xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:head>Introduction</ns0:head><ns0:p>The coronavirus disease 2019 (COVID-19) pandemic, caused by severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2), is an ongoing global health crisis <ns0:ref type='bibr' target='#b58'>(Yuen et al. 2020)</ns0:ref>, directly and indirectly impacting all spheres of human life <ns0:ref type='bibr' target='#b42'>(Ozili & Arun 2020)</ns0:ref>. More than 15,300,000 confirmed cases including more than 630,000 deaths have been documented worldwide, affecting 213 countries and territories around the world (https://covid19.who.int/).</ns0:p><ns0:p>Determining the factors influencing the severity of COVID-19 is important <ns0:ref type='bibr' target='#b2'>(Armengaud et al. 2020)</ns0:ref>. Although COVID-19 disease does not only affect elderly people, the severity of symptoms increases with age (https://www.cdc.gov/coronavirus/2019-ncov/need-extraprecautions/older-adults.html). In the United States, people 65 years and older account for 45% of hospitalizations, 53% of admissions to intensive care units and 80% of deaths <ns0:ref type='bibr' target='#b32'>(Le Couteur, Anderson & Newman 2020)</ns0:ref>.</ns0:p><ns0:p>Several other risk factors have been found for severe COVID-19, such as comorbidities, dyspnea, chest pain, cough, expectoration, decreased lymphocytes, and increased inflammation indicators <ns0:ref type='bibr'>(Li et al. 2020)</ns0:ref>. Hypertension, diabetes, coronary heart disease, cerebrovascular illness, chronic obstructive pulmonary disease and kidney dysfunction worsen clinical outcomes <ns0:ref type='bibr' target='#b29'>(Ji et al. 2020)</ns0:ref>. Low socioeconomic status is an additional risk factor, given the common association between poverty and health issues, low access to healthy food, overcrowding, and lack of labor privileges to practice social distancing <ns0:ref type='bibr' target='#b57'>(Yancy 2020)</ns0:ref>.</ns0:p><ns0:p>In response to the increasing numbers of cases of infection and deaths, numerous nonpharmaceutical interventions have been implemented, including social distancing, border closures, school closures, measures to isolate symptomatic individuals and their contacts, and large-scale lockdowns of populations <ns0:ref type='bibr' target='#b7'>(Courtemanche et al. 2020;</ns0:ref><ns0:ref type='bibr'>Flaxman et al. 2020)</ns0:ref>. Some pharmacological measures have also (often controversially) been proposed in order to prevent COVID-19 disease or reduce its severity, such as the use of remdesivir (e.g., <ns0:ref type='bibr' target='#b5'>Beigel et al. 2020)</ns0:ref>, dexamethasone (RECOVERY Collaborative <ns0:ref type='bibr'>Group 2020)</ns0:ref>, adjunctive therapies (https://files.covid19treatmentguidelines.nih.gov/guidelines/section/section_85.pdf), Vitamin D (e.g., <ns0:ref type='bibr' target='#b25'>Grant et al. 2020)</ns0:ref>, drugs (e.g., <ns0:ref type='bibr' target='#b28'>Jakhar and Kaur 2020;</ns0:ref><ns0:ref type='bibr' target='#b56'>Wright et al. 2020)</ns0:ref> and COVID-19 vaccines (e.g., <ns0:ref type='bibr' target='#b24'>Graham 2020</ns0:ref>, https://www.who.int/publications/m/item/draft-landscape-of-covid-19-candidate-vaccines.</ns0:p><ns0:p>The term 'heterologous immunity' is applied when an infection by one pathogen can induce and/or alter the immune response against another unrelated pathogen. Heterologous immunity can improve or decrease protective immunity against a given pathogen, and/or cause severe immunopathology or tolerance to self-antigens <ns0:ref type='bibr'>(Goodridge et al. 2016</ns0:ref><ns0:ref type='bibr' target='#b1'>, Agrawal 2019)</ns0:ref>. Arokiaraj <ns0:ref type='bibr' target='#b44'>(2020)</ns0:ref> reported a negative correlation between influenza vaccination rates and COVID-19 related mortality and morbidity. Marín-Hernández et al. ( <ns0:ref type='formula'>2020</ns0:ref>) also showed epidemiological evidence of an association between higher influenza vaccine uptake by elderly people and lower percentage of COVID-19 deaths in Italy (R 2 = 0.35, p = 0.0051). In a study analyzing 92,664 clinically and molecularly confirmed COVID-19 cases in Brazil, <ns0:ref type='bibr' target='#b11'>Fink et al. (2020)</ns0:ref> reported that patients who received a recent flu vaccine experienced on average 8% lower odds of needing intensive care treatment (95% CIs [0.86, 0.99]), 18% lower odds of requiring invasive respiratory support (0.74, 0.88), and 17% lower odds of death (0.75, 0.89).</ns0:p><ns0:p>Moreover, <ns0:ref type='bibr'>Pawlowski et al. (2020)</ns0:ref> analyzed the immunization records of 137,037 individuals who tested positive in a SARS-CoV-2 PCR. They found that polio, Hemophilus influenzae type-B (HIB), measles-mumps-rubella (MMR), varicella, pneumococcal conjugate (PCV13), geriatric flu, and hepatitis A / hepatitis B (HepA-HepB) vaccines, which had been administered in the past one, two, and five years, were associated with decreased SARS-CoV-2 infection rates, even after taking into account the geographic SARS-CoV-2 incidence and testing rates, demographics, comorbidities, and a number of other vaccinations. In addition, age, race/ethnicity, and blood group stratified analyses showed a significantly lower rate of SARS-CoV-2 infection among black individuals who had been given the PCV13 vaccine, with a relative risk of 0.45 at the 5year time horizon analyzed (n = 653, 95% CI = (0.32, 0.64), p = 0.000069).</ns0:p><ns0:p>By contrast, in a study with 6,120 subjects, Wolff <ns0:ref type='bibr' target='#b44'>(2020)</ns0:ref> reported that influenza vaccination was significantly associated with a higher risk of some other respiratory diseases, due to virus interference. In a specific examination of non-influenza viruses, the odds of coronavirus infection (but not the COVID-19 virus) in vaccinated individuals were significantly higher, when compared to unvaccinated individuals (odds ratio = 1.36).</ns0:p><ns0:p>Given that heterologous immunity could improve protective immunity against COVID-19 and, thus, prevent COVID-19 deaths in the future, the aim in this study was to analyze the possible association between COVID-19 deaths and the influenza vaccination rate in elderly people worldwide, with a negative association expected.</ns0:p></ns0:div>
<ns0:div><ns0:head>Materials & Methods</ns0:head><ns0:p>To look for an association between COVID-19 deaths and influenza vaccination, available data sets from 39 countries, each with ≥ 0.5 million inhabitants, were analyzed. In smaller states (i.e., < 0.5 million inhabitants), the rate of erroneous identification of COVID-19 deaths may be particularly high due to the lack of expertise, measuring devices and experience. Moreover, in such microstates small absolute changes in COVID-19 deaths may result in extreme values of relative indices, such as COVID-19 deaths per million inhabitants (DP) and COVID-19 Case Fatality Ratio (CFR). The variables DP and CFR, based on confirmed infected people per million inhabitants (CIP) in 2020, COVID-19 tests per million inhabitants, and influenza vaccination rate (IVR) (%) in people ≥ 65 years old in 2019 or latest available data were analyzed (Table <ns0:ref type='table'>1</ns0:ref>). The DP, CIP and CFR data were recorded from the public web site https://www.worldometers.info/coronavirus/. CFR was calculated as the rate of DP per CIP. IVR data were also taken from https://data.oecd.org/healthcare/influenza-vaccination-rates.htm, https://oecdcode.org/disclaimers/israel.html and https://www.statista.com/chart/16575/global-fluimmunization-rates-vary/ (retrieved on July 25, 2020). Vietnam's 2017 IVR was recorded from <ns0:ref type='bibr' target='#b40'>Nguyen et al. (2020)</ns0:ref>, and Singapore's 2016/2017 IVR from https://www.todayonline.com/commentary/why-singapores-adult-vaccination-rate-so-low.</ns0:p><ns0:p>As the relationship between DP and the number of people tested for COVID-19 was not statistically significant, the DP data set was not modified (corrected). To analyze the data, the non-parametric Spearman's correlation coefficient (r s ) and its R S 2 and respective p-value (2tailed) were first applied to determine any association between DP and CFR with IVR, using R (R Core Team 2017). Regression curves were then created by Generalized additive model (GAM) using the 'ggplot2' package and function (method = 'gam') <ns0:ref type='bibr' target='#b53'>(Wickham et al. 2013)</ns0:ref>, also in R.</ns0:p><ns0:p>As the analysis included countries with different socioeconomic status, demographic structure, urban/rural settings, time of arrival of the pandemic and national control strategies, there may be PeerJ reviewing <ns0:ref type='table'>PDF | (2020:08:51704:1:2:NEW 26 Aug 2020)</ns0:ref> Manuscript to be reviewed complex interactions between IVR and other correlated predictor variables. With the aim of accurately estimating the influence of IVR on DP and CFR and mitigating the effects of confounding variables, variable importance ranking was performed, including as predictor variables IVR and some potentially important geographical, socioeconomic and nonpharmaceutical-intervention variables <ns0:ref type='bibr' target='#b10'>(Escobar et al. 2020)</ns0:ref>. The centroid longitudes (º) and latitudes (º) of each country were used as geographical variables calculated by the 'rgeos' and 'rworldmap' packages, along with the 'getMap' and 'gCentroid' functions, implemented in R (version 3.3.4; R Development Core Team, 2017). For each country considered, the study recorded socioeconomic variables as the degree of urbanization (DUR) in 2020 (https://www.cia.gov/library/publications/the-world-factbook/fields/349.html), the population density (PD) in 2018 (https://data.worldbank.org/indicator/EN.POP.DNST), the Human Development Index (HDI) in 2018 (http://hdr.undp.org/en/composite/HDI) and the percentage of elderly people (PEP) in 2019 (https://data.worldbank.org/indicator/SP.POP.65UP.TO.ZS?name_desc=false), which were all retrieved on July 13, 2020 (Table <ns0:ref type='table'>2</ns0:ref>). Finally, two aspects were recorded as Covid-19 prevention measures, i.e. the degree of requirement to use masks (mask) in public (with three degrees: none, parts of country, full country) (https://masks4all.co/what-countries-require-masks-in-public/) and the lockdown degree (lockdown) (with three levels: no lockdown, partial lockdown, nationwide lockdown); all of these sources and the noted in Table <ns0:ref type='table'>3</ns0:ref> were consulted on Aug 13, 2020.</ns0:p><ns0:p>Variable importance ranking was carried out using the 'party' package and the non-parametric random forest function 'cforest', along with Out of bag (OOB) score (with the default option 'controls = cforest_unbiased' and the conditional permutation importance 'varimp(obj, conditional = TRUE)'). Following the permutation principle of the 'mean decrease in accuracy' importance, this machine learning algorithm guarantees unbiased variable importance for predictor variables of different types <ns0:ref type='bibr' target='#b50'>(Strobl et al. 2008)</ns0:ref>.</ns0:p><ns0:p>With the aim of mitigating the effects of confounding factors, IVR, DP and CFR evaluations were also conducted for countries with similar social conditions (> 50% of DUR, HDI of > 0.80, > 15% of PEP, and PD between 25 and 350 inhabitants per km 2 ) <ns0:ref type='bibr' target='#b10'>(Escobar et al. 2020)</ns0:ref> Manuscript to be reviewed countries with similar longitudes (10 -20º, parts of Europe and 100 -140º, East and Southeast Asia along with Australia and New Zealand).</ns0:p><ns0:p>As IVR and the other eight predictor variables were not strongly correlated (|r s | ≤ 0.57; r s (IVR x DUR) = +0.52; r s (IVR x Long) = -0.46; r s (IVR x HDI) = 0.36), these variables were therefore included in non-parametric Random Forest (RF) models of DP and CFR, including a 5-fold cross validation approach, repeated 30 times using the package 'caret' together with the function 'train' <ns0:ref type='bibr'>(Venables and Ripley, 1999;</ns0:ref><ns0:ref type='bibr' target='#b54'>Williams et al., 2018,</ns0:ref> http://topepo.github.io/caret/index.html) in R software. The goodness-of-fit of the regression model was evaluated using the (pseudo) coefficient of determination (R 2 ) and the root mean square error (RMSE). At worldwide level (39 countries studied), the positive associations between DP and IVR were also statistically significant (r s (IVR x DP) = +0.49 with p = 0.0016, R s 2 (IVR x DP) = 0.24) (Fig. <ns0:ref type='figure' target='#fig_7'>3, Table 5</ns0:ref>). However, the relationships between IVR and CFR were not statistically significant.</ns0:p></ns0:div>
<ns0:div><ns0:head>Results</ns0:head></ns0:div>
<ns0:div><ns0:head>For</ns0:head><ns0:p>In the IVR interval from 7 to 50 %, the association was not significant, although a trend for DP and CFR to be positively associated with IVR was observed. DP and CFR varied strongly when IVR was 50% or higher (Figs. <ns0:ref type='figure' target='#fig_5'>1-3</ns0:ref>).</ns0:p><ns0:p>Worldwide, the unbiased ranking showed the degree of importance of each variable analyzed.</ns0:p><ns0:p>The variables Long (with 55.9 and 52.3 %) and IVR (with 36.3 and 24.5 %) were by far the most important of the nine variables used to predict DP and CFR, respectively. The degree of urbanization (DUR) in 2020 was the third most important variable, with an importance of 5.7% for predicting DP. The percentage of elderly people (PEP) in 2019 was the third most important variable (11.5%) in the CFR model (Figs. <ns0:ref type='figure' target='#fig_7'>4 and 5</ns0:ref>). The nine predictor variables considered in this study explained 63% of the variation in DP (RMSE =161.9) and 43 % of the variation in CFR (RMSE = 0.039).</ns0:p></ns0:div>
<ns0:div><ns0:head>Discussion</ns0:head><ns0:p>Contrary to expectations, the present worldwide analysis and European sub-analysis do not support the previously reported negative association between COVID-19 deaths (DP) and influenza vaccination rate (IVR) in elderly people, observed in studies in Brazil and Italy <ns0:ref type='bibr' target='#b11'>(Fink et al. 2020;</ns0:ref><ns0:ref type='bibr' target='#b36'>Marín-Hernández et al. 2020)</ns0:ref>. Previous studies attributed the beneficial effect of influenza vaccination in reducing severity of COVID-19 disease to better prevention of potential influenza-SARS-CoV-2 coinfections <ns0:ref type='bibr' target='#b3'>(Arokiaraj 2020)</ns0:ref> and, more likely, to changes in innate immunity <ns0:ref type='bibr' target='#b39'>(Netea et al. 2020)</ns0:ref>. The innate immune response induced by recent vaccination could result in more rapid and efficient SARS-CoV-2 clearance, preventing progressive dissemination into lower areas of lung tissues <ns0:ref type='bibr' target='#b11'>(Fink et al. 2020)</ns0:ref>.</ns0:p><ns0:p>The negative association between the proportion of DP and IVR found in Italy was explained as probably caused by i) a higher influenza vaccine rate occurring in higher economic groups with overall better health, ii) chance, iii) a relationship with seasonal respiratory virus infections, or iv) an unrelated mechanistic association <ns0:ref type='bibr' target='#b36'>(Marín-Hernández et al. 2020)</ns0:ref>. However, the induction of cross-neutralizing antibodies and T-cells that directly target other RNA viruses like SARS-CoV-2 and cross-protection seem unlikely, given the extraordinary diversity of influenza viruses <ns0:ref type='bibr' target='#b11'>(Fink et al. 2020)</ns0:ref>. Therefore, the above-mentioned arguments cannot explain the positive, direct or indirect relationship between IVR and both DP and CFR found in this study, which was confirmed by an unbiased ranking variable importance (Figs. <ns0:ref type='figure' target='#fig_7'>4 and 5</ns0:ref>) using RF models. The influenza vaccine may increase influenza immunity at the expense of reduced immunity to SARS-CoV-2 by some unknown biological mechanism, as suggested by <ns0:ref type='bibr' target='#b8'>Cowling et al. (2012)</ns0:ref> for non-influenza respiratory virus. Alternatively, weaker temporary, non-specific immunity after influenza virus infection could cause this positive association due to stimulation of the innate immune response Manuscript to be reviewed during and for a short time after infection <ns0:ref type='bibr' target='#b37'>(McGill et al. 2009;</ns0:ref><ns0:ref type='bibr'>Khaitov et al. 2009)</ns0:ref>. People who had received the influenza vaccination would have been protected against influenza but not against other virus infections, due to reduced non-specific immunity in the following weeks <ns0:ref type='bibr' target='#b8'>(Cowling et al. 2012)</ns0:ref>, probably caused by virus interference <ns0:ref type='bibr' target='#b27'>(Isaacs & Lindenmann 1957;</ns0:ref><ns0:ref type='bibr' target='#b48'>Seppälä et al. 2011;</ns0:ref><ns0:ref type='bibr' target='#b55'>Wolff 2020)</ns0:ref>. Although existing human vaccine adjuvants have a high level of safety, specific adjuvants in influenza vaccines should also be tested for adverse reactions <ns0:ref type='bibr' target='#b41'>(Petrovsky 2015)</ns0:ref>.</ns0:p><ns0:p>The strong variation in DP and CFR from an IVR of about 50% or larger may be the result of interactions among the different measures applied in the analyzed countries (Figs. <ns0:ref type='figure' target='#fig_5'>1-3</ns0:ref>), e.g. initiation of interventions, emergency plans and health systems against COVID-19. For example, Australia and South Korea had a very low DP and CFR compared with Belgium and United Kingdom (Table <ns0:ref type='table'>1</ns0:ref>).</ns0:p><ns0:p>The high correlation between the longitude of the country centroid and DP and CFR emphasize a significant increase in CP and CFR from eastern to western regions in the world <ns0:ref type='bibr'>(Table 5,</ns0:ref><ns0:ref type='bibr'>Figs. 4</ns0:ref> and 5), as confirmed by <ns0:ref type='bibr' target='#b34'>Leung et al. (2020)</ns0:ref> and <ns0:ref type='bibr' target='#b49'>Skórka et al. (2020)</ns0:ref>. Longitude could act as a proxy for variables such as lifestyle, social behavior, genetics, geographically isolated and remote populations, which may also be associated with CP and CFR. In the severe 1918-1919 influenza pandemic, remote or isolated populations were also most affected, at least partly because of a lack of prior immunity in locations that had not been recently affected by any form of influenza <ns0:ref type='bibr' target='#b51'>(Mathews et al. 2009)</ns0:ref>. Therefore, crossing geographical and ecological barriers is also is a key factor in spreading diseases <ns0:ref type='bibr' target='#b26'>(Hallatschek & Fisher 2014;</ns0:ref><ns0:ref type='bibr' target='#b38'>Murray et al. 2015)</ns0:ref>.</ns0:p><ns0:p>Both DP and CFR were weakly and positively correlated (p < 0.05) with the absolute value of geographical latitude (abs(Lat)), degree of urbanization (DUR), percentage of elderly people (PEP) and population density (Tables <ns0:ref type='table'>4 and 5</ns0:ref>). In a global analysis, <ns0:ref type='bibr' target='#b10'>Escobar et al. (2020)</ns0:ref> also found positive associations between COVID-19 mortality and the percentage of population aged ≥ 65 years and urbanization, but still more strongly with the Human Development Index. <ns0:ref type='bibr' target='#b34'>Leung et al. (2020)</ns0:ref> also reported positive associations between latitude, temperature by week and by month prior to the first reported COVID-19 case. Lower temperature at northern latitudes was a strong independent predictor of national COVID-19 mortality.</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:08:51704:1:2:NEW 26 Aug 2020)</ns0:p></ns0:div>
<ns0:div><ns0:head>Manuscript to be reviewed</ns0:head><ns0:p>Although countywide lockdowns and use of face masks by the general public should reduce COVID-19 transmission <ns0:ref type='bibr' target='#b6'>(Conyon, He & Thomsen 2020;</ns0:ref><ns0:ref type='bibr'>Eikenberry et al. 2020)</ns0:ref>, the variables lockdown degree and the degree of requirement for mask use in public were not associated with DP and CFR in the present study (Tables <ns0:ref type='table'>4 and 5, Figs 4 and 5</ns0:ref>). <ns0:ref type='bibr' target='#b33'>Leffler et al. (2020)</ns0:ref> reported in a global study that internal lockdown requirements were not associated with mortality, but that in countries that recommended use of face masks early on at the national level, the COVID-19 death rate was lower than expected.</ns0:p><ns0:p>As is often the case, the success of these two interventions could be a matter of the details.</ns0:p><ns0:p>Although countywide lockdowns were proclaimed in many countries, the restrictive measures and their implementations differed in degree, strictness and implementation date in relation to the advance of the disease (see references in Table <ns0:ref type='table'>3</ns0:ref>). Also, although the use of face masks may be similar in many countries, the mask quality and correct use may differ from country to country. <ns0:ref type='bibr' target='#b13'>Fischer et al. (2020)</ns0:ref> found that speaking through some types of face mask seemed to disperse the largest droplets into a multitude of smaller droplets which remain in the air longer than large droplets. This explains the apparent rise in droplet count relative to no mask. Thus, the use of ineffective masks could be counterproductive.</ns0:p></ns0:div>
<ns0:div><ns0:head>Conclusions</ns0:head><ns0:p>Given the positive relationship between influenza vaccination rate and the number of deaths per million found in this study, further exploration would be valuable to explain these findings and to make conclusions. Additional work on this line of research may yield results to improve prevention of COVID-19 deaths. = the requirement degree of using masks in public (with three degrees: none, parts of country, full country), lockdown = lockdown degree (with three levels: no lockdown, partial lockdown, nationwide lockdown) of each country, at worldwide level (39 countries studied). Manuscript to be reviewed </ns0:p></ns0:div><ns0:figure xml:id='fig_0'><ns0:head /><ns0:label /><ns0:figDesc>and for PeerJ reviewing PDF | (2020:08:51704:1:2:NEW 26 Aug 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_1'><ns0:head /><ns0:label /><ns0:figDesc>the 26 European countries considered , the results indicated that COVID-19 deaths per million inhabitants (DP) and the COVID-19 Case Fatality Ratio (CFR) were positively and statistically significantly associated with influenza vaccination rate (IVR) in people ≥ 65 yearsold in 2019 or latest data available (r s (IVR x DP) = +0.62 with p = 0.0008, R s 2 (IVR x DP) = 0.38; r s (IVR x CFR) = +0.50 with p = 0.01, R S 2 (IVR x CFR) = 0.25) (Figs. 1 and 2, Table 4). In evaluations including only countries with similar social conditions, r s (IVR x DP) was equal to +0.65 (p = 0.002, N = 20) and r s (IVR x CFR) +0.48 (p = 0.03, N = 20). In analyses including only countries with similar longitude of the country centroid (Long), r s (IVR x DP) was equal to +0.83 (p = 0.003, N = 10) (Long from 10 to 20º) and r s (IVR x DP) +0.76 (p = 0.046, N = 7) (Long from 100 to 140º).</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_2'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:08:51704:1:2:NEW 26Aug 2020) </ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_3'><ns0:head>Figure 1</ns0:head><ns0:label>1</ns0:label><ns0:figDesc>Figure 1</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_4'><ns0:head>Figure 2</ns0:head><ns0:label>2</ns0:label><ns0:figDesc>Figure 2</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_5'><ns0:head>Figure 3</ns0:head><ns0:label>3</ns0:label><ns0:figDesc>Figure 3</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_6'><ns0:head>Figure 4 Unbiased</ns0:head><ns0:label>4</ns0:label><ns0:figDesc>Figure 4</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_7'><ns0:head>Figure 5</ns0:head><ns0:label>5</ns0:label><ns0:figDesc>Figure 5</ns0:figDesc></ns0:figure>
<ns0:note place='foot'>PeerJ reviewing PDF | (2020:08:51704:1:2:NEW 26Aug 2020) </ns0:note>
</ns0:body>
" | "Instituto de Silvicultura e Industria de la Madera
Universidad Juárez del Estado de Durango
Antonio Palazón-Bru
Academic Editor
PeerJ
Carretera Mazatlán Km 5.5
C.P. 34120 Durango
México
Tel.: +52-618-827-12-15
Fax: ++52-618-825-18-86
E-mail:wehenkel@ujed.mx
08-22-2020
Dear Dr. Palazón-Bru,
I have again revised my paper “Positive association between COVID-19 deaths and
influenza vaccination rates in elderly people worldwide” ((#2020:08:51704:0:1:REVIEW)) in
response to the reviewers' comments. I thank you for your time taken to review
the manuscript and agree that the comments certainly have improved the
manuscript’s content. Along with this letter, you will find a document describing
the way each of your comments were addressed and a revised copy of the
manuscript.
Yours sincerely,
Dr. Christian Wehenkel
Positive association of COVID-19 deaths with influenza vaccination in
Europe and worldwide
(#2020:08:51704:0:1:REVIEW)
Response to the Reviewer 1 (Anonymous)
Basic reporting
This manuscript meets all the four requirements:
(1) Literature references, sufficient field background/context provided.
(2) The article should include sufficient introduction and background to demonstrate how the
work fits into the broader field of knowledge. Relevant prior literature should be appropriately
referenced.
(3) Professional article structure, figures, tables. Raw data shared.
(4) Self-contained with relevant results to hypotheses.
Many thanks!
Experimental design
This manuscript meets all the four requirements, except the third:
(1) Original primary research within Aims and Scope of the journal.
(2) Research question well defined, relevant & meaningful. It is stated how research fills an
identified knowledge gap.
(3) Rigorous investigation performed to a high technical & ethical standard.
(4) Methods should be described with sufficient information to be reproducible by another
investigator.
This study was not performed to a high technical standard. The death rate of COVID-19 is highly
associated with age distribution, prevention measures, and geographical distribution. Only the
influence of these factors has been excluded, the conclusion that the influenza vaccine coverage
is positively associated with the death rate of COVID-19 is reliable. Therefore, the authors need
to collect the data regarding age distribution, prevention measures, and geographical
distribution of the countries, and then use proper statistical methods to investigate the
relationship between the influenza vaccine coverage and the death rate of COVID-19.
See below the modifications.
Validity of the findings
Please see my comments above.
Comments for the Author
This manuscript is well written, and the topic is interesting, and the authors have found some
novel data. However, the death rate of COVID-19 is highly associated with age distribution,
prevention measures, and geographical distribution. Only the influence of these factors has been
excluded, the conclusion that the influenza vaccine coverage is positively associated with the
death rate of COVID-19 is somehow reliable. Therefore, the authors need to collect the data
regarding age distribution, prevention measures, and geographical distribution of the
investigated countries (not difficult), and then use proper statistical methods to investigate the
relationship between the influenza vaccine coverage and the death rate of COVID-19. Otherwise,
the conclusion could be wrong and misleading.
I added (lines 122-163): “As the analysis included countries with different socioeconomic status,
demographic structure, urban/rural settings, time of arrival of the pandemic and national control
strategies, there may be complex interactions between IVR and other correlated predictor
variables. With the aim of accurately estimating the influence of IVR on DP and CFR and mitigating
the effects of confounding variables, variable importance ranking was performed, including as
predictor variables IVR and some potentially important geographical, socioeconomic and nonpharmaceutical-intervention variables (Escobar et al. 2020). The centroid longitudes (º) and
latitudes (º) of each country were used as geographical variables calculated by the “rgeos” and
“rworldmap” packages, along with the “getMap” and “gCentroid” functions, implemented in R
(version 3.3.4; R Development Core Team, 2017). For each country considered, the study recorded
socioeconomic variables as the degree of urbanization (DUR) in 2020
(https://www.cia.gov/library/publications/the-world-factbook/fields/349.html), the population
density (PD) in 2018 (https://data.worldbank.org/indicator/EN.POP.DNST), the Human
Development Index (HDI) in 2018 (http://hdr.undp.org/en/composite/HDI) and the percentage of
elderly people (PEP) in 2019
(https://data.worldbank.org/indicator/SP.POP.65UP.TO.ZS?name_desc=false), which were all
retrieved on July 13, 2020 (Table 2). Finally, two aspects were recorded as Covid-19 prevention
measures, i.e. the degree of requirement to use masks (mask) in public (with three degrees: none,
parts of country, full country) (https://masks4all.co/what-countries-require-masks-in-public/) and
the lockdown degree (lockdown) (with three levels: no lockdown, partial lockdown, nationwide
lockdown); all of these sources and the noted in Table 3 were consulted on Aug 13, 2020.
Variable importance ranking was carried out using the “party” package and the non-parametric
random forest function “cforest”, along with Out of bag (OOB) score (with the default option
“controls = cforest_unbiased” and the conditional permutation importance “varimp(obj,
conditional = TRUE)”). Following the permutation principle of the “mean decrease in accuracy”
importance, this machine learning algorithm guarantees unbiased variable importance for
predictor variables of different types (Strobl et al. 2008).
With the aim of mitigating the effects of confounding factors, IVR, DP and CFR evaluations were
also conducted for countries with similar social conditions (>50% of DUR, HDI of > 0.80, > 15% of
PEP, and PD between 25 and 350 inhabitants per km2) (Escobar et al. 2020) and for countries with
similar longitudes (10 - 20º, parts of Europe and 100 - 140º, East and Southeast Asia along with
Australia and New Zealand).
As IVR and the other eight predictor variables were not strongly correlated (ǀrsǀ ≤ 0.57; rs (IVR x
DUR) = +0.52; rs (IVR x Long) = -0.46; rs (IVR x HDI) = 0.36), these variables were therefore included
in non-parametric Random Forest (RF) models of DP and CFR, including a 5-fold cross validation
approach, repeated 30 times using the package “caret” together with the function “train”
(Venables and Ripley, 1999; Williams et al., 2018, http://topepo.github.io/caret/index.html) in R
software. The goodness-of-fit of the regression model was evaluated using the (pseudo) coefficient
of determination (R2) and the root mean square error (RMSE).”
(Lines 170-175): “In evaluations including only countries with similar social conditions, rs (IVR x DP)
was equal to +0.65 (p = 0.002, N = 20) and rs (IVR x CFR) +0.48 (p = 0.03, N = 20). In analyses
including only countries with similar longitude of the country centroid (Long), rs (IVR x DP) was
equal to +0.83 (p = 0.003, N = 10) (Long from 10 to 20º) and rs (IVR x DP) +0.76 (p = 0.046, N = 7)
(Long from 100 to 140º).”
(Lines 180-186): “Worldwide, the unbiased ranking showed the degree of importance of each
variable analyzed. The variables Long (with 55.9 and 52.3 %) and IVR (with 36.3 and 24.5 %) were
by far the most important of the nine variables used to predict DP and CFR, respectively. The
degree of urbanization (DUR) in 2020 was the third most important variable, with an importance of
5.7% for predicting DP. The percentage of elderly people (PEP) in 2019 was the third most
important variable (11.5%) in the CFR model (Figs. 4 and 5). The nine predictor variables
considered in this study explained 63% of the variation in DP (RMSE =161.9) and 43 % of the
variation in CFR (RMSE = 0.039).
(lines 228 -260): “The high correlation between the longitude of the country centroid and DP and
CFR emphasize a significant increase in CP and CFR from eastern to western regions in the world
(Table 5, Figs. 4 and 5), as confirmed by Leung et al. (2020) and Skórka et al. (2020). Longitude
could act as a proxy for variables such as lifestyle, social behavior, genetics, geographically isolated
and remote populations, which may also be associated with CP and CFR. In the severe 1918–1919
influenza pandemic, remote or isolated populations were also most affected, at least partly
because of a lack of prior immunity in locations that had not been recently affected by any form of
influenza (Mathews et al. 2009). Therefore, crossing geographical and ecological barriers is also is a
key factor in spreading diseases (Hallatschek & Fisher 2014; Murray et al. 2015).
Both DP and CFR were weakly and positively correlated (p < 0.05) with the absolute value of
geographical latitude (abs(Lat)), degree of urbanization (DUR), percentage of elderly people (PEP)
and population density (Tables 4 and 5). In a global analysis, Escobar et al. (2020) also found
positive associations between COVID-19 mortality and the percentage of population aged ≥ 65
years and urbanization, but still more strongly with the Human Development Index. Leung et al.
(2020) also reported positive associations between latitude, temperature by week and by month
prior to the first reported COVID-19 case. Lower temperature at northern latitudes was a strong
independent predictor of national COVID-19 mortality.
Although countywide lockdowns and use of face masks by the general public should reduce COVID19 transmission (Conyon, He &Thomsen 2020; Eikenberry et al. 2020), the variables lockdown
degree and the degree of requirement for mask use in public were not associated with DP and CFR
in the present study (Tables 4 and 5, Figs 4 and 5). Leffler et al. (2020) reported in a global study
that internal lockdown requirements were not associated with mortality, but that in countries that
recommended use of face masks early on at the national level, the COVID-19 death rate was lower
than expected.
As is often the case, the success of these two interventions could be a matter of the details.
Although countywide lockdowns were proclaimed in many countries, the restrictive measures and
their implementations differed in degree, strictness and implementation date in relation to the
advance of the disease (see references in Table 3). Also, although the use of face masks may be
similar in many countries, the mask quality and correct use may differ from country to country.
Fischer et al. (2020) found that speaking through some types of face mask seemed to disperse the
largest droplets into a multitude of smaller droplets which remain in the air longer than large
droplets. This explains the apparent rise in droplet count relative to no mask. Thus, the use of
ineffective masks could be counterproductive.”
Response to the Reviewer 2 (Dr. Daniela Marín-Hernández)
Basic reporting
The introduction needs to be improved using recent published papers, summarized and directed
to address the justification.
Done.
The English language should be improved to ensure that an international audience can clearly
understand your text.
The English was improved.
Line 15: Please consider using caused instead of provoked.
Done.
Line 17: Please consider eliminating potentially, using pharmacological instead of medicinal and
using proposed instead of discussed.
Done.
Line 18: Please consider eliminating disease, since it is included in the acronym COVID-19.
Done.
Is Influenza vaccination a protective factor or a correlation between Influenza vaccination and
COVID-19 deaths have been reported?
I changed (Line 70): “…a negative correlation”.
Line 19: The justification needs to be adapted to the type of analysis the author did, the author is
expanding the knowledge of this reported correlation but not explaining the cause.
I added (Line 19): “…, perhaps by heterologous immunity or changes in innate immunity.”
Line 20: Please consider using preventing COVID-19 deaths instead of save human lives.
Done.
Line 21: Please consider using in the elderly instead of older persons and eliminating in Europe
since you are doing a worldwide analysis.
Done.
Line 23: Please consider describing the statistical analysis you performed.
I added (lines 27-30): “To accurately estimate the influence of IVR on COVID-19 deaths and
mitigate effects of confounding variables, a sophisticated ranking of the importance of different
variables was performed, including as predictor variables IVR and some potentially important
geographical and socioeconomic variables as well as variables related to non-pharmaceutical
intervention.”
Line 27: Please consider using ≥ 65 yo instead of older persons, using studies are needed to
prevent COVID-19 deaths.
I added: “…people ≥ 65 years-old. Further exploration is needed to explain these findings, and
additional work on this line of research may lead to prevention of deaths associated with COVID19.”
Line 28: Please consider using caused instead of provoked
I modified the sentence, now: “Further exploration is needed to explain these findings, and
additional work on this line of research may lead to prevention of deaths associated with COVID19.”
Line 36: Please consider using could influence COVID-19 severity instead of influence COVID-19
disease.
Done.
Line 37: Please consider rephrasing, COVID-19 is not a disease of only older people, but severity
increases with age, using as reference: https://www.cdc.gov/coronavirus/2019-ncov/needextra-precautions/older-adults.html. Please consider eliminating for example.
Done.
Line 36-47: Please consider rewriting the risk factors section into one paragraph and summarizing
it in a coherent and straight-forward way.
I modified the text, now: “Several other risk factors have been found for severe COVID-19, such as
comorbidities, dyspnea, chest pain, cough, expectoration, decreased lymphocytes, and increased
inflammation indicators (Li et al. 2020). Hypertension, diabetes, coronary heart disease,
cerebrovascular illness, chronic obstructive pulmonary disease and kidney dysfunction worsen
clinical outcomes (Ji et al. 2020). Low socioeconomic status is an additional risk factor, given the
common association between poverty and health issues, low access to healthy food, overcrowding,
and lack of labor privileges to practice social distancing (Yancy 2020).”
Line 48: Please consider using the term non-pharmaceutical interventions and rephrasing this
sentence using as reference: Flaxman S, Mishra S, Gandy A, et al. Estimating the effects of nonpharmaceutical interventions on COVID-19 in Europe [published online ahead of print, 2020 Jun
8]. Nature. 2020;10.1038/s41586-020-2405-7. doi:10.1038/s41586-020-2405-7
I added (lines 55 -58): “In response to the increasing numbers of cases of infection and deaths,
numerous non-pharmaceutical interventions have been implemented, including social distancing,
border closures, school closures, measures to isolate symptomatic individuals and their contacts,
and large-scale lockdowns of populations (Courtemanche et al. 2020; Flaxman et al. 2020).”
Line 49: Please consider eliminating potentially and using pharmaceutical interventions instead
of medical measures
Done.
Line 48-52: Please consider rewriting this pharmaceutical interventions paragraph including as
references:
-Beigel JH, Tomashek KM, Dodd LE, et al. Remdesivir for the Treatment of Covid-19 - Preliminary
Report [published online ahead of print, 2020 May 22]. N Engl J Med.
2020;10.1056/NEJMoa2007764. doi:10.1056/NEJMoa2007764
-RECOVERY Collaborative Group, Horby P, Lim WS, et al. Dexamethasone in Hospitalized Patients
with Covid-19 - Preliminary Report [published online ahead of print, 2020 Jul 17]. N Engl J Med.
2020;10.1056/NEJMoa2021436. doi:10.1056/NEJMoa2021436
-https://files.covid19treatmentguidelines.nih.gov/guidelines/section/section_85.pdf
I added (lines 60-65): “…the use of remdesivir (e.g., Beigel et al. 2020), dexamethasone (RECOVERY
Collaborative Group 2020), adjunctive therapies
(https://files.covid19treatmentguidelines.nih.gov/guidelines/section/section_85.pdf), Vitamin D
(e.g., Grant et al. 2020), drugs (e.g., Jakhar and Kaur 2020; Wright et al. 2020) and COVID-19
vaccines (e.g., Graham 2020, https://www.who.int/publications/m/item/draft-landscape-of-covid19-candidate-vaccines.”
Line 52: Please consider specifying that you are referring to COVID-19 vaccines and using as
reference: https://www.who.int/publications/m/item/draft-landscape-of-covid-19-candidatevaccines
Done.
Line 53: Please consider introducing the concept of heterologous effects or non-specific effects of
vaccines, using as reference:
- Agrawal, Babita. “Heterologous Immunity: Role in Natural and Vaccine-Induced Resistance to
Infections.” Frontiers in immunology vol. 10 2631. 8 Nov. 2019, doi:10.3389/fimmu.2019.02631
- Goodridge, Helen S et al. “Harnessing the beneficial heterologous effects of vaccination.”
Nature reviews. Immunology vol. 16,6 (2016): 392-400. doi:10.1038/nri.2016.43
I added (lines 66 -69): “The term 'heterologous immunity' is applied when an infection by one
pathogen can induce and/or alter the immune response against another unrelated pathogen.
Heterologous immunity can improve or decrease protective immunity against a given pathogen,
and/or cause severe immunopathology or tolerance to self-antigens (Goodridge et al. 2016,
Agrawal 2019).”
Please consider including the results and using as a reference: Pawlowski C., Puranik A.,
Venkatakrishnan AJ., et al. Exploratory analysis of immunization records highlights decreased
SARS-CoV-2 rates in individuals with recent non-COVID-19 vaccinations. medRxiv
2020.07.27.20161976; doi: https://doi.org/10.1101/2020.07.27.20161976
I added (lines 78 -87): “Moreover, Pawlowski et al. (2020) analyzed the immunization records of
137,037 individuals who tested positive in a SARS-CoV-2 PCR. They found that polio, Hemophilus
influenzae type-B (HIB), measles-mumps-rubella (MMR), varicella, pneumococcal conjugate
(PCV13), geriatric flu, and hepatitis A / hepatitis B (HepA-HepB) vaccines, which had been
administered in the past one, two, and five years, were associated with decreased SARS-CoV-2
infection rates, even after taking into account the geographic SARS-CoV-2 incidence and testing
rates, demographics, comorbidities, and a number of other vaccinations. In addition, age,
race/ethnicity, and blood group stratified analyses showed a significantly lower rate of SARS-CoV-2
infection among black individuals who had been given the PCV13 vaccine, with a relative risk of
0.45 at the 5-year time horizon analyzed (n = 653, 95% CI = (0.32, 0.64), p = 0.000069).”
Line 54: Please consider avoiding the use of protective factor, it is speculative since Arokiaraj
performed a statistical test for correlation which does not necessarily suggest causation.
I modified (lines 69 -71): “Arokiaraj (2020) reported a negative correlation between influenza
vaccination rates and COVID-19 related mortality and morbidity.”
Line 61-65: Please consider summarizing the important point of the results from Wolff’s paper.
I improved the text (lines 88 – 92): “By contrast, in a study with 6,120 subjects, Wolff (2020)
reported that influenza vaccination was significantly associated with a higher risk of some other
respiratory diseases, due to virus interference. In a specific examination of non-influenza viruses,
the odds of coronavirus infection (but not the COVID-19 virus) in vaccinated individuals were
significantly higher, when compared to unvaccinated individuals (odds ratio = 1.36).”
Line 66-67: Please consider rephrasing or eliminating this sentence.
Done. Sentence eliminated.
Line 68: Please consider rewriting the justification of the analysis, the author is expanding the
knowledge of this reported correlation but not explaining the cause.
I modified the sentence (line 93 - 94): “Given that heterologous immunity could improve protective
immunity against COVID-19 and, thus, prevent COVID-19 deaths in the future, …”
Line 70: Please consider using in the elderly or in people ≥ 65 yo instead of older persons and
eliminating in Europe since you are doing a worldwide analysis.
Done.
Experimental design
Since you are analyzing different countries, with different socioeconomic status, demographic
structure, urban/rural settings, time of arrival of pandemic and national control strategies, you
should consider mitigating effects of potentially confounding factors. The results might be
different if you take into account all of these aspects, or only analyze similar countries that meet
the same criteria.
Line 72-94: Please consider using active voice instead of passive voice. Since you are analyzing
different countries, with different socioeconomic status, demographic structure, urban/rural
settings, time of arrival of pandemic and national control strategies, consider mitigating effects
of potentially confounding factors. Consider reviewing Escobar LE, Molina-Cruz A, Barillas-Mury
C. BCG vaccine protection from severe coronavirus disease 2019 (COVID-19). Proc Natl Acad Sci U
S A. 2020;117(30):17720-17726. doi:10.1073/pnas.2008410117
I added (lines 122-163): “As the analysis included countries with different socioeconomic status,
demographic structure, urban/rural settings, time of arrival of the pandemic and national control
strategies, there may be complex interactions between IVR and other correlated predictor
variables. With the aim of accurately estimating the influence of IVR on DP and CFR and mitigating
the effects of confounding variables, variable importance ranking was performed, including as
predictor variables IVR and some potentially important geographical, socioeconomic and nonpharmaceutical-intervention variables (Escobar et al. 2020). The centroid longitudes (º) and
latitudes (º) of each country were used as geographical variables calculated by the “rgeos” and
“rworldmap” packages, along with the “getMap” and “gCentroid” functions, implemented in R
(version 3.3.4; R Development Core Team, 2017). For each country considered, the study recorded
socioeconomic variables as the degree of urbanization (DUR) in 2020
(https://www.cia.gov/library/publications/the-world-factbook/fields/349.html), the population
density (PD) in 2018 (https://data.worldbank.org/indicator/EN.POP.DNST), the Human
Development Index (HDI) in 2018 (http://hdr.undp.org/en/composite/HDI) and the percentage of
elderly people (PEP) in 2019
(https://data.worldbank.org/indicator/SP.POP.65UP.TO.ZS?name_desc=false), which were all
retrieved on July 13, 2020 (Table 2). Finally, two aspects were recorded as Covid-19 prevention
measures, i.e. the degree of requirement to use masks (mask) in public (with three degrees: none,
parts of country, full country) (https://masks4all.co/what-countries-require-masks-in-public/) and
the lockdown degree (lockdown) (with three levels: no lockdown, partial lockdown, nationwide
lockdown); all of these sources and the noted in Table 3 were consulted on Aug 13, 2020.
Variable importance ranking was carried out using the “party” package and the non-parametric
random forest function “cforest”, along with Out of bag (OOB) score (with the default option
“controls = cforest_unbiased” and the conditional permutation importance “varimp(obj,
conditional = TRUE)”). Following the permutation principle of the “mean decrease in accuracy”
importance, this machine learning algorithm guarantees unbiased variable importance for
predictor variables of different types (Strobl et al. 2008).
With the aim of mitigating the effects of confounding factors, IVR, DP and CFR evaluations were
also conducted for countries with similar social conditions (>50% of DUR, HDI of > 0.80, > 15% of
PEP, and PD between 25 and 350 inhabitants per km2) (Escobar et al. 2020) and for countries with
similar longitudes (10 - 20º, parts of Europe and 100 - 140º, East and Southeast Asia along with
Australia and New Zealand).
As IVR and the other eight predictor variables were not strongly correlated (ǀrsǀ ≤ 0.57; rs (IVR x
DUR) = +0.52; rs (IVR x Long) = -0.46; rs (IVR x HDI) = 0.36), these variables were therefore included
in non-parametric Random Forest (RF) models of DP and CFR, including a 5-fold cross validation
approach, repeated 30 times using the package “caret” together with the function “train”
(Venables and Ripley, 1999; Williams et al., 2018, http://topepo.github.io/caret/index.html) in R
software. The goodness-of-fit of the regression model was evaluated using the (pseudo) coefficient
of determination (R2) and the root mean square error (RMSE).”
(Lines 170-175): “In evaluations including only countries with similar social conditions, rs (IVR x DP)
was equal to +0.65 (p = 0.002, N = 20) and rs (IVR x CFR) +0.48 (p = 0.03, N = 20). In analyses
including only countries with similar longitude of the country centroid (Long), rs (IVR x DP) was
equal to +0.83 (p = 0.003, N = 10) (Long from 10 to 20º) and rs (IVR x DP) +0.76 (p = 0.046, N = 7)
(Long from 100 to 140º).”
(Lines 180-186): “Worldwide, the unbiased ranking showed the degree of importance of each
variable analyzed. The variables Long (with 55.9 and 52.3 %) and IVR (with 36.3 and 24.5 %) were
by far the most important of the nine variables used to predict DP and CFR, respectively. The
degree of urbanization (DUR) in 2020 was the third most important variable, with an importance of
5.7% for predicting DP. The percentage of elderly people (PEP) in 2019 was the third most
important variable (11.5%) in the CFR model (Figs. 4 and 5). The nine predictor variables
considered in this study explained 63% of the variation in DP (RMSE =161.9) and 43 % of the
variation in CFR (RMSE = 0.039).
(lines 228 -260): “The high correlation between the longitude of the country centroid and DP and
CFR emphasize a significant increase in CP and CFR from eastern to western regions in the world
(Table 5, Figs. 4 and 5), as confirmed by Leung et al. (2020) and Skórka et al. (2020). Longitude
could act as a proxy for variables such as lifestyle, social behavior, genetics, geographically isolated
and remote populations, which may also be associated with CP and CFR. In the severe 1918–1919
influenza pandemic, remote or isolated populations were also most affected, at least partly
because of a lack of prior immunity in locations that had not been recently affected by any form of
influenza (Mathews et al. 2009). Therefore, crossing geographical and ecological barriers is also is a
key factor in spreading diseases (Hallatschek & Fisher 2014; Murray et al. 2015).
Both DP and CFR were weakly and positively correlated (p < 0.05) with the absolute value of
geographical latitude (abs(Lat)), degree of urbanization (DUR), percentage of elderly people (PEP)
and population density (Tables 4 and 5). In a global analysis, Escobar et al. (2020) also found
positive associations between COVID-19 mortality and the percentage of population aged ≥ 65
years and urbanization, but still more strongly with the Human Development Index. Leung et al.
(2020) also reported positive associations between latitude, temperature by week and by month
prior to the first reported COVID-19 case. Lower temperature at northern latitudes was a strong
independent predictor of national COVID-19 mortality.
Although countywide lockdowns and use of face masks by the general public should reduce COVID19 transmission (Conyon, He &Thomsen 2020; Eikenberry et al. 2020), the variables lockdown
degree and the degree of requirement for mask use in public were not associated with DP and CFR
in the present study (Tables 4 and 5, Figs 4 and 5). Leffler et al. (2020) reported in a global study
that internal lockdown requirements were not associated with mortality, but that in countries that
recommended use of face masks early on at the national level, the COVID-19 death rate was lower
than expected.
As is often the case, the success of these two interventions could be a matter of the details.
Although countywide lockdowns were proclaimed in many countries, the restrictive measures and
their implementations differed in degree, strictness and implementation date in relation to the
advance of the disease (see references in Table 3). Also, although the use of face masks may be
similar in many countries, the mask quality and correct use may differ from country to country.
Fischer et al. (2020) found that speaking through some types of face mask seemed to disperse the
largest droplets into a multitude of smaller droplets which remain in the air longer than large
droplets. This explains the apparent rise in droplet count relative to no mask. Thus, the use of
ineffective masks could be counterproductive.”
Line 71: Please consider not repeating the results of the papers you have already described.
I eliminate “based on the results of Fink et al. (2020) and Marín-Hernández et al. (2020)”.
Line 75: Please consider explaining why you only chose to analyze countries with at least 0.5
million inhabitants.
Done.
Line 76: Please consider using COVID-19 deaths per million instead of dead persons by COVID-19
per Million inhabitants as the depend variable.
Done.
Line 77: Please consider using Case Fatality Ratio instead of proportion of COVID-19 deaths of
confirmed infected persons.
Done.
Line 78: Please consider using COVID-19 tests per million inhabitants instead of number of tested
persons for COVID-19 per million inhabitants.
Done
Line 79: Please consider using in people ≥ 65 years old instead of persons aged 65 and older.
Done
Line 82: Please consider eliminating also on July 25, 2020, since this data is not continuously
updating and using IVRs were taken from instead of the data of IVR were taken.
Thank you. I modified the writing.
Line 85-86: Please consider using 2017 Vietnam’s and Singapore’s IVR were obtained from xx and
xx, respectively.
A modified writing was used to avoid starting the sentence with a number.
Line 88-89: Please consider explaining the statistical test you performed and if by not modified
are you referring to a cofounder.
I added (lines 122-163): “As the analysis included countries with different socioeconomic status,
demographic structure, urban/rural settings, time of arrival of the pandemic and national control
strategies, there may be complex interactions between IVR and other correlated predictor
variables. With the aim of accurately estimating the influence of IVR on DP and CFR and mitigating
the effects of confounding variables, variable importance ranking was performed, including as
predictor variables IVR and some potentially important geographical, socioeconomic and nonpharmaceutical-intervention variables (Escobar et al. 2020). The centroid longitudes (º) and
latitudes (º) of each country were used as geographical variables calculated by the “rgeos” and
“rworldmap” packages, along with the “getMap” and “gCentroid” functions, implemented in R
(version 3.3.4; R Development Core Team, 2017). For each country considered, the study recorded
socioeconomic variables as the degree of urbanization (DUR) in 2020
(https://www.cia.gov/library/publications/the-world-factbook/fields/349.html), the population
density (PD) in 2018 (https://data.worldbank.org/indicator/EN.POP.DNST), the Human
Development Index (HDI) in 2018 (http://hdr.undp.org/en/composite/HDI) and the percentage of
elderly people (PEP) in 2019
(https://data.worldbank.org/indicator/SP.POP.65UP.TO.ZS?name_desc=false), which were all
retrieved on July 13, 2020 (Table 2). Finally, two aspects were recorded as Covid-19 prevention
measures, i.e. the degree of requirement to use masks (mask) in public (with three degrees: none,
parts of country, full country) (https://masks4all.co/what-countries-require-masks-in-public/) and
the lockdown degree (lockdown) (with three levels: no lockdown, partial lockdown, nationwide
lockdown); all of these sources and the noted in Table 3 were consulted on Aug 13, 2020.
Variable importance ranking was carried out using the “party” package and the non-parametric
random forest function “cforest”, along with Out of bag (OOB) score (with the default option
“controls = cforest_unbiased” and the conditional permutation importance “varimp(obj,
conditional = TRUE)”). Following the permutation principle of the “mean decrease in accuracy”
importance, this machine learning algorithm guarantees unbiased variable importance for
predictor variables of different types (Strobl et al. 2008).
With the aim of mitigating the effects of confounding factors, IVR, DP and CFR evaluations were
also conducted for countries with similar social conditions (>50% of DUR, HDI of > 0.80, > 15% of
PEP, and PD between 25 and 350 inhabitants per km2) (Escobar et al. 2020) and for countries with
similar longitudes (10 - 20º, parts of Europe and 100 - 140º, East and Southeast Asia along with
Australia and New Zealand).
As IVR and the other eight predictor variables were not strongly correlated (ǀrsǀ ≤ 0.57; rs (IVR x
DUR) = +0.52; rs (IVR x Long) = -0.46; rs (IVR x HDI) = 0.36), these variables were therefore included
in non-parametric Random Forest (RF) models of DP and CFR, including a 5-fold cross validation
approach, repeated 30 times using the package “caret” together with the function “train”
(Venables and Ripley, 1999; Williams et al., 2018, http://topepo.github.io/caret/index.html) in R
software. The goodness-of-fit of the regression model was evaluated using the (pseudo) coefficient
of determination (R2) and the root mean square error (RMSE).”
Line 104: please consider using was instead of were.
Done.
Validity of the findings
Line 113-116: please consider using contrary to what was expected instead of contrary to was
expected and rephrasing the paragraph, for example, our worldwide analysis and European subanalysis do not support the previously reported negative association between COVID-19 deaths
and influenza vaccination rate.
Done. Thank you.
134-140: Please consider rephrasing, the main idea is not clear.
I modified this text (lines 216-222): “People who had received the influenza vaccination would have
been protected against influenza but not against other virus infections, due to reduced non-specific
immunity in the following weeks (Cowling et al. 2012), probably caused by virus interference
(Isaacs & Lindenmann 1957; Seppälä et al. 2011; Wolff 2020). Although existing human vaccine
adjuvants have a high level of safety, specific adjuvants in influenza vaccines should also be tested
for adverse reactions (Petrovsky 2015).”
141-144: Please consider rephrasing, the current phrasing makes comprehension difficult. Please
consider using the great variation instead of strong variation and initiation of interventions
instead of reaction speed.
Done.
148-149: please consider using studies are needed to prevent COVID-19 deaths instead of
exploration is valuable to save human lives.
Done.
149: Please consider specifying which type of studies are suitable to explain your findings. Maybe
clinical trials?
Since there are many potential lines of research, I opted to leave open the research lines.
Images: Consider adding the p value so it is possible to quickly know if it is statistically significant
or not
Done.
Figure legend: Consider adding the results of the statistical analysis.
Done.
Comments for the Author
Your most important issue is the fact of not considering the differences between countries. Since
you are analyzing different countries, with different socioeconomic status, demographic
structure, urban/rural settings, time of arrival of pandemic and national control strategies, you
should consider mitigating effects of potentially confounding factors. The results might be
different if you take into account all of these aspects, or only analyze similar countries that meet
the same criteria.
Was included, see above
The next most important item is that the introduction needs to be improved using recent
published papers, summarized and directed to address the justification.
See above.
The least important is that the English language should be improved to ensure that an
international audience can clearly understand your text.
The English language was improved.
153-154: Please consider rewriting this paragraph, for example: I would like to express my
gratitude to xxx and xxx for their careful review and insightful comments
I used the second part of the suggestion.
Many thanks for good comments and recommendation!
" | Here is a paper. Please give your review comments after reading it. |
9,744 | "<ns0:abstract xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:p>Background. The coronavirus disease 2019 (COVID-19) pandemic, caused by severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2), is an ongoing global health crisis, directly and indirectly impacting all spheres of human life. Some pharmacological measures have been proposed to prevent COVID-19 or reduce its severity, such as vaccinations. Previous reports indicate that influenza vaccination appears to be negatively correlated with COVID-19-associated mortality, perhaps as a result of heterologous immunity or changes in innate immunity. The understanding of such trends in correlations could prevent deaths from COVID-19 in the future. The aim of this study was therefore to analyze the association between COVID-19 related deaths and influenza vaccination rate (IVR) in elderly people worldwide. Methods. To determine the association between COVID-19 deaths and influenza vaccination, available data sets from countries with more than 0.5 million inhabitants were analyzed (in total 39 countries).To accurately estimate the influence of IVR on COVID-19 deaths and mitigate effects of confounding variables, a sophisticated ranking of the importance of different variables was performed, including as predictor variables IVR and some potentially important geographical and socioeconomic variables as well as variables related to non-pharmaceutical intervention. The associations were measured by non-parametric Spearman rank correlation coefficients and random forest functions. Results. The results showed a positive association between COVID-19 deaths and IVR of people ≥ 65 years-old. Further exploration is needed to explain these findings, and additional work on this line of research may lead to prevention of deaths associated with COVID-19.</ns0:p></ns0:div>
</ns0:abstract>
<ns0:body xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:head>Introduction</ns0:head><ns0:p>The coronavirus disease 2019 (COVID-19) pandemic, caused by severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2), is an ongoing global health crisis <ns0:ref type='bibr' target='#b52'>(Yuen et al. 2020)</ns0:ref>, directly and indirectly impacting all spheres of human life <ns0:ref type='bibr' target='#b37'>(Ozili & Arun 2020)</ns0:ref>. More than 15,300,000 confirmed cases including more than 630,000 deaths have been documented worldwide, affecting 213 countries and territories around the world (https://covid19.who.int/).</ns0:p><ns0:p>Determining the factors influencing the severity of COVID-19 is important <ns0:ref type='bibr' target='#b1'>(Armengaud et al. 2020)</ns0:ref>. Although COVID-19 disease does not only affect elderly people, the severity of symptoms increases with age (https://www.cdc.gov/coronavirus/2019-ncov/need-extraprecautions/older-adults.html; <ns0:ref type='bibr' target='#b25'>Le Couteur, Anderson & Newman 2020)</ns0:ref>. Several other risk factors have been found for severe COVID-19, such as comorbidities, dyspnea, chest pain, cough, expectoration, decreased lymphocytes, and increased inflammation indicators <ns0:ref type='bibr'>(Li et al. 2020)</ns0:ref>. Low socioeconomic status is an additional risk factor <ns0:ref type='bibr' target='#b51'>(Yancy 2020)</ns0:ref>.</ns0:p><ns0:p>In response to the increasing numbers of COVID-19 cases and deaths, numerous nonpharmaceutical interventions have been implemented, including social distancing, border closures, school closures, measures to isolate symptomatic individuals and their contacts, and large-scale lockdowns of populations <ns0:ref type='bibr' target='#b7'>(Courtemanche et al. 2020;</ns0:ref><ns0:ref type='bibr'>Flaxman et al. 2020)</ns0:ref>. Some pharmacological measures have also (often controversially) been proposed in order to prevent COVID-19 disease or reduce its severity, such as the use of remdesivir (e.g., <ns0:ref type='bibr' target='#b4'>Beigel et al. 2020)</ns0:ref>, dexamethasone (RECOVERY Collaborative <ns0:ref type='bibr'>Group 2020)</ns0:ref>, adjunctive therapies (https://files.covid19treatmentguidelines.nih.gov/guidelines/section/section_85.pdf) and COVID-19 candidate vaccines (e.g., <ns0:ref type='bibr' target='#b19'>Graham 2020</ns0:ref>, https://www.who.int/publications/m/item/draftlandscape-of-covid-19-candidate-vaccines.</ns0:p><ns0:p>The term 'heterologous immunity' is applied when an infection by one pathogen can induce and/or alter the immune response against another unrelated pathogen. Heterologous immunity can improve or decrease protective immunity against a given pathogen, and/or cause severe immunopathology or tolerance to self-antigens. Heterologous immunity can also result in nonspecific effects (also called 'heterologous effects') of vaccines which affect unrelated infections and diseases, such as extending the protective outcomes of vaccinations <ns0:ref type='bibr'>(Goodridge et al. 2016</ns0:ref> Manuscript to be reviewed Agrawal 2019). Arokiaraj <ns0:ref type='bibr' target='#b39'>(2020)</ns0:ref> reported a negative correlation between influenza vaccination rates and COVID-19 related mortality and morbidity. Marín-Hernández et al. ( <ns0:ref type='formula'>2020</ns0:ref>) also showed epidemiological evidence of an association between higher influenza vaccine uptake by elderly people and lower percentage of COVID-19 deaths in Italy. In a study analyzing 92,664 clinically and molecularly confirmed COVID-19 cases in Brazil, <ns0:ref type='bibr' target='#b11'>Fink et al. (2020)</ns0:ref> reported that patients who received a recent flu vaccine experienced on average 17% lower odds of death. Moreover, <ns0:ref type='bibr'>Pawlowski et al. (2020)</ns0:ref> analyzed the immunization records of 137,037 individuals who tested positive in a SARS-CoV-2 PCR. They found that polio, Hemophilus influenzae type-B (HIB), measles-mumps-rubella (MMR), varicella, pneumococcal conjugate (PCV13), geriatric flu, and hepatitis A / hepatitis B (HepA-HepB) vaccines, which had been administered in the past one, two, and five years, were associated with decreased SARS-CoV-2 infection rates.</ns0:p><ns0:p>By contrast, in a study with 6,120 subjects, Wolff <ns0:ref type='bibr' target='#b39'>(2020)</ns0:ref> reported that influenza vaccination was significantly associated with a higher risk of some other respiratory diseases, due to virus interference. In a specific examination of non-influenza viruses, the odds of coronavirus infection (but not the COVID-19 virus) in vaccinated individuals were significantly higher, when compared to unvaccinated individuals (odds ratio = 1.36).</ns0:p><ns0:p>Given that heterologous immunity could improve protective immunity against COVID-19 and, thus, prevent COVID-19 deaths in the future, the aim in this study was to analyze the possible association between COVID-19 deaths and the influenza vaccination rate in elderly people worldwide, with a negative association expected.</ns0:p></ns0:div>
<ns0:div><ns0:head>Materials & Methods</ns0:head><ns0:p>To look for an association between COVID-19 deaths and influenza vaccination, I analyzed available data sets from 39 countries, each with ≥ 0.5 million inhabitants. In smaller states (i.e., < 0.5 million inhabitants), the rate of erroneous identification of COVID-19 deaths may be particularly high due to the lack of expertise, measuring devices and experience. Moreover Manuscript to be reviewed influenza vaccination rate (IVR) (%) in people ≥ 65 years old in 2019 or latest available data (Table <ns0:ref type='table'>1</ns0:ref>). I recorded the DPMI, CPMI and CFR data from the public web site https://www.worldometers.info/coronavirus/. Then, I calculated CFR as the rate of DPMI per CPMI. IVR data were also taken from https://data.oecd.org/healthcare/influenza-vaccinationrates.htm, https://oecdcode.org/disclaimers/israel.html and https://www.statista.com/chart/16575/global-flu-immunization-rates-vary/ (retrieved on July 25, 2020). Vietnam's 2017 IVR was recorded from Nguyen et al. <ns0:ref type='bibr' target='#b39'>(2020)</ns0:ref>, and Singapore's 2016/2017 IVR from https://www.todayonline.com/commentary/why-singapores-adult-vaccination-rate-solow.</ns0:p><ns0:p>To analyze the data, I first calculated the non-parametric Spearman rank correlation coefficient (r s ) and its R S 2 and respective p-value (2-tailed) to determine any association between DPMI and CFR with IVR, using R (R Core Team 2017). As the relationship between DPMI and the number of people tested for COVID-19 was not statistically significant based on r s and its p-value, I did not modified (corrected) the DPMI data set. Then, I created regression curves by Generalized additive model (GAM) using the 'ggplot2' package and function (method = 'gam') <ns0:ref type='bibr' target='#b48'>(Wickham et al. 2013)</ns0:ref>, also in R.</ns0:p><ns0:p>As the analysis included countries with different socioeconomic status, demographic structure, urban/rural settings, time of arrival of the pandemic and national control strategies, there may be complex interactions between IVR and other correlated predictor variables. With the aim of accurately estimating the influence of IVR on DPMI and CFR and mitigating the effects of confounding variables, I performed variable importance ranking, including as predictor variables IVR and some potentially important geographical, socioeconomic and non-pharmaceuticalintervention variables <ns0:ref type='bibr' target='#b10'>(Escobar et al. 2020)</ns0:ref>. I used the centroid longitudes (º) and latitudes (º) of each country as geographical variables calculated by the 'rgeos' and 'rworldmap' packages, along with the 'getMap' and 'gCentroid' functions, implemented in R (version 3.3.4; R Development Core Team, 2017). For each country considered, the study recorded socioeconomic variables as the degree of urbanization (DUR) in 2020 <ns0:ref type='table'>2020:08:51704:2:0:NEW 9 Sep 2020)</ns0:ref> Manuscript to be reviewed (https://data.worldbank.org/indicator/SP.POP.65UP.TO.ZS?name_desc=false), which were all retrieved on July 13, 2020 (Table <ns0:ref type='table'>2</ns0:ref>). Finally, I recorded two aspects as Covid-19 prevention measures, i.e. the degree of requirement to use masks (mask) in public (with three degrees: none, parts of country, full country) (https://masks4all.co/what-countries-require-masks-in-public/) and the lockdown degree (lockdown) (with three levels: no lockdown, partial lockdown, nationwide lockdown); all of these sources and the noted in Table <ns0:ref type='table'>3</ns0:ref> were consulted on Aug 13, 2020.</ns0:p><ns0:p>Variable importance ranking was carried out using the 'party' package and the non-parametric random forest function 'cforest', along with Out of bag (OOB) score (with the default option 'controls = cforest_unbiased' and the conditional permutation importance 'varimp(obj, conditional = TRUE)'). Following the permutation principle of the 'mean decrease in accuracy' importance, this machine learning algorithm guarantees unbiased variable importance for predictor variables of different types <ns0:ref type='bibr' target='#b46'>(Strobl et al. 2008)</ns0:ref>.</ns0:p><ns0:p>To mitigate the effects of confounding factors, IVR, DPMI and CFR evaluations were also conducted for countries with similar social conditions (> 50% of DUR, HDI of > 0.80, > 15% of PEP, and PD between 25 and 350 inhabitants per km 2 ) <ns0:ref type='bibr' target='#b10'>(Escobar et al. 2020</ns0:ref>) and for countries with similar longitudes (10 -20º in parts of Europe and 100 -140º, East and Southeast Asia along with Australia and New Zealand).</ns0:p><ns0:p>As IVR and the other eight predictor variables were not strongly correlated (|r s | ≤ 0.57; r s (IVR x DUR) = +0.52; r s (IVR x Long) = -0.46; r s (IVR x HDI) = 0.36), therefore, I included these variables in non-parametric Random Forest (RF) models of DPMI and CFR, including a 5-fold cross validation approach, repeated 30 times using the package 'caret' together with the function 'train' <ns0:ref type='bibr'>(Venables and Ripley, 1999;</ns0:ref><ns0:ref type='bibr' target='#b49'>Williams et al., 2018,</ns0:ref> http://topepo.github.io/caret/index.html) in R software. Finally, I evaluated the goodness-of-fit of the regression model using the (pseudo) coefficient of determination (R 2 ) and the root mean square error (RMSE).</ns0:p></ns0:div>
<ns0:div><ns0:head>Results</ns0:head><ns0:p>For the 26 European countries considered , the results indicated that COVID-19 deaths per million inhabitants (DPMI) and the COVID-19 Case Fatality Ratio (CFR) were positively and <ns0:ref type='table'>4</ns0:ref>).</ns0:p><ns0:p>In evaluations including only countries with similar social conditions, r s (IVR x DPMI) was equal to +0.65 (p = 0.002, N = 20) and r s (IVR x CFR) +0.48 (p = 0.03, N = 20). In analyses including only countries with similar longitude of the country centroid (Long), r s (IVR x DPMI) was equal to +0.83 (p = 0.003, N = 10) (Long from 10 to 20º) and r s (IVR x DPMI) +0.76 (p = 0.046, N = 7) (Long from 100 to 140º).</ns0:p><ns0:p>At worldwide level (39 countries studied), the positive associations between DPMI and IVR were also statistically significant (r s (IVR x DPMI) = +0.49 with p = 0.0016, R s 2 (IVR x DPMI) = 0.24) (Fig. <ns0:ref type='figure' target='#fig_10'>3, Table 5</ns0:ref>). However, the relationships between IVR and CFR were not statistically significant.</ns0:p><ns0:p>In the IVR interval from 7 to 50 %, the association was not significant, although a trend for DPMI and CFR to be positively associated with IVR was observed. DPMI and CFR varied strongly when IVR was 50% or higher (Figs. <ns0:ref type='figure' target='#fig_8'>1-3</ns0:ref>).</ns0:p><ns0:p>Worldwide, the unbiased ranking showed the degree of importance of each variable analyzed.</ns0:p><ns0:p>The variables Long (with 55.9 and 52.3 %) and IVR (with 36.3 and 24.5 %) were by far the most important of the nine variables used to predict DPMI and CFR, respectively. The degree of urbanization (DUR) in 2020 was the third most important variable, with an importance of 5.7% for predicting DPMI. The percentage of elderly people (PEP) in 2019 was the third most important variable (11.5%) in the CFR model (Figs. <ns0:ref type='figure' target='#fig_10'>4 and 5</ns0:ref>). The nine predictor variables considered in this study explained 63% of the variation in DPMI (RMSE =161.9) and 43 % of the variation in CFR (RMSE = 0.039).</ns0:p></ns0:div>
<ns0:div><ns0:head>Discussion</ns0:head><ns0:p>Contrary to expectations, the present worldwide analysis and European sub-analysis do not support the previously reported negative association between COVID-19 deaths (DPMI) and influenza vaccination rate (IVR) in elderly people, observed in studies in Brazil and Italy <ns0:ref type='bibr' target='#b11'>(Fink et al. 2020;</ns0:ref><ns0:ref type='bibr' target='#b29'>Marín-Hernández et al. 2020)</ns0:ref>. Previous studies attributed the beneficial effect of influenza vaccination in reducing severity of COVID-19 disease to better prevention of potential influenza-SARS-CoV-2 coinfections <ns0:ref type='bibr' target='#b2'>(Arokiaraj 2020)</ns0:ref> and, more likely, to changes in innate Manuscript to be reviewed immunity <ns0:ref type='bibr' target='#b33'>(Netea et al. 2020)</ns0:ref>. The innate immune response induced by recent vaccination could result in more rapid and efficient SARS-CoV-2 clearance, preventing progressive dissemination into lower areas of lung tissues <ns0:ref type='bibr' target='#b11'>(Fink et al. 2020)</ns0:ref>.</ns0:p><ns0:p>The negative association between the proportion of DPMI and IVR found in Italy was explained as probably caused by i) a higher influenza vaccine rate occurring in higher economic groups with overall better health, ii) chance, iii) a relationship with seasonal respiratory virus infections, or iv) an unrelated mechanistic association <ns0:ref type='bibr' target='#b29'>(Marín-Hernández et al. 2020)</ns0:ref>. However, the induction of cross-neutralizing antibodies and T-cells that directly target other RNA viruses like SARS-CoV-2 and cross-protection seem unlikely, given the extraordinary diversity of influenza viruses <ns0:ref type='bibr' target='#b11'>(Fink et al. 2020)</ns0:ref>. Therefore, the above-mentioned arguments cannot explain the positive, direct or indirect relationship between IVR and both DPMI and CFR found in this study, which was confirmed by an unbiased ranking variable importance (Figs. <ns0:ref type='figure' target='#fig_10'>4 and 5</ns0:ref>) using RF models. The influenza vaccine may increase influenza immunity at the expense of reduced immunity to SARS-CoV-2 by some unknown biological mechanism, as suggested by <ns0:ref type='bibr' target='#b8'>Cowling et al. (2012)</ns0:ref> for non-influenza respiratory virus. Alternatively, weaker temporary, non-specific immunity after influenza viral infection could cause this positive association due to stimulation of the innate immune response during and for a short time after infection <ns0:ref type='bibr' target='#b31'>(McGill et al. 2009;</ns0:ref><ns0:ref type='bibr'>Khaitov et al. 2009)</ns0:ref>. People who had received the influenza vaccination would have been protected against influenza but not against other viral infections, due to reduced non-specific immunity in the following weeks <ns0:ref type='bibr' target='#b8'>(Cowling et al. 2012)</ns0:ref>, probably caused by virus interference <ns0:ref type='bibr' target='#b21'>(Isaacs & Lindenmann 1957;</ns0:ref><ns0:ref type='bibr' target='#b44'>Seppälä et al. 2011;</ns0:ref><ns0:ref type='bibr' target='#b50'>Wolff 2020)</ns0:ref>. Although existing human vaccine adjuvants have a high level of safety, specific adjuvants in influenza vaccines should also be tested for adverse reactions, such as additionally increased inflammation indicators <ns0:ref type='bibr' target='#b35'>(Petrovsky 2015)</ns0:ref> in Covid-19 patients with already strongly increased inflammation <ns0:ref type='bibr'>(Qin et al. 2020)</ns0:ref>.</ns0:p><ns0:p>The strong variation in DPMI and CFR from an IVR of about 50% or larger may be the result of interactions among the different measures applied in the analyzed countries (Figs. <ns0:ref type='figure' target='#fig_8'>1-3</ns0:ref>), e.g. initiation of interventions, emergency plans and health systems against COVID-19. For example, Australia and South Korea had a very low DPMI and CFR compared with Belgium and United Kingdom (Table <ns0:ref type='table'>1</ns0:ref>).</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:08:51704:2:0:NEW 9 Sep 2020)</ns0:p></ns0:div>
<ns0:div><ns0:head>Manuscript to be reviewed</ns0:head><ns0:p>The high correlation between the longitude of the country centroid and DPMI and CFR emphasize a significant increase in CP and CFR from eastern to western regions in the world (Table <ns0:ref type='table'>5</ns0:ref>, Figs. <ns0:ref type='figure' target='#fig_10'>4 and 5</ns0:ref>), as confirmed by <ns0:ref type='bibr' target='#b27'>Leung et al. (2020)</ns0:ref> and <ns0:ref type='bibr' target='#b45'>Skórka et al. (2020)</ns0:ref>. Longitude could act as a proxy for variables such as lifestyle, social behavior, genetics, geographically isolated and remote populations, which may also be associated with CP and CFR. In the severe 1918-1919 influenza pandemic, remote or isolated populations were also affected, at least partly because of the lack of prior immunity in locations that had not been recently affected by any form of influenza <ns0:ref type='bibr' target='#b30'>(Mathews et al. 2009)</ns0:ref>. Therefore, crossing geographical and ecological barriers also is a key factor in spreading diseases <ns0:ref type='bibr' target='#b20'>(Hallatschek & Fisher 2014;</ns0:ref><ns0:ref type='bibr' target='#b32'>Murray et al. 2015)</ns0:ref>.</ns0:p><ns0:p>Both DPMI and CFR were weakly and positively correlated (p < 0.05) with the absolute value of geographical latitude (abs(Lat)), degree of urbanization (DUR), percentage of elderly people (PEP) and population density (Tables <ns0:ref type='table'>4 and 5</ns0:ref>). In a global analysis, <ns0:ref type='bibr' target='#b10'>Escobar et al. (2020)</ns0:ref> also found positive associations between COVID-19 mortality and the percentage of population aged ≥ 65 years and urbanization, but still more strongly with the Human Development Index. <ns0:ref type='bibr' target='#b27'>Leung et al. (2020)</ns0:ref> also reported positive associations between latitude, temperature by week and by month prior to the first reported COVID-19 case. Lower temperature at northern latitudes was a strong independent predictor of national COVID-19 mortality.</ns0:p><ns0:p>Although countywide lockdowns and use of face masks by the general public should reduce COVID-19 transmission <ns0:ref type='bibr' target='#b5'>(Conyon, He & Thomsen 2020;</ns0:ref><ns0:ref type='bibr'>Eikenberry et al. 2020)</ns0:ref>, the variables lockdown degree and the degree of requirement for mask use in public were not associated with DPMI and CFR in the present study (Tables <ns0:ref type='table'>4 and 5</ns0:ref>, Figs <ns0:ref type='figure' target='#fig_10'>4 and 5</ns0:ref>). <ns0:ref type='bibr' target='#b26'>Leffler et al. (2020)</ns0:ref> reported in a global study that internal lockdown requirements were not associated with mortality, but that in countries that recommended use of face masks early on at the national level, the COVID-19 death rate was lower than expected.</ns0:p><ns0:p>Although countywide lockdowns were proclaimed in many countries, the restrictive measures and their implementations differed in degree, strictness and implementation date in relation to the advance of the disease (see references in Table <ns0:ref type='table'>3</ns0:ref>). Also, although many countries have required masks in public, the mask quality and correct use may differ from country to country. In this regard, <ns0:ref type='bibr' target='#b13'>Fischer et al. (2020)</ns0:ref> found that the use of ineffective masks could be counterproductive. </ns0:p></ns0:div>
<ns0:div><ns0:head>Manuscript to be reviewed</ns0:head><ns0:p>This could explain the non-significant differences between the means of DPMI among countries with and without one or both requirements, lockdown and masks.</ns0:p><ns0:p>Finally, the study is limited by the fact that I didn't normalize the time of arrival of the pandemic. Moreover, the associations found may change in the future because the COVID-19 pandemic was not over at the end of the study.</ns0:p></ns0:div>
<ns0:div><ns0:head>Conclusions</ns0:head><ns0:p>Given the positive relationship between influenza vaccination rate and the number of deaths per million found in this study, further exploration would be valuable to explain these findings and to make conclusions. Additional work on this line of research may also yield results to improve prevention of COVID-19 deaths. = the requirement degree of using masks in public (with three degrees: none, parts of country, full country), lockdown = lockdown degree (with three levels: no lockdown, partial lockdown, nationwide lockdown) of each country, at worldwide level (39 countries studied). Manuscript to be reviewed </ns0:p></ns0:div><ns0:figure xml:id='fig_0'><ns0:head>,</ns0:head><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:08:51704:2:0:NEW 9 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_1'><ns0:head /><ns0:label /><ns0:figDesc>, in such microstates small absolute changes in COVID-19 deaths may result in extreme values of relative indices, such as COVID-19 deaths per million inhabitants (DPMI) and COVID-19 Case Fatality Ratio (CFR). I analyzed the variables DPMI and CFR, based on documented COVID-19 cases per million inhabitants (CPMI) in 2020, COVID-19 tests per million inhabitants, and PeerJ reviewing PDF | (2020:08:51704:2:0:NEW 9 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_2'><ns0:head>(</ns0:head><ns0:label /><ns0:figDesc>https://www.cia.gov/library/publications/the-world-factbook/fields/349.html), the population density (PD) in 2018 (https://data.worldbank.org/indicator/EN.POP.DNST), the Human Development Index (HDI) in 2018 (http://hdr.undp.org/en/composite/HDI) and the percentage of elderly people (PEP) in 2019 PeerJ reviewing PDF | (</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_3'><ns0:head /><ns0:label /><ns0:figDesc>statistically significantly associated with influenza vaccination rate (IVR) in people ≥ 65 yearsold in 2019 or latest data available (r s (IVR x DPMI) = +0.62 with p = 0.0008, R s 2 (IVR x DPMI) = 0.38; r s (IVR x CFR) = +0.50 with p = 0.01, R S 2 (IVR x CFR) = 0.25) (Figs. 1 and 2, Table</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_4'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:08:51704:2:0:NEW 9 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_5'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:08:51704:2:0:NEW 9 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_6'><ns0:head>Figure 1</ns0:head><ns0:label>1</ns0:label><ns0:figDesc>Figure 1</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_7'><ns0:head>Figure 2</ns0:head><ns0:label>2</ns0:label><ns0:figDesc>Figure 2</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_8'><ns0:head>Figure 3</ns0:head><ns0:label>3</ns0:label><ns0:figDesc>Figure 3</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_9'><ns0:head>Figure 4 Unbiased</ns0:head><ns0:label>4</ns0:label><ns0:figDesc>Figure 4</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_10'><ns0:head>Figure 5</ns0:head><ns0:label>5</ns0:label><ns0:figDesc>Figure 5</ns0:figDesc></ns0:figure>
<ns0:note place='foot'>PeerJ reviewing PDF | (2020:08:51704:2:0:NEW 9 Sep 2020)Manuscript to be reviewed</ns0:note>
</ns0:body>
" | "Instituto de Silvicultura e Industria de la Madera
Universidad Juárez del Estado de Durango
Antonio Palazón-Bru
Academic Editor
PeerJ
Carretera Mazatlán Km 5.5
C.P. 34120 Durango
México
Tel.: +52-618-827-12-15
Fax: ++52-618-825-18-86
E-mail:wehenkel@ujed.mx
09-08-2020
Dear Dr. Palazón-Bru,
I have again revised my paper “Positive association between COVID-19 deaths and
influenza vaccination rates in elderly people worldwide” in response to the reviewers 2'
comments. I thank you again for your time taken to review the manuscript and
agree that the comments certainly have improved the manuscript’s content. Along
with this letter, you will find a document describing the way each of your
comments were addressed and a revised copy of the manuscript.
Yours sincerely,
Dr. Christian Wehenkel
Response to Reviewer 1
Many thanks.
Response to Reviewer 2 (Daniela Marín-Hernández)
I appreciate the careful review and insightful comments of Dra. Marín-Hernández.
Line 97-163: Please consider using active voice instead of passive voice.
Done.
Line 100-105: Are you referring to states or countries? Please add the references for your
justification.
All references (see links to the public web sites) noted “countries”.
Please consider changing your abbreviation DP to a more suitable.
I modified the “DP” to “DPMI”.
Line 105: Please consider rewriting this sentence, it is not clear what do you mean by based on
confirmed infected people.
I modified the term to “documented COVID-19 cases”.
Line 115: Please consider explaining the statistical test you performed to conclude that the
relationship between DP and the number of people tested was not statistically significant.
I added: “based on rs and its p-value”. Above it was noted that rs is Spearman’s correlation
coefficient.
Line 116-121: Please consider rephrasing, maybe starting with: I calculated the Spearman rank
correlation coefficients by comparing….
Done!
Line 23-24: Please consider eliminating and a negative association was expected.
Done!
Line 25-31: The description of the statistical method you performed is still missing.
I added: “The associations were measured by non-parametric Spearman rank correlation
coefficients and random forest functions.”
Line 42-54: Please consider rewriting the risk factors section into one paragraph and summarizing
it in a coherent and straight-forward way. Since this section is not crucial for the justification of
your study, it shouldn’t be two paragraphs of your introduction.
I modified the text, now: “Determining the factors influencing the severity of COVID-19 is
important (Armengaud et al. 2020). Although COVID-19 disease does not only affect elderly people,
the severity of symptoms increases with age (https://www.cdc.gov/coronavirus/2019-ncov/needextra-precautions/older-adults.html; Le Couteur, Anderson & Newman 2020). Several other risk
factors have been found for severe COVID-19, such as comorbidities, dyspnea, chest pain, cough,
expectoration, decreased lymphocytes, and increased inflammation indicators (Ji et al. 2020; Li et
al. 2020). Low socioeconomic status is an additional risk factor (Yancy 2020).”
Line 55: Please consider using COVID-19 cases and deaths instead of cases of infection.
Done!
Line 62: Please consider eliminating Vitamin D, since it is considered an adjunctive therapy.
Done!
Line 63: Please consider eliminating drugs since you are already referring in this paragraph to the
only two approved ones.
Done!
Line 64: Please consider using COVID-19 candidate vaccines.
Done!
Line 66-77: You didn’t introduce the concept of heterologous effects or non-specific-effects of
vaccines.
I added: “Heterologous immunity can also result in non-specific effects (also called 'heterologous
effects') of vaccines which affect unrelated infections and diseases, such as extending the
protective outcomes of vaccinations (Goodridge et al. 2016, Agrawal 2019).”
Line 35-96: Please consider summarizing your introduction. You don’t need to specify the
numeric results from all the papers you are referring to.
Done!
Line 97-163: Please consider using active voice instead of passive voice.
Done.
Line 100-105: Are you referring to states or countries? Please add the references for your
justification.
All references (see links) noted “countries”.
Please consider changing your abbreviation DP to a more suitable.
I modified the “DP” to “DPMI”.
Line 105: Please consider rewriting this sentence, it is not clear what do you mean by based on
confirmed infected people.
I modified the term to “documented COVID-19 cases”.
Line 115: Please consider explaining the statistical test you performed to conclude that the
relationship between DP and the number of people tested was not statistically significant.
I added: “based on rs and its p-value”. Above it was noted that rs is Spearman rank correlation
coefficient.
Line 116-121: Please consider rephrasing, maybe starting with: I calculated the Spearman rank
correlation coefficients by comparing….
Done!
Line 193-260: Please consider including as an important limitation the fact that you didn’t
normalize the time of arrival of the pandemic.
I wrote (lines 251-253): “Finally, the study is limited by the fact that I didn’t normalize the time of
arrival of the pandemic. Moreover, the associations found may change in the future because the
COVID-19 pandemic was not over at the end of the study.”
Line 218: Please consider using viral instead of virus infections
Done!
Line 219: Please consider using viral interference instead of virus interference
Done!
Line 233: Please consider eliminating most
Done!
Line 234: Please consider using the lack instead of a lack, using that weren’t recently affected
instead of that had not been recently affected.
Done: “a lack” was changed to “the lack”.
Line 235: Please consider using are instead of is
“is” applies to the “crossing”, not to the barriers.
Line 236: Please consider eliminating is
Done!
Line 252: Please consider rephrasing, the main idea is not clear
Was eliminated!
Line 252- 260: Please consider summarizing this paragraph and emphasize the importance of this
idea to your findings.
I modified and summarized the text as follows: “Although countywide lockdowns were proclaimed
in many countries, the restrictive measures and their implementations differed in degree,
strictness and implementation date in relation to the advance of the disease (see references in
Table 3). Also, although many countries have required masks in public, the mask quality and
correct use may differ from country to country. In this regard, Fischer et al. (2020) found that the
use of ineffective masks could be counterproductive. This could explain the non-significant
differences between the means of DPMI among countries with and without one or both
requirements, lockdown and masks.”
" | Here is a paper. Please give your review comments after reading it. |
9,745 | "<ns0:abstract xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:p>Background. Human brucellosis imposes a heavy burden on the health and economy of endemic regions. Since 2011, China has reported at least 35,000 human brucellosis cases annually, with more than 90% of these cases reported in the northern. Given the alarmingly high incidence and variation in the geographical distribution of human brucellosis cases, there is an urgent need to decipher the causes of such variation in geographical distribution. Method. We conducted a retrospective epidemiological study in Shaanxi Province from January 1, 2005 to December 31, 2018 to investigate the association between meteorological factors and transmission of human brucellosis according to differences in geographical distribution and seasonal fluctuation in northwestern China for the first time. Results. Human brucellosis cases were mainly distributed in the Shaanbei upland plateau before 2008 and then slowly extended towards the southern region with significant seasonal fluctuation. The results of quasi-Poisson generalized additive mixed model (GAMM) indicated that air temperature, sunshine duration, rainfall, relative humidity, and evaporation with maximum lag time within 7 months played crucial roles in the transmission of human brucellosis with seasonal fluctuation. Compared with the Shaanbei upland plateau, Guanzhong basin had more obvious fluctuations in the occurrence of human brucellosis due to changes in meteorological factors. Additionally, the established GAMM model showed high accuracy in predicting the occurrence of human brucellosis based on the meteorological factors.</ns0:p></ns0:div>
<ns0:div><ns0:head>Conclusion.</ns0:head><ns0:p>These findings may be used to predict the seasonal fluctuations of human brucellosis and to develop reliable and cost-effective prevention strategies in Shaanxi Province and other areas with similar environmental conditions.</ns0:p></ns0:div>
</ns0:abstract>
<ns0:body xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:head>Introduction</ns0:head><ns0:p>Brucellosis is a bacterial zoonosis caused by genus Brucella, including B. abortus, B. canis, B. suis, B. ceti, B. pinnipedialis and B. inopinata <ns0:ref type='bibr' target='#b0'>(Dadar et al., 2019)</ns0:ref>. Compared with the individual-to-individual transmission, the environment-to-individual transmission is more common <ns0:ref type='bibr' target='#b1'>(Li et al., 2017)</ns0:ref>, which means that most people are infected by contacting with Brucella in the environment, such as contaminated forage, water, grass, liquids, products, raw milk and the uterine fluids from infected animals <ns0:ref type='bibr' target='#b1'>(Li et al., 2017;</ns0:ref><ns0:ref type='bibr'>Nematollahi et al., 2017;</ns0:ref><ns0:ref type='bibr' target='#b4'>Chen et al., 2013;</ns0:ref><ns0:ref type='bibr' target='#b5'>Liu et al., 2020)</ns0:ref>. Thus, shepherds, breeders, abattoirs workers, veterinarians and laboratory personnel who potentially contact the bacteria are at high risks <ns0:ref type='bibr' target='#b6'>(Chen et al., 2016)</ns0:ref>. Infected people present a variety of symptoms such as night sweats, arthralgia, undulant fever, hepatomegaly, headaches, myalgia, and personality changes <ns0:ref type='bibr' target='#b7'>(Wang et al., 2020;</ns0:ref><ns0:ref type='bibr' target='#b8'>Lou et al., 2016;</ns0:ref><ns0:ref type='bibr' target='#b10'>Li et al., 2013)</ns0:ref> . Since most infected people go to the clinic for treatment only when they have clinical symptoms such as undulant fever, asymptomatic infections often missed and/or misdiagnosed <ns0:ref type='bibr' target='#b11'>(Zhen et al., 2013)</ns0:ref>. In addition, low awareness, ineffective preventative measures, high initial treatment failure, substantial residual disability, and relapse rates have contributed to the heavy burden of this disease on the health and economy of endemic regions <ns0:ref type='bibr' target='#b12'>(Lai et al., 2017)</ns0:ref>.</ns0:p><ns0:p>Globally, the World Health Organization (WHO) and the Food and Agriculture Organization of the United Nations (FAO) announced that brucellosis is one of the most significantly neglected zoonotic diseases in the world <ns0:ref type='bibr' target='#b0'>(Dadar et al., 2019)</ns0:ref>. The virus results in tremendous economic losses and health threats in countries whose economies are dominated by livestock keeping. Human brucellosis is epidemic in many countries especially in Latin America, Middle East, and South and Central Asia, with a total of cases more than 500,000 cases each year <ns0:ref type='bibr' target='#b14'>(Zhao et al., 2019)</ns0:ref>. Since 2011, 31 provinces in China have reported at least 35,000 human brucellosis cases annually, with more than 90% of these cases reported in northern China <ns0:ref type='bibr' target='#b15'>(Wang et al., 2013)</ns0:ref>. Nationwide, the incidence of human brucellosis shows an apparent geographic expansion from northern pastureland provinces to the adjacent grassland and agricultural areas, then to southern coastal and southwestern areas <ns0:ref type='bibr' target='#b16'>(Guan et al., 2018;</ns0:ref><ns0:ref type='bibr' target='#b17'>Yang et al., 2020)</ns0:ref>. From 2005 to 2018, a total of 12,671 confirmed human brucellosis cases were reported in Shaanxi Province, with approximately 70.36% of cases reported in Shaanbei upland plateau, and the average annual incidence reached 11.50/100,000, which was higher than the national average incidence in 2018 (2.73/100,000) <ns0:ref type='bibr'>(Pang et al., 2020)</ns0:ref>. Given the alarmingly high incidence and variation in the geographical distribution of human brucellosis cases, there is an urgent need to decipher the causes of such variation in geographical distribution.</ns0:p><ns0:p>The aim of the current study was to examine the spatial and temporal distributions of human brucellosis in Shaanxi Province between 2005 and 2018. We also explored the associations between meteorological factors and the environment-to-individual transmission of human brucellosis using a quasi-Poisson generalized additive mixed model (GAMM), and then analyzed the driving effect of meteorological factors on the distribution of human brucellosis and predicted the short-term incidence trend in the main epidemic areas of Shaanxi, northwestern China.</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:05:49509:1:1:NEW 15 Aug 2020)</ns0:p><ns0:p>Manuscript to be reviewed</ns0:p></ns0:div>
<ns0:div><ns0:head>Materials & Methods</ns0:head></ns0:div>
<ns0:div><ns0:head>Study Region</ns0:head><ns0:p>Shaanxi is a northwestern administrative province of China, extending from 31°42′-39°35′N to 105°29′-115°15′E. It consists of 11 prefecture-level cities including 30 municipal districts and 77 counties with an area of 205,800 km² and a total population of 3.86 million in 2018 (http://tjj.shaanxi.gov.cn/index.htm). According to the typical climate and landforms, Shaanxi Province can be divided into three distinct natural sub-regions: the Shaanbei upland plateau (northern Shaanxi), the Guanzhong basin (middle part of Shaanxi), and the Shaannan mountainous region (southern Shaanxi) (Figure <ns0:ref type='figure'>S1</ns0:ref>). The Shaanbei upland plateau includes two cities (Yulin and Yan'an) and is part of the transitional landscape spanning from the Maowusu Desert to the Loess Plateau, which belongs to a typical crisscross zone of livestock keeping and farming. This region has constantly been one of the most severely affected areas by brucellosis in China <ns0:ref type='bibr' target='#b4'>(Chen et al., 2013)</ns0:ref>, and the nearby provinces (Inner Mongolia, Gansu, Ningxia, and Shanxi) have all reported a high prevalence of brucellosis <ns0:ref type='bibr' target='#b12'>(Lai et al., 2017;</ns0:ref><ns0:ref type='bibr'>Wang et al., 2013)</ns0:ref>. The Guanzhong basin is located in the middle of Shaanxi province and mainly comprises plains and river landforms. This region has a large population with a huge demand for milk and meat products; therefore, transportation and livestock trade are active in this region. In recent years, the epidemic of brucellosis showed a rapid increasing trend and mainly affected people with occupational exposures, such as farmers, market workers, slaughterhouse workers, and veterinarians <ns0:ref type='bibr' target='#b6'>(Chen et al., 2016;</ns0:ref><ns0:ref type='bibr' target='#b14'>Zhao et al., 2019)</ns0:ref>. The Shaannan mountainous region is characterized by mountains and hills with a subtropical and continental monsoon climate. Sporadic human brucellosis cases have been reported in the region.</ns0:p></ns0:div>
<ns0:div><ns0:head>Data Collection</ns0:head><ns0:p>In China, human brucellosis is listed as a class B notifiable infectious disease, and reporting of data for diagnosed cases (including age, gender, occupation, address and date of brucellosis onset) to the local Center for Disease Control and Prevention (CDC) through the National Notifiable Infectious Diseases Reporting Information System is mandatory. For this study, we obtained the data of human brucellosis cases in Shaanxi from January 1, 2005 to December 31, 2018 from the CDC of Shaanxi Province. All cases were diagnosed based on a combination of epidemiologic exposures, clinical features (fever lasting several days or weeks, sweating, fatigue, and muscle or joint pain, e.g.), and serological test results. Confirmatory tests including the standard plate agglutination test (PAT), rose bengal plate test, serum agglutination test, or isolation of Brucella spp were further conducted to confirm the human brucellosis cases. In addition, the demographic data from the 6th census conducted by the National Bureau of Statistics of China in 2010 were used to calculate the incidence of human brucellosis. The local monthly meteorological data for air temperature (°C), evaporation (mm), rainfall (mm), sunshine duration (h) and relative humidity (%) for the study period were obtained from the Chinese Bureau of Meteorology (http://data.cma.cn/). Manuscript to be reviewed</ns0:p></ns0:div>
<ns0:div><ns0:head>Statistical Analysis</ns0:head><ns0:p>We determined the spatiotemporal distributions of human brucellosis in Shaanxi Province between 2005 and 2018 with stratification for three sub-regions, and the monthly accumulative incidence of brucellosis was geo-referenced to each county on a digital map. In addition, we used a time-series analysis method to identify the relationships and potential effects between meteorological factors and the incidence of human brucellosis. Specifically, we utilized the cross-correlation analysis to assess the associations between meteorological factors and human brucellosis for a range of lags up to 7 months and presented meteorological factors with the maximum correlation coefficients in areas with high incidences (Shaanbei upland plateau and Guanzhong basin). Finally, to identify potential non-linear relationships between meteorological factors and the monthly incidence of human brucellosis, we applied cubic spline analysis including those variables from the quasi-Poisson GAMM model; the model was performed to examine the independent contribution of meteorological factors to the transmission of human brucellosis <ns0:ref type='bibr' target='#b19'>(Cao et al., 2020;</ns0:ref><ns0:ref type='bibr' target='#b20'>Sun et al., 2018)</ns0:ref>. To adjust for long-term trends, seasonality, and over-dispersion, we selected an appropriate degree of freedom (df) and lag for each variable in the model <ns0:ref type='bibr' target='#b21'>(Duan et al., 2017;</ns0:ref><ns0:ref type='bibr' target='#b22'>Zhang et al., 2010)</ns0:ref>. The structure of the model used in the current study is shown below:</ns0:p><ns0:formula xml:id='formula_0'>(1) log (Y t ) = 𝛽 0 + 𝐺𝑟𝑜𝑢𝑝 + 𝑠 ( 𝑖𝑛𝑐𝑖𝑑𝑒𝑛𝑐𝑒,𝑚𝑜𝑛𝑡ℎ ) + ∑𝑠(𝑣𝑎𝑟𝑖𝑎𝑏𝑙𝑒𝑠 𝑡 -e ,𝑏𝑦 = Group,df),</ns0:formula><ns0:p>where is the number of human brucellosis cases in month t; is the intercept; is used R squared value, deviance explained value, and generalized cross-validation (GCV) estimation value were used to determine the most appropriate model. The statistical analyses were performed using R software version 3.6.0 with the packages of 'mgcv' and 'itsadug'. The spatial analyses were performed using ArcGIS 10.2 Software (ESRI Inc.; Redlands, CA, USA). All statistical tests were two-sided, and a p value <0.05 was considered statistically significant. <ns0:ref type='figure'>1</ns0:ref>). When stratified by the three sub-regions, the annual number of cases in Shaanbei upland plateau showed bimodal peaks with a major peak in 2008 (1,104 cases, 19.93 per 100,000 persons) and a minor peak in 2014 (794 cases, 14.34 per 100,000 persons); however, the annual number of cases of Guanzhong basin gradually increased from 2008, peaking in 2014 with 729 cases (3.12 per 100,000 persons). The annual number of cases of Shaannan mountainous region slowly increased as well, peaking in 2015 with 83 cases (0.99 per 100,000 persons) (Figure <ns0:ref type='figure' target='#fig_5'>2</ns0:ref>). The spatial distribution of human brucellosis showed that most cases occurred in Shaanbei upland plateau before 2008, with gradual expansion to middle and southern region (Figure <ns0:ref type='figure' target='#fig_6'>3A</ns0:ref>). Notably, Dali county and Chengcheng county, which were located in low incidence areas, were sites of outbreaks in 2014. The monthly accumulative incidence showed seasonal fluctuations (Figure <ns0:ref type='figure' target='#fig_6'>3B</ns0:ref>), with 60.36% (7,648) of cases occurring between March and July. Sporadic human brucellosis cases occurred in numerous regions of Shaannan mountainous region without significant seasonal fluctuations. Further analysis found that the peak of human brucellosis in Shaanbei upland plateau occurred in March to July, but Guanzhong basin occurred in April to July (Figure <ns0:ref type='figure' target='#fig_6'>3C and 3D</ns0:ref>). In addition, Zizhou County reported the highest annual incidence (153.46 per 100,000 persons) in 2008 and the highest monthly accumulative incidence (104.03 per 100,000 persons) in April.</ns0:p></ns0:div>
<ns0:div><ns0:head>Results</ns0:head></ns0:div>
<ns0:div><ns0:head>Spatial and Temporal Distributions of Human Brucellosis</ns0:head></ns0:div>
<ns0:div><ns0:head>Correlations between Human Brucellosis Incidence and Meteorological Factors</ns0:head><ns0:p>Shaanbei upland plateau and the Guanzhong basin had the highest numbers of human brucellosis cases in Shaanxi as major endemic area. Figure S 2A&2B showed the monthly incidence of human brucellosis in Shaanbei upland plateau was periodic and seasonal fluctuations. In contrast, the periodic and seasonal fluctuations of monthly incidence in Guanzhong basin gradually became obvious, as the incidence increased. The mean air temperature, sunshine duration, evaporation illustrated showed a similar fluctuation with the monthly incidence, while for rainfall and relative humidity an opposite trend is observed. Therefore, correlations between meteorological factors and the number of human brucellosis cases were further explored in these two regions (Table <ns0:ref type='table'>1</ns0:ref>). The maximum correlation coefficients for meteorological factors, including air temperature, rainfall, relative humidity, sunshine duration and evaporation, showed lag time of 4 months, 3 months, 2 months, 6 months, and 5 months in Shaanbei upland plateau, respectively. However, the maximum correlation coefficients for the abovementioned meteorological factors showed lag time of 5 months, 4 months, 3 months, 7 months, and 6 months in the Guanzhong basin, respectively, which all lagged 1 month behind those in Shaanbei upland plateau. According to our correlation analysis between meteorological factors and monthly number of reported cases in the two regions, air temperature had a relatively stronger correlation than other factors, with Spearman correlation coefficients of 0.56 and 0.36 in Shaanbei upland plateau and the Guanzhong basin, respectively ( </ns0:p></ns0:div>
<ns0:div><ns0:head>Estimation of Meteorological Effects on Human Brucellosis in Shaanbei Upland Plateau and Guanzhong Basin</ns0:head><ns0:p>To analyze the seasonal fluctuation of human brucellosis of Shaanbei upland plateau and the Guanzhong basin, a quasi-Poisson GAMM model was used. After controlling for autocorrelation, seasonality, and the lag effect, we found that the monthly number of human brucellosis cases was significantly associated with previous cases and meteorological factors (most p values <0.05), including air temperature, relative humidity, rainfall, evaporation, and sunshine duration (Table <ns0:ref type='table' target='#tab_2'>3</ns0:ref>). According to the values for R square (0.90), deviance explained (91.00%), and GCV principles (4.32), the most appropriate GAMM model was fitted and selected. The observed and fitted cases from the final model matched relatively well for Shaanbei upland plateau and the Guanzhong basin, and the respective goodness of fit (R square) values were 91.43% and 87.83%, respectively. A good fit between observed cases and predicted cases was achieved, using the 24-month observations, and the goodness-of-fit analyses showed that the residuals did not cause significant auto-correlation in the final model (Figures <ns0:ref type='figure' target='#fig_8'>4A&4B and 5A&5B</ns0:ref>). In addition, we found the effects of meteorological factors on the occurrence of human brucellosis were different in Shaanbei upland plateau and Guanzhong basin through Figure <ns0:ref type='figure' target='#fig_9'>6A</ns0:ref>-6E and Table <ns0:ref type='table' target='#tab_2'>3</ns0:ref>. Compared with the Shaanbei upland plateau, Guanzhong basin had more obvious fluctuations in the occurrence of human brucellosis due to changes in meteorological factors, such as air temperature, relative humidity, rainfall, sunshine duration, and evaporation. The summed effects of the meteorological factors parameters terms in Guanzhong basin was 3.48 (95% CI: 2.84-4.12), while the Shaanbei upland plateau was 2.94 (95% CI: 2.70-3.20). The effect of the meteorological factors indicated that except for the air temperature of the current month and rainfall of the previous month showed an upward trend, and previous sunshine duration was U-shaped, the rest of the meteorological factors was a downward trend in Guanzhong basin. High air temperature, low relative humidity, less rainfall, short sunshine duration, and suitable evaporation in this region all had positive effects on the incidence of human brucellosis. In addition, such positive effects were observed for low air temperature and long sunshine duration of the previous month. However, only a weak trend can be observed in Shaanbei upland plateau. Furthermore, the identified interactions between two meteorological factors in association with human brucellosis are shown in Figure <ns0:ref type='figure' target='#fig_7'>S4 A-J</ns0:ref>. The result showed the interaction between high air temperature and lower relative humidity, higher air temperature and shorter sunshine duration, lower relative humidity and shorter sunshine duration, lower relative humidity and suitable evaporation, and less rainfall and suitable evaporation were all obviously associated with high incidence of human brucellosis. Especially in the environment of high air temperature, low relative humidity and short sunshine duration, the risk of human brucellosis is higher.</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:05:49509:1:1:NEW 15 Aug 2020)</ns0:p><ns0:p>Manuscript to be reviewed</ns0:p></ns0:div>
<ns0:div><ns0:head>Discussion</ns0:head><ns0:p>To our best knowledge, this is the first comprehensive study to analyze the association between meteorological factors and transmission of human brucellosis according to differences in geographical distribution and seasonal fluctuation. We found that human brucellosis cases were distributed mainly in the northern part of Shaanxi province before 2008 and then slowly extended towards south with obvious seasonal fluctuations. The quasi-Poisson GAMM model suggested that air temperature, sunshine duration, rainfall, relative humidity, and evaporation with lag time within 7 months may play a crucial role in the transmission, especially environment-to-individual transmission, of human brucellosis and the variation in geographical distribution, and the model had great accuracy in predicting the occurrence of human brucellosis.</ns0:p><ns0:p>The results suggested that the established GAMM can accurately forecast the short-term incidence over 24 months based on meteorological factors, which has important public health implications for developing reliable and cost-effective prevention strategies, including vaccination time, reservoir surveillance, environment disinfection frequency, elimination rates of infected animals and medical resource allocation.</ns0:p><ns0:p>The higher incidences of human brucellosis in Shaanbei upland plateau corroborated with the observation that the northern part of China has historically experienced severe endemic area of brucellosis <ns0:ref type='bibr' target='#b15'>(Wang et al., 2013)</ns0:ref>. In fact, the cases of human brucellosis reported in the northern part of China account for more than 90% of the total reported cases every year <ns0:ref type='bibr' target='#b4'>(Chen et al., 2013;</ns0:ref><ns0:ref type='bibr' target='#b12'>Lai et al., 2017;</ns0:ref><ns0:ref type='bibr' target='#b15'>Wang et al., 2013;</ns0:ref><ns0:ref type='bibr' target='#b23'>Shi et al., 2017)</ns0:ref>. Our study utilized a spatial-temporal overview and analysis of potential interactions between meteorological factors and found that the number of reported cases in Shaanbei upland plateau decreased gradually over the study period, while the number of reported cases in the Guanzhong basin increased rapidly. In addition, we observed that the peak of incidence and lag period of meteorological factors was 1 month earlier in Shaanbei upland plateau than Guanzhong basin. Therefore, we assumed that the Guanzhong basin has become a new endemic region of brucellosis, and imported cases have led to the occurrence of hysteresis.</ns0:p><ns0:p>In addition to the geographic difference, we observed seasonal fluctuation in the human brucellosis in the study (Figure <ns0:ref type='figure' target='#fig_5'>2</ns0:ref> and Figure <ns0:ref type='figure' target='#fig_5'>S2</ns0:ref>). Consistent with the previous finding that meteorological factors play a crucial role in the seasonal fluctuation and transmission of human brucellosis between reservoir and susceptible populations <ns0:ref type='bibr' target='#b10'>(Li et al., 2013;</ns0:ref><ns0:ref type='bibr' target='#b14'>Zhao et al., 2019;</ns0:ref><ns0:ref type='bibr' target='#b19'>Cao et al., 2020)</ns0:ref>, our study found that air temperature, sunshine duration, relative humidity, rainfall, and evaporation were associated with the seasonal fluctuation and geographic variation in the incidence of human brucellosis. The metrological factors may influence the dynamics of reservoirs and viral transmission within a susceptible population; other factors such as types of land use, vegetation, and patterns of agricultural production are also possible contributors accounting for the observed association (Figure <ns0:ref type='figure' target='#fig_6'>S3</ns0:ref>). Short sunshine duration and low relative humidity may have contributed to the human brucellosis incidence through a few pathways. First, appropriate sunshine duration and relative humidity have significant effects on estrus of reservoirs, which is critical to the transmission among susceptible herbs of brucellosis <ns0:ref type='bibr' target='#b10'>(Li et al., 2013;</ns0:ref><ns0:ref type='bibr' target='#b19'>Cao et al., 2020)</ns0:ref>. Second, most animals infected with brucellosis do not show obvious signs, and transmission to the susceptible population occurs mainly through abortion and secretion infection. With the arrival of the production season, a large number of Brucella are excreted into the environment through the abortive secretion of reservoirs. Meanwhile, a large number of susceptible lambs are bred and a dry environment suitable for transmission of Brucella is established, which may lead to higher risks of exposing to susceptible livestock and humans <ns0:ref type='bibr' target='#b20'>(Yang et al., 2018)</ns0:ref>. In addition, low rainfall and suitable evaporation have shown to exert negative impacts on vegetation growth that serves as food for livestock <ns0:ref type='bibr'>(Cotterill et al., 2018)</ns0:ref>. Therefore, our findings may suggest that vegetation plays an important role in the transmission and variation of geographical distribution of human brucellosis.</ns0:p><ns0:p>We hypothesized that several factors contributed to the observed trends in human brucellosis cases. Shaanbei upland plateau has been a traditional endemic area for human brucellosis and has experienced several outbreaks. Thus, public and governmental sectors may have strengthened awareness of prevention and have adopted effective and targeted preventive measures. For example, they may have promoted prevention and control of brucellosis in the public and adopted strengthened quarantine practices in endemic areas. As a result, all these measures may have led to the blockage of the transmission routes <ns0:ref type='bibr' target='#b1'>(Li et al., 2017)</ns0:ref>, the reduction in the incidence, and the reduction of the impact of indirect factors such as meteorological factors. However, the Guanzhong basin is a new endemic area for human brucellosis, and thus, local sectors may have lacked experience and awareness in the prevention and control of the epidemic. In addition, transportation and farming are active in this area, which may have influence on the seasonal fluctuation. In the winter, the livestock and production activities closely related to infected livestock such as cows and sheep are less active. In comparison, in the warm spring and summer seasons, the activities related to farming and production (e.g.; lamb delivery, lamb breeding, handling aborted placenta, shearing wool, and processing and trading meat products) that are closely related to infected reservoirs may increase the exposure frequency of the local population to the virus existing in infected reservoirs and contaminated products. Furthermore, high air temperature has been shown to have a positive effect on the persistence and transmission of Brucella <ns0:ref type='bibr'>(Lee et al., 2013)</ns0:ref>, which is likely to infect susceptible herbs and cause epidemics when protection and awareness are insufficient. Some limitations of the current study merit consideration. Although the data quality from the National Notifiable Disease Surveillance System is expected to be highly credible, underreporting of cases due to mild or unnoticeable clinical symptoms may still exist. In addition, the forecast model constructed in the current study was based on meteorological factors only and used to predict short-term incidence. Third, we did not have data on other factors that may influence the transmission of human brucellosis such as dietary habits, tourism, and traveling <ns0:ref type='bibr'>(Zhu et al., 2017)</ns0:ref>; thus, residual confounding factors may exist that could affect the interpretation of the results. Finally, that is necessary to confirm part areas of Guanzhong basin Manuscript to be reviewed whether has become endemic region of brucellosis through testing of susceptible animals in key areas.</ns0:p></ns0:div>
<ns0:div><ns0:head>Conclusions</ns0:head><ns0:p>The current study implies that meteorological factors may make important contributions to the transmission and variation in geographical distribution of human brucellosis by affecting the external environment, susceptible livestock and/or human populations, activity of reservoirs or the susceptible population, and vegetation. The established GAMM forecast model was shown to be accurate and applicable for predicting the seasonal fluctuation of human brucellosis. Further studies are warranted to validate our findings and develop reliable and cost-effective prevention strategies to combat human brucellosis. Manuscript to be reviewed Manuscript to be reviewed Manuscript to be reviewed Manuscript to be reviewed Manuscript to be reviewed </ns0:p><ns0:note type='other'>Figure 2</ns0:note><ns0:note type='other'>Figure 3</ns0:note><ns0:note type='other'>Figure 4</ns0:note><ns0:note type='other'>Figure 5</ns0:note><ns0:note type='other'>Figure 6</ns0:note></ns0:div><ns0:figure xml:id='fig_0'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:05:49509:1:1:NEW 15 Aug 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_1'><ns0:head /><ns0:label /><ns0:figDesc>difference; exploits the link between incidence and s ( 𝑖𝑛𝑐𝑖𝑑𝑒𝑛𝑐𝑒,𝑚𝑜𝑛𝑡ℎ ) month to control random effects; denotes the cubic spline of ∑s(𝑣𝑎𝑟𝑖𝑎𝑏𝑙𝑒𝑠 𝑡e ,𝑏𝑦 = Group,df) meteorological factors, including air temperature, evaporation, rainfall, sunshine duration and relative humidity, in the previous e months with corresponding for different region. The df = 12</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_2'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:05:49509:1:1:NEW 15 Aug 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_3'><ns0:head /><ns0:label /><ns0:figDesc>Guanzhong basin in Shaanxi Province, China, 2005-2018. *Summary of the GAMM smooth terms, their p-values and effective degrees of freedom are listed. We constructed the final model based on current monthly meteorological factors, temperature lag of 4 months, rainfall lag of 3 months, relative humidity lag of 2 months, sunshine duration lag of 6 months and evaporation lag of 5 months in Shaanbei upland plateau. Meanwhile, we constructed the final model for the Guanzhong basin based on current monthly meteorological factors and previous meteorological factors with lag times 1 month greater than those in Shaanbei upland plateau.</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_4'><ns0:head>Figure 1 Figure 1 .</ns0:head><ns0:label>11</ns0:label><ns0:figDesc>Figure 1</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_5'><ns0:head>Figure 2 .</ns0:head><ns0:label>2</ns0:label><ns0:figDesc>Figure 2. Temporal distribution of human brucellosis in Shaanbei upland plateau , Guanzhong basin, and Shaannan mountainous region, 2005-2018.</ns0:figDesc><ns0:graphic coords='21,42.52,199.12,525.00,294.75' type='bitmap' /></ns0:figure>
<ns0:figure xml:id='fig_6'><ns0:head>Figure 3 .</ns0:head><ns0:label>3</ns0:label><ns0:figDesc>Figure 3. Spatial and temporal distribution of human brucellosis in Shaanxi Province.</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_7'><ns0:head>Figure 4 .</ns0:head><ns0:label>4</ns0:label><ns0:figDesc>Figure 4. Observed and predicted numbers of human brucellosis cases (A), and the scatterplot (B) in Shaanbei upland plateau, 2005-2018.</ns0:figDesc><ns0:graphic coords='23,42.52,199.12,525.00,189.75' type='bitmap' /></ns0:figure>
<ns0:figure xml:id='fig_8'><ns0:head>Figure 5 .</ns0:head><ns0:label>5</ns0:label><ns0:figDesc>Figure 5. Observed and predicted numbers of human brucellosis cases (A), and the scatterplot (B) in Guanzhong basin, 2005-2018.</ns0:figDesc><ns0:graphic coords='24,42.52,199.12,525.00,189.75' type='bitmap' /></ns0:figure>
<ns0:figure xml:id='fig_9'><ns0:head>Figure 6 .</ns0:head><ns0:label>6</ns0:label><ns0:figDesc>Figure 6. Relationship between number of human brucellosis cases and temperature (A), relative humidity (B), rainfall (C), sunshine duration (D), and evaporation (E) in different regions.</ns0:figDesc></ns0:figure>
<ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='26,42.52,70.87,292.83,672.95' type='bitmap' /></ns0:figure>
<ns0:figure type='table' xml:id='tab_1'><ns0:head /><ns0:label /><ns0:figDesc>Table 1 and Figure S 3A&3B). The mean values for the monthly number of human brucellosis cases and meteorological factors in Shaanbei upland plateau and the Guanzhong basin are presented in Table 2. From 2005 to 2018 in Shaanbei upland plateau, the mean values for monthly cases, air temperature, relative humidity, cumulative rainfall, cumulative evaporation, and cumulative sunshine duration were 53.07 cases, 9.90°C, 55.53%, 40.78 mm, 122.86 mm, and 215.87 h, respectively. The corresponding mean values in the Guanzhong basin were 21.16 cases, 12.10°C, 65.16%, 50.79 mm, 101.97 mm, and 171.90 h, respectively.</ns0:figDesc><ns0:table /></ns0:figure>
<ns0:figure type='table' xml:id='tab_2'><ns0:head>Table 3 (on next page)</ns0:head><ns0:label>3</ns0:label><ns0:figDesc /><ns0:table /></ns0:figure>
<ns0:figure type='table' xml:id='tab_3'><ns0:head>Table 3 .</ns0:head><ns0:label>3</ns0:label><ns0:figDesc>Approximate significance of smooth terms in Shaanbei upland plateau and</ns0:figDesc><ns0:table /></ns0:figure>
<ns0:figure type='table' xml:id='tab_4'><ns0:head>Table 3 .</ns0:head><ns0:label>3</ns0:label><ns0:figDesc>Approximate significance of smooth terms in Shaanbei upland plateau and Guanzhong basin in Shaanxi Province,China, 2005China, - 2018. . </ns0:figDesc><ns0:table><ns0:row><ns0:cell>Variables</ns0:cell><ns0:cell>Regions</ns0:cell><ns0:cell>Effective degrees of freedom</ns0:cell><ns0:cell>F</ns0:cell><ns0:cell>p</ns0:cell></ns0:row><ns0:row><ns0:cell>s(Air temperature)</ns0:cell><ns0:cell>Guanzhong basin Shaanbei upland plateau</ns0:cell><ns0:cell>6.41 1.00</ns0:cell><ns0:cell>1.80 0.86</ns0:cell><ns0:cell>0.08 0.35</ns0:cell></ns0:row><ns0:row><ns0:cell>s(Lag in air temperature)</ns0:cell><ns0:cell>Guanzhong basin Shaanbei upland plateau</ns0:cell><ns0:cell>7.73 1.59</ns0:cell><ns0:cell>1.49 4.34</ns0:cell><ns0:cell>0.15 0.01</ns0:cell></ns0:row><ns0:row><ns0:cell>s(Relative humidity)</ns0:cell><ns0:cell>Guanzhong basin Shaanbei upland plateau</ns0:cell><ns0:cell>5.44 1.00</ns0:cell><ns0:cell>3.61 0.12</ns0:cell><ns0:cell><0.001 0.73</ns0:cell></ns0:row><ns0:row><ns0:cell>s(Lag in relative humidity)</ns0:cell><ns0:cell>Guanzhong basin Shaanbei upland plateau</ns0:cell><ns0:cell>6.62 1.00</ns0:cell><ns0:cell>4.06 4.21</ns0:cell><ns0:cell><0.001 0.04</ns0:cell></ns0:row><ns0:row><ns0:cell>s(Sunshine duration)</ns0:cell><ns0:cell>Guanzhong basin Shaanbei upland plateau</ns0:cell><ns0:cell>2.37 1.75</ns0:cell><ns0:cell>2.71 3.18</ns0:cell><ns0:cell>0.04 0.04</ns0:cell></ns0:row><ns0:row><ns0:cell>s(Lag in sunshine duration)</ns0:cell><ns0:cell>Guanzhong basin Shaanbei upland plateau</ns0:cell><ns0:cell>9.41 1.00</ns0:cell><ns0:cell>1.66 4.64</ns0:cell><ns0:cell>0.09 0.03</ns0:cell></ns0:row><ns0:row><ns0:cell>s(Evaporation)</ns0:cell><ns0:cell>Guanzhong basin Shaanbei upland plateau</ns0:cell><ns0:cell>3.47 2.31</ns0:cell><ns0:cell>4.60 1.63</ns0:cell><ns0:cell><0.001 0.21</ns0:cell></ns0:row><ns0:row><ns0:cell>s(Lag in evaporation)</ns0:cell><ns0:cell>Guanzhong basin Shaanbei upland plateau</ns0:cell><ns0:cell>1.30 1.00</ns0:cell><ns0:cell>6.37 4.01</ns0:cell><ns0:cell>0.01 0.05</ns0:cell></ns0:row><ns0:row><ns0:cell>s(Rainfall)</ns0:cell><ns0:cell>Guanzhong basin Shaanbei upland plateau</ns0:cell><ns0:cell>8.61 1.00</ns0:cell><ns0:cell>2.38 0.12</ns0:cell><ns0:cell>0.01 0.73</ns0:cell></ns0:row><ns0:row><ns0:cell>s(Lag in rainfall)</ns0:cell><ns0:cell>Guanzhong basin Shaanbei upland plateau</ns0:cell><ns0:cell>8.23 1.00</ns0:cell><ns0:cell>1.75 0.42</ns0:cell><ns0:cell>0.08 0.52</ns0:cell></ns0:row></ns0:table></ns0:figure>
<ns0:note place='foot'>PeerJ reviewing PDF | (2020:05:49509:1:1:NEW 15 Aug 2020)</ns0:note>
</ns0:body>
" | "July 29, 2020
Dear Dr. Timothy Driscoll,
Thank you for your kind comments on our manuscript entitled “Spatiotemporal expansion of human brucellosis in Shaanxi Province, Northwestern China and model for risk prediction (peerj-49509)”. We are very glad to learn that our manuscript is acceptable for publication in your Journal with minor revisions. All the comments of the reviewers as well as the PeerJ Editorial Board have been fully considered, and we have carefully revised the manuscript as well. Our point-to-point responses are provided as following.
Thanks again,
Sincerely yours,
Kun Liu, Ph.D.
Department of Epidemiology, School of Public Health,
Air Force Medical University
Xi’an, 710032
People's Republic of China
Responses to Editor:
The reviewers and I agree that your manuscript is well-written and requires only minor revisions to be acceptable for publication. Please see the detailed reviewer comments for these changes. Perhaps the most significant revision is to remove Table 4, which has only a single row, and report those results directly in the text.
Response: Thanks for your kindness. We have carefully revised the manuscript, and all the comments of the reviewers as well as the PeerJ Editorial Board have been fully considered.
Responses to Reviewer 1:
This is an interesting and important paper. In terms of content it is well written and makes an important contribution to the body of work on brucellosis and public health. The manuscript is well written. Some grammar, however needs to be checked: For example:
1. Line 26, grammar – with more than 90% of these cases were reported in the northern – “were” needs to be removed.
Response: Thanks, we have revised the mistake in the revised manuscript (Line 26).
2. Line 52 – please remove “allergic” as this brings a different connotation.
Response: We appreciate the reviewer’s comment, and we have removed “allergic” in the revised manuscript (Line 52).
3. Line 54 and 55 – what is common about clinical presentation of brucellosis is the and not necessarily asymptomatic as the author puts it. Please revise this.
Response: We agree with the reviewer’s very important comment. We have changed the sentence “Since most infected people go to the clinic for treatment only when they have clinical symptoms such as undulant fever, asymptomatic infections often missed and/or misdiagnosed” in the revised manuscript (Line 55-57).
4. Line 62 – please replace animal husbandry to livestock keeping.
Response: We appreciate the reviewer’s important comment. We have replaced animal husbandry to livestock keeping in the revised manuscript (Line 67).
5. Line 63 -64 – there is problem with grammar, “including many countries in Latin America, Middle East, and South and Central Asia, which occurred more than 64 500,000 cases every year”. You may consider breaking this from line 62 at livestock keeping.
Response: Sorry for our grammar mistakes. We have revised this sentence “The virus results in tremendous economic losses and health threats in countries whose economies are dominated by livestock keeping. Human brucellosis is endemic in many countries especially in Latin America, Middle East, and South and Central Asia, with a total of cases more than 500,000 cases each year” in lines 66-69 of the revised manuscript with tracks.
6. Line 65 – “with more than 90% of these cases were reported in northern” – remove “were”.
Response: We are very sorry for this error, and we have revised this error in the revised manuscript (Line 71).
7. Line 66-67 – the author mentions the: apparent geographic expansion” but does not follow up to say a bit more about what this is. So the statement is just hanging.
Response: We appreciate the reviewer’s important comment. We have added more about apparent geographic expansion of human brucellosis in China, such as “Nationwide, the incidence of human brucellosis shows an apparent geographic expansion from northern pastureland provinces to the adjacent grassland and agricultural areas, then to southern coastal and southwestern areas” in the revised manuscript (Line 73-75).
Responses to Reviewer 2:
The authors conducted a retrospective epidemiological study in Shaanxi Province from 2005 to 2018 to explore the association between meteorological factors and transmission of human brucellosis according to differences in geographical distribution and seasonal fluctuation in northwestern China. In China, the affected area of brucellosis did expand from northern pastureland provinces to the adjacent grassland and agricultural areas, then to southern coastal and southwestern areas. The findings may be used to predict the seasonal fluctuations of human brucellosis and to develop target prevention and control strategy in the study area.
1. Line 116, in the part of 'Data collection', given that there are several temperature indicators provided by the Chinese Bureau of Meteorology (http://data.cma.cn/), the authors should indicate the exact temperature index involved, e.g., average air temperature or maximum air temperature.
Response: Thank you for your reminding. We sincerely apologize for the confusion of exact temperature index, we have revised as “average air temperature” in the revised manuscript.
2. Only one row of data cannot form Table 4, thus there is no need to indicate the GAMM model result in the format of table, text description is enough.
Response: We appreciate the reviewer’s important comment. We have deleted Table 4, and described the GAMM model results in the section of Result in the revised manuscript (Line 216-218).
The following typos should be corrected.
1) Line 171, 'Zizhou country' should be 'Zizhou county';
Response: We are very sorry for this error, and we have revised this error in the revised manuscript (Line 180).
2) Lines 187, 244 and the legend of Table 3, 'lag times' should be 'lag time';
Response: We appreciate the reviewer’s comment, and we have revised all the “lag times” to “lag time” through the manuscript and Tables.
3) The article number for reference [2] should be added.
Response: We appreciate the reviewer’s very important comment. We have proofread the manuscript carefully, but we haven’t found place where should added reference [2] (Line 48, Line 302).
4) Title page, the authors should check the second affiliation of First author, now it is indicated as 'Centre for Disease Prevent and Control ……'，is here 'prevent' should be 'prevention'?
Response: Thanks, we have revised the error in the revised manuscript.
" | Here is a paper. Please give your review comments after reading it. |
9,746 | "<ns0:abstract xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:p>Porcine epidemic diarrhea virus (PEDV) is a leading cause of diarrhea in pigs worldwide.</ns0:p><ns0:p>Virus isolation and genetic evolutionary analysis allow investigations into the prevalence of epidemic strains and provide data for the clinical diagnosis and vaccine development. In this study, we investigated the genetic characteristics of PEDV circulation in Asia through virus isolation and comparative genomics analysis. A PEDV strain designated HB2018 was isolated from a pig in a farm experiencing a diarrhea outbreak. The complete genome sequence of HB2018 was 28,138 bp in length. Phylogenetic analysis of HB2018 and 207 PEDVs in Asia showed that most PEDV strains circulating in Asia after 2010 belong to genotype GII, particularly GII-a. The PEDV vaccine strain CV777 belonged to GI, and thus, unmatched genotypes between CV777 and GII-a variants might partially explain incomplete protection by the CV777-derived vaccine against PEDV variants in China. In addition, we found the S protein of variant strains contained numerous mutations compared to the S protein of CV777, and these mutations occurred in the N-terminal domain of the S protein. These mutations may influence the antigenicity, pathogenicity, and neutralization properties of the variant strains.</ns0:p></ns0:div>
</ns0:abstract>
<ns0:body xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:head>INTRODUCTION</ns0:head><ns0:p>Porcine epidemic diarrhea (PED) is a high contagious and devastating disease resulting in the watery diarrhea in suckling pigs with high mortality and morbidity <ns0:ref type='bibr'>[1]</ns0:ref>. The causative agent of PED, the porcine epidemic diarrhea virus (PEDV), is an enveloped, single-stranded, positive-sense RNA virus belonging to the genus Alphacoronavirus in the family Coronaviridae <ns0:ref type='bibr' target='#b1'>[2]</ns0:ref>. PEDV possesses a 28-kb genome which encodes seven proteins including ORF1a, ORF1b, spike (S) glycoprotein, ORF3 hypothetical protein, envelop (E) protein, membrane (M) protein and nucleocapsid protein <ns0:ref type='bibr' target='#b2'>[3]</ns0:ref>. Among these proteins, the S protein plays a key role in interaction between the virus and host cells. S protein consists of 1383-amino acids <ns0:ref type='bibr' target='#b3'>[4]</ns0:ref>, and amino acid changes in S protein may lead to antigenic variations and affect the virus virulence <ns0:ref type='bibr' target='#b4'>[5,</ns0:ref><ns0:ref type='bibr' target='#b5'>6]</ns0:ref>. Therefore, this protein is commonly used as an important target for analyzing genetic variations and molecular epidemiology of PEDV <ns0:ref type='bibr' target='#b6'>[7]</ns0:ref>. PED outbreaks have been reported continuously in China since 1973. PED was well controlled since administration of a CV777-derived vaccine <ns0:ref type='bibr' target='#b7'>[8,</ns0:ref><ns0:ref type='bibr' target='#b8'>9]</ns0:ref>. However, recent outbreaks of PED in China since 2010 was due to the re-emergence of PEDV, and the continuous spread of the virus during the last 10 years has resulted in serious economic losses in the pig industry in Asian countries <ns0:ref type='bibr' target='#b9'>[10]</ns0:ref>. In these outbreaks, inactivated vaccines and attenuated live vaccines, which were derived from CV777, were used to control the disease but neither of them provided effective protection <ns0:ref type='bibr' target='#b10'>[11,</ns0:ref><ns0:ref type='bibr' target='#b12'>12]</ns0:ref>. Moreover, the virus has evolved since 2010 <ns0:ref type='bibr' target='#b2'>[3,</ns0:ref><ns0:ref type='bibr' target='#b13'>13,</ns0:ref><ns0:ref type='bibr' target='#b14'>14]</ns0:ref>, and acquisition of whole genome features of PEDV provides a convenient tool for the tracking of PEDV epidemiology <ns0:ref type='bibr' target='#b15'>[15]</ns0:ref>.</ns0:p><ns0:p>In addition, virus isolation and genetic analysis allow investigations on the prevalence of epidemic strains and will provide information for diagnosis and vaccine developments <ns0:ref type='bibr' target='#b16'>[16]</ns0:ref>. In this study, we isolated a highly pathogenic PEDV strain HB2018 from a pig in a farm experiencing PED PeerJ reviewing PDF | (2020:03:47007:1:1:NEW 17 Jul 2020)</ns0:p><ns0:p>Manuscript to be reviewed outbreaks in Hubei province, China, and determined its complete genome sequence. By comparing the HB2018 genome sequence with the sequences of 207 PEDV isolates circulating in Asia, which were publicly available in the Genbank data base, this study also aims to elucidate the evolutionary and genetic characteristics of PEDV currently circulation in different regions of Asia.</ns0:p></ns0:div>
<ns0:div><ns0:head>MATERIALS AND METHODS</ns0:head></ns0:div>
<ns0:div><ns0:head>Virus detection and isolation</ns0:head><ns0:p>In 2018, an outbreak of diarrhea occurred in a CV777-vaccinated pig farm (numbers of sows Manuscript to be reviewed</ns0:p><ns0:formula xml:id='formula_0'>CCCCGGTATTGAATATACCACAGT-3', R: 5'-TTTCTGTTGGCCACCCTTTAGT-3'),</ns0:formula><ns0:p>respectively. Vero cells (Purchased from ATCC, Manassas, VA, USA) were used for virus isolation. In brief, homogenate supernatants and trypsin (5 µg/ml) were inoculated into monolayers of Vero cells, which were then incubated in a 37 °C incubator supplemented with 5% CO 2 . Cells with obvious cytopathic effects (CPEs) were harvested, thawed, and refrozen multiple times. The harvested virus suspension was then inoculated into newly prepared Vero cells for passages, and the propagation was continuously performed for 20 passages (F20). Virus RNA was extracted every five passage for RT-PCR detection of the virus nucleic acids.</ns0:p></ns0:div>
<ns0:div><ns0:head>Virus titration and serum neutralization</ns0:head><ns0:p>Virus titers were measured on 96-well plates using 10-fold serial dilutions of culture supernatant in triplicate per dilution to determine the quantity of viruses required to produce CPEs in 50% of cells. After incubating for enough time, no more CPEs appeared, and TCID50 was calculated using the Reed-Muench method <ns0:ref type='bibr' target='#b17'>[17]</ns0:ref>. The virus titer was also determined by plaque assay using Vero cells and expressed as plaque-forming units (PFU) per mL. The serum neutralization (SN) test was performed in 96-well plates with inactivated serum collected from the guinea pigs infected with the vaccine strain CV777. The virus was diluted in serum-free DMEM to make 200 TCID in a 50 μL volume, and mixed with 50 μL of 2-fold serial dilution serum. The mixture was added to cells cultured in 96-well plates and incubated at 37℃ for 1 h. After removing the mixture and thoroughly washing three times with PBS, the cells were incubated at 37 ℃ with 5% CO 2 for 2 days. Neutralization titers were calculated as the reciprocal of the highest dilution of serum that inhibits CPEs.</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:03:47007:1:1:NEW 17 Jul 2020)</ns0:p><ns0:p>Manuscript to be reviewed</ns0:p></ns0:div>
<ns0:div><ns0:head>Genome sequencing and annotation</ns0:head><ns0:p>Genomic RNA was extracted using the TAKARA RNA extraction kit (Takara, Kusatsu, Shiga, Japan) following the manufacture instruction. The quantity and quality of the extracted RNA were measured by using a Nanodrop spectrophotometer (Thermo, Waltham, MA, USA). The RNA was then subjected to reverse transcription for cDNA using a cDNA synthesis kit (Thermo, Waltham, MA, USA). Genome sequencing was performed with a paired-end library constructed by using a NEB-Next® DNA Library Prep Master Mix Set for Illumina (NEB, Ipswich, MA, USA)</ns0:p><ns0:p>and subsequently sequenced on an Illumina NextSeq 500 with 2 × 150 paired end sequencing chemistry. After filtering, the clean reads were assembled using SPAdes v3.10.1 <ns0:ref type='bibr' target='#b18'>[18]</ns0:ref> and assembled sequences were mapped to the reference genome. The prediction of the genes and proteins were conducted with Prokka v1.12 and RAST Serve (http://rast.nmpdr.org) <ns0:ref type='bibr' target='#b3'>[4]</ns0:ref>. The complete genome sequence as well as its annotations were deposited into NCBI GenBank under the accession number MT166307.</ns0:p></ns0:div>
<ns0:div><ns0:head>Comparative genomics and bioinformatical analysis</ns0:head><ns0:p>The NCBI data was search for 'porcine epidemic diarrhea virus' and a total of 207 complete genome sequences were publicly available for PEDV isolates representing different parts of Asia (See Table <ns0:ref type='table' target='#tab_1'>S1</ns0:ref> in supplementary materials). All of these 207 sequences were downloaded for further analysis. The average nucleotide sequence identity between the genomes of HB2018 and CV777 was calculated by ANI calculator <ns0:ref type='bibr' target='#b19'>[19]</ns0:ref>. Sequence alignments were performed using MAFFT v7.4.02 <ns0:ref type='bibr' target='#b20'>[20]</ns0:ref>. Nucleotide sequence similarity was assessed by SimPlot v.3.5.1 <ns0:ref type='bibr' target='#b21'>[21]</ns0:ref>, with a sliding window size of 500 bp, step size of 100 nucleotides, and 1,000 bootstrap replicates, using gap-stripped alignments and the F84 (ML) distance model. Phylogenetic trees based on complete genome sequences were generated by using MEGA X software with 1,000 bootstrapping <ns0:ref type='bibr' target='#b22'>[22]</ns0:ref> and</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:03:47007:1:1:NEW 17 Jul 2020)</ns0:p><ns0:p>Manuscript to be reviewed were visualized using iTOL v.4 (Interactive Tree of Life, http://itol.embl.de/). The evolutionary history was inferred by using the Maximum Likelihood method and Tamura-Nei model (Tamura & Nei 1993). Initial tree(s) for the heuristic search were obtained automatically by applying</ns0:p><ns0:p>Neighbor-Join and BioNJ algorithms to a matrix of pairwise distances estimated using the Maximum Composite Likelihood (MCL) approach, and then selecting the topology with superior log likelihood value. The tree is drawn to scale, with branch lengths measured in the number of substitutions per site. Protein structure was generated using SWISS-MODEL (Waterhouse et al.</ns0:p></ns0:div>
<ns0:div><ns0:head>2018</ns0:head><ns0:p>). Protein N-glycosylation sites were predicted using online software (http://www.cbs.dtu.dk/services/NetNGlyc/). Threshold values of greater than 0.5 and Jury agreement 9/9 were used for the high-specificity N-glycosylation sites determination <ns0:ref type='bibr' target='#b23'>[23]</ns0:ref>.</ns0:p></ns0:div>
<ns0:div><ns0:head>RESULTS</ns0:head></ns0:div>
<ns0:div><ns0:head>Isolation of PEDV HB2018 and its genomic characteristics</ns0:head><ns0:p>RT-PCR detection of the viral nucleic acids revealed that the intestinal samples from pigs suffered and died from severe watery diarrhea were positive for PEDV but negative for TGEV and PoRV (Figure <ns0:ref type='figure'>S1</ns0:ref> in supplementary materials). Through virus isolation and purification using Vero cells and determination of PEDV nucleic acids using RT-PCR, a PEDV strain was finally recovered and designated HB2018. The TCID50/0.1 mL value of HB2018 was 10 Manuscript to be reviewed Phylogenetic analysis based on the complete genome sequence showed that HB2018 was phylogenetically distinct from the vaccine strain CV777 (Figure <ns0:ref type='figure' target='#fig_3'>1A</ns0:ref>). According to the genotyping system based on a full-length genomic sequence analysis <ns0:ref type='bibr' target='#b2'>[3,</ns0:ref><ns0:ref type='bibr' target='#b7'>8]</ns0:ref>, HB2018 and CV777 belonged to two different genotype: HB2018 was assigned as a type GII strain while CV777 was a GI strain (Figure <ns0:ref type='figure' target='#fig_3'>1A</ns0:ref>). The average nucleotide identity between the genomes of HB2018 and CV777</ns0:p><ns0:p>(GenBank accession no. AF353511) was 96.06% (Figure <ns0:ref type='figure'>S2</ns0:ref> in supplementary materials). The ORF1, ORF3, E, M, and N genes of HB2018 as well as their encoding proteins were highly homologous to those of CV777 (nucleotide identity ≥ 95% for genes; amino acid similarity ≥ 95% for proteins) (Figure <ns0:ref type='figure'>1B</ns0:ref>; Table <ns0:ref type='table' target='#tab_1'>1</ns0:ref>). However, the identity of the S genes and proteins between the two strains was relatively low: the homology for nucleotide and amino acid sequences between HB2018 and CV777 were 93.76% and 93.44%, respectively (Figures <ns0:ref type='figure'>1B & 1C</ns0:ref>; Table <ns0:ref type='table' target='#tab_1'>1</ns0:ref>).</ns0:p><ns0:p>Compared to the S protein of CV777, the S protein of HB2018 had changes, deletions, and/or insertions of amino acids at multiple sites (Table <ns0:ref type='table' target='#tab_3'>S2</ns0:ref> and Figure <ns0:ref type='figure'>S3</ns0:ref> in supplementary materials).</ns0:p><ns0:p>Notably, most of these mutations occurred in the N-terminal domain (NTD, 19-233aa) of the S protein (Figure <ns0:ref type='figure'>1D</ns0:ref>; Figure <ns0:ref type='figure'>S3</ns0:ref> in supplementary materials). Interestingly, some of these mutations were located within the neutralizing epitopes of PEDV [COE (499-638), SS2 (748-755), SS6</ns0:p><ns0:p>(764-771) and 2C10 (1368-1374)]. In addition, these mutations led to a structural change at some parts of the HB2018 S protein compared to the CV777 S protein (Figure <ns0:ref type='figure'>1D</ns0:ref>).</ns0:p></ns0:div>
<ns0:div><ns0:head>Phylogenetic analysis of Aisan PEDV isolates</ns0:head><ns0:p>To explore the phylogenic relationships of the PEDVs currently circulating in Asia, a</ns0:p><ns0:p>Neighbor-joining tree was generated using the complete genome sequence (Figure <ns0:ref type='figure' target='#fig_3'>2A</ns0:ref>). The result revealed that the 208 PEDV strains in Asia, representing the 207 genome sequences publicly available in NCBI and the HB2018 sequence was divided into two genogroups: GI (classical) and</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:03:47007:1:1:NEW 17 Jul 2020)</ns0:p><ns0:p>Manuscript to be reviewed GII (variant). Interestingly, isolates in China before 2010 and the vaccine strain CV777 were included within the GI genogroup. However, most of the PEDV isolates from China as well as the other Asian countries after 2010 belonged to GII genogroup (Figure <ns0:ref type='figure' target='#fig_3'>2A</ns0:ref>). The phylogenetic tree also showed that the two genogroups consisted of several subgroups: the genogroup GI was divided into two subgroups, GI-a and GI-b, while the genogroups GII was divided into three subgroups, GII-a, GII-b, and GII-c (Figure <ns0:ref type='figure' target='#fig_3'>2A</ns0:ref>). The GI-a and GI-b subgroups included isolates from China before 2010 and several Chinese isolates between 2010 and 2015 (Figure <ns0:ref type='figure' target='#fig_3'>2A</ns0:ref>). Most of the GII isolates from China and South Korea and all GII isolates from Japan were included within the GII-a subgroup, while less proportion of the Chinese GII isolates and most of the GII-a isolates from Southeast Asia (Vietnam and Thailand) were included within GII-b subgroup (Figure <ns0:ref type='figure' target='#fig_3'>2A</ns0:ref>). Interestingly, the GII-c subgroup only consisted of isolates from China (Figure <ns0:ref type='figure' target='#fig_3'>2A</ns0:ref>). By analyzing isolation years and genogroups of PEDVs, the history of PEDV and the evolution in China are speculated. Between 1986 and 2008, only five PEDV strains were sequenced in China, and all of them belonged to G1 (Figure <ns0:ref type='figure' target='#fig_3'>2A</ns0:ref>). However, the number of PEDV sequences increased significantly after 2010 (Figure <ns0:ref type='figure'>2B</ns0:ref>). (Q→H) (Figure <ns0:ref type='figure'>4</ns0:ref>). Interestingly, these amino acid changes were also found in the ORF3 proteins of many GII-c isolates after 2016. Sequence comparisons revealed that there were no common INDELs or mutations in E proteins of one subgroup of GII strains compared to E proteins of other subgroups of GII strains (Figure <ns0:ref type='figure' target='#fig_3'>5A</ns0:ref>). M proteins of most GII-a strains had a glutamine (Q) at site 13; however, all GII-b strains isolated between 2011 and 2012 had a glutamic acid (E) at the same position in their M proteins, and this amino acid change (Q→E) occurred frequently in M proteins of GII-b since 2013 (Figure <ns0:ref type='figure'>5B</ns0:ref>). A similar phenomenon was also observed in the M proteins of the GII-c strains, as most of the GII-c strains isolated before 2016 had a glutamine (Q) at position 13 in their M proteins, but a Q→E change at position 13 was seen in the M proteins of more frequently in strains isolated after 2016. In addition, amino acid changes at positions 192 (G→S) and 214 (S→A)</ns0:p><ns0:p>appeared simultaneously in M proteins of some GII-b and GII-c strains. Similarly, in N proteins, amino acid changes at positions 216 (M→V) and 241 (R→K) appeared simultaneously in many GII strains (Figure <ns0:ref type='figure'>5C</ns0:ref>).</ns0:p></ns0:div>
<ns0:div><ns0:head>Discussion</ns0:head><ns0:p>As an infectious virus attracted great intention, the PEDV strains were frequently reported and isolated in Asia. The virus isolation and genetic analysis will provide important information for</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:03:47007:1:1:NEW 17 Jul 2020)</ns0:p><ns0:p>Manuscript to be reviewed PEDV research and vaccine developments. In this article, isolation and genetic features of Asian PEDV strains were reviewed. These mutations in HB2018, especially in the S protein, might be the pathogenic determinants for it, because some deletions and insertions in the S protein may change the antigenicity, pathogenicity and neutralization properties <ns0:ref type='bibr' target='#b8'>[9,</ns0:ref><ns0:ref type='bibr' target='#b23'>23,</ns0:ref><ns0:ref type='bibr' target='#b25'>25]</ns0:ref>. The presence of these mutations in the NTD of S protein in HB2018 might have an effect on the viral pathogenicity since the S-NTD domain is proposed to be the region relevant to the virulence of PEDV <ns0:ref type='bibr' target='#b5'>[6,</ns0:ref><ns0:ref type='bibr' target='#b27'>[26]</ns0:ref><ns0:ref type='bibr' target='#b28'>[27]</ns0:ref><ns0:ref type='bibr' target='#b29'>[28]</ns0:ref>. In addition, the structural changes led by these mutations in S protein of HB2018 might influence the immunogenicity. The distinct phylogenetic relationship between our isolation HB2018 and CV777 might partly explain why vaccination of pigs with CV777 did not provide effective protection against the infection of HB2018 in the vaccinated pig farm. The analysis based on isolation years and genogroups of PEDVs in Asia might also revealed the vaccine CV777 didn't match with the pandemic PEDV isolations. Before 2010, all the strains in China belonged to GI. During this period, PEDV was well controlled in China due to the use of CV777 which was the GI-based vaccine <ns0:ref type='bibr' target='#b8'>[9,</ns0:ref><ns0:ref type='bibr' target='#b9'>10]</ns0:ref>. The phylogenetic analysis of Asian PEDV isolates showed that most of the PEDV isolates from Asia after 2010 belonged to GII genogroup, while the vaccine CV777 were included within GI genogroup. These findings agree with the results of the other studies <ns0:ref type='bibr' target='#b2'>[3,</ns0:ref><ns0:ref type='bibr' target='#b7'>8]</ns0:ref>. The unmatched genotypes between CV777 and PEDV epidemic strains in Asia after 2010 could explain why vaccination with CV777 could not stop the outbreak of PED in many Asian countries after 2010 and provide effective protection against the current epidemic strains <ns0:ref type='bibr' target='#b8'>[9,</ns0:ref><ns0:ref type='bibr' target='#b10'>11,</ns0:ref><ns0:ref type='bibr' target='#b12'>12,</ns0:ref><ns0:ref type='bibr' target='#b30'>29]</ns0:ref>.</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:03:47007:1:1:NEW 17 Jul 2020)</ns0:p></ns0:div>
<ns0:div><ns0:head>Manuscript to be reviewed</ns0:head><ns0:p>With the most reported numbers of PEDV strains, China has more genogroups than other countries. The GII-c subgroup only consisted of isolates from China, these findings are also in agreement with previous studies <ns0:ref type='bibr' target='#b2'>[3,</ns0:ref><ns0:ref type='bibr' target='#b7'>8]</ns0:ref>, suggesting that the genotypes of PEDV strains circulating in China might be more heterogeneous than those of the isolates in other Asian countries. These findings may also explain why PEDV vaccines developed in China contain more than one strains that generally include CV777 and at least one more local GII isolate (http://vdts.ivdc.org.cn:8081/cx/#). The new emerged PEDV in 2010 might accelerate numerous isolations and sequencing of PEDVs <ns0:ref type='bibr' target='#b9'>[10,</ns0:ref><ns0:ref type='bibr' target='#b31'>30]</ns0:ref>. In this article, it was found that the number of PEDV sequences increased significantly after 2010 and most sequences were GII strains. These results are in good agreement with the findings of the PEDV epidemiological investigations in China <ns0:ref type='bibr' target='#b8'>[9,</ns0:ref><ns0:ref type='bibr' target='#b32'>31]</ns0:ref>. It has been reported that PEDV GII isolates were more virulent than GI isolates <ns0:ref type='bibr' target='#b33'>[32]</ns0:ref>. This might in part explain why the traditional vaccines had no to little effect on the control and spread of PEDV in China after 2010. It is noteworthy that PEDV GII strains are also responsible for the recent outbreaks of PED in North America and Europe <ns0:ref type='bibr' target='#b34'>[33]</ns0:ref>. These findings suggest the circulation of PEDV GII strains also pose a problem to the global pig industry. S protein is the most variable protein of PEDV, the amino acid changes in this protein may lead to virus variation and affect the virus virulence <ns0:ref type='bibr' target='#b4'>[5,</ns0:ref><ns0:ref type='bibr' target='#b5'>6]</ns0:ref>. The mutations between were found between GI-a strains and GI-b strains, it is still uncertain whether these mutations between them has a biological significance. While, the mutations occurred in S-NTD of the S protein between GI strains and GII strains might in part explain why do the PEDV GII isolates be more pathogenic than the GI isolates <ns0:ref type='bibr' target='#b33'>[32]</ns0:ref>, as S-NTD is proposed to be the region relevant to the virulence of PEDV <ns0:ref type='bibr' target='#b5'>[6,</ns0:ref><ns0:ref type='bibr' target='#b27'>[26]</ns0:ref><ns0:ref type='bibr' target='#b28'>[27]</ns0:ref><ns0:ref type='bibr' target='#b29'>[28]</ns0:ref>. It is worthy note that PEDVs with insertions of amino acids at 167-168 and deletions of amino acids at 55-58 and 144 in their S proteins are called S-INDEL strains <ns0:ref type='bibr' target='#b35'>[34]</ns0:ref>. A previous study has found infection of the S-INDEL strains could induce pro-inflammatory cytokines through the non-canonical NF-κB signaling pathway by activating RIG-I; however, infection of the non-S-INDEL strains suppresses the induction of pro-inflammatory cytokines and type-I interferon production by down-regulation of TLRs and downstream signaling molecules <ns0:ref type='bibr' target='#b36'>[35]</ns0:ref>. Whether the continuous deletion of 194 amino acids occurred in Japanese strains will affect the virulence of these strains are unknown and warrant further exploration. A previous study however has found that a Japanese strain Tottori2, which had the same deletion, had non-lethal effects in piglets <ns0:ref type='bibr' target='#b37'>[36]</ns0:ref>.</ns0:p><ns0:p>The mutations also were found in C domain of S protein, since the NTD and C-domain both can bind to the host cell receptor and function as the receptor-binding domain, the amino acid changes in their sequences may have important role for the virus <ns0:ref type='bibr' target='#b38'>[37]</ns0:ref>. It has been reported the N-linked glycosylation sites on the S protein of some coronaviruses such as SARS-CoV play a critical role in the viral entry <ns0:ref type='bibr' target='#b39'>[38]</ns0:ref>. The phylogenetic and N-linked glycosylation sites analysis of S protein may offer reasons for further studies. There was no too many mutations were found in the ORF3-E-M-N proteins, it might be because some of them, such as E protein, do not bear too much immune selective pressure since it has no effect on the host cell growth or cell cycle <ns0:ref type='bibr' target='#b40'>[39]</ns0:ref>.</ns0:p></ns0:div>
<ns0:div><ns0:head>CONCLUSIONS</ns0:head><ns0:p>PeerJ reviewing PDF | (2020:03:47007:1:1:NEW 17 Jul 2020)</ns0:p><ns0:p>Manuscript to be reviewed</ns0:p><ns0:p>In conclusion, through virus isolation and complete genome sequencing, we obtained PEDV HB2018 strain. Using this virus, we investigated the genetic and phylogenetic characteristics of PEDV isolates in China as well as in Asia in this study. Phylogenetic analysis revealed heterogeneous genotypes of PEDVs circulate in Asia, but GII particularly GII-a genotype represents the main epidemic genotype in the continent. Our study also revealed that most of the PEDVs currently prevalent in Asian countries displayed a different genotype as well as a distant relationship from the conventional vaccine strain CV777. This finding might explain why CV777derived vaccine provided poor protection against PEDV epidemics (variant strains) since 2010. In addition, we also identified many mutations in the S, ORF3, E, M, N proteins of the variant strains (GII) compared to those of the classical strains <ns0:ref type='bibr' target='#b36'>[35]</ns0:ref>. The presence of these mutations, particularly those determined in the S proteins, may affect the antigenicity, pathogenicity, and neutralization properties of the variant strains.</ns0:p></ns0:div>
<ns0:div><ns0:head>ADDITIONAL INFORMATION AND DECLARATIONS</ns0:head><ns0:note type='other'>Figure 1</ns0:note><ns0:p>Phylogenetic and genetic characteristics of PEDV strain HB2018. Manuscript to be reviewed</ns0:p><ns0:note type='other'>Figure 2</ns0:note><ns0:p>Genotyping of the 208 Asian PEDV strains and the non-Asian reference strains based on full-length genomic sequences. Manuscript to be reviewed High-specificity N-glycosylation sites predicted in Asian strains.</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:03:47007:1:1:NEW 17 Jul 2020)</ns0:p><ns0:p>Manuscript to be reviewed Manuscript to be reviewed</ns0:p><ns0:formula xml:id='formula_1'>CV777 - NTS A • • - • • • • • - - - - - - - - - - HB2018 • • - • • • - - • • - - - - - - - - - - LZC - - • • • • • • • • • - - - - - - - - NLT G DR13 • • - • • NSS N - - • • • - - - - - - - - SD-M NSSS NTS A • • - • • • • • - - - - - - - - - - AJ1102 • • - • • • • - • • • - - - - - - - - - FJZZ1 • • - • • • • - • • • - - - - - - - - - LS - - - • • • - - • • • NVT R - - - - - - - CHS - • • • • • • • • • - - - NFT D - - - - - China CHHNQX- 314 - • • • • • - - • • • - - - - NLT A NCT E - - - CHYJ130330 • • • • NST N • - NIT I • • - - - - - - - - - CBR1 • • • • - • • - • • • - - - - - - - - - Thailand AVCT12 - - • • • • • • • • - - - - - - - - NYT A - Taiwan PT-P5 • • - • • NSS N • - • NKT R • - - - - - - - - KNU-1709 • • - • • NSS N - - • • • - - NST V - - - NIS S - - KNU-1702 NSSS - • • • • - - • • • - - - - - - - - -</ns0:formula></ns0:div>
<ns0:div><ns0:head>South</ns0:head><ns0:formula xml:id='formula_2'>14PED96 N S VN/JFP1013 • • • • - • - - • • • - - - - - - - - - Tottori2 - - • • • • • • • • - - - - - - - NYT A - OKY-1 - - • • • • • • • • • - - - - - - - - - Japan IBR-7 • • - • • NSS N - - • • • - - - - - - - - - 1</ns0:formula><ns0:p>The high-specificity N-glycosylation sites and their amino acids are summarized, the representative strains from each country/regions are listed.' •' means the strain has this high-specificity N-glycosylation site; '-' means the strain has no this high-specificity N-glycosylation site; the amino acid sequences means the strain has a high-specificity N-glycosylation in this sits , but the amino acid sequences are different with the common sequences.</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:03:47007:1:1:NEW 17 Jul 2020)</ns0:p></ns0:div><ns0:figure xml:id='fig_0'><ns0:head>≥</ns0:head><ns0:label /><ns0:figDesc>100) in Hubei Province in China. Many pigs in the farm suffered from severe watery diarrhea, and some of them died. Samples of intestinal tissues were collected from dead pigs and sent to the Veterinary Diagnostic Laboratory of Hubei Academy of Agricultural Sciences in Wuhan, China, for diagnosis. Tissues were immersed with Dulbecco's modified Eagle medium (DMEM; Gibco, Grand Island, NY, USA), and were then homogenized using a QIAGEN TissueLyser II (QIAGEN, Dusseldorf, Nordrhein-Westfalen, Germany). The sample homogenates were then frozen at −80 °C and thawed for three times. After that, the supernatants were filtered through a 0.22-μm membrane and were harvested for RNA and virus isolation. Total RNAs were extracted using TRIzol (Thermo, Waltham, MA, USA) and were reverse transcribed to cDNA using a Thermo Scientific First Strand cDNA Synthesis kit (Thermo, Waltham, MA, USA). Viral nucleic acids were detected by RT-PCR assays using the cDNA as templates and the primers specific for PEDV(F: 5'-TTCGGTTCTATTCCCGTTGATG-3', R: 5'-CCCATGAAGCACTTTCTCACTATC-3'), TGEV (transmissible gastroenteritis virus) (F: 5'-TTACAAACTCGCTATCGCATGG-3', R: 5'-TCTTGTCACATCACCTTTACCTGC-3') and PoRV (porcine rotavirus) (F: 5'-PeerJ reviewing PDF | (2020:03:47007:1:1:NEW 17 Jul 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_1'><ns0:head /><ns0:label /><ns0:figDesc>While several PEDV sequences belonged to genogroup GI after 2010, most sequences from China were GII strains (Figures 2A & 2B). of another GII-a strain (NW17) had a continuous deletion of 6 amino acids (DLYLAI) at positions 168-173; while ORF3 proteins of six GII-b strains (YN15, YN30, YN60, YN90, YN144, YN200) had a continuous deletion of 79 amino acids at their C-terminal (positions 146-224) (Figure 4). Compared to ORF3 proteins of the GII-a isolates, more than half of the GII-b isolates had amino acid changes at positions 25 (L→S), 70 (I→V), 80 (V→F), 107 (C→F), 168 (D→N), and 182</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_2'><ns0:head>(</ns0:head><ns0:label /><ns0:figDesc>A) Phylogenetic analysis of HB2018 and the other PEDV strains based on the whole genome sequence; (B) Nucleotide similarity of the complete genome sequences between PEDV strains HB2018 and CV777; (C) Sequence alignment of the S-NTD regions of PEDV strains HB2018 and CV777; (D) Modelling the 3D structure of the S proteins of HB2018 and CV777. PeerJ reviewing PDF | (2020:03:47007:1:1:NEW 17 Jul 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_3'><ns0:head>( A )</ns0:head><ns0:label>A</ns0:label><ns0:figDesc>The evolutionary history was inferred by using the Maximum Likelihood method based on the Jukes-Cantor model. The tree with the highest log likelihood (-123851.65) is shown. Initial tree(s) for the heuristic search were obtained automatically by applying Neighbor-Join and BioNJ algorithms to a matrix of pairwise distances estimated using the Maximum Composite Likelihood (MCL) approach, and then selecting the topology with superior log likelihood value.The tree is drawn to scale, with branch lengths measured in the number of substitutions per site. GenBank accession numbers of strains, years, places of isolation, genogroups, andsubgroups are shown. (B) Line chart shows the number of PEDV sequences obtained by gene subgroup and year of sampling. Yearly percentages of samples positive for PEDV are indicated by different colored lines.</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_5'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:03:47007:1:1:NEW 17 Jul 2020)</ns0:figDesc></ns0:figure>
<ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='25,42.52,229.87,525.00,324.75' type='bitmap' /></ns0:figure>
<ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='26,42.52,178.87,525.00,201.00' type='bitmap' /></ns0:figure>
<ns0:figure type='table' xml:id='tab_1'><ns0:head>Table 1 .</ns0:head><ns0:label>1</ns0:label><ns0:figDesc>Sequence comparisons of different ORF regions between HB2018 and CV777.</ns0:figDesc><ns0:table><ns0:row><ns0:cell>ORFs</ns0:cell><ns0:cell>HB2018 vs. CV777 Amino acid similarity (%)</ns0:cell><ns0:cell>DNA identity (%)</ns0:cell></ns0:row><ns0:row><ns0:cell>ORF1</ns0:cell><ns0:cell>97.73</ns0:cell><ns0:cell>97.11</ns0:cell></ns0:row><ns0:row><ns0:cell>S</ns0:cell><ns0:cell>93.44</ns0:cell><ns0:cell>93.76</ns0:cell></ns0:row><ns0:row><ns0:cell>ORF3</ns0:cell><ns0:cell>95.98</ns0:cell><ns0:cell>96.44</ns0:cell></ns0:row><ns0:row><ns0:cell>E</ns0:cell><ns0:cell>97.40</ns0:cell><ns0:cell>96.97</ns0:cell></ns0:row><ns0:row><ns0:cell>M</ns0:cell><ns0:cell>99.12</ns0:cell><ns0:cell>97.80</ns0:cell></ns0:row><ns0:row><ns0:cell>N</ns0:cell><ns0:cell>96.60</ns0:cell><ns0:cell>95.48</ns0:cell></ns0:row></ns0:table><ns0:note>2PeerJ reviewing PDF | (2020:03:47007:1:1:NEW 17 Jul 2020)</ns0:note></ns0:figure>
<ns0:figure type='table' xml:id='tab_2'><ns0:head>Table 2 (on next page)</ns0:head><ns0:label>2</ns0:label><ns0:figDesc /><ns0:table /></ns0:figure>
<ns0:figure type='table' xml:id='tab_3'><ns0:head>Table 2 .</ns0:head><ns0:label>2</ns0:label><ns0:figDesc>High-specificity N-glycosylation sites predicted in Asian strains.</ns0:figDesc><ns0:table><ns0:row><ns0:cell>1</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell cols='4'>High-specificity N-glycosylation sites 1</ns0:cell></ns0:row><ns0:row><ns0:cell>57</ns0:cell><ns0:cell>112</ns0:cell><ns0:cell>127</ns0:cell><ns0:cell>212</ns0:cell><ns0:cell>320</ns0:cell><ns0:cell>347</ns0:cell><ns0:cell>510</ns0:cell><ns0:cell>552</ns0:cell><ns0:cell>777</ns0:cell><ns0:cell>1245</ns0:cell><ns0:cell>1257</ns0:cell></ns0:row></ns0:table></ns0:figure>
</ns0:body>
" | "Dear Dr. Silva,
On behalf of my co-authors, we thank you very much for giving us an opportunity to revise our manuscript, we appreciate editor and reviewers very much for their positive and constructive comments and suggestions on our manuscript entitled “Isolation and evolutionary analyses of porcine epidemic diarrhea virus in Asia”. (ID: peerj-reviewing-47007-v0(1); submission ID: 2020:03:47007).
We have studied reviewer’s comments carefully and have made revisions which marked in red in the paper. We have tried our best to revise our manuscript according to the comments. Attached please find the revised version, which we would like to submit for your kind consideration.
We would like to express our great appreciation to you and reviewers for comments on our paper. Looking forward to hearing from you.
Thank you and best regards.
Yours sincerely,
Corresponding author:
Prof. Yongxiang Tian
E-mails: tyxanbit@163.com
Summary and General Comments
---------------------------------------------------------------------------------------------
Reviewer: Babatunde Motayo
Basic reporting
The authors have adhered to the journal's basic reporting guidelines.
Experimental design
No comment
Validity of the findings
No comment
Comments for the author
The paper describes the isolation and characterization of a virulent strain of PEDV from a pig farm in China. The paper is well structured and the methods are detailed and reproducible enough. The have reported a distinct strain of PEDV different from the vaccine strain with mutations in the S protein which may be responsible for incomplete protection from the vaccine strain. The paper is well written but there are some typographical errors that should be looked into and corrected.
.
SPECIFIC COMMENTS: Abstract line 25, the authors school recast the phrase as thus. In this study we the genetic characteristics of PEDV in Asia. The phylogenetic should be deleted, phylogeny is also a part of genetic analysis in this context.
Response: Thanks for your suggestion. We have revised the points accordingly. See line 25 in the resubmitted version.
Also one isolate reported from just one farm is insufficient for the strong conclusions drawn and should be included as a limitation in the study.
Response: Thanks for pointing out this to us. While we just have one isolate for sequencing, we included the genomic sequences of all PEDV isolates in Asia which are publicly available in GenBank at the time of writing. Therefore, we think our study could still draw some conclusions. We have modified our statements in the conclusion section. See line 337-351. Thank you.
Figure legend 1 line 472 (B) The legend does not correspond with the figure, the legend states that Sequence alignment on the complete genomes of PEDV strains HB2018 and CV777. While the figure is a similarity plot of the two PEDV strains illustrated on SimPlot. The authors are advised to make the appropriate correction.
Response: Thanks for pointing out this to us. It has been revised accordingly. Please see line 518.
Once again, we appreciate your time and suggestions on our manuscript.
Reviewer 2
Basic reporting
Dear editor
The paper by Liang et al brings a new PEDV genome recently isolated in China. I think it is an interesting study which uses phylogenetic analysis to call attention to the possible inefficiency of the current PEDV vaccine against viruses from other “genogroups”.
Overall, my major criticism is over the phylogenetic methods chosen by the authors. While I think that the distinction between strains HB2018 and CV777 (the “vaccine strain”) is clear enough to appear whichever method was used, there are several issues (see below) concerning the phylogenetic methods chosen by the authors.
These concerns are easy to tackle and only require more data analysis.
Major issues
- Phylogenetic analysis
1) I'm very concerned about the phylogenetic analysis performed here. NJ is a 'quick and dirty' heuristic that usually recovers the minimum evolution tree. However, distance-based methods are usually worse than character-based methods (such as ML - or Bayesian methods, which also rely on a likelihood function). Based on the current dataset, this NJ tree is also hard to root. Two non-mutually exclusive alternatives would be a) include an outgroup, such as another related alphacoronaviridae, or b) use a molecular-clock method based on sample dates (such as Beast), which will generate a rooted tree. There is also no information about the genetic distance used and a justification for that (note that even in MEGA X there is a module to compare the fit of different evolutionary nucleotide models). Simple genetic distances such as F84, JC, K2P may be unable to account for multiple hits in the alignment, which may have profound impacts on tree topology. I do not question that HB2018 and CV777 do belong to different genotypes, neither that the low effectiveness of the PEDV vaccine has to do with this issue. However, I strongly suggest that that the authors reanalyze their data using a more rigorous phylogenetic framework.
Response: Thanks for your suggestion. We agree and we have re-generated a phylogenetic tree using the ML method. Please see the new figure 2A and lines 134-143.
2) Why does the tree in Figure 1 have so few terminals? How were the 207 sequences filtered out? Does DR13 belong to any 'genogroup'?
Response: Thanks for pointing out this to us. DR13 belongs to GI. We have modified the Figure. Please see the new figure 1A.
3) What is the point of Figure 1C? Do the authors believe that S protein has a different underlying tree than the rest of the genome? If so, which process would have caused this? Recombination? Or is it just an artifact of a higher mutation rate which increases phylogenetic 'noise' in the region? Or is it just noise due to the lower absolute number of mutations available for phylogeny estimation? How is it possible to discriminate among these alternatives?
Response: We agree and we have removed this panel from Figure 1 accordingly. Thank you.
4) Similarly, in Figure 2, there is no information about the ML methodology in the Material and Methods, no justification for the use of JC model, and no justification for the rooting position. Note that several genotypes are not clades. While this can be a real phenomenon, a poorly reconstructed phylogeny can also play a role in this.
Response: Thanks for pointing out this to us. We have added the required contents in the resubmitted version. Please see lines 134-143, and lines 523-533.
5) Are 'genogroup' assignments a result from this study, or was the phylogenetic structure in PEDV known? This could be given in Table S1. On the other hand, if it is an original result, how it was performed, given that genogroups are not clades in the phylogenetic tree? It is important to clarify this issue.
Response: Thank you for your comment. We define the genogroups according to previous publications [Wang et al., Virus Res, 2016 (PMID: 27261169); Guo et al., Transbound Emerg Dis., 2019 (PMID: 30102851)]. We used the same strategy to genotype the PEDV strains.
- Other
6) Please be aware of the use of 'homology' when discussing 'identity'. Homology is a qualitative term that refers (in this case) to the origin of a given genome position. The position is homologous irrespective of the nucleotide present in two given sequences. Newly inserted positions may lack a homologous equivalent (though the ORF itself is homologous), but this is not what is being computed. Also avoid referring to 'homology' in lines 152 and 154.
Response: Thanks for pointing out this to us. We have changed the writing accordingly. See lines 171-174.
7) I wonder if modelling the 3D structure of the S protein and understanding its topology and polarity may give other insights about the effectiveness of a general vaccine against PEDV.
Response: Thanks for your suggestion. We have generated a 3D model accordingly. Please see the new Figure 1D. The relative contents please see lines 188-191.
8) Overall, I think that the discussion part of the 'results and discussion' section could be shortened, as some ideas are repeated several times (such as the inference that the CV777-based vaccine is ineffective against GII strains).
Response: Thanks for your suggestion. We have shorted this part accordingly. Please see the resubmitted version.
9) Most of sections 'Analysis of the S protein' and 'Analysis of the ORF3-E-M-N proteins' is only descriptive. There are some interesting insights, but they get lost amidst idiosyncratic substitutions of uncertain biological meaning. I suggest the authors to rephrase this section trying to resume and discuss only the most important results. The complete list of list of mutations can be given as a supplementary material.
Response: Thanks for your suggestion. We have revised this part accordingly. Please see lines 297-321 the resubmitted version.
Minor issues
line 52. substitute 'China in 2010' for 'China since 2010'
Response: It has been revised accordingly. See line 54. Thank you.
line 54. I'm no native English speaker, but the use of 'nor' in this sentence sounds weird to me. I'd replace it by 'and'
Response: Actually, this manuscript has been re-edited by one of our co-authors, Prof. Yoo, a native English speak and a virologist at UIUC in the United States. However, thank you for pointing this out and we have revised this point accordingly. Please see line 57. Thank you.
line 61. '...and determined its complete genome sequence'
Response: It has been revised accordingly. See line 65. Thank you.
line 63. I'd substitute this sentence, which is quite vague, by a more formal description of the study objectives
Response: It has been revised accordingly. See lines 66-69. Thank you.
line 116. remove 'of'
Response: It has been revised accordingly. See line 124. Thank you.
line 118. Are these complete genomes?
Response: We have revised this place and made it clearer. See line 127. Thank you.
line 157. I see no reason to give the full list of S changes in the text
Response: Thank you for your suggestion. We have removed them from the main text and put them in Table in supplementary materials. Please see lines 175-177, 237-239, 240-243, and Tables S2~S4 in supplementary materials.
line 185. I find this sentence a little bit misleading. Please note that 'ALL' isolates <2008 included in the analyses are only 5 isolates.
Response: It has been revised accordingly. See line 197. Thank you.
line 202. substitute 'strains' by 'strain'
Response: It has been revised accordingly. See line 217. Thank you.
line 204. Please rephrase this sentence
Response: We have removed this sentence. See line 219. Thank you.
line 285. What is 'relatively highly'?
Response: It has been revised accordingly. See line 297. Thank you.
Figures have low resolution and are hard to read (at least in the proof pdf)
Response: Thank you for your comment. All figures submitted are vector graphs. They could be enlarged for read clearer.
Experimental design
no comment (see basic reporting)
Validity of the findings
no comment (see basic reporting)
Comments for the author
no comment (see basic reporting)
Once again, we appreciate your time and suggestions on our manuscript.
Reviewer 3
Basic reporting
The manuscript describes the results of PEDV HB2018 genome comparative analysis. The PEDV HB2018 has been isolated and sequenced, and the evolutionary relationships with PEDV strains have been investigated. The results are of interest but should be improved by additional analysis or data.
Experimental design
The PEDV HB2018 phylogenetic, statistical and evolutionary analysis has been done with widely used software and algorithms. However, taking into account a number of analyzed sequences some additional investigations should've been performed ( e.g. recombination analysis, pairwise matrix, minority variants, dN/dS, positive selection sites, etc). I believe that the genetic analysis of one viral sequence is no longer acceptable as an original article.
Validity of the findings
The number of figures should be minimized and the quality improved. For instance, in Figure 2, it is almost impossible to see the name of the PEDV strains on the circular tree. The authors also should refrain from using print screens as figures.
The authors argue that the mutations in PEDV HB2018 strain may affect the antigenicity, pathogenicity, and neutralization properties of the variant strains, however, no results confirming this proposition are presented. The results of the virus neutralization test or comparative growing curve will definitely benefit the study.
Comments for the author
The authors present the genetic analysis of novel PEDV HB2018 strain, which is important for animal coronavirus research, but I would recommend to reconsider the manuscript as a short communication or a brief report.
Response: Thank you for your comment. However, we still prefer to be published as a research article. Although we only sequenced one isolates from China, our study has included the genomic sequences of all PEDV isolates (207) in Asia which are publicly available in GenBank at the time of writing. We did a lot of analyses in this study and it is not just a brief report of the isolation and complete genome sequencing of one strain. Please consider our time and work on the bioinformatical analysis.
" | Here is a paper. Please give your review comments after reading it. |
9,747 | "<ns0:abstract xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:p>Porcine epidemic diarrhea virus (PEDV) is a leading cause of diarrhea in pigs worldwide.</ns0:p><ns0:p>Virus isolation and genetic evolutionary analysis allow investigations into the prevalence of epidemic strains and provide data for the clinical diagnosis and vaccine development. In this study, we investigated the genetic characteristics of PEDV circulation in Asia through virus isolation and comparative genomics analysis. A PEDV strain designated HB2018 was isolated from a pig in a farm experiencing a diarrhea outbreak. The complete genome sequence of HB2018 was 28,138 bp in length. Phylogenetic analysis of HB2018 and 207 PEDVs in Asia showed that most PEDV strains circulating in Asia after 2010 belong to genotype GII, particularly GII-a. The PEDV vaccine strain CV777 belonged to GI, and thus, unmatched genotypes between CV777 and GII-a variants might partially explain incomplete protection by the CV777-derived vaccine against PEDV variants in China. In addition, we found the S protein of variant strains contained numerous mutations compared to the S protein of CV777, and these mutations occurred in the N-terminal domain of the S protein. These mutations may influence the antigenicity, pathogenicity, and neutralization properties of the variant strains.</ns0:p></ns0:div>
</ns0:abstract>
<ns0:body xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:head>INTRODUCTION</ns0:head><ns0:p>Porcine epidemic diarrhea (PED) is a high contagious and devastating disease resulting in the watery diarrhea in suckling pigs with high mortality and morbidity <ns0:ref type='bibr' target='#b43'>(Zhang et al. 2019)</ns0:ref>. The causative agent of PED, the porcine epidemic diarrhea virus (PEDV), is an enveloped, singlestranded, positive-sense RNA virus belonging to the genus Alphacoronavirus in the family</ns0:p><ns0:p>Coronaviridae <ns0:ref type='bibr' target='#b40'>(Woo et al. 2012)</ns0:ref>. PEDV possesses a 28-kb genome which encodes seven proteins including ORF1a, ORF1b, spike (S) glycoprotein, ORF3 hypothetical protein, envelop (E) protein, membrane (M) protein and nucleocapsid protein <ns0:ref type='bibr' target='#b9'>(Guo et al. 2019</ns0:ref>). Among these proteins, the S protein plays a key role in interaction between the virus and host cells. S protein consists of 1383amino acids <ns0:ref type='bibr'>(Aziz et al.)</ns0:ref>, and amino acid changes in S protein may lead to antigenic variations and affect the virus virulence <ns0:ref type='bibr' target='#b7'>(Gong et al. 2018;</ns0:ref><ns0:ref type='bibr' target='#b32'>Suzuki et al. 2018)</ns0:ref>. Therefore, this protein is commonly used as an important target for analyzing genetic variations and molecular epidemiology of PEDV <ns0:ref type='bibr' target='#b13'>(Hsueh et al. 2020</ns0:ref>). PED outbreaks have been reported continuously in China since 1973. PED was well controlled since administration of a CV777-derived vaccine <ns0:ref type='bibr' target='#b2'>(Chen et al. 2019a;</ns0:ref><ns0:ref type='bibr' target='#b36'>Wang et al. 2016a</ns0:ref>). However, recent outbreaks of PED in China since 2010 was due to the re-emergence of PEDV, and the continuous spread of the virus during the last 10 years has resulted in serious economic losses in the pig industry in Asian countries <ns0:ref type='bibr' target='#b42'>(Yang et al. 2013)</ns0:ref>. In these outbreaks, inactivated vaccines and attenuated live vaccines, which were derived from CV777, were used to control the disease but neither of them provided effective protection <ns0:ref type='bibr' target='#b30'>(Sun et al. 2012;</ns0:ref><ns0:ref type='bibr' target='#b45'>Zhou et al. 2012)</ns0:ref>. Moreover, the virus has evolved since 2010 <ns0:ref type='bibr' target='#b9'>(Guo et al. 2019;</ns0:ref><ns0:ref type='bibr' target='#b12'>Hsu et al. 2018;</ns0:ref><ns0:ref type='bibr' target='#b31'>Sun et al. 2019)</ns0:ref>, and acquisition of whole genome features of PEDV provides a convenient tool for the tracking of PEDV epidemiology <ns0:ref type='bibr' target='#b3'>(Chen et al. 2019b</ns0:ref>). In addition, virus isolation and genetic analysis allow</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:03:47007:2:0:NEW 13 Aug 2020)</ns0:p><ns0:p>Manuscript to be reviewed investigations on the prevalence of epidemic strains and will provide information for diagnosis and vaccine developments <ns0:ref type='bibr' target='#b16'>(Li et al. 2018)</ns0:ref>. In this study, we isolated a highly pathogenic PEDV strain HB2018 from a pig in a farm experiencing PED outbreaks in Hubei province, China, and determined its complete genome sequence. By comparing the HB2018 genome sequence with the sequences of 207 PEDV isolates circulating in Asia, which were publicly available in the Genbank data base, this study also aims to elucidate the evolutionary and genetic characteristics of PEDV currently circulation in different regions of Asia.</ns0:p></ns0:div>
<ns0:div><ns0:head>MATERIALS AND METHODS</ns0:head></ns0:div>
<ns0:div><ns0:head>Virus detection and isolation</ns0:head><ns0:p>In 2018, an outbreak of diarrhea occurred in a CV777-vaccinated pig farm (numbers of sows ≥ 100) in Hubei Province in China. Many pigs in the farm suffered from severe watery diarrhea, and some of them died. Samples of intestinal tissues were collected from dead pigs and sent to the Veterinary Diagnostic Laboratory of Hubei Academy of Agricultural Sciences in Wuhan, China, for diagnosis. Tissues were immersed with Dulbecco's modified Eagle medium (DMEM; Gibco, Grand Island, NY, USA), and were then homogenized using a QIAGEN TissueLyser II <ns0:ref type='bibr'>(QIAGEN, Dusseldorf, Nordrhein-Westfalen, Germany)</ns0:ref>. The sample homogenates were then frozen at −80 °C and thawed for three times. After that, the supernatants were filtered through a 0.22-μm membrane and were harvested for RNA and virus isolation. Total RNAs were extracted using TRIzol (Thermo, Waltham, MA, USA) and were reverse transcribed to cDNA using a Thermo Scientific First Strand cDNA Synthesis kit (Thermo, Waltham, MA, USA). Viral nucleic acids were detected by RT-PCR assays using the cDNA as templates and the primers specific for PEDV (F: 5'-TTCGGTTCTATTCCCGTTGATG-3', R: 5'-CCCATGAAGCACTTTCTCACTATC-3'), incubated at 37 ℃ with 5% CO 2 for 2 days. Neutralization titers were calculated as the reciprocal of the highest dilution of serum that inhibits CPEs.</ns0:p></ns0:div>
<ns0:div><ns0:head>Genome sequencing and annotation</ns0:head><ns0:p>Genomic RNA was extracted using the TAKARA RNA extraction kit (Takara, Kusatsu, Shiga, Japan) following the manufacture instruction. The quantity and quality of the extracted RNA were measured by using a Nanodrop spectrophotometer (Thermo, Waltham, MA, USA). The RNA was then subjected to reverse transcription for cDNA using a cDNA synthesis kit (Thermo, Waltham, MA, USA). Genome sequencing was performed with a paired-end library constructed by using a NEB-Next® DNA Library Prep Master Mix Set for Illumina (NEB, Ipswich, MA, USA)</ns0:p><ns0:p>and subsequently sequenced on an Illumina NextSeq 500 with 2 × 150 paired end sequencing chemistry. After filtering, the clean reads were assembled using SPAdes v3.10.1 <ns0:ref type='bibr' target='#b1'>(Bankevich et al. 2012</ns0:ref>) and assembled sequences were mapped to the reference genome. The prediction of the genes and proteins were conducted with Prokka v1.12 and RAST Serve (http://rast.nmpdr.org) <ns0:ref type='bibr' target='#b0'>(Aziz et al. 2008</ns0:ref>). The complete genome sequence as well as its annotations were deposited into NCBI GenBank under the accession number MT166307.</ns0:p></ns0:div>
<ns0:div><ns0:head>Comparative genomics and bioinformatical analysis</ns0:head><ns0:p>The NCBI data was search for 'porcine epidemic diarrhea virus' and a total of 207 complete genome sequences were publicly available for PEDV isolates representing different parts of Asia (See Table <ns0:ref type='table' target='#tab_0'>S1</ns0:ref> in supplementary materials). All of these 207 sequences were downloaded for further analysis. The average nucleotide sequence identity between the genomes of HB2018 and CV777 was calculated by ANI calculator <ns0:ref type='bibr' target='#b8'>(Goris et al. 2007)</ns0:ref>. Sequence alignments were performed using MAFFT v7.4.02 <ns0:ref type='bibr' target='#b14'>(Katoh & Standley 2013)</ns0:ref>. Nucleotide sequence similarity was assessed by</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:03:47007:2:0:NEW 13 Aug 2020)</ns0:p><ns0:p>Manuscript to be reviewed SimPlot v.3.5.1 <ns0:ref type='bibr' target='#b19'>(Lole et al. 1999)</ns0:ref>, with a sliding window size of 500 bp, step size of 100 nucleotides, and 1,000 bootstrap replicates, using gap-stripped alignments and the F84 (ML) distance model. Phylogenetic trees based on complete genome sequences were generated by using MEGA X software with 1,000 bootstrapping <ns0:ref type='bibr' target='#b15'>(Kumar et al. 2018</ns0:ref>) and were visualized using iTOL v.4 (Interactive Tree of Life, http://itol.embl.de/). The evolutionary history was inferred by using the Maximum Likelihood method and Tamura-Nei model <ns0:ref type='bibr' target='#b33'>(Tamura & Nei 1993)</ns0:ref>. Initial tree(s) for the heuristic search were obtained automatically by applying Neighbor-Join and BioNJ algorithms to a matrix of pairwise distances estimated using the Maximum Composite Likelihood (MCL) approach, and then selecting the topology with superior log likelihood value. The tree is drawn to scale, with branch lengths measured in the number of substitutions per site. Single nucleotide polymorphisms (SNPs) between two genome sequences were determined by the MAUVE package (version 2.4.0) <ns0:ref type='bibr' target='#b5'>(Darling et al. 2004)</ns0:ref>, and the coding effect of these SNPs were analyzed using a previously reported local Perl command <ns0:ref type='bibr' target='#b21'>(Peng et al. 2016)</ns0:ref>. Protein structure was generated using SWISS-MODEL <ns0:ref type='bibr'>(Waterhouse et al. 2018)</ns0:ref>. Protein N-glycosylation sites were predicted using online software (http://www.cbs.dtu.dk/services/NetNGlyc/). Threshold values of greater than 0.5 and Jury agreement 9/9 were used for the high-specificity N-glycosylation sites determination <ns0:ref type='bibr' target='#b26'>(Sagesser et al. 1997)</ns0:ref>.</ns0:p></ns0:div>
<ns0:div><ns0:head>RESULTS</ns0:head></ns0:div>
<ns0:div><ns0:head>Isolation of PEDV HB2018 and its genomic characteristics</ns0:head><ns0:p>RT-PCR detection of the viral nucleic acids revealed that the intestinal samples from pigs suffered and died from severe watery diarrhea were positive for PEDV but negative for TGEV and</ns0:p><ns0:p>PoRV (Figure <ns0:ref type='figure' target='#fig_3'>S1</ns0:ref> in supplementary materials). Through virus isolation and purification using Vero to 27,692). Phylogenetic analysis based on the complete genome sequence showed that HB2018 was phylogenetically distinct from the vaccine strain CV777 (Figure <ns0:ref type='figure' target='#fig_6'>1A</ns0:ref>). According to the genotyping system based on a full-length genomic sequence analysis <ns0:ref type='bibr' target='#b9'>(Guo et al. 2019;</ns0:ref><ns0:ref type='bibr' target='#b36'>Wang et al. 2016a</ns0:ref>), HB2018 and CV777 belonged to two different genotype: HB2018 was assigned as a type GII strain while CV777 was a GI strain (Figure <ns0:ref type='figure' target='#fig_6'>1A</ns0:ref>). The average nucleotide identity between the genomes of HB2018 and CV777 (GenBank accession no. AF353511) was 96.06% (Figure <ns0:ref type='figure' target='#fig_4'>S2</ns0:ref> in supplementary materials). The ORF1, ORF3, E, M, and N genes of HB2018 as well as their encoding proteins were highly homologous to those of CV777 (nucleotide identity ≥ 95% for genes; amino acid similarity ≥ 95% for proteins) (Figure <ns0:ref type='figure' target='#fig_6'>1B</ns0:ref>; Table <ns0:ref type='table' target='#tab_0'>1</ns0:ref>). However, the identity of the S genes and proteins between the two strains was relatively low: the homology for nucleotide and amino acid sequences between HB2018 and CV777 were 93.76% and 93.44%, respectively (Figures <ns0:ref type='figure' target='#fig_6'>1B & 1C</ns0:ref>; Table <ns0:ref type='table' target='#tab_0'>1</ns0:ref>). SNP analysis determined a total of 946 SNPs in the genome sequence of HB2018 when compared to the genome sequence of the reference strain CV777. Among these SNPs, 925 SNPs including 262 non-synonymous substitutions and 663 synonymous substitutions were located with the ORF regions, with an overall ratio of nonsynonymous to synonymous substitutions (dN/dS) of 0.39 (Table <ns0:ref type='table' target='#tab_1'>2</ns0:ref>). The dN/dS ratios in each of the ORFs encoded by the PeerJ reviewing PDF | (2020:03:47007:2:0:NEW 13 Aug 2020)</ns0:p><ns0:p>Manuscript to be reviewed HB2018 genome ranged from 0.15 to 0.61, with the S protein had the highest dN/dS ratio (Table <ns0:ref type='table' target='#tab_1'>2</ns0:ref>).</ns0:p><ns0:p>Compared to the S protein of CV777, the S protein of HB2018 had changes, deletions, and/or insertions of amino acids at multiple sites (Table <ns0:ref type='table' target='#tab_1'>S2</ns0:ref> and Figure <ns0:ref type='figure' target='#fig_5'>S3</ns0:ref> in supplementary materials).</ns0:p><ns0:p>Notably, most of these mutations occurred in the N-terminal domain (NTD, 19-233aa) of the S protein (Figure <ns0:ref type='figure' target='#fig_6'>1D</ns0:ref>; Figure <ns0:ref type='figure' target='#fig_5'>S3</ns0:ref> in supplementary materials). Interestingly, some of these mutations were located within the neutralizing epitopes of PEDV [COE (499-638), SS2 (748-755), SS6</ns0:p><ns0:p>(764-771) and 2C10 (1368-1374)]. In addition, these mutations led to a structural change at some parts of the HB2018 S protein compared to the CV777 S protein (Figure <ns0:ref type='figure' target='#fig_6'>1D</ns0:ref>).</ns0:p></ns0:div>
<ns0:div><ns0:head>Phylogenetic analysis of Aisan PEDV isolates</ns0:head><ns0:p>To explore the phylogenic relationships of the PEDVs currently circulating in Asia, a</ns0:p><ns0:p>Neighbor-joining tree was generated using the complete genome sequence (Figure <ns0:ref type='figure'>2A</ns0:ref>). The result revealed that the 208 PEDV strains in Asia, representing the 207 genome sequences publicly available in NCBI and the HB2018 sequence was divided into two genogroups: GI (classical) and GII (variant). Interestingly, isolates in China before 2010 and the vaccine strain CV777 were included within the GI genogroup. However, most of the PEDV isolates from China as well as the other Asian countries after 2010 belonged to GII genogroup (Figure <ns0:ref type='figure'>2A</ns0:ref>). The phylogenetic tree also showed that the two genogroups consisted of several subgroups: the genogroup GI was divided into two subgroups, GI-a and GI-b, while the genogroups GII was divided into three subgroups, GII-a, GII-b, and GII-c (Figure <ns0:ref type='figure'>2A</ns0:ref>). The GI-a and GI-b subgroups included isolates from China before 2010 and several Chinese isolates between 2010 and 2015 (Figure <ns0:ref type='figure'>2A</ns0:ref>). Most of the GII isolates from China and South Korea and all GII isolates from Japan were included within the GII-a subgroup, while less proportion of the Chinese GII isolates and most of the GII-a 494NVTS497, 549NCTE552, 587NISI590, 1055NKTL1058, and 1067NRTG1070 were retained in these two strains.</ns0:p></ns0:div>
<ns0:div><ns0:head>Analysis of the ORF3-E-M-N proteins</ns0:head><ns0:p>Unlike the S protein, ORF3 is a conserved protein among the PEDV isolates <ns0:ref type='bibr' target='#b37'>(Wang et al. 2016b</ns0:ref>). However, several PEDVs were found to have characteristic amino acid mutations in the ORF3 (Figure <ns0:ref type='figure'>4</ns0:ref>). Compared to ORF3 proteins of many G1 strains and all GII strains, nine GI-a</ns0:p><ns0:p>PEDVs (Q→H) (Figure <ns0:ref type='figure'>4</ns0:ref>). Interestingly, these amino acid changes were also found in the ORF3 proteins of many GII-c isolates after 2016.</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:03:47007:2:0:NEW 13 Aug 2020)</ns0:p><ns0:p>Manuscript to be reviewed</ns0:p><ns0:p>Sequence comparisons revealed that there were no common INDELs or mutations in E proteins of one subgroup of GII strains compared to E proteins of other subgroups of GII strains (Figure <ns0:ref type='figure'>5A</ns0:ref>). M proteins of most GII-a strains had a glutamine (Q) at site 13; however, all GII-b strains isolated between 2011 and 2012 had a glutamic acid (E) at the same position in their M proteins, and this amino acid change (Q→E) occurred frequently in M proteins of GII-b since 2013 (Figure <ns0:ref type='figure'>5B</ns0:ref>). A similar phenomenon was also observed in the M proteins of the GII-c strains, as most of the GII-c strains isolated before 2016 had a glutamine (Q) at position 13 in their M proteins, but a Q→E change at position 13 was seen in the M proteins of more frequently in strains isolated after 2016. In addition, amino acid changes at positions 192 (G→S) and 214 (S→A)</ns0:p><ns0:p>appeared simultaneously in M proteins of some GII-b and GII-c strains. Similarly, in N proteins, amino acid changes at positions 216 (M→V) and 241 (R→K) appeared simultaneously in many GII strains (Figure <ns0:ref type='figure'>5C</ns0:ref>).</ns0:p></ns0:div>
<ns0:div><ns0:head>Discussion</ns0:head><ns0:p>As an infectious virus attracted great intention, the PEDV strains were frequently reported and isolated in Asia. The virus isolation and genetic analysis will provide important information for PEDV research and vaccine developments. In this study, we isolated a GII-a strain HB2018 and determined its genomic characteristics. Comparative genomic analysis revealed that the ORF1, ORF3, E, M, and N genes of HB2018 as well as their encoding proteins were highly homologous to those of CV777 (Figure <ns0:ref type='figure' target='#fig_6'>1B</ns0:ref>; Table <ns0:ref type='table' target='#tab_0'>1</ns0:ref>). However, a number of SNPs were determined within these ORFs, with the S proteins showed the highest dN/dS ratio (Table <ns0:ref type='table' target='#tab_0'>1</ns0:ref>). Since dN/dS ratio is commonly used as a measure of purifying versus diversifying selection <ns0:ref type='bibr' target='#b25'>(Rocha et al. 2006</ns0:ref>), the PeerJ reviewing PDF | (2020:03:47007:2:0:NEW 13 Aug 2020)</ns0:p><ns0:p>Manuscript to be reviewed highest dN/dS ratio suggests S protein is under diversifying selection, and this diversifying selection might be associated with its frequent interaction with host cells. Sequence alignments determined many mutations in the genome sequence of HB2018 compared to that of the reference strain CV777. These mutations, especially in the S protein, might be the pathogenic determinants for it, because some deletions and insertions in the S protein may change the antigenicity, pathogenicity and neutralization properties <ns0:ref type='bibr' target='#b2'>(Chen et al. 2019a;</ns0:ref><ns0:ref type='bibr' target='#b26'>Sagesser et al. 1997;</ns0:ref><ns0:ref type='bibr' target='#b44'>Zhang et al. 2015)</ns0:ref>. The presence of these mutations in the NTD of S protein in HB2018</ns0:p><ns0:p>might have an effect on the viral pathogenicity since the S-NTD domain is proposed to be the region relevant to the virulence of PEDV <ns0:ref type='bibr' target='#b11'>(Hou et al. 2017;</ns0:ref><ns0:ref type='bibr' target='#b27'>Su et al. 2018;</ns0:ref><ns0:ref type='bibr' target='#b28'>Su et al. 2019;</ns0:ref><ns0:ref type='bibr' target='#b32'>Suzuki et al. 2018</ns0:ref>). In addition, the structural changes led by these mutations in S protein of HB2018 might influence the immunogenicity. The distinct phylogenetic relationship between our isolation HB2018 and CV777 might partly explain why vaccination of pigs with CV777 did not provide effective protection against the infection of HB2018 in the vaccinated pig farm. The analysis based on isolation years and genogroups of PEDVs in Asia might also revealed the vaccine CV777 didn't match with the pandemic PEDV isolations. Before 2010, all the strains in China belonged to GI.</ns0:p><ns0:p>During this period, PEDV was well controlled in China due to the use of CV777 which was the GI-based vaccine <ns0:ref type='bibr' target='#b2'>(Chen et al. 2019a;</ns0:ref><ns0:ref type='bibr' target='#b42'>Yang et al. 2013)</ns0:ref>. The phylogenetic analysis of Asian PEDV isolates showed that most of the PEDV isolates from Asia after 2010 belonged to GII genogroup, while the vaccine CV777 were included within GI genogroup. These findings agree with the results of the other studies <ns0:ref type='bibr' target='#b9'>(Guo et al. 2019;</ns0:ref><ns0:ref type='bibr' target='#b36'>Wang et al. 2016a)</ns0:ref>. The unmatched genotypes between</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:03:47007:2:0:NEW 13 Aug 2020)</ns0:p><ns0:p>Manuscript to be reviewed CV777 and PEDV epidemic strains in Asia after 2010 could explain why vaccination with CV777</ns0:p><ns0:p>could not stop the outbreak of PED in many Asian countries after 2010 and provide effective protection against the current epidemic strains <ns0:ref type='bibr' target='#b2'>(Chen et al. 2019a;</ns0:ref><ns0:ref type='bibr' target='#b23'>Puranaveja et al. 2009;</ns0:ref><ns0:ref type='bibr' target='#b30'>Sun et al. 2012;</ns0:ref><ns0:ref type='bibr' target='#b45'>Zhou et al. 2012)</ns0:ref>.</ns0:p><ns0:p>With the most reported numbers of PEDV strains, China has more genogroups than other countries. The GII-c subgroup only consisted of isolates from China, these findings are also in agreement with previous studies <ns0:ref type='bibr' target='#b9'>(Guo et al. 2019;</ns0:ref><ns0:ref type='bibr' target='#b36'>Wang et al. 2016a)</ns0:ref>, suggesting that the genotypes of PEDV strains circulating in China might be more heterogeneous than those of the isolates in other Asian countries. These findings may also explain why PEDV vaccines developed in China contain more than one strains that generally include CV777 and at least one more local GII isolate (http://vdts.ivdc.org.cn:8081/cx/#). The new emerged PEDV in 2010 might accelerate numerous isolations and sequencing of PEDVs <ns0:ref type='bibr' target='#b18'>(Li et al. 2012;</ns0:ref><ns0:ref type='bibr' target='#b42'>Yang et al. 2013)</ns0:ref>. In this article, it was found that the number of PEDV sequences increased significantly after 2010 and most sequences were GII strains. These results are in good agreement with the findings of the PEDV epidemiological investigations in China <ns0:ref type='bibr' target='#b2'>(Chen et al. 2019a;</ns0:ref><ns0:ref type='bibr' target='#b29'>Sun et al. 2018)</ns0:ref>. It has been reported</ns0:p><ns0:p>that PEDV GII isolates were more virulent than GI isolates <ns0:ref type='bibr' target='#b35'>(Vlasova et al. 2014)</ns0:ref>. This might in part explain why the traditional vaccines had no to little effect on the control and spread of PEDV in China after 2010. It is noteworthy that PEDV GII strains are also responsible for the recent outbreaks of PED in North America and Europe <ns0:ref type='bibr' target='#b4'>(Choudhury et al. 2016</ns0:ref>). These findings suggest the circulation of PEDV GII strains also pose a problem to the global pig industry.</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:03:47007:2:0:NEW 13 Aug 2020)</ns0:p><ns0:p>Manuscript to be reviewed S protein is the most variable protein of PEDV, the amino acid changes in this protein may lead to virus variation and affect the virus virulence <ns0:ref type='bibr' target='#b7'>(Gong et al. 2018;</ns0:ref><ns0:ref type='bibr' target='#b32'>Suzuki et al. 2018</ns0:ref>). The mutations between were found between GI-a strains and GI-b strains, it is still uncertain whether these mutations between them has a biological significance. While, the mutations occurred in S-NTD of the S protein between GI strains and GII strains might in part explain why do the PEDV GII isolates be more pathogenic than the GI isolates <ns0:ref type='bibr' target='#b35'>(Vlasova et al. 2014)</ns0:ref>, as S-NTD is proposed to be the region relevant to the virulence of PEDV <ns0:ref type='bibr' target='#b11'>(Hou et al. 2017;</ns0:ref><ns0:ref type='bibr' target='#b27'>Su et al. 2018;</ns0:ref><ns0:ref type='bibr' target='#b28'>Su et al. 2019;</ns0:ref><ns0:ref type='bibr' target='#b32'>Suzuki et al. 2018)</ns0:ref>. It is worthy note that PEDVs with insertions of amino acids at 167-168 and deletions of amino acids at 55-58 and 144 in their S proteins are called S-INDEL strains <ns0:ref type='bibr'>(Wang et al. 2014)</ns0:ref>. A previous study has found infection of the S-INDEL strains could induce proinflammatory cytokines through the non-canonical NF-κB signaling pathway by activating RIG-I; however, infection of the non-S-INDEL strains suppresses the induction of pro-inflammatory cytokines and type-I interferon production by down-regulation of TLRs and downstream signaling molecules <ns0:ref type='bibr' target='#b34'>(Temeeyasen et al. 2018)</ns0:ref>. Whether the continuous deletion of 194 amino acids occurred in Japanese strains will affect the virulence of these strains are unknown and warrant further exploration. A previous study however has found that a Japanese strain Tottori2, which had the same deletion, had non-lethal effects in piglets <ns0:ref type='bibr' target='#b20'>(Masuda et al. 2015)</ns0:ref>. The mutations also were Manuscript to be reviewed S protein of some coronaviruses such as SARS-CoV play a critical role in the viral entry <ns0:ref type='bibr' target='#b10'>(Han et al. 2007</ns0:ref>). The phylogenetic and N-linked glycosylation sites analysis of S protein may offer reasons for further studies. There was no too many mutations were found in the ORF3-E-M-N proteins, it might be because some of them, such as E protein, do not bear too much immune selective pressure since it has no effect on the host cell growth or cell cycle <ns0:ref type='bibr' target='#b41'>(Xu et al. 2013)</ns0:ref>.</ns0:p></ns0:div>
<ns0:div><ns0:head>CONCLUSIONS</ns0:head><ns0:p>In conclusion, through virus isolation and complete genome sequencing, we obtained PEDV HB2018 strain. Using this virus, we investigated the genetic and phylogenetic characteristics of PEDV isolates in China as well as in Asia in this study. Phylogenetic analysis revealed heterogeneous genotypes of PEDVs circulate in Asia, but GII particularly GII-a genotype represents the main epidemic genotype in the continent. Our study also revealed that most of the PEDVs currently prevalent in Asian countries displayed a different genotype as well as a distant relationship from the conventional vaccine strain CV777. This finding might explain why CV777derived vaccine provided poor protection against PEDV epidemics (variant strains) since 2010. In addition, we also identified many mutations in the S, ORF3, E, M, N proteins of the variant strains (GII) compared to those of the classical strains <ns0:ref type='bibr'>(Temeeyasen et al.)</ns0:ref>. The presence of these mutations, particularly those determined in the S proteins, may affect the antigenicity, pathogenicity, and neutralization properties of the variant strains.</ns0:p></ns0:div>
<ns0:div><ns0:head>ADDITIONAL INFORMATION AND DECLARATIONS</ns0:head></ns0:div>
<ns0:div><ns0:head>Funding</ns0:head><ns0:p>PeerJ reviewing PDF | (2020:03:47007:2:0:NEW 13 Aug 2020)</ns0:p><ns0:p>Manuscript to be reviewed Manuscript to be reviewed</ns0:p><ns0:formula xml:id='formula_0'>CV777 - NTS A • • - • • • • • - - - - - - - - - - HB2018 • • - • • • - - • • - - - - - - - - - - LZC - - • • • • • • • • • - - - - - - - - NLT G DR13 • • - • • NSS N - - • • • - - - - - - - - SD-M NSSS NTS A • • - • • • • • - - - - - - - - - - AJ1102 • • - • • • • - • • • - - - - - - - - - FJZZ1 • • - • • • • - • • • - - - - - - - - - LS - - - • • • - - • • • NVT R - - - - - - - CHS - • • • • • • • • • - - - NFT D - - - - - China CHHNQX- 314 - • • • • • - - • • • - - - - NLT A NCT E - - - CHYJ130330 • • • • NST N • - NIT I • • - - - - - - - - - CBR1 • • • • - • • - • • • - - - - - - - - - Thailand AVCT12 - - • • • • • • • • - - - - - - - - NYT A - Taiwan PT-P5 • • - • • NSS N • - • NKT R • - - - - - - - - KNU-1709 • • - • • NSS N - - • • • - - NST V - - - NIS S - - KNU-1702 NSSS - • • • • - - • • • - - - - - - - - -</ns0:formula></ns0:div>
<ns0:div><ns0:head>South</ns0:head><ns0:formula xml:id='formula_1'>14PED96 N S VN/JFP1013 • • • • - • - - • • • - - - - - - - - - Tottori2 - - • • • • • • • • - - - - - - - NYT A - OKY-1 - - • • • • • • • • • - - - - - - - - - Japan IBR-7 • • - • • NSS N - - • • • - - - - - - - - - 1</ns0:formula><ns0:p>The high-specificity N-glycosylation sites and their amino acids are summarized, the representative strains from each country/regions are listed.' •' means the strain has this high-specificity N-glycosylation site; '-' means the strain has no this high-specificity N-glycosylation site; the amino acid sequences means the strain has a high-specificity N-glycosylation in this sits , but the amino acid sequences are different with the common sequences.</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:03:47007:2:0:NEW 13 Aug 2020)</ns0:p></ns0:div><ns0:figure xml:id='fig_0'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:03:47007:2:0:NEW 13 Aug 2020)Manuscript to be reviewed cells and determination of PEDV nucleic acids using RT-PCR, a PEDV strain was finally recovered and designated HB2018. The TCID50/0.1 mL value of HB2018 was 10 5.3 . The complete genome sequence of PEDV strain HB2018 was 28,138 bp in length. This 2.8-kb genome contained seven open reading frames (ORFs): ORF1a (nucleotide positions 281 to 12634), ORF1b (positions 12,664 to 20,625), S gene (positions 24,782 to 25,456), ORF3 (positions 25,675 to 25,667), E gene (positions 25,437 to 25,667), M gene (positions 25,675 to 26,355), and N gene (positions 26,367</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_1'><ns0:head /><ns0:label /><ns0:figDesc>(ZJUG12013, 85-7, 85-7-mutant 1, 85-7-mutant 2, 85-7-mutant 3, 85-7-mutant 4, 85-7mutant 5, 85-7-A40, and 85-7-C40) had a continuous deletion of 70 amino acids at their N-terminal (positions 1-70) of ORF3; two GI-a strains (85-7-mutant 2 and 85-7-mutant 4) had a continuous deletion of 47 amino acids at their C-terminal (positions 178-224) of ORF3; while nine GI-b strains (JS2008, AH-M, SD-M, SQ2014, SC1402, HLJBY, PEDV-SX, JSLS-12015 and JS-22015) had a continuous deletion of 133 amino acids at their C-terminal (positions 92-224) of ORF3 (Figure 4). Compared to ORF3 proteins of many GI and GII strains, ORF3 protein of a GII-a strain (CHSXYL2016) had a continuous deletion of 14 amino acids at positions 211-224; ORF3 protein of another GII-a strain (NW17) had a continuous deletion of 6 amino acids (DLYLAI) at positions 168-173; while ORF3 proteins of six GII-b strains (YN15, YN30, YN60, YN90, YN144, YN200) had a continuous deletion of 79 amino acids at their C-terminal (positions 146-224) (Figure 4). Compared to ORF3 proteins of the GII-a isolates, more than half of the GII-b isolates had amino acid changes at positions 25 (L→S), 70 (I→V), 80 (V→F), 107 (C→F), 168 (D→N), and 182</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_2'><ns0:head /><ns0:label /><ns0:figDesc>found in C domain of S protein, since the NTD and C-domain both can bind to the host cell receptor and function as the receptor-binding domain, the amino acid changes in their sequences may have important role for the virus<ns0:ref type='bibr' target='#b17'>(Li 2012)</ns0:ref>. It has been reported the N-linked glycosylation sites on the PeerJ reviewing PDF | (2020:03:47007:2:0:NEW 13 Aug 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_3'><ns0:head>Figure S1 :</ns0:head><ns0:label>S1</ns0:label><ns0:figDesc>Figure S1: RT-PCR detection of the viral nucleic acids from the intestinal samples of the pigs</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_4'><ns0:head>Figure S2 :</ns0:head><ns0:label>S2</ns0:label><ns0:figDesc>Figure S2: Sequence alignments on the whole genomes of HB2018 and CV777.</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_5'><ns0:head>Figure S3 :</ns0:head><ns0:label>S3</ns0:label><ns0:figDesc>Figure S3: Sequence alignments on the S proteins of HB2018 and CV777.</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_6'><ns0:head>Figure 1</ns0:head><ns0:label>1</ns0:label><ns0:figDesc>Figure 1</ns0:figDesc></ns0:figure>
<ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='28,42.52,229.87,525.00,324.75' type='bitmap' /></ns0:figure>
<ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='29,42.52,178.87,525.00,201.00' type='bitmap' /></ns0:figure>
<ns0:figure type='table' xml:id='tab_0'><ns0:head>Table 1 .</ns0:head><ns0:label>1</ns0:label><ns0:figDesc>Sequence comparisons of different ORF regions between HB2018 and CV777.</ns0:figDesc><ns0:table><ns0:row><ns0:cell>ORFs</ns0:cell><ns0:cell>HB2018 vs. CV777 Amino acid similarity (%)</ns0:cell><ns0:cell>DNA identity (%)</ns0:cell></ns0:row><ns0:row><ns0:cell>ORF1</ns0:cell><ns0:cell>97.73</ns0:cell><ns0:cell>97.11</ns0:cell></ns0:row><ns0:row><ns0:cell>S</ns0:cell><ns0:cell>93.44</ns0:cell><ns0:cell>93.76</ns0:cell></ns0:row><ns0:row><ns0:cell>ORF3</ns0:cell><ns0:cell>95.98</ns0:cell><ns0:cell>96.44</ns0:cell></ns0:row><ns0:row><ns0:cell>E</ns0:cell><ns0:cell>97.40</ns0:cell><ns0:cell>96.97</ns0:cell></ns0:row><ns0:row><ns0:cell>M</ns0:cell><ns0:cell>99.12</ns0:cell><ns0:cell>97.80</ns0:cell></ns0:row><ns0:row><ns0:cell>N</ns0:cell><ns0:cell>96.60</ns0:cell><ns0:cell>95.48</ns0:cell></ns0:row></ns0:table></ns0:figure>
<ns0:figure type='table' xml:id='tab_1'><ns0:head>Table 2 Single nucleotide polymorphism (SNP) analysis and dN/dS ratios of PEDV strains HB2018 and CV777.</ns0:head><ns0:label>2</ns0:label><ns0:figDesc /><ns0:table><ns0:row><ns0:cell /><ns0:cell>ORFs</ns0:cell><ns0:cell>Sum</ns0:cell><ns0:cell>Non-synonymous</ns0:cell><ns0:cell>Synonymous</ns0:cell><ns0:cell>dN/dS</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>Total</ns0:cell><ns0:cell>925</ns0:cell><ns0:cell>262</ns0:cell><ns0:cell>663</ns0:cell><ns0:cell>0.395</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>ORF1a</ns0:cell><ns0:cell>397</ns0:cell><ns0:cell>118</ns0:cell><ns0:cell>279</ns0:cell><ns0:cell>0.423</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>ORF1b</ns0:cell><ns0:cell>188</ns0:cell><ns0:cell>26</ns0:cell><ns0:cell>162</ns0:cell><ns0:cell>0.160</ns0:cell></ns0:row><ns0:row><ns0:cell>HB2018 vs.</ns0:cell><ns0:cell>S</ns0:cell><ns0:cell>233</ns0:cell><ns0:cell>88</ns0:cell><ns0:cell>145</ns0:cell><ns0:cell>0.607</ns0:cell></ns0:row><ns0:row><ns0:cell>CV777</ns0:cell><ns0:cell>ORF3</ns0:cell><ns0:cell>24</ns0:cell><ns0:cell>9</ns0:cell><ns0:cell>15</ns0:cell><ns0:cell>0.600</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>E</ns0:cell><ns0:cell>8</ns0:cell><ns0:cell>2</ns0:cell><ns0:cell>6</ns0:cell><ns0:cell>0.333</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>M</ns0:cell><ns0:cell>15</ns0:cell><ns0:cell>2</ns0:cell><ns0:cell>13</ns0:cell><ns0:cell>0.154</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>N</ns0:cell><ns0:cell>60</ns0:cell><ns0:cell>17</ns0:cell><ns0:cell>43</ns0:cell><ns0:cell>0.395</ns0:cell></ns0:row></ns0:table></ns0:figure>
<ns0:figure type='table' xml:id='tab_2'><ns0:head>Table 3 .</ns0:head><ns0:label>3</ns0:label><ns0:figDesc>High-specificity N-glycosylation sites predicted in Asian strains.</ns0:figDesc><ns0:table><ns0:row><ns0:cell>High-specificity N-glycosylation sites 1</ns0:cell></ns0:row></ns0:table></ns0:figure>
</ns0:body>
" | "Editor comments (Pedro Silva)
MINOR REVISIONS
Please address the final request from reviewer #2 ('It is still not clear why did the authors use the TN93 model. I assume that they used the “model selection” module of MegaX, mas this should be stated explicitly, as well as the criteria for choosing this model (BIC, AICc, etc).') I would also appreciate it if you could add the dN/dS and positive sites requested earlier by reviewer #3, which I believe would strongly improve your paper.
Response: Thank you for your comments. For the reason why we select the TN model. Actually, this model was developed for “estimating the number of transitional and transversional substitutions per site, as well as the total number of nucleotide substitutions” (PMID: 8336541). This model is a common selection in the “model/method selection” module of MEGA X, and it is widely used to perform phylogenetic analysis and generate phylogenetic trees in many genomic and epidemiological studies of bacteria or virus (e.g. PMID: 30834329; PMID: 11371577; PMID: 16485483; PMID: 21943222; PMID: 27241307; PMID: 12951272; etc.).
For the second requirement, we have added the contents associated with dN/dS and positive sites in our revised manuscript. Please see lines 175-185, 292-297, and the newly included Table 2.
Reviewer 1 (Babatunde Motayo)
Basic reporting
The authors have adhered to the journals reporting format and the paper is well structured. I have no comment.
Experimental design
No comment
Validity of the findings
No further comment
Comments for the Author
The article is well written the authors have satisfied all my concerns, no further comment
Response: We sincerely acknowledge your contributions on our work. Thank you very much again.
Reviewer 2 (Anonymous)
Basic reporting
Dear editor
Thank you for the opportunity to evaluate the new version of Liang et al manuscript. Overall, I think that the authors did a great job in the revision and that this I a much improved version of the study.
However, I have two brief considerations to make:
Experimental design
1) It is still not clear why did the authors use the TN93 model. I assume that they used the “model selection” module of MegaX, mas this should be stated explicitly, as well as the criteria for choosing this model (BIC, AICc, etc).
Response: Thank you for your comments. The “Tamura-Nei” model was developed for “estimating the number of transitional and transversional substitutions per site, as well as the total number of nucleotide substitutions” (PMID: 8336541). This model is a common selection in the “model/method selection” module of MEGA X, and it is widely used to perform phylogenetic analysis and generate phylogenetic trees in many genomic and epidemiological studies of bacteria or virus (e.g. PMID: 30834329; PMID: 11371577; PMID: 16485483; PMID: 21943222; PMID: 27241307; PMID: 12951272; etc.).
Validity of the findings
2) I didn’t like the splitting between Results and Discussion in the new version. First, I think that there is a lot of discussion going on in the Results section (I think that whenever you put your results in context considering the literature you are “discussing” them). Second, I think that combining results and discussion improved readability, without the problems of repeating ideas in both sections.
Response: Thank you for your comments. It is the editor’s requirement to separate the two sections.
" | Here is a paper. Please give your review comments after reading it. |
9,748 | "<ns0:abstract xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:p>Porcine epidemic diarrhea virus (PEDV) is a leading cause of diarrhea in pigs worldwide.</ns0:p><ns0:p>Virus isolation and genetic evolutionary analysis allow investigations into the prevalence of epidemic strains and provide data for the clinical diagnosis and vaccine development. In this study, we investigated the genetic characteristics of PEDV circulation in Asia through virus isolation and comparative genomics analysis. A PEDV strain designated HB2018 was isolated from a pig in a farm experiencing a diarrhea outbreak. The complete genome sequence of HB2018 was 28,138 bp in length. Phylogenetic analysis of HB2018 and 207 PEDVs in Asia showed that most PEDV strains circulating in Asia after 2010 belong to genotype GII, particularly GII-a. The PEDV vaccine strain CV777 belonged to GI, and thus, unmatched genotypes between CV777 and GII-a variants might partially explain incomplete protection by the CV777-derived vaccine against PEDV variants in China. In addition, we found the S protein of variant strains contained numerous mutations compared to the S protein of CV777, and these mutations occurred in the N-terminal domain of the S protein. These mutations may influence the antigenicity, pathogenicity, and neutralization properties of the variant strains.</ns0:p></ns0:div>
</ns0:abstract>
<ns0:body xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:head>INTRODUCTION</ns0:head><ns0:p>Porcine epidemic diarrhea (PED) is a high contagious and devastating disease resulting in the watery diarrhea in suckling pigs with high mortality and morbidity <ns0:ref type='bibr' target='#b46'>(Zhang et al. 2019)</ns0:ref>. The causative agent of PED, the porcine epidemic diarrhea virus (PEDV), is an enveloped, singlestranded, positive-sense RNA virus belonging to the genus Alphacoronavirus in the family</ns0:p><ns0:p>Coronaviridae <ns0:ref type='bibr' target='#b43'>(Woo et al. 2012)</ns0:ref>. PEDV possesses a 28-kb genome which encodes seven proteins including ORF1a, ORF1b, spike (S) glycoprotein, ORF3 hypothetical protein, envelop (E) protein, membrane (M) protein and nucleocapsid protein <ns0:ref type='bibr' target='#b11'>(Guo et al. 2019</ns0:ref>). Among these proteins, the S protein plays a key role in interaction between the virus and host cells. S protein consists of 1383amino acids <ns0:ref type='bibr'>(Aziz et al.)</ns0:ref>, and amino acid changes in S protein may lead to antigenic variations and affect the virus virulence <ns0:ref type='bibr' target='#b9'>(Gong et al. 2018;</ns0:ref><ns0:ref type='bibr' target='#b35'>Suzuki et al. 2018)</ns0:ref>. Therefore, this protein is commonly used as an important target for analyzing genetic variations and molecular epidemiology of PEDV <ns0:ref type='bibr' target='#b15'>(Hsueh et al. 2020</ns0:ref>). PED outbreaks have been reported continuously in China since 1973. PED was well controlled since administration of a CV777-derived vaccine <ns0:ref type='bibr' target='#b3'>(Chen et al. 2019a;</ns0:ref><ns0:ref type='bibr' target='#b40'>Wang et al. 2016a</ns0:ref>). However, recent outbreaks of PED in China since 2010 was due to the re-emergence of PEDV, and the continuous spread of the virus during the last 10 years has resulted in serious economic losses in the pig industry in Asian countries <ns0:ref type='bibr' target='#b45'>(Yang et al. 2013)</ns0:ref>. In these outbreaks, inactivated vaccines and attenuated live vaccines, which were derived from CV777, were used to control the disease but neither of them provided effective protection <ns0:ref type='bibr' target='#b33'>(Sun et al. 2012;</ns0:ref><ns0:ref type='bibr' target='#b48'>Zhou et al. 2012)</ns0:ref>. Moreover, the virus has evolved since 2010 <ns0:ref type='bibr' target='#b11'>(Guo et al. 2019;</ns0:ref><ns0:ref type='bibr' target='#b14'>Hsu et al. 2018;</ns0:ref><ns0:ref type='bibr' target='#b34'>Sun et al. 2019)</ns0:ref>, and acquisition of whole genome features of PEDV provides a convenient tool for the tracking of PEDV epidemiology <ns0:ref type='bibr' target='#b5'>(Chen et al. 2019b</ns0:ref>). In addition, virus isolation and genetic analysis allow Manuscript to be reviewed investigations on the prevalence of epidemic strains and will provide information for diagnosis and vaccine developments <ns0:ref type='bibr' target='#b19'>(Li et al. 2018)</ns0:ref>. In this study, we isolated a highly pathogenic PEDV strain HB2018 from a pig in a farm experiencing PED outbreaks in Hubei province, China, and determined its complete genome sequence. By comparing the HB2018 genome sequence with the sequences of 207 PEDV isolates circulating in Asia, which were publicly available in the Genbank data base, this study also aims to elucidate the evolutionary and genetic characteristics of PEDV currently circulation in different regions of Asia.</ns0:p></ns0:div>
<ns0:div><ns0:head>MATERIALS AND METHODS</ns0:head></ns0:div>
<ns0:div><ns0:head>Virus detection and isolation</ns0:head><ns0:p>In 2018, an outbreak of diarrhea occurred in a CV777-vaccinated pig farm (numbers of sows ≥ 100) in Hubei Province in China. Many pigs in the farm suffered from severe watery diarrhea, and some of them died. Samples of intestinal tissues were collected from dead pigs and sent to the Veterinary Diagnostic Laboratory of Hubei Academy of Agricultural Sciences in Wuhan, China, for diagnosis. Tissues were immersed with Dulbecco's modified Eagle medium (DMEM; Gibco, Grand Island, NY, USA), and were then homogenized using a QIAGEN TissueLyser II <ns0:ref type='bibr'>(QIAGEN, Dusseldorf, Nordrhein-Westfalen, Germany)</ns0:ref>. The sample homogenates were then frozen at −80 °C and thawed for three times. After that, the supernatants were filtered through a 0.22-μm membrane and were harvested for RNA and virus isolation. Total RNAs were extracted using TRIzol (Thermo, Waltham, MA, USA) and were reverse transcribed to cDNA using a Thermo Scientific First Strand cDNA Synthesis kit (Thermo, Waltham, MA, USA). Viral nucleic acids were detected by RT-PCR assays using the cDNA as templates and the primers specific for PEDV (F: 5'-TTCGGTTCTATTCCCGTTGATG-3', R: 5'-CCCATGAAGCACTTTCTCACTATC-3'), incubated at 37 ℃ with 5% CO 2 for 2 days. Neutralization titers were calculated as the reciprocal of the highest dilution of serum that inhibits CPEs.</ns0:p></ns0:div>
<ns0:div><ns0:head>Genome sequencing and annotation</ns0:head><ns0:p>Genomic RNA was extracted using the TAKARA RNA extraction kit (Takara, Kusatsu, Shiga, Japan) following the manufacture instruction. The quantity and quality of the extracted RNA were measured by using a Nanodrop spectrophotometer (Thermo, Waltham, MA, USA). The RNA was then subjected to reverse transcription for cDNA using a cDNA synthesis kit (Thermo, Waltham, MA, USA). Genome sequencing was performed with a paired-end library constructed by using a NEB-Next® DNA Library Prep Master Mix Set for Illumina (NEB, Ipswich, MA, USA)</ns0:p><ns0:p>and subsequently sequenced on an Illumina NextSeq 500 with 2 × 150 paired end sequencing chemistry. After filtering, the clean reads were assembled using SPAdes v3.10.1 <ns0:ref type='bibr' target='#b1'>(Bankevich et al. 2012</ns0:ref>) and assembled sequences were mapped to the reference genome. The prediction of the genes and proteins were conducted with Prokka v1.12 and RAST Serve (http://rast.nmpdr.org) <ns0:ref type='bibr' target='#b0'>(Aziz et al. 2008</ns0:ref>). The complete genome sequence as well as its annotations were deposited into NCBI GenBank under the accession number MT166307.</ns0:p></ns0:div>
<ns0:div><ns0:head>Comparative genomics and bioinformatical analysis</ns0:head><ns0:p>The NCBI data was search for 'porcine epidemic diarrhea virus' and a total of 207 complete genome sequences were publicly available for PEDV isolates representing different parts of Asia (See Table <ns0:ref type='table' target='#tab_0'>S1</ns0:ref> in supplementary materials). All of these 207 sequences were downloaded for further analysis. The average nucleotide sequence identity between the genomes of HB2018 and CV777 was calculated by ANI calculator <ns0:ref type='bibr' target='#b10'>(Goris et al. 2007)</ns0:ref>. Sequence alignments were performed using MAFFT v7.4.02 <ns0:ref type='bibr' target='#b16'>(Katoh & Standley 2013)</ns0:ref>. Nucleotide sequence similarity and the putative Manuscript to be reviewed recombination sites was assessed by SimPlot v.3.5.1 <ns0:ref type='bibr' target='#b23'>(Lole et al. 1999)</ns0:ref>, with a sliding window size of 500 bp, step size of 100 nucleotides, and 1,000 bootstrap replicates, using gap-stripped alignments and the F84 (ML) distance model. Phylogenetic trees based on complete genome sequences were generated by using MEGA X software with 1,000 bootstrapping <ns0:ref type='bibr' target='#b17'>(Kumar et al. 2018</ns0:ref>).The evolutionary history was inferred by using the Maximum Likelihood method and Tamura-Nei model <ns0:ref type='bibr' target='#b36'>(Tamura & Nei 1993)</ns0:ref>. Initial tree(s) for the heuristic search were obtained automatically by applying Neighbor-Join and BioNJ algorithms to a matrix of pairwise distances estimated using the Maximum Composite Likelihood (MCL) approach, and then selecting the topology with superior log likelihood value. The tree is drawn to scale, with branch lengths measured in the number of substitutions per site. A maximum likelihood tree was also generated using the BEAST 2 package (version 2.6.3) <ns0:ref type='bibr' target='#b2'>(Bouckaert et al. 2019)</ns0:ref>. Gamma correction for site heterogeneity and the GTR model <ns0:ref type='bibr' target='#b8'>(Gatto et al. 2007)</ns0:ref> were selected for the tree generation. Both of the trees were annotated and visualized by using the iTOL v.4 online tool (Interactive Tree of Life, http://itol.embl.de/) <ns0:ref type='bibr' target='#b18'>(Letunic & Bork 2019)</ns0:ref>. Single nucleotide polymorphisms (SNPs) between two genome sequences were determined by the MAUVE package (version 2.4.0) <ns0:ref type='bibr' target='#b7'>(Darling et al. 2004)</ns0:ref>, and the coding effect of these SNPs were analyzed using a previously reported local Perl command <ns0:ref type='bibr' target='#b25'>(Peng et al. 2016)</ns0:ref>. Protein structure was generated using SWISS-MODEL <ns0:ref type='bibr'>(Waterhouse et al. 2018)</ns0:ref>. Protein N-glycosylation sites were predicted using online software (http://www.cbs.dtu.dk/services/NetNGlyc/). Threshold values of greater than 0.5 and Jury agreement 9/9 were used for the high-specificity N-glycosylation sites determination <ns0:ref type='bibr' target='#b29'>(Sagesser et al. 1997)</ns0:ref>. Phylogenetic analysis based on the complete genome sequence showed that HB2018 was phylogenetically distinct from the vaccine strain CV777 (Figure <ns0:ref type='figure' target='#fig_13'>1A</ns0:ref>). According to the genotyping system based on a full-length genomic sequence analysis <ns0:ref type='bibr' target='#b11'>(Guo et al. 2019;</ns0:ref><ns0:ref type='bibr' target='#b40'>Wang et al. 2016a</ns0:ref>), HB2018 and CV777 belonged to two different genotype: HB2018 was assigned as a type GII strain while CV777 was a GI strain (Figure <ns0:ref type='figure' target='#fig_13'>1A</ns0:ref>). The average nucleotide identity between the genomes of HB2018 and CV777 (GenBank accession no. AF353511) was 96.06% (Figure <ns0:ref type='figure' target='#fig_14'>S2</ns0:ref> in supplementary materials). The ORF1, ORF3, E, M, and N genes of HB2018 as well as their encoding proteins were highly homologous to those of CV777 (nucleotide identity ≥ 95% for genes; amino acid similarity ≥ 95% for proteins) (Figure <ns0:ref type='figure' target='#fig_13'>1B</ns0:ref>; Table <ns0:ref type='table' target='#tab_0'>1</ns0:ref>). However, the identity of the S genes and proteins between the two strains was relatively low: the homology for nucleotide and amino acid sequences between HB2018 and CV777 were 93.76% and 93.44%, respectively (Figures <ns0:ref type='figure' target='#fig_13'>1B & 1C</ns0:ref>; Table <ns0:ref type='table' target='#tab_0'>1</ns0:ref>). SNP analysis determined a total of 946 SNPs in the genome sequence of HB2018 when compared to the genome sequence of the reference strain CV777. Among these Manuscript to be reviewed SNPs, 925 SNPs including 262 non-synonymous substitutions and 663 synonymous substitutions were located with the ORF regions, with an overall ratio of nonsynonymous to synonymous substitutions (dN/dS) of 0.39 (Table <ns0:ref type='table' target='#tab_1'>2</ns0:ref>). The dN/dS ratios in each of the ORFs encoded by the HB2018 genome ranged from 0.15 to 0.61, with the S protein had the highest dN/dS ratio (Table <ns0:ref type='table' target='#tab_1'>2</ns0:ref>).</ns0:p></ns0:div>
<ns0:div><ns0:head>RESULTS</ns0:head></ns0:div>
<ns0:div><ns0:head>Isolation of PEDV HB2018 and its genomic characteristics</ns0:head><ns0:p>Compared to the S protein of CV777, the S protein of HB2018 had changes, deletions, and/or insertions of amino acids at multiple sites (Table <ns0:ref type='table' target='#tab_1'>S2</ns0:ref> and Figure <ns0:ref type='figure'>S3</ns0:ref> in supplementary materials).</ns0:p><ns0:p>Notably, most of these mutations occurred in the N-terminal domain (NTD, 19-233aa) of the S protein (Figure <ns0:ref type='figure' target='#fig_13'>1C</ns0:ref>; Figure <ns0:ref type='figure'>S3</ns0:ref> in supplementary materials). Interestingly, some of these mutations were located within the neutralizing epitopes of PEDV [COE (499-638), SS2 (748-755), SS6</ns0:p><ns0:p>(764-771) and 2C10 (1368-1374)]. In addition, these mutations led to a structural change at some parts of the HB2018 S protein compared to the CV777 S protein (Figures <ns0:ref type='figure' target='#fig_13'>1D&1E</ns0:ref>).</ns0:p></ns0:div>
<ns0:div><ns0:head>Phylogenetic analysis of Aisan PEDV isolates</ns0:head><ns0:p>To explore the phylogenic relationships of the PEDVs currently circulating in Asia, we generated two maximum likelihood trees based on the whole genome sequences, either by using the MEGA X software with the Tamura-Nei model (Figure <ns0:ref type='figure' target='#fig_14'>2A</ns0:ref>) or by using the BEAST 2 package with the GTR model (Figure <ns0:ref type='figure' target='#fig_14'>2B</ns0:ref>). Both of the results revealed that the 208 PEDV strains in Asia, representing the 207 genome sequences publicly available in NCBI and the HB2018 sequence was divided into two genogroups: GI (classical) and GII (variant). Interestingly, isolates in China before 2010 and the vaccine strain CV777 were included within the GI genogroup. However, most of the PEDV isolates from China as well as the other Asian countries after 2010 belonged to GII genogroup (Figures 2A&2B). </ns0:p></ns0:div>
<ns0:div><ns0:head>Analysis on the S protein</ns0:head><ns0:p>Compared to the S proteins of the Chinese GI-a strains, amino acid changes, deletions, and/or insertions were observed at multiple sites within the S proteins of the Chinese GI-b strains (Table <ns0:ref type='table' target='#tab_2'>S3</ns0:ref> and Txt S1 in supplementary materials). Compared to the S proteins of the Chinese GI strains, the S proteins of the Chinese GII strains commonly had amino acid changes, deletions, and/or insertions at several sites (Table <ns0:ref type='table'>S4 and Txt</ns0:ref> (Table <ns0:ref type='table' target='#tab_2'>3</ns0:ref>). Among these sites, 212NVTS215, 777NISI780, and 1245NKTL1248 were conserved in almost all of the 208 Asian strains. However, the N-glycosylation sites at sites 57-60 of some strains were 'NSSS' rather than 'NSTW'. This is because the S proteins of these strains had the 59QGVN62 deletion compared to the S proteins of the other strains. Due to the amino acids changes, the N-glycosylation sites at sites 347-350 in some strains also changed from 'NSSD' to 'NSSN' or 'NSTN'. The 510NITV513 and 552NVTN555 N-glycosylation sites were missing in </ns0:p></ns0:div>
<ns0:div><ns0:head>Analysis of the ORF3-E-M-N proteins</ns0:head><ns0:p>Unlike the S protein, ORF3 is a conserved protein among the PEDV isolates <ns0:ref type='bibr' target='#b41'>(Wang et al. 2016b</ns0:ref>). However, several PEDVs were found to have characteristic amino acid mutations in the ORF3 (Figure <ns0:ref type='figure'>4</ns0:ref>). Compared to ORF3 proteins of many G1 strains and all GII strains, nine GI-a PEDVs (ZJUG12013, 85-7, 85-7-mutant 1, 85-7-mutant 2, 85-7-mutant 3, 85-7-mutant 4, 85-7- Manuscript to be reviewed (Q→H) (Figure <ns0:ref type='figure'>4</ns0:ref>). Interestingly, these amino acid changes were also found in the ORF3 proteins of many GII-c isolates after 2016. Sequence comparisons revealed that there were no common INDELs or mutations in E proteins of one subgroup of GII strains compared to E proteins of other subgroups of GII strains (Figure <ns0:ref type='figure'>5A</ns0:ref>). M proteins of most GII-a strains had a glutamine (Q) at site 13; however, all GII-b strains isolated between 2011 and 2012 had a glutamic acid (E) at the same position in their M proteins, and this amino acid change (Q→E) occurred frequently in M proteins of GII-b since 2013 (Figure <ns0:ref type='figure'>5B</ns0:ref>). A similar phenomenon was also observed in the M proteins of the GII-c strains, as most of the GII-c strains isolated before 2016 had a glutamine (Q) at position 13 in their M proteins, but a Q→E change at position 13 was seen in the M proteins of more frequently in strains isolated after 2016. In addition, amino acid changes at positions 192 (G→S) and 214 (S→A)</ns0:p><ns0:p>appeared simultaneously in M proteins of some GII-b and GII-c strains. Similarly, in N proteins, amino acid changes at positions 216 (M→V) and 241 (R→K) appeared simultaneously in many GII strains (Figure <ns0:ref type='figure'>5C</ns0:ref>).</ns0:p></ns0:div>
<ns0:div><ns0:head>Discussion</ns0:head><ns0:p>As an infectious virus attracted great intention, the PEDV strains were frequently reported and isolated in Asia. The virus isolation and genetic analysis will provide important information for PEDV research and vaccine developments. In this study, we isolated a GII-a strain HB2018 and determined its genomic characteristics (Figure <ns0:ref type='figure' target='#fig_13'>1A</ns0:ref>). Comparative genomic analysis revealed that the ORF1, ORF3, E, M, and N genes of HB2018 as well as their encoding proteins were highly homologous to those of CV777 (Figure <ns0:ref type='figure' target='#fig_13'>1B</ns0:ref>; Table <ns0:ref type='table' target='#tab_0'>1</ns0:ref>). However, a number of SNPs were Manuscript to be reviewed determined within these ORFs, with the S proteins showed the highest dN/dS ratio (Table <ns0:ref type='table' target='#tab_1'>2</ns0:ref>). Since dN/dS ratio is commonly used as a measure of purifying versus diversifying selection <ns0:ref type='bibr' target='#b28'>(Rocha et al. 2006</ns0:ref>), the highest dN/dS ratio suggests S protein is under diversifying selection, and this diversifying selection might be associated with its frequent interaction with host cells. Sequence alignments determined many mutations in the genome sequence of HB2018 compared to that of the reference strain CV777. These mutations, especially in the S protein, might be the pathogenic determinants for it, because some deletions and insertions in the S protein may change the antigenicity, pathogenicity and neutralization properties <ns0:ref type='bibr' target='#b3'>(Chen et al. 2019a;</ns0:ref><ns0:ref type='bibr' target='#b29'>Sagesser et al. 1997;</ns0:ref><ns0:ref type='bibr' target='#b47'>Zhang et al. 2015)</ns0:ref>. The presence of these mutations in the NTD of S protein in HB2018</ns0:p><ns0:p>might have an effect on the viral pathogenicity since the S-NTD domain is proposed to be the region relevant to the virulence of PEDV <ns0:ref type='bibr' target='#b13'>(Hou et al. 2017;</ns0:ref><ns0:ref type='bibr' target='#b30'>Su et al. 2018;</ns0:ref><ns0:ref type='bibr' target='#b31'>Su et al. 2019;</ns0:ref><ns0:ref type='bibr' target='#b35'>Suzuki et al. 2018</ns0:ref>). In addition, the structural changes led by these mutations in S protein of HB2018 might influence the immunogenicity. The distinct phylogenetic relationship between our isolation HB2018 and CV777 might partly explain why vaccination of pigs with CV777 did not provide effective protection against the infection of HB2018 in the vaccinated pig farm (Figures <ns0:ref type='figure' target='#fig_14'>2A&2B</ns0:ref>).</ns0:p><ns0:p>The analysis based on isolation years and genogroups of PEDVs in Asia might also revealed the vaccine CV777 did not match with the pandemic PEDV isolations. Before 2010, all the strains in China belonged to GI (Figures 2A&2B). During this period, PEDV was well controlled in China due to the use of CV777 which was the GI-based vaccine <ns0:ref type='bibr' target='#b3'>(Chen et al. 2019a;</ns0:ref><ns0:ref type='bibr' target='#b45'>Yang et al. 2013)</ns0:ref>.</ns0:p><ns0:p>The phylogenetic analysis of Asian PEDV isolates showed that most of the PEDV isolates from Manuscript to be reviewed Asia after 2010 belonged to GII genogroup, while the vaccine CV777 were included within GI genogroup. These findings agree with the results of the other studies <ns0:ref type='bibr' target='#b11'>(Guo et al. 2019;</ns0:ref><ns0:ref type='bibr' target='#b40'>Wang et al. 2016a)</ns0:ref>. The unmatched genotypes between CV777 and PEDV epidemic strains in Asia after 2010 could explain why vaccination with CV777 could not stop the outbreak of PED in many Asian countries after 2010 and provide effective protection against the current epidemic strains <ns0:ref type='bibr' target='#b3'>(Chen et al. 2019a;</ns0:ref><ns0:ref type='bibr' target='#b26'>Puranaveja et al. 2009;</ns0:ref><ns0:ref type='bibr' target='#b33'>Sun et al. 2012;</ns0:ref><ns0:ref type='bibr' target='#b48'>Zhou et al. 2012)</ns0:ref>.</ns0:p><ns0:p>With the most reported numbers of PEDV strains, China has more genogroups than other countries. The GII-c subgroup only consisted of isolates from China, these findings are also in agreement with previous studies <ns0:ref type='bibr' target='#b11'>(Guo et al. 2019;</ns0:ref><ns0:ref type='bibr' target='#b40'>Wang et al. 2016a)</ns0:ref>, suggesting that the genotypes of PEDV strains circulating in China might be more heterogeneous than those of the isolates in other Asian countries. These findings may also explain why PEDV vaccines developed in China contain more than one strains that generally include CV777 and at least one more local GII isolate (http://vdts.ivdc.org.cn:8081/cx/#). The new emerged PEDV in 2010 might accelerate numerous isolations and sequencing of PEDVs <ns0:ref type='bibr' target='#b22'>(Li et al. 2012;</ns0:ref><ns0:ref type='bibr' target='#b45'>Yang et al. 2013)</ns0:ref>. In this article, it was found that the number of PEDV sequences increased significantly after 2010 and most sequences were GII strains. These results are in good agreement with the findings of the PEDV epidemiological investigations in China <ns0:ref type='bibr' target='#b3'>(Chen et al. 2019a;</ns0:ref><ns0:ref type='bibr' target='#b32'>Sun et al. 2018)</ns0:ref>. It has been reported</ns0:p><ns0:p>that PEDV GII isolates were more virulent than GI isolates <ns0:ref type='bibr' target='#b39'>(Vlasova et al. 2014)</ns0:ref>. This might in part explain why the traditional vaccines had no to little effect on the control and spread of PEDV in China after 2010. It is noteworthy that PEDV GII strains are also responsible for the recent Manuscript to be reviewed outbreaks of PED in North America and Europe <ns0:ref type='bibr' target='#b6'>(Choudhury et al. 2016</ns0:ref>). These findings suggest the circulation of PEDV GII strains also pose a problem to the global pig industry. S protein is the most variable protein of PEDV, the amino acid changes in this protein may lead to virus variation and affect the virus virulence <ns0:ref type='bibr' target='#b9'>(Gong et al. 2018;</ns0:ref><ns0:ref type='bibr' target='#b35'>Suzuki et al. 2018)</ns0:ref>. The mutations between were found between GI-a strains and GI-b strains, it is still uncertain whether these mutations between them has a biological significance. While, the mutations occurred in S-NTD of the S protein between GI strains and GII strains might in part explain why do the PEDV GII isolates be more pathogenic than the GI isolates <ns0:ref type='bibr' target='#b39'>(Vlasova et al. 2014)</ns0:ref>, as S-NTD is proposed to be the region relevant to the virulence of PEDV <ns0:ref type='bibr' target='#b13'>(Hou et al. 2017;</ns0:ref><ns0:ref type='bibr' target='#b30'>Su et al. 2018;</ns0:ref><ns0:ref type='bibr' target='#b31'>Su et al. 2019;</ns0:ref><ns0:ref type='bibr' target='#b35'>Suzuki et al. 2018)</ns0:ref>. It is worthy note that PEDVs with insertions of amino acids at 167-168 and deletions of amino acids at 55-58 and 144 in their S proteins are called S-INDEL strains <ns0:ref type='bibr'>(Wang et al. 2014)</ns0:ref>. A previous study has found infection of the S-INDEL strains could induce proinflammatory cytokines through the non-canonical NF-κB signaling pathway by activating RIG-I; however, infection of the non-S-INDEL strains suppresses the induction of pro-inflammatory cytokines and type-I interferon production by down-regulation of TLRs and downstream signaling molecules <ns0:ref type='bibr' target='#b37'>(Temeeyasen et al. 2018)</ns0:ref>. Whether the continuous deletion of 194 amino acids occurred in Japanese strains will affect the virulence of these strains are unknown and warrant further exploration. A previous study however has found that a Japanese strain Tottori2, which had the same deletion, had non-lethal effects in piglets <ns0:ref type='bibr' target='#b24'>(Masuda et al. 2015)</ns0:ref>. The mutations also were Manuscript to be reviewed and function as the receptor-binding domain, the amino acid changes in their sequences may have important role for the virus <ns0:ref type='bibr' target='#b20'>(Li 2012)</ns0:ref>. It has been reported the N-linked glycosylation sites on the S protein of some coronaviruses such as SARS-CoV play a critical role in the viral entry <ns0:ref type='bibr' target='#b12'>(Han et al. 2007</ns0:ref>). The phylogenetic and N-linked glycosylation sites analysis of S protein may offer reasons for further studies. There was no too many mutations were found in the ORF3-E-M-N proteins, it might be because some of them, such as E protein, do not bear too much immune selective pressure since it has no effect on the host cell growth or cell cycle <ns0:ref type='bibr' target='#b44'>(Xu et al. 2013)</ns0:ref>.</ns0:p></ns0:div>
<ns0:div><ns0:head>CONCLUSIONS</ns0:head><ns0:p>In conclusion, through virus isolation and complete genome sequencing, we obtained PEDV HB2018 strain. Using this virus, we investigated the genetic and phylogenetic characteristics of PEDV isolates in China as well as in Asia in this study. Phylogenetic analysis revealed heterogeneous genotypes of PEDVs circulate in Asia, but GII particularly GII-a genotype represents the main epidemic genotype in the continent. Our study also revealed that most of the PEDVs currently prevalent in Asian countries displayed a different genotype as well as a distant relationship from the conventional vaccine strain CV777. This finding might explain why CV777derived vaccine provided poor protection against PEDV epidemics (variant strains) since 2010. In addition, we also identified many mutations in the S, ORF3, E, M, N proteins of the variant strains (GII) compared to those of the classical strains <ns0:ref type='bibr'>(Temeeyasen et al.)</ns0:ref>. The presence of these mutations, particularly those determined in the S proteins, may affect the antigenicity, pathogenicity, and neutralization properties of the variant strains.</ns0:p></ns0:div>
<ns0:div><ns0:head>ADDITIONAL INFORMATION AND DECLARATIONS</ns0:head><ns0:p>PeerJ reviewing PDF | (2020:03:47007:3:1:NEW 10 Sep 2020)</ns0:p><ns0:p>Manuscript to be reviewed Manuscript to be reviewed</ns0:p><ns0:formula xml:id='formula_0'>CV777 - NTS A • • - • • • • • - - - - - - - - - - HB2018 • • - • • • - - • • - - - - - - - - - - LZC - - • • • • • • • • • - - - - - - - - NLT G DR13 • • - • • NSS N - - • • • - - - - - - - - SD-M NSSS NTS A • • - • • • • • - - - - - - - - - - AJ1102 • • - • • • • - • • • - - - - - - - - - FJZZ1 • • - • • • • - • • • - - - - - - - - - LS - - - • • • - - • • • NVT R - - - - - - - CHS - • • • • • • • • • - - - NFT D - - - - - China CHHNQX- 314 - • • • • • - - • • • - - - - NLT A NCT E - - - CHYJ130330 • • • • NST N • - NIT I • • - - - - - - - - - CBR1 • • • • - • • - • • • - - - - - - - - - Thailand AVCT12 - - • • • • • • • • - - - - - - - - NYT A - Taiwan PT-P5 • • - • • NSS N • - • NKT R • - - - - - - - - KNU-1709 • • - • • NSS N - - • • • - - NST V - - - NIS S - - KNU-1702 NSSS - • • • • - - • • • - - - - - - - - -</ns0:formula></ns0:div>
<ns0:div><ns0:head>South</ns0:head><ns0:formula xml:id='formula_1'>14PED96 N S VN/JFP1013 • • • • - • - - • • • - - - - - - - - - Tottori2 - - • • • • • • • • - - - - - - - NYT A - OKY-1 - - • • • • • • • • • - - - - - - - - - Japan IBR-7 • • - • • NSS N - - • • • - - - - - - - - - 1</ns0:formula><ns0:p>The high-specificity N-glycosylation sites and their amino acids are summarized, the representative strains from each country/regions are listed.' •' means the strain has this high-specificity N-glycosylation site; '-' means the strain has no this high-specificity N-glycosylation site; the amino acid sequences means the strain has a high-specificity N-glycosylation in this sits , but the amino acid sequences are different with the common sequences.</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:03:47007:3:1:NEW 10 Sep 2020)</ns0:p></ns0:div><ns0:figure xml:id='fig_0'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:03:47007:3:1:NEW 10 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_1'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:03:47007:3:1:NEW 10 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_2'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:03:47007:3:1:NEW 10 Sep 2020) Manuscript to be reviewed RT-PCR detection of the viral nucleic acids revealed that the intestinal samples from pigs suffered and died from severe watery diarrhea were positive for PEDV but negative for TGEV and PoRV (Figure S1 in supplementary materials). Through virus isolation and purification using Vero cells and determination of PEDV nucleic acids using RT-PCR, a PEDV strain was finally recovered and designated HB2018. The TCID50/0.1 mL value of HB2018 was 10 5.3 . The complete genome sequence of PEDV strain HB2018 was 28,138 bp in length. This 2.8-kb genome contained seven open reading frames (ORFs): ORF1a (nucleotide positions 281 to 12634), ORF1b (positions 12,664 to 20,625), S gene (positions 24,782 to 25,456), ORF3 (positions 25,675 to 25,667), E gene (positions 25,437 to 25,667), M gene (positions 25,675 to 26,355), and N gene (positions 26,367 to 27,692).</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_3'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:03:47007:3:1:NEW 10 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_4'><ns0:head /><ns0:label /><ns0:figDesc>The phylogenetic trees also showed that the two genogroups consisted of several subgroups: the genogroup GI was divided into two subgroups, GI-a and GI-b, while the genogroups GII was divided into three subgroups, GII-a, GII-b, and GII-c (Figures 2A&2B).The GI-a and GI-b subgroups included isolates from China before 2010 and several Chinese isolates between 2010 and 2015 (Figures 2A&2B). Most of the GII isolates from China and South Korea and all GII isolates from Japan were included within the GII-a subgroup, while less proportion of the Chinese GII isolates and most of the GII-a isolates from Southeast Asia (Vietnam and Thailand) were included within GII-b subgroup (Figures 2A&2B).Interestingly, the GII-c subgroup only consisted of isolates from China (Figures 2A&2B).</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_5'><ns0:head /><ns0:label /><ns0:figDesc>By analyzing isolation years and genogroups of PEDVs, the history of PEDV and the evolution in China are speculated. Between 1986 and 2008, only five PEDV strains were sequenced in China, and all of them belonged to G1 (Figures 2A&2B).However, the number of PEDV sequences increased significantly after 2010 (Figure 2C). While several PEDV sequences belonged to genogroup GI after 2010, most sequences from China were GII strains (Figures 2A&2B&2C).</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_6'><ns0:head /><ns0:label /><ns0:figDesc>S1 in supplementary materials). Most these mutations occurred in S-NTD (19-233aa) of the S protein (Figure3A; Txt S1 in supplementary materials).Compared to S proteins of the Chinese GII-a strains, S proteins of most isolates from Japan, South Korea, and Vietnam did not contain characteristic amino acid mutations, with the exception of S proteins of two Japanese strains (NIG-2/JPN/2014 and KMM-1/JPN/2014) which had aPeerJ reviewing PDF | (2020:03:47007:3:1:NEW 10 Sep 2020) Manuscript to be reviewed continuous deletion of 194 amino acids at sites 23-216 (Figure 3B; Txt S2 in supplementary materials). In addition, four isolates from Japan (GenBank accession num. LC063844~LC063847) and three isolates from South Korea (KNU-141112-S DEL5, KNU-141112-S DEL5ORF3, KNU-1406-1) had a continuous deletion of 5 amino acids (GENQG) at sites 56-60 in their S proteins compared to the S proteins of the Chinese GII-a strains (Figure 3B; Txt S2 in supplementary materials). In addition to the GII-a strains, several isolates from China, South Korea, Thailand, and Vietnam were GII-b strains (Figure 1). Compared to S proteins of most of the Chinese GII-b strains, S proteins of the GII-b isolates from Thailand and Vietnam had amino acid changes at sites 130-131 (SI→DN), 182 (Y→H), 287 (I→M), 324 (N→D), 327 (S→A), 358 (A→T), 367 (I→T), 433 (D→G), 558-559 (TN→PT), 1287 (E→K), and 1317 (L→F) (Txt S3 in supplementary materials). The N-glycosylation sites in the S proteins of the Asian strains studied were investigated herein. Bioinformatical analysis revealed that most of the 208 Asian isolates contained 7~9 highspecificity N-glycosylation sites in their S proteins. When combining all high-specificity Nglycosylation sites determined and deleting the duplicates, eleven sites appeared in most isolates, including 57NSTW60, 112NATA115, 127NKTL130, 212NVTS215, 320NDTS323, 347NSSD350, 510NITV513, 552NVTN555, 777NISI780, 1245NKTL1248, and 1257NRTG1260</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_7'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:03:47007:3:1:NEW 10 Sep 2020) Manuscript to be reviewed S proteins of the isolates from South Korea and Vietnam. Due to the large deletion of amino acids at sites 23-216, the 57NSTW60, 112NATA115, 127NKTL130, and 212NVTS215 Nglycosylation sites were missing in S proteins of two Japanese strains NIG-2/JPN/2014 and KMM-1/JPN/2014, and seven N-glycosylation sites, including 130NDTS133, 157NSSN160, 494NVTS497, 549NCTE552, 587NISI590, 1055NKTL1058, and 1067NRTG1070 were retained in these two strains.</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_8'><ns0:head /><ns0:label /><ns0:figDesc>mutant 5, 85-7-A40, and 85-7-C40) had a continuous deletion of 70 amino acids at their N-terminal (positions 1-70) of ORF3; two GI-a strains (85-7-mutant 2 and 85-7-mutant 4) had a continuous deletion of 47 amino acids at their C-terminal (positions 178-224) of ORF3; while nine GI-b strains (JS2008, AH-M, SD-M, SQ2014, SC1402, HLJBY, PEDV-SX, JSLS-12015 and JS-22015) had a continuous deletion of 133 amino acids at their C-terminal (positions 92-224) of ORF3 (Figure 4). Compared to ORF3 proteins of many GI and GII strains, ORF3 protein of a GII-a strain (CHSXYL2016) had a continuous deletion of 14 amino acids at positions 211-224; ORF3 protein of another GII-a strain (NW17) had a continuous deletion of 6 amino acids (DLYLAI) at positions 168-173; while ORF3 proteins of six GII-b strains (YN15, YN30, YN60, YN90, YN144, YN200) had a continuous deletion of 79 amino acids at their C-terminal (positions 146-224) (Figure 4). Compared to ORF3 proteins of the GII-a isolates, more than half of the GII-b isolates had amino acid changes at positions 25 (L→S), 70 (I→V), 80 (V→F), 107 (C→F), 168 (D→N), and 182 PeerJ reviewing PDF | (2020:03:47007:3:1:NEW 10 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_9'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:03:47007:3:1:NEW 10 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_10'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:03:47007:3:1:NEW 10 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_11'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:03:47007:3:1:NEW 10 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_12'><ns0:head /><ns0:label /><ns0:figDesc>found in C domain of S protein, since the NTD and C-domain both can bind to the host cell receptor PeerJ reviewing PDF | (2020:03:47007:3:1:NEW 10 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_13'><ns0:head>Figure 1</ns0:head><ns0:label>1</ns0:label><ns0:figDesc>Figure 1</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_14'><ns0:head>Figure 2</ns0:head><ns0:label>2</ns0:label><ns0:figDesc>Figure 2</ns0:figDesc></ns0:figure>
<ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='28,42.52,229.87,525.00,324.75' type='bitmap' /></ns0:figure>
<ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='29,42.52,178.87,525.00,201.00' type='bitmap' /></ns0:figure>
<ns0:figure type='table' xml:id='tab_0'><ns0:head>Table 1 .</ns0:head><ns0:label>1</ns0:label><ns0:figDesc>Sequence comparisons of different ORF regions between HB2018 and CV777.</ns0:figDesc><ns0:table><ns0:row><ns0:cell>ORFs</ns0:cell><ns0:cell>HB2018 vs. CV777 Amino acid similarity (%)</ns0:cell><ns0:cell>DNA identity (%)</ns0:cell></ns0:row><ns0:row><ns0:cell>ORF1</ns0:cell><ns0:cell>97.73</ns0:cell><ns0:cell>97.11</ns0:cell></ns0:row><ns0:row><ns0:cell>S</ns0:cell><ns0:cell>93.44</ns0:cell><ns0:cell>93.76</ns0:cell></ns0:row><ns0:row><ns0:cell>ORF3</ns0:cell><ns0:cell>95.98</ns0:cell><ns0:cell>96.44</ns0:cell></ns0:row><ns0:row><ns0:cell>E</ns0:cell><ns0:cell>97.40</ns0:cell><ns0:cell>96.97</ns0:cell></ns0:row><ns0:row><ns0:cell>M</ns0:cell><ns0:cell>99.12</ns0:cell><ns0:cell>97.80</ns0:cell></ns0:row><ns0:row><ns0:cell>N</ns0:cell><ns0:cell>96.60</ns0:cell><ns0:cell>95.48</ns0:cell></ns0:row></ns0:table></ns0:figure>
<ns0:figure type='table' xml:id='tab_1'><ns0:head>Table 2 Single nucleotide polymorphism (SNP) analysis and dN/dS ratios of PEDV strains HB2018 and CV777.</ns0:head><ns0:label>2</ns0:label><ns0:figDesc /><ns0:table><ns0:row><ns0:cell /><ns0:cell>ORFs</ns0:cell><ns0:cell>Sum</ns0:cell><ns0:cell>Non-synonymous</ns0:cell><ns0:cell>Synonymous</ns0:cell><ns0:cell>dN/dS</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>Total</ns0:cell><ns0:cell>925</ns0:cell><ns0:cell>262</ns0:cell><ns0:cell>663</ns0:cell><ns0:cell>0.395</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>ORF1a</ns0:cell><ns0:cell>397</ns0:cell><ns0:cell>118</ns0:cell><ns0:cell>279</ns0:cell><ns0:cell>0.423</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>ORF1b</ns0:cell><ns0:cell>188</ns0:cell><ns0:cell>26</ns0:cell><ns0:cell>162</ns0:cell><ns0:cell>0.160</ns0:cell></ns0:row><ns0:row><ns0:cell>HB2018 vs.</ns0:cell><ns0:cell>S</ns0:cell><ns0:cell>233</ns0:cell><ns0:cell>88</ns0:cell><ns0:cell>145</ns0:cell><ns0:cell>0.607</ns0:cell></ns0:row><ns0:row><ns0:cell>CV777</ns0:cell><ns0:cell>ORF3</ns0:cell><ns0:cell>24</ns0:cell><ns0:cell>9</ns0:cell><ns0:cell>15</ns0:cell><ns0:cell>0.600</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>E</ns0:cell><ns0:cell>8</ns0:cell><ns0:cell>2</ns0:cell><ns0:cell>6</ns0:cell><ns0:cell>0.333</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>M</ns0:cell><ns0:cell>15</ns0:cell><ns0:cell>2</ns0:cell><ns0:cell>13</ns0:cell><ns0:cell>0.154</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>N</ns0:cell><ns0:cell>60</ns0:cell><ns0:cell>17</ns0:cell><ns0:cell>43</ns0:cell><ns0:cell>0.395</ns0:cell></ns0:row></ns0:table></ns0:figure>
<ns0:figure type='table' xml:id='tab_2'><ns0:head>Table 3 .</ns0:head><ns0:label>3</ns0:label><ns0:figDesc>High-specificity N-glycosylation sites predicted in Asian strains.</ns0:figDesc><ns0:table><ns0:row><ns0:cell>High-specificity N-glycosylation sites 1</ns0:cell></ns0:row></ns0:table></ns0:figure>
</ns0:body>
" | "Editor comments (Pedro Silva)
MAJOR REVISIONS
Reviewer #2's requests for model selection are, in my view, extremely relevant. I was also disappointed that many of the additional analyses (' e.g. recombination analysis, pairwise matrix, minority variants, dN/dS, positive selection sites, etc).' ) requested by reviewer #3 in their original review remain unperformed, even though I called attention to that omission in my previous decisions.
Response: Thank you very much, Dr. Silva, for both your and the reviewer’s constructive comments and suggestions. Following reviewer’s suggestion, we have generated another ML tree by using the BEAST 2 package with the GTR model and Gamma correction for site heterogeneity (the new Figure 2B in the revised manuscript), in addition to our ML tree by using the MEGAX software with the TN93 model (Figure 2A in the revised manuscript), even though the two tree are similar (Figure 2A vs. 2B in the revised manuscript). According to the review’s suggestions, we have also added those contents associated with the tree generation in the materials and methods section (lines 143-147 in the clean version), in the results section (lines 194-197 in the clean version), and the discussion section (lines 314 and 317). For the suggestions on of the additional analyses such as the SNP, the dN/dS, and the positive selection sites, we have added the required contents in the revised manuscript (clean version; see lines 147-150 in the materials and methods section, lines 178-184 and Table 2 in the results section, and lines 298-302 in the discussion section). We hope our revised manuscript meets the requirements this time.
Once again, we sincerely acknowledge your contribution to our manuscript.
Reviewer 2 (Anonymous)
Basic reporting
Dear editor/authors
Even though I think that the main conclusions of the paper are robust to phylogenetic model choice, I don't buy the explanation that it is ok to use a given model because others have use it before. Model selection in Mega X takes less than 5 minutes to run, and would give the authors a better justification for their selected model. GTR is also a widely used model. Gamma correction for site heterogeneity is also widely used. Because a given model is adequate for other datasets, it doesn't mean it is adequate for THIS dataset, and that's why model selection should be used in phylogenetic studies.
There are other, much more complex issues concerning model selection, which I don't think are relevant here. However, I still believe that a minimal justification is needed. MEGA X provides a relatively easy way of doing this. However, I leave the final decision for the editor as I don't think this impacts the conclusions of the study.
Your sincerely
Experimental design
no comment
Validity of the findings
no comment
Response: Thank you very much for your constructive comments and suggestions. We agree with all your suggestions. Therefore, in this revised manuscript, we have generated another ML tree by using the BEAST 2 package with the GTR model and Gamma correction for site heterogeneity (the new Figure 2B in the revised manuscript). The results revealed that the topology of the tree generated through this methodology was similar to the ML tree generated by using the MEGAX software with the TN93 model (Figure 2A in the revised manuscript). According to your suggestions, we have included both trees (Figures 2A&2B) in the revised manuscript, and have added those contents associated with the tree generation in the materials and methods section (lines 143-147 in the clean version), in the results section (lines 194-197 in the clean version), and the discussion section (lines 314 and 317).
Once again, we thank you for your work and contributions to our manuscript.
" | Here is a paper. Please give your review comments after reading it. |
9,749 | "<ns0:abstract xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:p>Nitrogen-fixers (diazotrophs) are often an important source nitrogen source to phytoplankton nutrient budgets in N-limited marine environments. Diazotrophic symbioses between cyanobacteria and diatoms can dominate nitrogen-fixation regionally, particularly in river plumes such as the Amazon River and in open ocean mesoscale blooms. This study reports the successful isolation and growth of multiple strains of a symbiotic cyanobacteria-diatom from the Gulf of Mexico in monocultures using a modified artificial seawater medium. We document the influence of light and nutrients on nitrogen fixation and growth rates of the host diatom Hemiaulus hauckii Grunow together with its diazotrophic endosymbiont Richelia intracellularis Schmidt, as well as less complete results on the Hemiaulus membranaceus -R. intracellularis symbiosis. These symbioses rates reported here are for the joint diatom-cyanobacteria unit.</ns0:p><ns0:p>Symbiont diazotrophy was sufficient to support both the host diatom and cyanobacteria symbiont, and the entire symbiosis replicated and grew without added nitrogen. Maximum growth rates of multiple strains of H. hauckii symbioses in N-free medium were 0.74-0.93 div d -1 . Growth rates followed light saturation kinetics in H. hauckii symbioses with a growth compensation light intensity (E C ) of 7-16 µmol m -2 sec -1 and saturation light level (E K ) of 84-110 µmol m -2 sec -1 . Nitrogen-fixation rates by the symbiont while within the host followed a diel pattern where rates increased from near-zero in the scotophase to a maximum 4-6 hours into the photophase. At the onset of the scotophase, nitrogen-fixation rates declined over several hours to near-zero values. Nitrogen fixation also exhibited light saturation kinetics, Maximum N 2 fixation rates (84 fmol N 2 heterocyst -1 h -1 ) in low light adapted cultures (50 µmol m -2 s -1) were approximately 40-50% of rates (144-154 fmol N 2 heterocyst -1 h -1 ) in high light (150 and 200 µmol m -2 s -1 ) adapted cultures. Maximum laboratory N 2 fixation rates were ~3 to 4-fold higher</ns0:p></ns0:div>
</ns0:abstract>
<ns0:body xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:p>Nitrogen-fixers (diazotrophs) are often an important source nitrogen source to phytoplankton nutrient budgets in N-limited marine environments. Diazotrophic symbioses between cyanobacteria and diatoms can dominate nitrogen-fixation regionally, particularly in river plumes such as the Amazon River and in open ocean mesoscale blooms. This study reports the successful isolation and growth of multiple strains of a symbiotic cyanobacteria-diatom from the Gulf of Mexico in monocultures using a modified artificial seawater medium. We document the influence of light and nutrients on nitrogen fixation and growth rates of the host diatom Hemiaulus hauckii Grunow together with its diazotrophic endosymbiont Richelia intracellularis Schmidt, as well as less complete results on the Hemiaulus membranaceus -R. intracellularis symbiosis. These symbioses rates reported here are for the joint diatom-cyanobacteria unit. Symbiont diazotrophy was sufficient to support both the host diatom and cyanobacteria symbiont, and the entire symbiosis replicated and grew without added nitrogen. Maximum growth rates of multiple strains of H. hauckii symbioses in N-free medium were 0.74-0.93 div d -1</ns0:p><ns0:p>. Growth rates followed light saturation kinetics in H. hauckii symbioses with a growth compensation light intensity (E C ) of 7-16 µmol m . Nitrogen-fixation rates by the symbiont while within the host followed a diel pattern where rates increased from near-zero in the scotophase to a maximum 4-6 hours into the photophase. At the onset of the scotophase, nitrogen-fixation rates declined over several hours to near-zero values. Nitrogen fixation also exhibited light saturation kinetics, Maximum N 2 fixation rates (84 fmol N 2 heterocyst ) adapted cultures. Maximum laboratory N 2 fixation rates were ~3 to 4-fold higher than literature-derived field rates of the H. hauckii symbiosis. In contrast to published results on the Rhizosolenia-Richelia symbiosis, the H. hauckii symbiosis did not use nitrate when added, although ammonium was consumed by the H. hauckii symbiosis. Symbiont-free host cell cultures could not be established; however, a symbiont-free H. hauckii strain was isolated directly from the field and grown on a nitratebased medium that would not support DDA growth. Our limited data raises the possibility that the asymbiotic H. hauckii are lines distinct from symbiotic H. hauckii with the latter part of an obligate symbiosis. While brief descriptions of successful culture isolation have been published, this report provides the first detailed description of the approaches, handling, and methodologies used for successful culture of this marine symbiosis. These techniques should permit a more widespread laboratory availability of these important marine symbioses.</ns0:p></ns0:div>
<ns0:div><ns0:head>Introduction</ns0:head><ns0:p>The phytoplankton flora of the open sea is a diverse assemblage of prokaryotic and eukaryotic cells that span a size range of ~1 to 2,000+ µm in diameter. Nitrogen is often a limiting nutrient in the open sea and planktonic nitrogen fixation(diazotrophy) occurs in many tropical and subtropical systems <ns0:ref type='bibr' target='#b20'>(Zehr 2011)</ns0:ref>. Diazotrophy occurs only in prokaryotic cells, but a variety of symbiotic associations between diazotrophic prokaryotes and host eukaryotes are known <ns0:ref type='bibr'>(Foster et al. 2006;</ns0:ref><ns0:ref type='bibr' target='#b21'>Foster & O'Mullan 2008;</ns0:ref><ns0:ref type='bibr' target='#b49'>Taylor 1982;</ns0:ref><ns0:ref type='bibr' target='#b55'>Villareal 1992</ns0:ref>) and cover the range from obligate symbioses to loosely associated consorts <ns0:ref type='bibr' target='#b8'>(Caputo et al. 2019;</ns0:ref><ns0:ref type='bibr' target='#b9'>Carpenter 2002;</ns0:ref><ns0:ref type='bibr'>Foster et al. 2006;</ns0:ref><ns0:ref type='bibr' target='#b21'>Foster & O'Mullan 2008)</ns0:ref>. Of these, diatom-diazotroph associations (DDAs) are the most visible with records dating back to the early 20 th century <ns0:ref type='bibr' target='#b40'>(Karsten 1905)</ns0:ref>. Two types of marine diatom-cyanobacteria symbioses are known: diatoms in the genera Neostreptotheca and Climacodium that host coccoid cyanobacteria <ns0:ref type='bibr' target='#b10'>(Carpenter & Janson 2000;</ns0:ref><ns0:ref type='bibr' target='#b30'>Hallegraeff & Jeffrey 1984)</ns0:ref>, and diatoms that host filamentous, heterocyst-forming cyanobacteria of the genera Richelia and Calothrix. Little is known about the characteristics of the coccoid symbionts. They can be found in a variety of diatoms as well as dinoflagellates and protozoa <ns0:ref type='bibr' target='#b8'>(Caputo et al. 2019;</ns0:ref><ns0:ref type='bibr'>Foster et al. 2006)</ns0:ref>. In some cases, the symbionts possess nitrogenase <ns0:ref type='bibr'>(Foster et al. 2006)</ns0:ref> and are capable of diazotrophy <ns0:ref type='bibr' target='#b20'>(Foster et al. 2011)</ns0:ref>. DDA symbioses involving heterocystous, diazotrophic cyanobacteria are abundant in both open ocean systems <ns0:ref type='bibr' target='#b13'>(Dore et al. 2008;</ns0:ref><ns0:ref type='bibr' target='#b57'>Villareal et al. 2011;</ns0:ref><ns0:ref type='bibr' target='#b63'>Wilson et al. 2008</ns0:ref>) and at intermediate salinities within the Amazon <ns0:ref type='bibr' target='#b22'>(Foster et al. 2007)</ns0:ref>, Mekong <ns0:ref type='bibr' target='#b28'>(Grosse et al. 2009</ns0:ref>) and Congo River plumes <ns0:ref type='bibr' target='#b23'>(Foster et al. 2009</ns0:ref>). These marine regions differ greatly in their characteristics, suggesting either a great plasticity in physiological responses to environmental variables or undocumented differentiation within these symbioses. Symbiont integration with the hosts varies as well. In the Manuscript to be reviewed</ns0:p><ns0:p>Rhizosolenia-Richelia DDA symbiosis, the symbiont is located in the periplasmic space between the frustule and plasmalemma and has limited contact with the external environment <ns0:ref type='bibr' target='#b38'>(Janson et al. 1999;</ns0:ref><ns0:ref type='bibr' target='#b52'>Villareal 1989</ns0:ref>). The Hemiaulus-Richelia DDA symbiont is appressed to the nucleus and truly intracellular <ns0:ref type='bibr' target='#b8'>(Caputo et al. 2019)</ns0:ref>, consistent with its reduced genome <ns0:ref type='bibr' target='#b33'>(Hilton et al. 2013)</ns0:ref>. The Chaetoceros-Calothrix DDA symbiont is completely extracellular to the host diatom <ns0:ref type='bibr' target='#b20'>(Foster et al. 2010)</ns0:ref>.</ns0:p><ns0:p>Despite their ubiquitous occurrence in tropical seas, the importance of Hemiaulus-Richelia symbioses were largely neglected due to the difficulty in seeing the symbiont until epifluorescence was a commonly available tool in oceanographic research <ns0:ref type='bibr' target='#b32'>(Heinbokel, 1986)</ns0:ref> and the documentation of N 2 fixation by the symbiont <ns0:ref type='bibr' target='#b54'>(Villareal 1991)</ns0:ref>. In addition to providing fixed N to the pelagic community, diatom-cyanobacteria symbioses also play an important role in the nitrogen and carbon cycles of oceanic systems by virtue of their potential to sequester carbon to the deep sea via aggregation and sinking <ns0:ref type='bibr' target='#b39'>(Karl et al. 2012;</ns0:ref><ns0:ref type='bibr' target='#b48'>Subramaniam et al. 2008</ns0:ref>). In the currency of oceanic nitrogen cycling, nitrogen derived from nitrogen fixation is generally balanced by a concurrent removal of atmospheric CO 2 <ns0:ref type='bibr' target='#b14'>(Eppley & Peterson 1979)</ns0:ref>. Thus, sinking material fueled by diazotrophy (either directly or indirectly) represents a net removal of CO 2 , and is a quantitatively important process in the transport of carbon to depth. DDAs, and particularly</ns0:p><ns0:p>Hemiaulus symbioses, are of particular oceanographic significance. Hemiaulus-Richelia symbioses bloom at ~ 10 3 cells L -1 frequently at the Hawai'i Ocean Time-series HOT <ns0:ref type='bibr' target='#b13'>(Dore et al. 2008;</ns0:ref><ns0:ref type='bibr' target='#b17'>Fong et al. 2008;</ns0:ref><ns0:ref type='bibr' target='#b45'>Scharek et al. 1999;</ns0:ref><ns0:ref type='bibr' target='#b61'>White et al. 2007)</ns0:ref>. At this location, they are the likely source of the summer export pulse that provides 20% of the annual carbon flux to 4,000 m in a 4-6-week window <ns0:ref type='bibr' target='#b39'>(Karl et al. 2012</ns0:ref>) and are regularly found on sinking particles <ns0:ref type='bibr' target='#b15'>(Farnelid et al. 2019)</ns0:ref>. Subtropical front blooms at ~28-30°N in the Pacific <ns0:ref type='bibr' target='#b51'>(Venrick 1974;</ns0:ref><ns0:ref type='bibr' target='#b58'>Villareal et al. 2012;</ns0:ref><ns0:ref type='bibr' target='#b62'>Wilson et al. 2004;</ns0:ref><ns0:ref type='bibr' target='#b63'>Wilson et al. 2008)</ns0:ref> and in waters west and north of HI <ns0:ref type='bibr' target='#b5'>(Brzezinski et al. 1998;</ns0:ref><ns0:ref type='bibr' target='#b57'>Villareal et al. 2011</ns0:ref>) suggest a basin scale significance. In the southwest Atlantic Ocean, Hemiaulus hauckii-Richelia blooms cover 10 5 + km 2 and sequester 1.7 Tmols of carbon annually <ns0:ref type='bibr' target='#b11'>(Carpenter et al. 1999;</ns0:ref><ns0:ref type='bibr' target='#b48'>Subramaniam et al. 2008)</ns0:ref> and CO 2 drawdown effects can extend to 10 6 km 2 <ns0:ref type='bibr' target='#b12'>(Cooley et al. 2007</ns0:ref>). The large size, chain-formation, and tendency to aggregate <ns0:ref type='bibr' target='#b45'>(Scharek et al. 1999;</ns0:ref><ns0:ref type='bibr' target='#b57'>Villareal et al. 2011)</ns0:ref> in the host Hemiaulus lead to an efficient export mechanism <ns0:ref type='bibr' target='#b64'>(Yeung et al. 2012</ns0:ref>) for both N and C. Culture studies on the growth and physiological characteristics of these symbioses are limited. The external symbiont Calothrix rhizosoleniae has been cultured without its host <ns0:ref type='bibr' target='#b20'>(Foster et al. 2010</ns0:ref>) in both natural and artificial seawater medium. Cultures of the Rhizosolenia-Richelia symbiosis using amended seawater has been reported in the literature with growth rates grows up to 0.8 div d -1 in N-free medium <ns0:ref type='bibr' target='#b53'>(Villareal 1990</ns0:ref>). In the Rhizosolenia-Richelia DDA growth of host and symbiont are uncoupled and symbiont-free host cells occur (but have reduced growth) even when no N is present, possibly through use of N excreted by Richelia into the medium. Addition of nitrate rapidly results in the loss of symbionts as the host uses the added nitrate to outgrow the symbiont <ns0:ref type='bibr' target='#b52'>(Villareal 1989;</ns0:ref><ns0:ref type='bibr' target='#b53'>Villareal 1990</ns0:ref>). Nitrogen fixation follows typical light saturation kinetics and can provide the entire N needs of the symbiosis <ns0:ref type='bibr' target='#b53'>(Villareal 1990</ns0:ref>). Although oceanographically more significant than other Rhizosolenia-Richelia DDA <ns0:ref type='bibr' target='#b32'>(Heinbokel 1986;</ns0:ref><ns0:ref type='bibr' target='#b48'>Subramaniam et al. 2008</ns0:ref><ns0:ref type='bibr' target='#b55'>, Villareal 1992)</ns0:ref>, there are no published culture-based data for the Hemiaulus-Richelia symbiosis.</ns0:p><ns0:p>Using nano-SIMS on field samples, <ns0:ref type='bibr' target='#b20'>Foster et al. (2011)</ns0:ref> were able to document the transport of recently fixed N is transported from the symbiont Richelia to the host Hemiaulus in sufficient quantities to support growth; however, it is not known whether Hemiaulus-Richelia can grow exclusively on diazotrophically fixed N. Regardless, the symbiont is clearly advantageous to the host since, where examined 80-100% of the Hemiaulus contain the symbiont <ns0:ref type='bibr' target='#b1'>(Bar-Zeev et al. 2008;</ns0:ref><ns0:ref type='bibr' target='#b32'>Heinbokel, 1986;</ns0:ref><ns0:ref type='bibr' target='#b54'>Villareal 1991;</ns0:ref><ns0:ref type='bibr' target='#b56'>Villareal 1994</ns0:ref>) and 85-100% of the total phytoplankton N needs in the Amazon River plume can be met by Hemiaulus DDA diazotrophy <ns0:ref type='bibr' target='#b11'>(Carpenter et al. 1999;</ns0:ref><ns0:ref type='bibr' target='#b59'>Weber et al. 2017</ns0:ref>). The symbiosis is not obligate for the host Rhizosolenia in DDA cultures <ns0:ref type='bibr' target='#b53'>(Villareal, 1990)</ns0:ref> and the field evidence suggests this may also be true for the host Hemiaulus <ns0:ref type='bibr' target='#b32'>(Heinbokel, 1986;</ns0:ref><ns0:ref type='bibr'>Kimor et al., 1978)</ns0:ref>. This latter hypothesis has not been tested due to the difficulty in growing the Hemiaulus-Richelia host-symbiont pair in vitro.</ns0:p><ns0:p>In this paper, we report the successful isolation of two species of the Hemiaulus-Richelia symbiosis into culture and expand on the brief culturing description reported in <ns0:ref type='bibr' target='#b46'>Schouten et al. (2013)</ns0:ref>. Using primarily H. hauckii-Richelia DDA strains, we document light-dependent growth rates, diel cycles of N 2 fixation, growth rate response to various forms of added nitrogen, and N 2 fixation rates. These parameters are essential to supporting modeling of DDA bloom formation and fate <ns0:ref type='bibr' target='#b16'>(Follett et al. 2018;</ns0:ref><ns0:ref type='bibr' target='#b47'>Stukel et al. 2014</ns0:ref>). In addition, key differences between the Hemiaulus and Rhizosolenia DDAs are noted. Manuscript to be reviewed text, a strain designation indicates a culture of a host diatom containing one or more symbionts.</ns0:p></ns0:div>
<ns0:div><ns0:head>Methods and Materials</ns0:head></ns0:div>
<ns0:div><ns0:head>Culturing</ns0:head><ns0:p>While Hemiaulus hauckii strain #91 was used for most of the experiments, a single strain for the entire suite of experiments was not possible due to loss of the strain or the periodic loss of vitality noted in the results. Strains used are identified in the text and in Table <ns0:ref type='table' target='#tab_2'>S1</ns0:ref>.</ns0:p><ns0:p>The Hemiaulus DDA could be isolated into MET-44 nutrient-amended sterile filtered seawater used for successful growth of the Rhizosolenia-Richelia symbiosis <ns0:ref type='bibr' target='#b53'>(Villareal 1990</ns0:ref>); however, this medium would not maintain the Hemiaulus-Richelia DDA for more than 1-2 weeks. The Hemiaulus-Richelia symbioses required re-isolation into the modified YBCII medium for successful maintenance >2-3 weeks. After isolation, cells were placed in a 25 °C incubator under cool white fluorescent illumination of 150-250 µmol m -2 sec -1 on a 12:12</ns0:p><ns0:p>Light:Dark (L:D) cycle. All cultures were grown as batch cultures. Cultures had a high rate of sudden decline and death when kept in medium longer than 7-10 days and careful attention was required to transfer the cultures to new medium within this time frame. Experiments were initiated within 6 months of culture isolation; cultures failed to make auxospores and were eventually lost after approximately 1-2 years in culture. No attempt was made to culture axenically; bacteria were rarely visible in the cultures until senescence when cell mortality was substantial. The H. hauckii DDA was the primary experimental tool. Hemiaulus membranaceus DDA cultures were examined for general characteristics but were not the subject of intensive experimentation. In March 2017, Hemiaulus chains were observed in the Port Aransas ship channel from the Imaging Flow Cytobot data stream <ns0:ref type='bibr' target='#b6'>(Campbell et al. 2010</ns0:ref><ns0:ref type='bibr' target='#b7'>(Campbell et al. , 2017))</ns0:ref>. Examination of net tow material noted numerous asymbiotic H. hauckii chains and no symbiotic cells. and H. membranaceus strain #82 utilized a high frequency time series approach in order to resolve changes occurring on an hourly basis or less. This approach used a series of individual measurements taken from a single vial over a period of up to ~12 hours and was utilized for two reasons. First, individual assays injections required 5-7 minutes to return to baseline. Triplicate measurements therefore required 15-21 minutes during which ethylene production was occurring at measurable rates, could not be consideration true replication of the ethylene measurement, and were unable to resolve rate changes on short time scales. The second was to minimize handling, agitation, and light/temperature variation of the sample. Six (H. hauckii) or 8 (H. membranaceus) paired vials were started at various time points in the diel cycle to permit overlap. Individual time series can be identified from the labelling in Table <ns0:ref type='table' target='#tab_2'>S1</ns0:ref>. Vials were sampled sequentially (1a, 1b, 2a, 2b, 3a,3b, then repeated) yielding approximately 1-1.5 hours between successive sampling of a single vial. The difference between successive measurements (ethylene per heterocyst) was normalized to the time difference between the two successive points (~1 -1.5 hours) and expressed as a rate (ethylene heterocyst -1 time -1 ). Eighty-nine (H. hauckii) and 78 (H.</ns0:p></ns0:div>
<ns0:div><ns0:head>Asymbiotic chains of</ns0:head><ns0:p>membranaceus) separate measurements were plotted against time using a 5-point running average (center point plus two on either side) to smooth the data. Rates from different vial series overlapped in time, thus the 5-point average has rates from independent time series.</ns0:p><ns0:p>Standard deviation was calculated on this 5-point series recognizing this is not a statistically useful value but only a metric for the noise in the data. Experimental cultures were adapted to at 25° C under 200 µmol m -2 s -1 illumination (cool-white fluorescence bulbs) on 12:12 LD cycle.</ns0:p><ns0:p>Experimental vials were incubated under these same conditions. Samples during the scotophase were collected/returned to the incubator in a darkened container and shielded from the dimmed laboratory lights during the assay.</ns0:p></ns0:div>
<ns0:div><ns0:head>Nutrient addition experiments</ns0:head><ns0:p>Nitrogen source experiments addressed the effect of various inorganic N sources on symbiosis growth and N 2 fixation. In these experiments, H. hauckii strain # 83 was transferred to three 2 L Manuscript to be reviewed to each day's run and at several points during the experiment. For each assay, 15 ml of culture sample was added to an acid-washed 25 ml incubation vial fitted with a grey chlorobutyl rubber serum stopper and crimped aluminum seals leaving 10 ml of headspace. Sterile-filtered medium was used as a control. A separate aliquot was retained for cell counts. One ml of acetylene generated from calcium carbide (Capone 1983) was introduced, gently swirled for 15-30 seconds to equilibrate while minimizing contact between the serum stopper and the culture, then 100 µL injected sampled with a Hamilton gas-tight syringe and injected into the GC. Each injection required 5-7 minutes after an injection to return to baseline.</ns0:p><ns0:p>Chlorophyll a was determined on methanol-extracted (24 hours, -20°C) samples (10-25 ml aliquot) collected on 0.4 µm pore size polycarbonate filters using a non-acidification method <ns0:ref type='bibr' target='#b60'>(Welschmeyer 1994)</ns0:ref>. Initial tests indicated the filters used did not leach fluorescent compounds in the methanol. When chl a cell -1 is referred to, it always includes both symbiont and host chl a.</ns0:p><ns0:p>Sample fluorescence was read on a TD-700 Fluorometer (Turner Designs, CA, USA).</ns0:p><ns0:p>For nutrients, a 25 mm, 0.22 µm pore-size membrane cellulose ester Millipore filter mounted on a syringe was rinsed with 5 ml of sample, filtrate discarded, and ten ml of sample medium filtered and frozen. A SEAL Analytical QuAAtro autoanalyzer was used to determine dissolved inorganic phosphate (DIP), nitrate +nitrite (N+N), ammonium (NH 4 + ), and silicate (SiO 4 -2 ) concentrations using the manufacturer's recommended chemistries. The chemistries are similar to automated analyses published in <ns0:ref type='bibr' target='#b27'>Grasshoff et al. (1999)</ns0:ref> with changes in reagent concentration and wetting agents specific to the manifold chemistries. Detection limits were ~0.05 µM for N+N, NH 4 + and P, and ~0.5 µM for Si.</ns0:p></ns0:div>
<ns0:div><ns0:head>Curve-fitting and Statistics</ns0:head><ns0:p>PeerJ reviewing PDF | (2019:08:40471:1:1:NEW 19 Jul 2020)</ns0:p><ns0:p>Manuscript to be reviewed</ns0:p><ns0:p>Light-dependent growth was fit to the Jassby-Platt hyperbolic tangent function (Jassby & Platt 1976) with a y-intercept term to permit calculation of compensation light intensity. The yintercept term was omitted for the N 2 fixation rates versus irradiance curves due to timedependent decline in dark N 2 fixation that became evident in the diel measurements. When not omitted, the time-dependent decline in dark N 2 fixation noted in the diel experiment at the beginning of the scotophase resulted in a highly variable initial slope as well as a significant yintercept (dark fixation rate) that was not consistent with the longer term rates after several hours in darkness. Delta Graph (Red Rocks Software) was used for graphics as well as curve fitting of the growth and N 2 -irradiance curves. T-tests were performed using the data analysis package in Microsoft Excel. Confidence intervals or standard deviations (noted in text) were calculated using Microsoft Excel software. Data from all figures are found in Table <ns0:ref type='table' target='#tab_2'>S1</ns0:ref>.</ns0:p><ns0:p>PeerJ reviewing PDF | (2019:08:40471:1:1:NEW 19 Jul 2020)</ns0:p><ns0:p>Manuscript to be reviewed</ns0:p></ns0:div>
<ns0:div><ns0:head>Results</ns0:head><ns0:p>Hemiaulus hauckii and Hemiaulus membranaceus with their symbiont Richelia intracellularis were successfully isolated repeated times. We found it was essential to remove the Hemiaulus from the net tow sample as quickly as possible (3-5 minutes after completion of the tow). Successful culturing resulted in rapidly growing chains of Hemiaulus reaching over 80 cells in length (Fig. <ns0:ref type='figure' target='#fig_11'>1</ns0:ref>). Multiple symbionts (usually 1-2, but never more than 4) were evident in the cells. Cultures were sensitive to handling, and swirling tubes to re-suspend chains resulted in chain breakage and decreased growth rates. Growth in undisturbed large volume containers (10 L+) resulted in complex aggregate formation. Strains were difficult to ship, and only one attempt out of approximately 15 resulted in successful establishment in another facility. A single auxospore-like structure was observed, but no cell diameter increases were observed in any of the cultures.</ns0:p><ns0:p>H. hauckii strains used in this study ranged from 12-17.5 µm (up to 30 µm observed) in diameter (pervalvar axis presented in broad girdle view) with a total cell volume range of 7,012 -23,574 µm 3 . H. membranaceus cells were not measured. Since auxosporulation did not occur, the strains gradually decreased in diameter over a period of 1-2 years and eventually died out.</ns0:p><ns0:p>Individual strains exhibited periods (weeks/months) of healthy growth (0.5-0.9 div d -1 ) with little care required. This growth pattern was interspersed with intervals (days/weeks) of low growth rates that required substantial attention and multiple backups to prevent loss of the culture. These cyclic patterns were not linked to batches of culture medium or glassware. While not enumerated, bacteria were rarely evident but certainly present since the cultures were not axenic.</ns0:p><ns0:p>Reasons for the observed growth pattern variability remain unknown.</ns0:p><ns0:p>Individual symbiosis strains were routinely maintained in modified YBC-II medium with no added nitrogen. High densities of Hemiaulus and its symbionts were possible with no N added to the synthetic seawater medium (residual combined inorganic N < 0.1 µM). Maximum cell counts of the host H. hauckii reached ~10,000 cells mL -1 with a maximum chl a concentration of 71 µg L -1 . Typical cell and chl a dynamics are shown in Fig. <ns0:ref type='figure' target='#fig_12'>2</ns0:ref>. High light (200 µmol m -2 s -1 ) chl a concentration reached a maximum approximately 3.5 times greater than the low light (50 µmol m -2 s -1 ) concentrations, although chl a per symbiosis (combined host and symbiont; multiple strains) remained approximately equal over time. In both light conditions, chl a per symbiosis was maximal (~4-5 pg chl a symbiosis -1 ) in early exponential growth and declined over time to ~2-3 pg chl a symbiosis -1 . Extensive chain formation resulted in a high degree of variation in measurements.</ns0:p><ns0:p>Growth rates of H. hauckii in N-deplete medium (Fig. <ns0:ref type='figure'>3</ns0:ref>) followed light saturation kinetics with host and symbiont growth rates highly correlated (r 2 = 0.98, p=0.05, t-test).</ns0:p><ns0:p>Photoinhibition was not observed at the maximum light level used (500 µmol m -2 s -1 ). A modified Jassby-Platt curve fit (Article S1) yielded a realized maximum growth rate µ from 0.74-0.93 div d -1 in replicated experiments (Fig. <ns0:ref type='figure'>3</ns0:ref>, Table <ns0:ref type='table' target='#tab_3'>1</ns0:ref>). Light saturated growth occurred with light-saturation (I k ) occurring at 84-110 µmol m -2 s -1 and an initial slope () of 0.009 div d -1 (µmol m -2 s -1 ) -1 in both irradiance curves. Compensation light intensity (I c ) calculated from the y-intercept and varied from 7-16 µmol m -2 s -1 . Nitrogen fixation rates estimated by acetylene reduction were tightly linked to the light:dark cycle (Fig. <ns0:ref type='figure' target='#fig_13'>4</ns0:ref>). The 5-point running average was necessary to smooth the variable Manuscript to be reviewed reduction rate in both H. hauckii and H. membranaceus DDA occurred approximately 4 hours into the photophase (12:12 photoperiod) with a broader maximum acetylene reduction rate extending for 4-6 hours. Acetylene reduction declined over several hours at photophase end to low (1-10% maximum values) but still measurable rates during the scotophase in both H. hauckii (Fig. <ns0:ref type='figure' target='#fig_13'>4a</ns0:ref>) and H. membranaceus (Fig. <ns0:ref type='figure' target='#fig_13'>4b</ns0:ref>) DDAs. Unlike the 4-hour discrete incubation diurnal pattern seen in Strain #22, H. hauckii strain #92 rates maintained high values until the end of the photophase (Fig. <ns0:ref type='figure' target='#fig_13'>4a</ns0:ref>). Hemiaulus membranaceus DDA rates were more symmetrically distributed around the middle of the photoperiod (Fig. <ns0:ref type='figure' target='#fig_13'>4b</ns0:ref>). In both data sets, the rates reached a maximum in the range of 45-55 fmol N 2 heterocyst -1 h -1 .</ns0:p><ns0:p>Nitrogen fixation-irradiance rates followed a light saturation curve (Fig. <ns0:ref type='figure'>5</ns0:ref>) fit to the hyperbolic tangent function. At the 150 and 200 µmol m -2 s -1 adaptation level (r 2 =0.95 and 0.97, respectively), the curve-fit maximum N 2 -fixation rates of 155 and 144 fmol N 2 heterocyst -1 h -1 , respectively. The maximum rates (light-saturated) at 150 and 200 µmol m -2 s -1 adaptation level were significantly (p<0.01, t-test) greater than the maximum (light-saturated) rate (86 fmol N 2 heterocyst -1 h -1 ) noted in cultures adapted to 50 µmol m -2 s -1 . The initial slope (light limited portion) of the N 2 fixation curve was approximately 75% higher in the 50 µmol m -2 s -1 adapted culture than the 150 and 200 µmol m -2 s -1 adaptation level. Preliminary experiments in 2010 found that H. hauckii strain #9 did not utilize nitrate (Table <ns0:ref type='table'>S2</ns0:ref>). Subsequent replication experiments found that 40 µM nitrate was not used by a different H. hauckii symbiosis strain (#83) in experiments conducted 1 year later (Fig. <ns0:ref type='figure' target='#fig_14'>6a</ns0:ref>). Ten µM added ammonium declined to ~0.4 µM in 13 days and then remained constant thereafter (Fig. <ns0:ref type='figure' target='#fig_13'>4b</ns0:ref>). Hemiaulus hauckii strain #83 drew down P and Si under all the available N sources at approximately equal rates. The addition of ammonium in an experimental comparison resulted in Manuscript to be reviewed higher percentages (up to 48%) of asymbiotic cells in exponential growth than when either nitrate (10-20%) was added or no N was present in the medium (10-20%) but the strain was not grown free of its symbiont.</ns0:p><ns0:p>A symbiont free strain of H. hauckii was maintained from March 2017 to August 2017 on a solely nitrate enriched, natural seawater medium. Ammonium concentrations in the aged stock seawater were 0.5 µM or less. It was periodically confirmed to be symbiont free by epifluorescence and maintained in a seawater-based culture medium (MET-44) that would not support the DDA strains. The strain was lost during Hurricane Harvey in August 2017 and no further information was collected.</ns0:p></ns0:div>
<ns0:div><ns0:head>Discussion</ns0:head><ns0:p>Physiology and rate measurements of Hemiaulus symbioses have previously been limited</ns0:p><ns0:p>to field collected and incubated samples. Using a modification of an artificial seawater medium, we have successfully and reproducibly cultured two species of Hemiaulus with its symbiont.</ns0:p><ns0:p>The inability to culture Hemiaulus in a seawater-based enrichment medium used for concurrent Rhizosolenia-Richelia cultures suggests that additional trace metal and chelation was required for sustained growth or that water quality issues are critical. <ns0:ref type='bibr' target='#b8'>Caputo et al. (2019)</ns0:ref> also reported brief success using an artificial medium. This difference highlights differing growth needs, sensitivities or tolerances of the Hemiaulus and Rhizosolenia DDAs that remain to be described. Greatest isolation success was found when the cells were rapidly removed from the net tow cod-end, suggesting sensitivity to the various exudates found in these concentrated sample.</ns0:p><ns0:p>In addition, the seawater was sterile filtered rather than autoclaved or pasteurized. Sterile filtration leaves the carbonate system and medium pH unaltered compared to heat treatment;</ns0:p><ns0:p>PeerJ reviewing PDF | (2019:08:40471:1:1:NEW 19 Jul 2020)</ns0:p><ns0:p>Manuscript to be reviewed however, viruses are not inactivated. Little is known of virus/DDA interactions, but viruses play a significant role in diatom mortality in general <ns0:ref type='bibr' target='#b42'>(Kranzler et al. 2019)</ns0:ref> and could be a problem for stable cultures. The oscillation between rapidly growing, apparently healthy cultures and less vigorous cultures is clearly an impediment to sustained culture as is the lack of auxospore formation. None of the isolations persisted for more than ~ 3 years making detailed work on model strains problematic at this time.</ns0:p><ns0:p>Previous estimates of N 2 fixation tracked 15 N isotope movement from the Richelia symbiont heterocysts to the host Hemiaulus cells using single-cell methods <ns0:ref type='bibr' target='#b20'>(Foster et al. 2011)</ns0:ref> and estimated that it was sufficient to support cell growth with a turnover time of up to 0.59 div <ns0:ref type='bibr'>Foster et al.'s (2011)</ns0:ref> rate measurements for H. hauckii-Richelia (n=17) averaged 20.4 ± 18.5 (std. dev.) fmol N heterocyst -1 h -1 (range 1.15-50.4 fmol heterocyst -1 h -1 ). These heterocyst normalized rates <ns0:ref type='bibr'>(Foster et al.'</ns0:ref>s Table <ns0:ref type='table' target='#tab_3'>1</ns0:ref> footnote), while lower, compare well to the rates observed in the cultures (up to 155 and 144 fmol N 2 heterocyst -1 h -1 ). Our culture growth rates (non-limiting conditions) suggest some degree of limitation in their field collections. Light history of the DDA cultures clearly affected rates with adaptation to 150-200 µmol m -2 s -1 PAR leading to a maximum N 2 fixation rate approximately 3 times higher than the maximum rates observed by <ns0:ref type='bibr' target='#b20'>Foster et al. (2011)</ns0:ref>. Data digitized (Plot Digitizer from SourceForge; http://plotdigitizer.sourceforge.net/) from Carpenter et al.'s (1999) Fig. <ns0:ref type='figure' target='#fig_12'>2</ns0:ref>, allows comparison of our symbiosis culture N 2 fixation rates to those from an Amazon River plume bloom where Hemiaulus DDA abundance reached 1.6 X 10 6 heterocysts L -1 . After extracting the N 2 fixation rates and concurrent heterocyst abundance from <ns0:ref type='bibr' target='#b38'>Carpenter et al.'s (1999)</ns0:ref> Fig. <ns0:ref type='figure' target='#fig_12'>2</ns0:ref>, we determined that their rates ranged from 0.6-40.3 fmol N 2 heterocyst -1 h -1 and were ~4 fold lower than the maximum rates seen in cultures.</ns0:p><ns0:formula xml:id='formula_0'>d -1 .</ns0:formula><ns0:p>PeerJ reviewing PDF | (2019:08:40471:1:1:NEW 19 Jul 2020)</ns0:p><ns0:p>Manuscript to be reviewed</ns0:p><ns0:p>These rates were generated from material collected by net and then prescreened to remove</ns0:p><ns0:p>Trichodesmium. Based on our isolation attempts, this handling probably adversely affected the rate. <ns0:ref type='bibr' target='#b11'>Carpenter et al. (1999)</ns0:ref> also reported undetectable nitrate uptake in the Hemiaulus DDA bloom, a result consistent with our culture observations that these DDA strains did not utilize nitrate.</ns0:p><ns0:p>Many growth characteristics of H. hauckii-R. intracellularis are similar to the Rhizosolenia clevei-R. intracellularis symbiosis. Maximum growth rates are slightly less than 1 div d -1 and are similar between the two DDAs despite their significant size difference. Growth rates are not photoinhibited up to 500 µmol photons m -2 s -1 . Rapidly growing cells form extensive chains. Culture agitation, albeit qualitatively measured, negatively affects chain formation and possibly growth rates. The diel pattern of nitrogen fixation in the Hemiaulus DDA cultures parallels the diel nifH nitrogenase gene expression seen in field samples of both Hemiaulus DDAs <ns0:ref type='bibr' target='#b65'>(Zehr et al. 2007</ns0:ref>), the Rhizosolenia DDA <ns0:ref type='bibr' target='#b31'>(Harke et al. 2019</ns0:ref>) and both gene expression and acetylene reduction in the Calothrix symbiont of the Chaetoceros DDA <ns0:ref type='bibr' target='#b20'>(Foster et al. 2010)</ns0:ref>.</ns0:p><ns0:p>The differential nitrate use by the Hemiaulus and Rhizosolenia DDAs is a significant difference between the two DDAs. Preferential NO 3utilization drove a higher host growth rate in a strain of the Rhizosolenia DDA, eventually leading to symbiont free host cultures <ns0:ref type='bibr' target='#b53'>(Villareal 1990</ns0:ref>) growing solely on NO 3 -. In field studies where N 2 appeared to be the primary N source, the Rhizosolenia host and symbiont DDA were tightly coupled <ns0:ref type='bibr' target='#b31'>(Harke et al. 2019)</ns0:ref>. There are at least two mechanisms that could produce this result in the Rhizosolenia DDA: downregulation of NO 3uptake. In contrast, ammonium was used and resulted in elevated percentages of symbiontfree hosts, but not a symbiont free culture. The free-living marine cyanobacterium</ns0:p><ns0:p>Trichodesmium can use NO 3either preferentially or concurrently during diazotrophy as an N source <ns0:ref type='bibr' target='#b35'>(Holl & Montoya 2005;</ns0:ref><ns0:ref type='bibr' target='#b41'>Klawonn et al. 2020;</ns0:ref><ns0:ref type='bibr' target='#b43'>Mulholland & Capone 2000)</ns0:ref> and other diazotrophs can simultaneously use N 2 and NO 3 - <ns0:ref type='bibr' target='#b36'>(Inomura et al. 2018</ns0:ref>). However, it is unusual for NO 3not to be used at all due to the higher overall energetic cost of nitrogen fixation added to the costs maintaining specialized cellular structures in diazotrophs <ns0:ref type='bibr' target='#b36'>(Inomura et al. 2018</ns0:ref>). The truly intracellular location of the Hemiaulus symbiont <ns0:ref type='bibr' target='#b8'>(Caputo et al. 2019)</ns0:ref> clearly limits its contact with the environment and the potential impact of NO 3 -, but our observations also require the host Hemiaulus to be unresponsive to external nitrate. In contrast, Hemiaulus spp. (no information on symbionts) has been reported with growth rates up to 2.2 div d -1 <ns0:ref type='bibr' target='#b26'>(Furnas 1991)</ns0:ref> in field experiments and 3.8 div d -1 in nitrate-based laboratory medium <ns0:ref type='bibr' target='#b3'>(Brand & Guillard 1981)</ns0:ref>. While the symbiont presence is undocumented but seems unlikely given the DDA growth rates reported in our paper of <1.0 div d -1 as well as by modelled symbiont diazotrophy <ns0:ref type='bibr' target='#b37'>(Inomura et al. 2020)</ns0:ref>. The <ns0:ref type='bibr' target='#b26'>Furnas (1991)</ns0:ref> and <ns0:ref type='bibr' target='#b3'>Brand & Guillard (1981)</ns0:ref> reports, as well as our briefly established asymbiotic strain on medium that would not support Hemiaulus DDA growth, all suggest that symbiont-free strains of Hemiaulus are extant in the modern ocean. Hemiaulus DDAs had an ancestral origin 50-100 million years ago Manuscript to be reviewed <ns0:ref type='bibr' target='#b8'>(Caputo et al. 2019)</ns0:ref>, but asymbiotic H. hauckii strains apparently still persist in the modern ocean. We suggest this data supports, but does not prove, that the Hemiaulus DDA, with its close metabolic coupling of the host-symbiont nitrogen metabolism <ns0:ref type='bibr' target='#b25'>(Foster & Zehr 2019;</ns0:ref><ns0:ref type='bibr' target='#b33'>Hilton et al. 2013)</ns0:ref>, is obligate and that symbiotic host Hemiaulus spp. are distinct from asymbiotic Hemiaulus strains. These asymbiotic strains should provide an invaluable tool for examining evolutionary processes in DDAs.</ns0:p><ns0:p>The growth rate and N 2 fixation results provide useful input to models examining the biogeochemical impact of the Hemiaulus DDA blooms in oceanic regions. The Amazon River plume is particularly noteworthy in that it has an explicit model describing the ecologicalbiogeochemical impacts. <ns0:ref type='bibr'>Stukel et al.'s (2014)</ns0:ref> model incorporated high generic N 2 -based DDA growth rates > 1 div d -1 with asymbiotic cells growing on ambient N at somewhat greater rates.</ns0:p><ns0:p>Our experimental results are much lower for growth on N 2 (maximum ~0.9 div d -1 ) and indicated no nitrate use. Non-diazotrophic, asymbiotic Hemiaulus growth rates from the literature are much higher than N 2 -based DDA rates. These are significant alterations in the input values available to <ns0:ref type='bibr' target='#b47'>Stukel et al. (2014)</ns0:ref>.</ns0:p><ns0:p>In addition, our results for H. hauckii DDAs found no evidence of growth rate photoinhibition at the highest light level used (500 µmol m -2 s -1 ). While instantaneous solar PAR may reach ~2,000 µmol m -2 s -1 <ns0:ref type='bibr' target='#b4'>(Björkman et al. 2015)</ns0:ref> <ns0:ref type='formula'>2011</ns0:ref>) for modelling applications. Our report presents the full range of data in that work and notes that rates can be ~ 0.2 div d -1 higher depending on the strain used. These higher rates are consistent with the mechanistic model of <ns0:ref type='bibr' target='#b37'>Inomura et al. (2020)</ns0:ref> in that host carbon fixation is substantial enough support to the symbiont N 2 fixation rates required for the unit DDA growth. This host derived carbon is likely to also be the reductant and energy source required to support the lengthy decline of N 2 fixation rates at the beginning of the scotophase noted in the diel experiment (Fig. <ns0:ref type='figure' target='#fig_13'>4</ns0:ref>). Further experimental verification is required. When comparing rates, the possibility of strain-specific variation between Foster et al. Results from the modified hyperbolic tangent function growth rate-irradiance and hyperbolic tangent function N 2 -fixation-irradiance curve fit.</ns0:p><ns0:p>The three N 2 -fixation experiments were adapted to the given light levels for 7 days. The growth rate experiments were adapted at each of the light levels for 7 days.</ns0:p><ns0:p>PeerJ reviewing PDF | (2019:08:40471:1:1:NEW 19 Jul 2020)</ns0:p><ns0:p>Manuscript to be reviewed 0.93 5 7 110 0.99 1 µmol m -2 s -1 2 (fmol N heterocyst -1 h -1 )(µmol m -2 s -1 ) -1 3 (div d -1 )(µmol m -2 s -1 ) -1 4 (fmol N heterocyst -1 h -1 ) 5 div d -1</ns0:p><ns0:p>PeerJ reviewing PDF | (2019:08:40471:1:1:NEW 19 Jul 2020)</ns0:p><ns0:p>Manuscript to be reviewed </ns0:p></ns0:div><ns0:figure xml:id='fig_2'><ns0:head /><ns0:label /><ns0:figDesc>filtered medium, the rinsed discarded, and then the tube filled with 15-20 mL of medium.Isolations were performed in a laboratory adjacent to the collection site. Multiple chains were rapidly isolated into sterile filtered seawater and transferred to a depression well containing 5-8 ml of medium. Using borosilicate pipets drawn to a fine diameter in a gas flame, these chains were then rinsed via serial transfer into wells containing sterile medium. Extensive rinsing (5-6 rinses) of a single chain before isolating the next chain was much less successful than 2-3 rinses. When isolated directly into the N-deplete modified YBCII medium, contaminant growth was minimal even with only 1-2 rinses.These techniques resulted in symbiosis isolation free of other eukaryotes or cyanobacteria in ~30-40% of the attempts. Preliminary experiments used strain #9 isolated during Spring 2010. Subsequent experiments used isolates established during Fall 2010 (strain #22) and Fall 2011 (strain #82, #83, #91, and #92). In all subsequent PeerJ reviewing PDF | (2019:08:40471:1:1:NEW 19 Jul 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_3'><ns0:head /><ns0:label /><ns0:figDesc>Hemiaulus hauckii were isolated into N-replete (40 µm NO3-) MET-44 amended sterile filtered seawater as noted above. Growth-rate and N 2 fixation versus irradiance experiments H. hauckii symbiosis strains #9 and #91, and H. membranaceus #82 were used for the irradiance-rate experiments. Initial experiments (Strain #9) used two light levels and are included for comparison. Detailed growth rates and N 2 fixation rates were measured in separate experiments using 7-8 different light levels (photosynthetic photon flux density) ranging from 15-600 µmol m -2 sec -1 measured by a QSP-170B irradiance meter (Biospherical Instruments;</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_4'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2019:08:40471:1:1:NEW 19 Jul 2020)Manuscript to be reviewed autoclaved glass Erlenmeyer flasks containing the maintenance medium listed above amended with one of the following nitrogen sources: no added nitrogen (control), added nitrate (40µM) or added ammonium (10µM). Samples were maintained at 25 °C and a salinity of 35. Reduced ammonium concentrations were used to avoid toxicity effects; the nitrate concentration duplicated work on the Rhizosolenia-Richelia symbiosis<ns0:ref type='bibr' target='#b52'>(Villareal 1989)</ns0:ref>. Nutrient concentrations and cell abundance were sampled 10 times throughout the duration of the 20-day experiment. Nutrient analyses and cell counts were done in duplicate.Analytical methodsCells were counted using a S52 Sedgewick-Rafter chamber on an Olympus BX51 epifluorescence microscope. Excitation/emission wavelengths for the epifluorescent filters used in counts and photography were 450 nm/680 nm (chlorophyll a), and 490 nm/ 565 nm (phycoerythrin). Both host cells and symbiont trichomes/heterocysts were enumerated. Percent symbiosis was calculated as the number of diatoms containing one or more Richelia trichomes divided by the total number of potential host cells. Growth rates were calculated using daily counts as the slope of the log of cell number over the change in time<ns0:ref type='bibr' target='#b29'>(Guillard 1973)</ns0:ref> with the 95% confidence interval around the slope of the line calculated in Microsoft Excel. Acetylene reduction assays (ARA) were performed as described inCapone (1993) corrected for ethylene solubility as described byBreitbarth et al. (2004) and assuming a mol ethylene reduced per mol N 2 conversion ratio of 4:1 (Jensen & Cox 1983 as modified by Capone 1993). An SRI 8610C gas chromatograph (SRI Instruments, Torrance, CA) equipped with a 30 cm silica gel column was used to quantify ethylene using a commercially prepared standard (GASCO Safeware Precision Gas Mixture, 10 and 100 ppm). Manufacturer-provided software (PeakSimple Chromatography Software) performed peak integrations. Standards were run prior PeerJ reviewing PDF | (2019:08:40471:1:1:NEW 19 Jul 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_5'><ns0:head /><ns0:label /><ns0:figDesc>point-to-point time series rates into a general diel curve. Two separate experimental treatments (the 4-hour incubations and the 5-point averaging series) indicated the maximum acetylene PeerJ reviewing PDF | (2019:08:40471:1:1:NEW 19 Jul 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_6'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2019:08:40471:1:1:NEW 19 Jul 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_7'><ns0:head /><ns0:label /><ns0:figDesc>symbiont diazotrophy by exposure to NO 3due to its extra-plasmalemma location and/or induction of host nitrate reductase pathways. The latter would result in diminished carbon flow PeerJ reviewing PDF | (2019:08:40471:1:1:NEW 19 Jul 2020)Manuscript to be reviewedto the symbiont in order to support nitrate assimilation into protein. Neither of these mechanisms appear to have occurred in the H. hauckii DDA strains we used. These results were replicated in individual experiments four years apart on different strains, excluding the possibility that the results were a laboratory condition artifact. For the Hemiaulus DDA, either nitrate cannot be used or diazotrophic supply exceeded any immediate N demand by the symbiosis and suppressed</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_8'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2019:08:40471:1:1:NEW 19 Jul 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_9'><ns0:head /><ns0:label /><ns0:figDesc>at Station ALOHA near Hawaii (22° 45' N 158° 00' W), average daily PAR incident at Sta. ALOHA over the diurnal is ~850 µmol m -2 s -1 from June-Aug. (calculated from Letelier et al. 2017). Vertical mixing rates will both reduce the time averaged PAR exposure exponentially with the depth of mixing and are rapid enough preclude general phytoplankton photoacclimation (Tomkins et al. 2020). Thus, it seems possible that in-situ PAR values would not photoinhibit these DDA strains. However, damaging effects by solar UV wavelengths (Zhu et al. 2020) require further examination. Follet al. (2018) and Inomura et al. (2020) utilized H. hauckii DDA growth rates extracted from Pyle (</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_10'><ns0:head /><ns0:label /><ns0:figDesc>Ocean collections, Carpenter et al.'s (1999) field collections and our Gulf of Mexico isolations cannot be excluded. Symbionts of the 3 diatom host clades have diverged and there is strong host specificity within the genera<ns0:ref type='bibr' target='#b24'>(Foster & Zehr 2006;</ns0:ref><ns0:ref type='bibr' target='#b38'>Janson et al. 1999</ns0:ref><ns0:ref type='bibr' target='#b1'>). Bar Zeev et al. (2008)</ns0:ref> noted evidence of seasonally varying Hemiaulus-DDA dominated Richelia clades in the Mediterranean but there is little data to assess how physiological characteristics vary with habitat. Rhizosolenia and Hemiaulus DDA symbionts appear limited to vertical exchange during division or possibly transmission during auxosporulation<ns0:ref type='bibr' target='#b25'>(Foster & Zehr 2019)</ns0:ref> raising the possibility of genetic drift of various degrees within populations (Bar-Zeev et al.2008).ConclusionsPeerJ reviewingPDF | (2019:08:40471:1:1:NEW 19 Jul 2020)Manuscript to be reviewedThe N 2 -fixation and growth rate data provided here are the first laboratory-based data for the Hemiaulus DDA. Two symbiotic associations between host diatoms and their intracellular heterocystous cyanobacterium (Hemiaulus hauckii -Richelia intracellularis and Hemiaulus membranaceus-Richelia intracellularis) have been successfully cultured. The culture methods provided here are based on an artificial seawater medium. The symbioses are sensitive to handling, requiring rapid collection and isolation for successful growth. Both symbioses grow without added nitrogen and are supported at maximum growth rates solely by symbiont nitrogen fixation. Maximum growth rates of the intact diatom-cyanobacterium symbiosis are < 1 div d -1 and are similar to the reported rates for another diatom-cyanobacterium symbiosis (Rhizosolenia clevei-Richelia intracellularis). Unlike the Rhizosolenia clevei-Richelia intracellularis symbiosis, the H. hauckii -Richelia intracellularis symbiosis does not assimilate nitrate.Nitrogen fixation by the heterocystous symbiont while within the host diatom has a clear diel pattern with maximum rates occurring during the photophase. Both growth and nitrogen fixation rates follow light saturation kinetics. The culture nitrogen fixation rates are consistent with field measured rates; however, maximum culture rates are 3-4 times field rates. Literature reports and isolation of nitrate-based Hemiaulus cultures are consistent with the existance both symbiont-free and symbiont-containing lines of the diatom Hemiaulus.Zehr JP. 2011. Nitrogen fixation by marine cyanobacteria. Trends in Microbiology 19:162-173. 10.1016/j.tim.2010.12.004 Zhu, Z., F. X. Fu, P. P. Qu, E. W. K. Mak, H. B. Jiang, R. F. Zhang, Z. Y. Zhu, K. S. Gao, and D. A. Hutchins. 2020. Interactions between ultraviolet radiation exposure and phosphorus limitation in the marine nitrogen-fixing cyanobacteria Trichodesmium and Crocosphaera. Limnology and Oceanography 65:363-376.</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_11'><ns0:head>Figure 1</ns0:head><ns0:label>1</ns0:label><ns0:figDesc>Figure 1</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_12'><ns0:head>Figure 2 Typical</ns0:head><ns0:label>2</ns0:label><ns0:figDesc>Figure 2</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_13'><ns0:head>Figure 4 .</ns0:head><ns0:label>4</ns0:label><ns0:figDesc>Figure 4</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_14'><ns0:head>Figure 6</ns0:head><ns0:label>6</ns0:label><ns0:figDesc>Figure 6</ns0:figDesc></ns0:figure>
<ns0:figure type='table' xml:id='tab_2'><ns0:head>Table S1 )</ns0:head><ns0:label>S1</ns0:label><ns0:figDesc>. For growth rates, cultures were grown at the 7 experimental light levels for 7 days and remained at the assigned light level through the duration of the experiments. Symbiosis</ns0:figDesc><ns0:table /><ns0:note>growth is used throughout this paper to refer to increases in host diatom numbers containing at least one symbiont. For N 2 fixation, strains were adapted to either 50, 150, or 200 (high light HL) µmol m -2 sec -1 at 25°C and a salinity of 35 under cool white fluorescent lighting for 7 days prior to the acetylene reduction assay. Each adaptation level was then exposed to 7-8 light levels for acetylene reduction assay Diel pattern of N 2 fixation H. hauckii strains #22 and #92, and H. membranaceus strain #82 were used for the diel study (12:12 L:D cycles at 200 µmol m -2 sec -1 ) examining the daily rhythm of N 2 fixation on culture medium with no added N. Initial experiments on H. hauckii strain #22 utilized a set of 6 discrete time points between 0600 to 2100. Each incubation lasted 4 hours with initial and final measurements taken in triplicate. Rates were normalized to heterocysts and used the center point of the 4 h incubation period as the time stamp. Subsequent experiments on H. hauckii strain #92</ns0:note></ns0:figure>
<ns0:figure type='table' xml:id='tab_3'><ns0:head>Table 1 (on next page)</ns0:head><ns0:label>1</ns0:label><ns0:figDesc /><ns0:table /></ns0:figure>
</ns0:body>
" | " 2 July 2020
Dear Editor,
Thank you for the chance to respond to the reviewers’ comments. As requested, I have addressed each point in the response using red italics. The comments have requested a considerable number of overlapping changes, particular the changes to the Methods section. Where this occurs, I have referred to the previous reviewer’s comments. The text is more cumbersome in places due to the need to carefully differentiate the host diatom Hemiaulus from the symbiosis Hemiaulus-Richelia (and several strains thereof). The confusion by the reviewers on this point on this was understandable when I re-read the paper. The methods have been extensively re-written to more clearly explain what we did. The Table S1 now includes the strains used in each figure as do the figure legends. The overall conclusions and data presentation has not changed. The extensive text editing and additions have resulted in a bit of a mess in the track changes version.
I appreciate the consideration of this manuscript and look forward to your response.
Regards,
Tracy Villareal
Corresponding Author.
Reviewer 1 (Anonymous)
Basic reporting
In general, the manuscript is written in clear, unambiguous, professional English. The intro and background show sufficient context and motivation for the experiments. This area of research is of great interest to the DDA community as there is very limited laboratory experimentation or published isolation methods for these important organisms. The figures were relevant although I note some comments regarding apparent inconsistencies (see line-by-line general comments). Raw data underlying individual figures was provided in a spreadsheet. It would be nice to know which strains were used with which raw data points, where applicable. I think this is mentioned in the text and figure legends but was hard to keep straight. Having it labeled in the figures and supplemental would help with clarity. Some data appears to be missing, for instance on Line 125 it is indicated that growth rates were measured at 7 different light levels, but much of this data is missing from the results and supplemental data. Perhaps this data was lost due to Hurricane Harvey?
The text on the light levels has been clarified that a range of 7+ light levels was used for each experiment. Each experiment has the data listed in the supplemental table; however, the individual light levels varied between experiments slightly due to bulbs aging, being replaced, or other minor changes between experiments in the walk-in incubator. The incubator failed several times over the course of the year+ required to perform the experiments and this created changes as well. The confusion may also be in part due to the preliminary data having only two light levels. This has been clarified. The text on strains used has been clarified although the major point was to note that a single strain could not be used for all the experiments. Since they are all dead and there is no archived material, there’s little extra information to be gained.
Experimental design
The research conducted herein is very relevant to the field and provides much-needed information on culturing, growth, n-fixation rates, and ecology of DDA’s formed by Hemiaulus-Richelia. In general, the methods were well written and relevant to the goals of the study. In some cases, however, the methods were a little hard to follow as written and some suggestions are listed below in the General Comments. In particular, a statement to the reason different strains were used would help clarify this point to readers. Also, more clarity as to which strains were used in which figure would help. A more detailed description of statistical methods to infer significance is needed.
Text has been added to clarify why different strains were used. All the reviewers commented on this and it has been discussed. The single statistical test used in the text has been identified as a t-test run in Excel.
Validity of the findings
As mentioned elsewhere, all underlying data for the figures are provided, but not all underlying experimental results mentioned in the methods. These data would be of interest to the readers I feel, assuming they were not lost in the hurricane. The discussion of results and intercomparing with prior field and laboratory observations, as well as cross genera comparisons, were robust and very well written. I truly enjoyed reading this manuscript. The manuscript will have a large impact on the field and is very relevant to our ecological understanding of these keystone organisms.
The experimental methods have been extensively re-written as per this general comment, specific comments to the author, and from other reviewers.
Comments for the Author
Lines 12 to 13 – looks like a space was between citations
Corrected
Line 35 – “Richealia” is spelled wrong
Corrected
Line 56 – there is a “cc” after “medium”
Corrected
Line 110 and 112 – it would be helpful to reviewers if a note is given as to why the use of different strains. In addition, rather than requiring the reader to go look at the figure legends or further in the text for environmental conditions of the experiments, it would be helpful if they were spelled out here as the N and light experiments are done. But maybe its just that sentence which is confusing.
This has been clarified. Most of the experiments were run with a single strain. However, preliminary experiments used a different strain and available data from other strains was added. The strains used are now included in the Table S1.
Line 134 – add ‘nm’ after 565
Done
Line 154 - italicize “a” in chlorophyll, here and elsewhere, also, how much volume for chl measurements?
Done
Line 159 – what is the 10 ml sample medium used for? I assume the nuts, but since it follows the method it reads as though it will be used for a different analysis. consider moving up.
Text has been added to clarify that this sample was for nutrients.
Line 160 – what stats were conducted with Excel? and on which data sets?
Excel was used to perform a t-test using the data analysis package. This has been added to the text.
Line 191 – I think a period is supposed to be here after the volume range
Corrected
Line 197 – Perhaps heterotrophic bacteria? Was any observations made on their numbers?
I’ve added some text here noting that bacteria were not apparent in the cultures, although we did not specifically enumerate them.
Line 205 – chlorophyll is now abbreviated where on line 204 it was not
I’ve changed chlorophyll a to chl a in the text except where it starts the sentence.
Line 236 – what stat?
t-test in Excel, added in text and methods
Line 242 – were these cultures axenic? If not, how might heterotrophic remineralization impact measured N+N values?
The cultures were not axenic. We only added nitrate in one experiment and at high concentration. It’s difficult to see how nitrification in oxic cultures could occur at sufficient rates to maintain the steady concentration seen over time. In other cultures that were part of this experiment, N+N buildup never occurred, so it seems reasonable to assume that no nitrification was occurring. Remineralization certainly could have been occurring, but there’s little evidence that it would be significant when 99% of bacteria cannot be cultured.
Figure 4 – How are the individual isolates represented in the figure? Are these the black squares? Not sure I see diamonds. If so, consider either using different colors to designate strain or additional shapes. If red squares are a 4 h incubation, how do the points span from 8am to 8pm?
The two strains are separated by color. 92 is black and 22 is red. The text has been reworded to include this. The text referring to diamonds has been removed. This was from an earlier version of the figure that included the raw data. That figure was too busy, so we removed the preliminary data but missed changing the figure legend. The methods text has been revised to address how the rates were calculated. The red squares (now red triangles for clarity) were from multiple discrete 4 hour incubations.
Figure 6 – Where is the 3rd N source? Or is this the -N? No closed squares are visible. If strain #9 is here, it should be called out in the figure as well as mentioned in the methods for this experiment. Also, legend says only 1 um of NH4 was added whereas methods and results state 10um.
The legend concentration has been corrected. The 3 N sources (N2, nitrate, ammonium) are now identified in the legend and are currently identified in the methods. Closed squares are not mentioned in the figure legend or on the figure so I don’t know how to answer this. They may be referring to seeing only two symbols in panel A and B. The near zero values for two measurements are superimposed and only the X is visible. Strain 9 was used in the preliminary experiment in 2010 (REU student) that identified this phenomenon of no nitrate use. The data is not shown on this figure. I have clarified this in the text.
Reviewer 2 (Anonymous)
Basic reporting
The language in this manuscript is clearly written. Take a closer look at the use of hyphens for compound adjectives. Because of the spotty use of some strains, communicating the experimental design sometimes became muddled. Comparisons to Rhizosolenia early in the manuscript led me to believe that there would be data presented from this strain, so I was left confused. The figures are clear, but in some cases require a legend or the captions need review (see specific comments below). Data are provided and I was hoping it could clarify questions I had regarding the N2 fixation calculations, but I now have more questions.
The reference to Rhizosolenia was only to put in context the limited information on these symbioses. It’s the only one to have been cultured as a symbiosis and is mentioned in only a few paragraphs in a 3 page introduction.
Experimental design
Isolating diatom-diazotroph associations from the field requires a valiant effort and I applaud you. The ephemeral nature of these cultures and the loss of cultures to Hurricane Harvey prevented the authors from conducting these experiments as thoroughly as they might want. The experimental design was unfortunately patchy, but there are important results here nonetheless. I have concerns about how the acetylene reduction assay was sampled and calculated. Unfortunately, the data provided did not answer them.
The methods have been reworked to describing the acetylene reduction assay.
Validity of the findings
So many strains have been lost that there is no opportunity to replicate the experiments. The nitrogen addition experiments seem robust. If there are any DNA or RNA samples stashed away, they could be helpful in confirming some conclusions by determining whether the cultures were affected by viruses or whether host and asymbiotic strains had diverged.
I am trying to understand how the N2 fixation numbers were calculated and reported. It is a well-documented assay so I don't think the methods need to be extended, but perhaps an explanation to the editors how the data were handled might be helpful.
I truly wish we had archived material but we have none. As noted above, I have added extra text to the methods regarding the acetylene reduction assay.
Comments for the Author
General comments:
It’s difficult to keep track of which species and strains were used in what experiment even though they are enumerated in the text. Please add a table listing the strains used and the measurements performed on each.
This information is now listed in the Table S1. Each figure legend lists which strains were used.
The culture isolation techniques described in the Results were very interesting, but I think they would be better served in the methods section. State what you learned, but leave the details in methods.
I have moved several points into the methods and reduced these two paragraphs to one. I note that these are details rarely, if ever, mentioned in culturing documentation. However, they were absolutely critical in establishing the cultures and having a context for understanding both the results and data from the field.
One confusing aspect of the paper is that Rhizosolenia is mentioned occasionally, but wasn’t used in the experiments. If there are no results to report, perhaps this paper should be written more clearly as a Hemiaulus paper.
Rhizosolenia-Richelia is the only DDA that has any lab-based culture rates available for the group. It was necessary to mention in this regard. The last paragraph of the Introduction clearly notes the focus of the paper as does the entirety of the abstract. I have tried to reword the text elsewhere to accommodate this confusion.
Italicize “a” in chlorophyll a.
Done.
Specific comments:
Abstract:
- No hyphen is required for “nitrogen fixers” or “nitrogen fixation.” Do use a hyphen for the compound adjective “nitrogen-fixing,” however. Check the use of hyphens here and throughout.
Done
- The phrase “a cyanobacteria-diatom symbiosis” sounds as if you only isolated one DDA but then you report two Hemiaulus species.
I have corrected this as well as several other places where the distinction between symbiosis and symbioses was possibly confusing. It became a difficult collective noun usage that I have tried to simplify where I can. American and British English differ on this, something I found interesting.
- The first time you report a species, spell out the entire genus, even if you’ve already used the genus for another species.
This appears to be an uncommon stylistic usage. According to Butcher’s Copy-editing, abbreviating the genus name for a second congener is acceptable. It can be part of journal’s stylistic requirements, so while I don’t see that in the PeerJ instructions, I’ve made the two changes to avoid a meaningless squabble.
- Hypenate “symbiont-free.”
Typo has been corrected.
Introduction:
-Line 35: “Richelia” is misspelled.
Corrected
-Line 56: Delete “cc” from “medium.”
Done
-Line 64: This entire paragraph is about Hemiaulus, so it’s better if you don’t bring in new information about Rhizosolenia.
The Rhizosolenia information is context for the nature of DDA symbioses. It sets up the hypothesis that we evaluate in the culture results.
Methods and Materials:
Lines 86 - 87: Use plural for “strains” when you list multiple strains.
In the review PDF, the word “strain” or “strains” is not used in these lines other than part of a designation (i.e. strain #82). I do not know how to address this other than look for singular-plural useage in general.
Lines 114 - 115: Were the cultures acclimated to the 2-L flasks or did you transfer them straight from the 50-mL tubes to the flasks?
They were transferred directly as far as I know, although it may have taken most of the tube. Since we measured the nutrient concentrations daily, there was no issue with any carryover (if that is the concern).
Lines 125 - 126: It is not clear to me which light levels were used in the spectrophotometer and which light levels were used to grow the cultures. Some of my comments elsewhere reflect that confusion. In addition, did you determine an optimal light level for each strain?
We did not mention a spectrophotometer anywhere in the paper, so this part of the comment is confusing. The light levels for maximum growth rate were determined from the results of this experiment and are discussed in the Results. As noted in the response to Reviewer #1, the text describing the light levels used has been modified to reflect the reality of how the incident light changed over time due to age and use of different incubators.
Line 139: Why the hyperbolic tangent? Is it the model that fit your data best? The hyperbolic tangent model is purely empirical with no theoretical biology supporting it. If it suits your data best, then use it. But if you haven’t tried one of the exponential functions, they can be derived from first principals. See Aalderink and Jovin, 1997, Journal of Plankton Research 19: 1713 - 1742.
We appreciate the reviewer’s perspective on this and agree that there are numerous models that can be used to fit data that has a light dependent slope and a saturation phase. In our case, we used an empirical model that has long historical use for its simplicity and generality to this type of curve. The reality is that modelers using our data will recast it in their favorite parameterization. In terms of the biology, while both the host and symbiont are photosynthetic (photosynthesis is the rate that Aalderink and Jovin paper describe), we did not measure photosynthesis. Our rate terms are growth rate and N2 fixation, so the first principals approach is not quite as clear as implied, particularly in a two organism system. It is certainly an interesting question, but far beyond the scope of the paper.
Line 148: How many different start times did you use in a 24-hour period?
This entire section has been rewritten. The start times are now identified in Table S1. The time points are identified with one of the 6 (H. hauckii) or 16 (H. membranaceus) individual time series used in the experiments
Line 155: “Nucleopore” is a brand. I presume these are polycarbonate? PC has a dye that extracts in methanol. Did you use a filter in your blank measurements?
This has been clarified in the methods. Our tests did not show any extractable fluorescence in the polycarbonate filters. However, these were from filters made nearly 10 years ago and the manufacturing may well have changed since then. In addition, the Welschmeyer non-acidification method uses slightly different wavelengths than the original acidification method. This may be part of the difference as well.
Line 160: What kind of filter? What material was it made from?
Cellulose acetate. The filter was initially rinsed with 5 ml of sample to remove contaminants. This text has been clarified.
Results:
Lines 174 - 175: Do you know what kinds of contaminants you had in your isolates? Were there phytoplankton other than the target diatom and symbiont cyanobacteria?
Text has been added to clarify this. There were no other eukaryotes or cyanobacteria. The cultures were not axenic nor did we think it wise to try and grow them that way due to the likely effects on antibiotics on the host and symbiont. All the results would have been called into question.
Lines 190 - 191: Did you measure cell height?
It must have been to generate cell volume, but I have no record of it. Nor do I have a record of the transapical axis dimension. Some pervalvar axis dimensions can be measured from the figure.
Line 206: “Both” light conditions? You listed 7 levels in your methods.
In the context of this paragraph and figure, only two light levels were being reported on for the chlorophyll: high light (200 µmol quanta m-2 s-1) and low light (50 200 µmol quanta m-2 s-1). This is noted in the figure legend, but I’ve added some text for clarity.
Lines 208 - 210: In the graph it seems that HL symbioses had more chl a than LL. This seems counter-intuitive to me when considering photoacclimation. What do you think is going on?
The chl per cell data shown in Fig. 2b does not support this statement. The chl cell-1 values are quite similar. It may be the reviewer is referring to Fig. 2a which shows total chl L-1, not chl cell-1. This is simply the result of a higher growth rate leading to more biomass. This is mentioned in the text in this paragraph.
Line 209: Define “symbiosis” earlier at the first use.
done
Lines 219 - 220: Could another curve work?
No. There was inadequate data for any curve with a number of the light levels simply dying. With a finite amount of time, the experiment was not attempted again and only maximum observed rate is reported. Some text has been added to clarify this.
Discussion:
Lines 258 - 261: This is a very interesting result and is in contrast to cyanobacteria, which fare better in natural oligotrophic seawater.
Thanks, although Trichodesmium is typically grown in ASW as well. .
Lines 263 - 272: Did you take any samples for DNA or RNA analysis? Perhaps a colleague could help you look for evidence of viruses.
Sadly, there are no archived samples for this type of analysis.
Lines 335 - 337: Do you have any genetic material preserved to determine whether these host and asymbiotic strains did diverge?
None.
Lines 345 - 346: What environment light levels are relevant? Did you go high enough?
This is, of course, a very big question. We were careful to phrase it to say that we saw no photoinhibition at the highest light level used. What phytoplankton are experiencing in the open sea is a good question that has been the subject of numerous papers addressing adaptation, degree and depth of mixing, spectral effects, integrated daily versus instantaneous light, and photoperiod effects. Some text has been added to address this, noting that in-situ light levels are not likely much more than our highest levels when total daily flux is considered as well as the effects of mixing.
Figures:
Fig. 1: Excellent micrographs. If possible, please include a higher magnification to better see the structure of the cyanobacteria.l
Epifluorescence or transmitted light microscopy is not the best way to resolve this. It’s really an electron microscopy problem due to the optical interference of the host cell. Thin sections are needed and we have no photographs of these.
Fig 3: Legend, please! And you used 7 light levels. Why are only three shown?
The Fig. 3 legend is in the PDF review file copy that I received, so there must have been some sort of error in the reviewer’s display of the PDF. Figure 3 of the same file shows the 7 light levels used in the experiment as well as the two from the preliminary experiment.
Fig 4: This is an astonishing number of points. How many aliquots did you take throughout the day to add acetylene? What do the error bars represent? Each point on the graph should represent multiple measurements in duplicate for one aliquot amended with acetylene. Multiply each point by 6 - 10, and it becomes an inhuman number of jabs in the GC. I’m not sure if the 5-point mean is meant to be a moving mean or something else. And the 4-h incubation is unclear to me.
The methods have been re-written to clarify this. As we note in this text, true replication for acetylene measurement is not possible due to the continuous evolution of acetylene and the finite time required for the GC to come back to baseline after the acetylene peak. Thus, we adapted a different approach. The supplemental data lists the sampling points. While numerous, it was possible to execute them in a single diel.
Figure 6: What was fit to a hyperbolic tangent function? There’s no P-E curve here. Figure 6 is a growth rate irradiance curve as noted in the figure legend.
I’m not sure what the question is. The shape of curve is typical for rates that have an initial light-limited phase and a light-saturation phase, hence the use of a general hyperbolic tangent function. The hyperbolic tangent function has been used for decades to describe the growth rate-irradiance relationship (Glover et al. 1987). We now note this in the methods. There are many equations that describe this general function (including the Michaelis-Menton curve if one considers light as a resource).
Reviewer 3 (Anonymous)
Basic reporting
Overall the text reads fine. I have suggested a few revisions.
Citations and discussion are heavily focusing on a narrow list of authors. An example of a relevant early reference that could be additionally included and discussed:
John F. Heinbokel 1986 OCCURRENCE OF RICHELIA INTACELLULARIS (CYANOPHYTA)WITHIN THE DIATOMS HEMIAULUS HAUKII ADN H. MEMBRANACEUS OFF HAWAII J Phycol
https://doi.org/10.1111/j.1529-8817.1986.tb00043.x
This reference was included in the original text on line 65. The sentence has been expanded and moved to a topic sentence. As far as the rest of it goes, the citations in these sections span 115 years and include every major lab that has worked or is working on these symbioses. There is little more that we can do without turning this into another review, most of which would not be particularly relevant to this paper.
The figures are ok. I have suggested a few edits to axes. One curve fit appears missing (see detail under General comments).
The manuscript is self-contained, but some additional detail should be provided. See below.
Experimental design
The manuscript reports important data on methods of isolation of a marine diazotroph symbiosis. There are not many known cultures of these associations, and very few laboratory observations.
I have some comments and suggestions regarding clarity of the data presentation. Several methods should be explained in more detail.
I suggest the authors include a table including key information about the specific strains they used in the experiments, to help reference these strains and experiments in any future studies.
Table S1 now lists the strains used in each figure and the strains are listed in the figure legends.. Additional information would be useful if the strains still existed and were in use in other experiments. However, as we note in the text, they have all perished. As we now note in the supplemental data table, most of the experiments used a single strain. The additional strains reported were used in preliminary or ancillary experiments. Since I expect no other laboratory data of this kind will be reported for a while, these data were included to present everything we were able to learn about these symbioses.
Validity of the findings
Overall the data should be useful to the N2 fixation research community
Replication is not always clear. Was there biological replication in the experiments or are the means measurements from a single culture over time?
‘The validity of using a ‘5-point running average’ for determining N2 fixation rates is not possible to assess without sufficient detail about the AR assay setup (see comments below).
As noted to other reviewers, this section has been considerably expanded and reworked to add clarity to this section.
One or two anecdotal statements without sufficient evidence are included in the discussion and should be removed.
See response under the specific comment.
Comments for the Author
Title: The title reads a bit cumbersome and unclear. I suggest re-wording such as:
Isolation, growth, and nitrogen fixation rates of the oceanic Hemiaulus-Richelia (diatom-cyanobacterium) symbiosis in culture
Fine with me. Oceanic was removed to reflect the dominance of this symbiosis in mid-salinity range waters of the Amazon River plume. It’s not strictly oceanic and is commonly found in shelf waters in tropical and temperate seas. The isolations themselves came from the coastal zone of Texas. These are low nutrient, but not oceanic regions.
Abstract:
The abstract should add a sentence to clarify whether Hemiaulus was cultured in a monoculture or whether the stated growth and N2 fixation rates were done while it was kept in the symbiosis with the host.
Done.
Diazotrophic symbiosis>reword. The symbiosis likely involves not just diazotrophy
Diazotrophic is an adjective that defines the specifics of this type symbiosis that give it a unique oceanographic significance. It may well include other relationships but that is not why these are so interesting to oceanographers.
Maximum growth rates of H. hauckii symbioses> remove the word ‘symbioses’
the H. membranaceus symbiosis> remove the words ‘the’ and ‘symbiosis’
The text as written is specifically referring to the growth of the entire symbiosis. To remove these words changes the meaning to only the host diatom. As we note, the host diatom would not grow without the symbiont. As noted earlier, we have modified the text throughout to clarify that all the rates refer to the host-symbiont unit. Neither was grown independently.
Narrative:
L2 prokaryotic and eukaryotic
done
L4 replace ‘these oligotrophic seas’ by ‘in the open ocean’
done
L7 Here and elsewhere: check the journal requirements about ordering citations. Typically you would mention the earliest citation first (here Villareal 1992).
The formatting in Endnote used the PeerJ style. Looking at other papers in the journal, the in-text citations appear to alphabetical, not chronological.
L55-56 The sentence about Calothrix is not well tied to the rest of the paragraph.
Text has been altered to this point.
L64 this may also BE true for…
done
L70 revise the sentence for clarity/remove redundancy regarding ‘growth rate’ and ‘growth rate response’.
revised
L72 ‘Modeling blooms’ is a bit ambiguous. Modeling bloom formation and fate?
revised
L84 uM > um. The former is a unit for micromoles.
Type corrected.
L87-90 State the purity grade of the chemicals.
done
L93 Unclear what you mean by ‘if required’.
This text has been clarified to note that short-term survival was possible in MET-44 medium, but the symbioses would not maintain in it.
L97 delete ‘quanta’ from the unit
I have done so globally. I know this is correct for the PPFD measured, but the QSP-100 units include quanta .
L106 Remove the weblink. Replace with a citation to a permanent source for the information (such as an external data repository) or remove entirely.
There is no permanent repoository for this data. I have deleted the link and substituted an additional reference provided by Dr. Campbell for data.
L110 Unclear what you mean by ‘given experiment’. Specify which experiment or state more clearly.
This entire section has been rewritten in response to the reviewers.
L113 The wording “symbioses growth” is ambiguous as the symbiosis is composed of two organisms. State more specifically. Do you mean, growth of Hemiaulus and Richelia and N2 fixation of the host-symbiont association?
The revised text has a specific definition of symbiosis growth included. N2 fixation necessarily is limited to the prokaryote symbiont. It is always normalized to numbers of heterocysts for that reason.
L114-115 Were there experimental replicates?
No, noted in text. However, the nitrate addition was a repeat of the preliminary experiment.
L115 Erlenmeyer
done
L116-117 Is there a reason why nitrate concentration was 4x the time that of ammonium?
Yes, ammonium is toxic at 40 µM. Forty µM was the stock addition for MET-44 nitrate used in the much earlier work on Rhizosolenia that showed the host Rhizosolenia would utilize nitrate and grow on it free of its symbiont Richelia.
L117 Was the salinity periodically checked? Evaporation is likely to have increased it over time.
Yes, text has been added earlier in the methods to address this.
L123-124 ‘Gradient’ here is misleading. Each treatment is a distinct light level, thus the experiment did not test the influence of a light gradient. The experiment tested the influence of distinct, constant light intensities.
During the methods reorganization, the word gradient was removed. However, it is a common usage but fortunately there is no need for an argument here.
L129 > “In addition, H. hauckii strain #92…”
Done
L132 Cell counts were performed > Cells were counted
done
L140-141 Do you mean the N2 fixation curves were forced to zero? Why was this different from the growth curves?
Growth rate curves are well understood to have a non-zero intercept that represents the compensation light intensity for growth. In the case of N2 fixation curves, it is not so straightforward. This is a short-term measurement where N2 fixation continues in the dark for some period of time as stored reductant is utilized for fixation. This is evident in the diel pattern of N2 fixation where dark fixation occurs at a declining rate. After several hours it is essentially zero. This is well understood to occur in the literature although rarely seen as clearly as in Fig 4a. Allowing a non-zero intercept permits an interpretation at first glance that steady-state N2 fixation occurs in the dark. We did not want this to be a citable point (Pyle et al. say that N2 fixation occurs in the dark) or for it to get into models. I have added some text to explain.
L140 This term>The y-intercept term
done
L142 More information is needed on how the assay was conducted. What was your vial size, liquid volume, gas volume, gas injection volume? What acetylene was used? How long did you incubate each vial and under what conditions? How much gas was removed from the vial per time point? How many times was each vial sampled?
This has been addressed in the methods re-write.
L144 ethylene conversion ratio 4:1> this needs to be stated more clearly. What does the 4:1 stand for?
Text has been added to define this.
L151 ‘center point of the time differential’> reword for clarity
This has been addressed in the methods re-write.
L151 ‘5 point mean of the multiple time series...’ Are you referring to replicate vials of the same treatment? Reword for clarity.
This has been addressed in the methods re-write.
L152-153 A citation to a MS thesis is not sufficient here to explain how the rates were calculated. Include the information in the methods and include a supplement if need be.
This has been addressed in the methods re-write and the citation deleted.
L153 How did you calibrate the GC? Did you run calibrations during the day?
Yes, added to text.
L154 Do not underline a in Chlorophyll a – it should be italized. Abbreviate consistently through the text as chl a (or Chl a).
corrected
L158-159 Nutrient methods need to be explained with appropriate citations, including detection limits.
The QuAAtro system, like all automated analyses uses proprietary reagent concentrations and additions in the manifold. However, they are all based on the standard methods for seawater analysis included in the now added reference. Detection limits have been added.
L170 what is sterile filtered seawater – explain how it was made
done
L175 >The use of these techniques resulted in…
done
L177 Delete ‘as well’
done
L179-180 Have been found by whom? Is this your personal experience? If so, revise sentence to indicate it.
This section has been revised and the sentence dropped. Context is provided by the revised text.
L181 Laboratory structure? What is this? A partially covered wetlab building? A device placed on top of the bin?
Clarified in revised text
L181-182 Unclear what ‘this site’ is referring to.
This has been clarified in the revised text
L190-191 Diameter? The cells are not spherical. Please state all the dimensions.
The diameter listed for Hemiaulus is the apical axis typically seen in gridle view in water mounts. This is standard usage; however, I have added descriptors to clarify this. The range of pervalvar and transapical measurements have been lost but are consistent with the dimensions seen in Fig. 1.
L204 time>times
done
L205 ‘chl per symbiosis’ is confusing. Do you mean per cell of Hemiaulus?
This is the combined chlorophyll of the Hemiaulus with its symbiont, now noted in the text.
L208-209 the definition of chl a per symbiosis should be stated earlier. In the Figure 2 you use chl a per cell though. This is confusing.
As per the other reviewer as well, the explanation of chl a per symbiosis has been moved up. The unit of the symbiosis is the host cell Hemiaulus. The chl a is normalized to the host Hemiaulus counts. However, a few percent of the cells always divide without the symbiont so there are asymbiotic Hemiaulus present with chl a. There’s no way to differentiate these asymbiotic Hemiaulus but given the low contribution of these cells and the imprecise nature of chlorophyll as biomass, it make no difference to the general point being made. I have reworded the figure legend title to describe this data as the symbiosis.
L228 scotophase?
It means dark phase, a term typically used in diel cycle studies.
L230-231 photophase or photoperiod? Would help the reader to stick to consistent terminology.
They are two different terms used for the same meaning in different contexts. To increase clarity, I have used only photophase here.
L249-253 The information about the symbiont free Hemiaulus is anecdotal with no other evidence shown except a statement in the narrative. Are there any observations of this strain that could be shared such as morphology, growth rates etc. Without such data the paragraph should probably be removed.
While this may not be quantitative data, it is an observation of great relevance. The relevant comparison is the nitrate based growth rates cited from the literature that are 3X + higher than the rates noted by diazotrophically supported Hemiaulus symbiosis cells. They make no sense when compared to the symbiosis data, but a great deal of sense when viewed in the context of asymbiotic lines of Hemiaulus. The field can choose to ignore this if it wishes. Knowing that the nitrate-utilizing, symbiont free strains exist in the wild is valuable information for future investigations. You can’t look for something if you don’t know it exists.
L283 Website citation here needs to be changed into a citation to a permanent source or company name, city, state, country.
This is a web-based company (name now included) that apparently does not have a physical address. It was included as a convenience to the reader. There are multiple ways to digitize a paper plot readily available online.
L285 What does ‘their rates’ mean? Did you convert their cell numbers to rates? Or did you take rates from the paper? If using actual rates from the paper it is unclear what the digitizing was used for.
As noted in the text, Carpenter et al.’s Fig. 2 presents data on N2 fixation and heterocyst abundance. There is no data table, so the values have to be extracted from the plot. Digitization refers to converting the X,Y coordinates on the plot to numerical values to allow an N2 fixation rate per heterocyst, values that we compare to our measured rates in culture. I’ve added some text that should clarify this.
L287 With ‘these rates’ are you referring to rates in Carpenter et al. 1999?
See above.
L299 nifH gene expression?
Common identifier in the field for the nitrogenase gene. Text added.
L306 >at least two mechanisms that..
corrected
L305-309 I suggest breaking this sentence into two to clarify the meaning.
Done
L314 utilized > used
done
L316 utilize > use
done
L331 The statement about origin of the organisms is unclear. The origin of the cyanobacterium is likely earlier than the host.
clarified
L339 ‘symbioses blooms’ This wording is awkward > reword.
clarified
Fig. 1. These are beautiful images. What filters were used for the fluorescence images? Are these images overlays of images with two filter sets? State in the figure legend. Remove ‘image credit’ – no need to credit yourself.
The epifluorescence filter sets are listed in the methods. The images are not overlays but paired images under transmitted light and epifluorescence as described in the figure legends. I’ve added some text to refer to the details in the methods. From my read of the Instructions to Authors, all images need a credit. I’ll let the editors call this one.
Fig. 2. Delete ‘quanta’. Chlorophyll > Chlorophyll a.
done
B: Specify whether you are showing chla per Hemiaulus or chla per Richelia cell?
Reworded legend as noted above.
Fig. 3. The strain with x- symbols does not appear to have a Growth-I curve associated with it. The light affinity appears substantially different for this strain. This result does not seem to be discussed in the text as you only refer to two G-I curves and a very similar alpha for the two.
The figure has been replotted to clarify this as per some of the other reviewers’ comments as well. The X-symbols (now solid squares) are only two data points and there is no reasonable way to curve fit them (strain #9). They are included only to provide some additional sense of how these growth rate-light data varied. The other two curves (both seen in the PDF review file that was returned to me) are from the same strain (#91).
Fig.4. ‘average… calculated from the slope’ > reword to clarify. How long were the incubations for the black symbols?
This entire section has been rewritten in the text and the legends amended.
Fig.5. ‘slope of successive measurements over a 4 h period’: This is confusing. Explain what you mean by a SLOPE of successive measurements. Are you referring to measurements representing a different cumulative incubation time from the same vial? If so, isn’t the comparison simply the rate/time from each of these measurements. This is not a slope.
This has been clarified in the methods and removed from the figure legends. Multiple measurements over time per vial yields ethylene per heterocyst on the y axis and time on the x axis. The linear fit to this over 4 hours is a slope (ethylene per heterocyst per time). The 95% CI is calculated on this slope in Excel.
Fig. 6. Should the y-axis be ‘% symbiotic Hemiaulus’? If not, the results do not correspond with the text narrative.
Yes, it should be. The legend has been corrected and figure.
" | Here is a paper. Please give your review comments after reading it. |
9,750 | "<ns0:abstract xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:p>Nitrogen fixers (diazotrophs) are often an important nitrogen source to phytoplankton nutrient budgets in N-limited marine environments. Diazotrophic symbioses between cyanobacteria and diatoms can dominate nitrogen fixation regionally, particularly in major river plumes and in open ocean mesoscale blooms. This study reports the successful isolation and growth in monocultures of multiple strains of a diatom-cyanobacteria symbiosis from the Gulf of Mexico using a modified artificial seawater medium. We document the influence of light and nutrients on nitrogen fixation and growth rates of the host diatom Hemiaulus hauckii Grunow together with its diazotrophic endosymbiont Richelia intracellularis Schmidt, as well as less complete results on the Hemiaulus membranaceus -R. intracellularis symbiosis. The symbioses rates reported here are for the joint diatom-cyanobacteria unit. Symbiont diazotrophy was sufficient to support both the host diatom and symbiotic cyanobacteria, and the entire symbiosis replicated and grew without added nitrogen. Maximum growth rates of multiple strains of H. hauckii symbioses in N-free medium with N 2 as the sole N source were 0.74-0.93 div d -1 .</ns0:p><ns0:p>Growth rates followed light saturation kinetics in H. hauckii symbioses with a growth compensation light intensity (E C ) of 7-16 µmol m -2 sec -1 and saturation light level (E K ) of 84-110 µmol m -2 sec -1 . Nitrogen fixation rates by the symbiont while within the host followed a diel pattern where rates increased from near-zero in the scotophase to a maximum 4-6 hours into the photophase. At the onset of the scotophase, nitrogen fixation rates declined over several hours to near-zero values. Nitrogen fixation also exhibited light saturation kinetics. Maximum N 2 fixation rates (84 fmol N 2 heterocyst -1 h -1 ) in low light adapted cultures (50 µmol m -2 s -1) were approximately 40-50% of rates (144-154 fmol N 2 heterocyst -1 h -1 ) in higher light (150 and 200 µmol m -2 s -1 ) adapted cultures. Maximum laboratory N 2 fixation rates were ~6 to 8-fold higher</ns0:p></ns0:div>
</ns0:abstract>
<ns0:body xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:p>Nitrogen fixers (diazotrophs) are often an important nitrogen source to phytoplankton nutrient budgets in N-limited marine environments. Diazotrophic symbioses between cyanobacteria and diatoms can dominate nitrogen-fixation regionally, particularly in major river plumes and in open ocean mesoscale blooms. This study reports the successful isolation and growth in monocultures of multiple strains of a diatom-cyanobacteria symbiosis from the Gulf of Mexico using a modified artificial seawater medium. We document the influence of light and nutrients on nitrogen fixation and growth rates of the host diatom Hemiaulus hauckii Grunow together with its diazotrophic endosymbiont Richelia intracellularis Schmidt, as well as less complete results on the Hemiaulus membranaceus-R.intracellularis symbiosis. The symbioses rates reported here are for the joint diatom-cyanobacteria unit. Symbiont diazotrophy was sufficient to support both the host diatom and cyanobacteria symbionts, and the entire symbiosis replicated and grew without added nitrogen. Maximum growth rates of multiple strains of H. hauckii symbioses in N-free medium with N 2 as the sole N source were 0.74-0.93 div d -1</ns0:p><ns0:p>. Growth rates followed light saturation kinetics in H. hauckii symbioses with a growth compensation light intensity (E C ) of 7-16 µmol m Nitrogen fixation rates by the symbiont while within the host followed a diel pattern where rates increased from near-zero in the scotophase to a maximum 4-6 hours into the photophase. At the onset of the scotophase, nitrogen-fixation rates declined over several hours to near-zero values. Nitrogen fixation also exhibited light saturation kinetics.</ns0:p><ns0:p>Maximum N 2 fixation rates (84 fmol N 2 heterocyst ) adapted cultures. Maximum laboratory N 2 fixation rates were ~6</ns0:p></ns0:div>
<ns0:div><ns0:head>Introduction</ns0:head><ns0:p>The phytoplankton flora of the open sea is a diverse assemblage of prokaryotic and eukaryotic cells that span a size range of ~1 to 2,000+ µm in diameter. Nitrogen is often a limiting nutrient in the open sea, and planktonic nitrogen fixation (diazotrophy) occurs in tropical, subtropical systems and high latitude systems <ns0:ref type='bibr' target='#b72'>(Zehr 2011;</ns0:ref><ns0:ref type='bibr' target='#b31'>Harding et al. 2018)</ns0:ref>.</ns0:p><ns0:p>However, nitrogen fixation can occur in a wide variety of deep-sea and benthic habitats not traditionally associated with nitrogen-limitation <ns0:ref type='bibr' target='#b71'>(Zehr and Capone 2020)</ns0:ref>. Diazotrophy occurs only in prokaryotic cells, but a variety of symbiotic associations between diazotrophic prokaryotes and host eukaryotes are known <ns0:ref type='bibr'>(Foster et al. 2006;</ns0:ref><ns0:ref type='bibr' target='#b21'>Foster & O'Mullan 2008;</ns0:ref><ns0:ref type='bibr' target='#b52'>Taylor 1982;</ns0:ref><ns0:ref type='bibr' target='#b59'>Villareal 1992;</ns0:ref><ns0:ref type='bibr' target='#b71'>Zehr and Capone 2020)</ns0:ref> and cover the range from obligate symbioses to loosely associated consorts <ns0:ref type='bibr' target='#b8'>(Caputo et al. 2019;</ns0:ref><ns0:ref type='bibr' target='#b9'>Carpenter 2002;</ns0:ref><ns0:ref type='bibr'>Foster et al. 2006;</ns0:ref><ns0:ref type='bibr' target='#b21'>Foster & O'Mullan 2008)</ns0:ref>. Of these, diatom-diazotroph associations (DDAs) are the most visible with records dating back to the early 20 th century <ns0:ref type='bibr' target='#b41'>(Karsten 1905)</ns0:ref>. Two types of marine diatom-cyanobacteria symbioses are known: diatoms in the genera Neostreptotheca and Climacodium that host coccoid cyanobacteria <ns0:ref type='bibr' target='#b10'>(Carpenter & Janson 2000;</ns0:ref><ns0:ref type='bibr' target='#b30'>Hallegraeff & Jeffrey 1984)</ns0:ref>, and diatoms that host filamentous, heterocyst-forming cyanobacteria of the genera Richelia and Calothrix. Little is known about the characteristics of the coccoid symbionts in diatoms, although the Climacodium symbiont is a diazotrophic Crocosphaera sp. <ns0:ref type='bibr' target='#b20'>(Foster et al. 2011)</ns0:ref>. DDA symbioses involving heterocystous, diazotrophic cyanobacteria are abundant in both open ocean systems <ns0:ref type='bibr' target='#b13'>(Dore et al. 2008;</ns0:ref><ns0:ref type='bibr' target='#b61'>Villareal et al. 2011;</ns0:ref><ns0:ref type='bibr' target='#b67'>Wilson et al. 2008</ns0:ref>) and at intermediate salinities within the Amazon <ns0:ref type='bibr' target='#b22'>(Foster et al. 2007</ns0:ref>), Mekong <ns0:ref type='bibr' target='#b28'>(Grosse et al. 2009</ns0:ref>) and Congo River plumes <ns0:ref type='bibr' target='#b23'>(Foster et al. 2009</ns0:ref>). These marine regions differ greatly in their characteristics, suggesting either a great plasticity in physiological responses to Manuscript to be reviewed environmental variables or undocumented differentiation within these symbioses. Symbiont integration with the hosts varies as well. In the Rhizosolenia-Richelia DDA symbiosis, the symbiont is located in the periplasmic space between the frustule and plasmalemma and has limited contact with the external environment <ns0:ref type='bibr' target='#b39'>(Janson et al. 1999;</ns0:ref><ns0:ref type='bibr' target='#b56'>Villareal 1989</ns0:ref>). The Hemiaulus-Richelia DDA symbiont is appressed to the nucleus and truly intracellular <ns0:ref type='bibr' target='#b8'>(Caputo et al. 2019)</ns0:ref>, consistent with its reduced genome <ns0:ref type='bibr' target='#b34'>(Hilton et al. 2013)</ns0:ref>. The Chaetoceros-Calothrix DDA symbiont is completely extracellular to the host diatom <ns0:ref type='bibr' target='#b19'>(Foster et al. 2010)</ns0:ref>.</ns0:p><ns0:p>Despite their ubiquitous occurrence in tropical seas, the Hemiaulus-Richelia symbiosis was largely overlooked until epifluorescence microscopy revealed the cryptic Richelia symbiont <ns0:ref type='bibr' target='#b33'>(Heinbokel, 1986)</ns0:ref> and N 2 fixation was documented in individually picked chains of the symbiosis <ns0:ref type='bibr' target='#b58'>(Villareal 1991)</ns0:ref>. In addition to providing fixed N to the pelagic community, diatomcyanobacteria symbioses play an important role in the nitrogen and carbon cycles of oceanic systems by virtue of their potential to sequester carbon to the deep sea via aggregation and sinking <ns0:ref type='bibr' target='#b40'>(Karl et al. 2012;</ns0:ref><ns0:ref type='bibr' target='#b51'>Subramaniam et al. 2008</ns0:ref>). In the currency of oceanic nitrogen cycling, nitrogen derived from photosynthetic nitrogen fixation is generally balanced by a concurrent removal of atmospheric CO 2 <ns0:ref type='bibr' target='#b14'>(Eppley & Peterson 1979)</ns0:ref>. Thus, sinking material fueled by phototrophic diazotrophy represents a net removal of CO 2 , and is a quantitatively important process in the transport of carbon to depth. DDAs, and particularly Hemiaulus symbioses, are of particular oceanographic significance. Hemiaulus-Richelia symbioses bloom at ~ 10 3 cells L -1 frequently at the Hawai'i Ocean Time-series HOT <ns0:ref type='bibr' target='#b13'>(Dore et al. 2008;</ns0:ref><ns0:ref type='bibr' target='#b17'>Fong et al. 2008;</ns0:ref><ns0:ref type='bibr' target='#b47'>Scharek et al. 1999;</ns0:ref><ns0:ref type='bibr' target='#b65'>White et al. 2007)</ns0:ref>. At this location, they are the likely source of the summer export pulse that provides 20% of the annual carbon flux to 4,000 m in a 4-6-week window <ns0:ref type='bibr' target='#b40'>(Karl et al. 2012</ns0:ref>) and are regularly found on sinking particles <ns0:ref type='bibr' target='#b15'>(Farnelid et al. 2019)</ns0:ref>. Subtropical front Manuscript to be reviewed blooms at ~28-30°N in the Pacific <ns0:ref type='bibr' target='#b55'>(Venrick 1974;</ns0:ref><ns0:ref type='bibr' target='#b62'>Villareal et al. 2012;</ns0:ref><ns0:ref type='bibr' target='#b66'>Wilson et al. 2004;</ns0:ref><ns0:ref type='bibr' target='#b67'>Wilson et al. 2008)</ns0:ref> and in waters west and north of HI <ns0:ref type='bibr' target='#b5'>(Brzezinski et al. 1998;</ns0:ref><ns0:ref type='bibr' target='#b61'>Villareal et al. 2011</ns0:ref>) suggest a basin scale significance. In the southwest Atlantic Ocean, Hemiaulus hauckii-Richelia blooms cover 10 5 + km 2 and sequester 1.7 Tmol of carbon annually <ns0:ref type='bibr' target='#b11'>(Carpenter et al. 1999;</ns0:ref><ns0:ref type='bibr' target='#b51'>Subramaniam et al. 2008)</ns0:ref> and CO 2 drawdown effects can extend to 10 6 km 2 <ns0:ref type='bibr' target='#b12'>(Cooley et al. 2007</ns0:ref>). The large size, chain-formation, and tendency to aggregate <ns0:ref type='bibr' target='#b47'>(Scharek et al. 1999;</ns0:ref><ns0:ref type='bibr' target='#b61'>Villareal et al. 2011)</ns0:ref> in the host Hemiaulus lead to an efficient export mechanism <ns0:ref type='bibr' target='#b68'>(Yeung et al. 2012</ns0:ref>) for both N and C. Culture studies on the growth and physiological characteristics of these symbioses are limited. The external symbiont Calothrix rhizosoleniae has been cultured without its host <ns0:ref type='bibr' target='#b19'>(Foster et al. 2010</ns0:ref>) in both natural and artificial seawater medium. Cultures of the Rhizosolenia-Richelia symbiosis using amended seawater have been reported in the literature with growth rates up to 0.8 div d -1 in fixed N-free medium <ns0:ref type='bibr' target='#b57'>(Villareal 1990</ns0:ref>). In the Rhizosolenia-Richelia DDA, host and symbiont growth can be independent and symbiont-free host cells occur (but have reduced growth rates) even when no fixed N is present, possibly through use of N excreted by Richelia into the medium. Addition of nitrate rapidly results in the loss of symbionts as asymbiotic Rhizosolenia uses the added nitrate, increases its growth rate, and out-competes symbiotic Rhizosolenia-Richelia <ns0:ref type='bibr' target='#b56'>(Villareal 1989;</ns0:ref><ns0:ref type='bibr' target='#b57'>Villareal 1990</ns0:ref>). Nitrogen fixation follows typical light saturation kinetics and can provide the entire N needs of the symbiosis <ns0:ref type='bibr' target='#b57'>(Villareal 1990</ns0:ref>). Although oceanographically more significant than other Rhizosolenia-Richelia DDA <ns0:ref type='bibr' target='#b33'>(Heinbokel 1986;</ns0:ref><ns0:ref type='bibr' target='#b51'>Subramaniam et al. 2008</ns0:ref><ns0:ref type='bibr' target='#b59'>, Villareal 1992)</ns0:ref>, there are no published culture-based data for the Hemiaulus-Richelia symbiosis.</ns0:p><ns0:p>PeerJ reviewing PDF | (2019:08:40471:2:0:NEW 8 Sep 2020)</ns0:p><ns0:p>Manuscript to be reviewed</ns0:p><ns0:p>Using nano-SIMS on field samples, <ns0:ref type='bibr' target='#b20'>Foster et al. (2011)</ns0:ref> were able to document the transport of recently fixed N from the symbiont Richelia to the host Hemiaulus in sufficient quantities to support growth; however, it is not known whether Hemiaulus-Richelia can grow exclusively on diazotrophically fixed N. Regardless, the symbiont is clearly advantageous to the host since, where examined, 80-100% of the Hemiaulus contain the symbiont <ns0:ref type='bibr' target='#b1'>(Bar-Zeev et al. 2008;</ns0:ref><ns0:ref type='bibr' target='#b33'>Heinbokel, 1986;</ns0:ref><ns0:ref type='bibr' target='#b58'>Villareal 1991;</ns0:ref><ns0:ref type='bibr' target='#b60'>Villareal 1994</ns0:ref>) and 85-100% of the total phytoplankton N needs in the Amazon River plume can be met by Hemiaulus DDA diazotrophy <ns0:ref type='bibr'>(Carpenter et al. 1999;</ns0:ref><ns0:ref type='bibr' target='#b63'>Weber et al. 2017</ns0:ref>). The symbiosis is not obligate for the host Rhizosolenia in DDA cultures <ns0:ref type='bibr' target='#b57'>(Villareal, 1990)</ns0:ref> and the field evidence suggests this may also be true for the host Hemiaulus <ns0:ref type='bibr' target='#b33'>(Heinbokel, 1986;</ns0:ref><ns0:ref type='bibr'>Kimor et al., 1978)</ns0:ref>. This latter hypothesis has not been tested due to the difficulty in growing the Hemiaulus-Richelia host-symbiont pair in vitro.</ns0:p><ns0:p>In this paper, we report the successful isolation of two species of the Hemiaulus-Richelia symbiosis into culture and expand on the brief culturing description reported in <ns0:ref type='bibr' target='#b48'>Schouten et al. (2013)</ns0:ref>. Using primarily H. hauckii-Richelia DDA strains, we document light-dependent growth rates, diel cycles of N 2 fixation, growth rate response to various forms of added nitrogen, and N 2 fixation rates. These parameters are essential to supporting modeling of DDA bloom formation and fate <ns0:ref type='bibr' target='#b16'>(Follett et al. 2018;</ns0:ref><ns0:ref type='bibr' target='#b50'>Stukel et al. 2014</ns0:ref>). In addition, key differences between the Throughout the text, N-free or fixed N-free medium will refer to culture medium that has no added organic or inorganic N, recognizing that dissolved N 2 will be abundant as an N source for diazotrophs. Sterile filtered medium and seawater were generated using commercially available sterile tissue culture towers and reservoirs (0.22 µm pore size filters). Sterile filtration units were rinsed with ~50 ml of medium prior to use for culture medium. Both nylon and methyl cellulose 0.22 µm pore size filters were used with no apparent difference in results. All chemicals were reagent grade or better. The modified YBCII medium was checked with a hand-held refractometer before each use and adjusted to a salinity of 35 as needed using 18 megaohm deionized water. Autoclaved tubes were rinsed with sterile filtered medium and then the tube filled with 15-20 mL of medium. This rinsing step was used for all flasks and tubes used for culturing. Isolations were performed within 5-10 minutes of collection. Using a stereomicroscope, multiple Hemiaulus chains were rapidly isolated from the net tow material using hand-held borosilicate pipets drawn to a fine diameter in a gas flame. In our work, the drawn-out pipets Manuscript to be reviewed decline and death when kept in medium longer than 7-10 days and careful attention was required to transfer the cultures to new medium within this time frame. Experiments were initiated within 6 months of culture isolation; cultures failed to make auxospores and were eventually lost after approximately 1-2 years in culture. No attempt was made to culture axenically; bacteria were rarely visible in the cultures under phase contrast or differential interference contrast optics until senescence when cell mortality was substantial. The H. hauckii DDA was the primary experimental tool. Hemiaulus membranaceus DDA cultures were examined for general characteristics but were not the subject of intensive experimentation. In March 2017, Hemiaulus chains were observed in the Port Aransas ship channel from the Imaging Flow Cytobot data stream <ns0:ref type='bibr' target='#b6'>(Campbell et al. 2010</ns0:ref><ns0:ref type='bibr' target='#b7'>(Campbell et al. , 2017))</ns0:ref>. Examination of net tow material noted numerous asymbiotic H. hauckii chains and no symbiotic cells. Asymbiotic chains of Hemiaulus hauckii were isolated into N-replete (40 µm NO 3 -) MET-44 amended sterile filtered seawater as noted above. Unless otherwise noted, all experiments were conducted using modifies YBCII medium with no added nitrogen. Dissolved N 2 was the only available nitrogen source.</ns0:p></ns0:div>
<ns0:div><ns0:head>Analytical methods</ns0:head><ns0:p>Cells were counted using a S52 Sedgewick-Rafter chamber on an Olympus BX51 epifluorescence microscope. Excitation/emission wavelengths for the epifluorescent filters used in counts and photography were 450 nm/680 nm (chlorophyll a), and 490 nm/ 565 nm (phycoerythrin). Both host cells and symbiont trichomes/heterocysts were enumerated. Percent symbiosis was calculated as the number of diatoms containing one or more Richelia trichomes divided by the total number of potential host cells. Growth rates (reported as div d -1 ) were calculated using daily counts as the slope of the log of cell number over the change in time Manuscript to be reviewed <ns0:ref type='bibr' target='#b29'>(Guillard 1973)</ns0:ref> with the 95% confidence interval around the slope of the line calculated in Microsoft Excel. Acetylene reduction assays (ARA) were performed as described in Capone <ns0:ref type='bibr'>(1993)</ns0:ref> corrected for ethylene solubility as described by Breitbarth et al. ( <ns0:ref type='formula'>2004</ns0:ref>) and assuming a mol ethylene reduced per mol N 2 conversion ratio of 4:1 (Jensen & Cox 1983 as modified by Capone <ns0:ref type='table'>1993</ns0:ref>). An SRI 8610C gas chromatograph (SRI Instruments, Torrance, CA) equipped with a 30 cm silica gel column was used to quantify ethylene using a commercially prepared standard (GASCO Safeware Precision Gas Mixture, 10 and 100 ppm). Manufacturer-provided software (PeakSimple Chromatography Software) performed peak integrations. Standards were run prior to each day's run and at several points during the experiment. For each assay, 15 ml of culture sample was added to an acid-washed 25 ml incubation vial fitted with a grey chlorobutyl rubber serum stopper and crimped aluminum seals leaving 10 ml of headspace. Sterile-filtered medium was used as a control. A separate aliquot was retained for cell counts. One ml of acetylene generated from calcium carbide (Capone 1983) was introduced, gently swirled for 15-30 seconds to equilibrate while minimizing contact between the serum stopper and the culture, then 100 µL of the vial headspace injected with a Hamilton gas-tight syringe and injected into the GC. Each injection required 5-7 minutes after an injection to return to baseline. Chlorophyll a was determined on methanol-extracted (24 hours, -20°C) samples (10-25 ml aliquot) collected on 0.4 µm pore size polycarbonate filters using a non-acidification method <ns0:ref type='bibr' target='#b64'>(Welschmeyer 1994)</ns0:ref>. Initial tests indicated the filters used did not leach fluorescent compounds in the methanol. When chl a cell -1 is referred to, it always includes both symbiont and host chl a.</ns0:p><ns0:p>Sample fluorescence was read on a TD-700 Fluorometer (Turner Designs, CA, USA).</ns0:p><ns0:p>PeerJ reviewing PDF | (2019:08:40471:2:0:NEW 8 Sep 2020)</ns0:p><ns0:p>For nutrients, a 25 mm, 0.22 µm pore-size membrane cellulose ester Millipore filter mounted on a syringe was rinsed with 5 ml of sample, filtrate discarded, and ten ml of sample medium was filtered and frozen. A SEAL Analytical QuAAtro autoanalyzer was used to determine dissolved inorganic phosphate (DIP), nitrate +nitrite (N+N), ammonium (NH 4 + ), and silicate (SiO 4 -2 ) concentrations using the manufacturer's recommended chemistries. The chemistries are similar to automated analyses published in <ns0:ref type='bibr' target='#b27'>Grasshoff et al. (1999)</ns0:ref> with changes in reagent concentration and wetting agents specific to the manifold chemistries. Detection limits were ~0.05 µM for N+N, NH 4 + and P, and ~0.5 µM for Si.</ns0:p><ns0:p>Growth-rate and N 2 fixation versus irradiance experiments H. hauckii symbiosis strains #9 and #91 were used for the irradiance-rate experiments.</ns0:p><ns0:p>Initial experiments (Strain #9) used two light levels and are included for comparison. Detailed growth rates and N 2 fixation rates were measured in separate experiments using 7-8 different light levels (photosynthetic photon flux density) ranging from 15-600 µmol m -2 sec -1 measured by a QSP-170B irradiance meter (Biospherical Instruments; Table <ns0:ref type='table'>S1</ns0:ref>). For growth rates, and H. membranaceus strain #82 utilized a high frequency time series approach in order to resolve changes occurring on an hourly basis or less. This approach used a series of individual measurements taken from a single vial over a period of up to ~12 hours and was utilized for two reasons. First, individual assays injections required 5-7 minutes to return to baseline. Triplicate measurements therefore required 15-21 minutes during which ethylene production was occurring at measurable rates, could not be consideration true replication of the ethylene measurement.</ns0:p><ns0:p>Averages of these triplicates would be unable to resolve rate changes on short time scales. The second reason for this approach was to minimize handling, agitation, and light/temperature variation of the samples. Six (H. hauckii) or 8 (H. membranaceus) paired vials were started at various time points in the diel cycle to permit overlap. Individual time series can be identified from the labelling in Table <ns0:ref type='table'>S1</ns0:ref>. Vials were sampled sequentially (1a, 1b, 2a, 2b, 3a,3b, then repeated) yielding approximately 1-1.5 hours between successive sampling of a single vial. The difference between successive measurements (ethylene per heterocyst) was normalized to the time difference between the two successive points (~1 -1.5 hours) and expressed as a rate (ethylene heterocyst -1 time -1 ). Eighty-nine (H. Manuscript to be reviewed beginning of the scotophase resulted in a highly variable initial slope as well as a significant yintercept (dark fixation rate) that was not consistent with the longer term rates after several hours in darkness. Delta Graph (Red Rocks Software) was used for graphics as well as curve fitting of the growth and N 2 -irradiance curves. T-tests were performed using the data analysis package in Microsoft Excel. Confidence intervals or standard deviations (noted in text) were calculated using Microsoft Excel software. Data from all figures are found in Table <ns0:ref type='table'>S1</ns0:ref>.</ns0:p><ns0:p>PeerJ reviewing PDF | (2019:08:40471:2:0:NEW 8 Sep 2020)</ns0:p><ns0:p>Manuscript to be reviewed</ns0:p></ns0:div>
<ns0:div><ns0:head>Results</ns0:head><ns0:p>Hemiaulus hauckii and Hemiaulus membranaceus with their symbiont Richelia intracellularis were successfully isolated multiple times. We found it was essential to remove the Hemiaulus from the net tow sample as quickly as possible (3-5 minutes after completion of the tow). Successful culturing resulted in rapidly growing chains of Hemiaulus reaching over 80 cells in length (Fig. <ns0:ref type='figure' target='#fig_14'>1</ns0:ref>). Multiple symbionts (usually 1-2, but never more than 4) were evident in the cells. Cultures were sensitive to handling, and swirling tubes to re-suspend chains resulted in chain breakage and decreased growth rates. Growth in undisturbed large volume containers (10 L+) resulted in complex aggregate formation. Strains were difficult to ship, and only one attempt out of approximately 15 resulted in successful establishment in another facility. A single auxospore-like structure was observed, but no cell diameter increases were observed in any of the cultures.</ns0:p><ns0:p>H. hauckii strains used in this study ranged from 12-17.5 µm (up to 30 µm observed) in diameter (pervalvar axis presented in broad girdle view) with a total cell volume range of 7,012 -23,574 µm 3 . H. membranaceus cells were not measured. Since auxosporulation did not occur, the strains gradually decreased in diameter over a period of 1-2 years and eventually died out.</ns0:p><ns0:p>Individual strains exhibited periods (weeks/months) of healthy growth (0.5-0.9 div d -1 ) with little care required. This growth pattern was interspersed with intervals (days/weeks) of low growth rates that required substantial attention and multiple backups to prevent loss of the culture. These cyclic patterns were not linked to batches of culture medium or glassware. While not enumerated, bacteria were rarely evident in light microscopy but certainly present since the cultures were not axenic. Reasons for the observed growth pattern variability remain unknown.</ns0:p><ns0:p>PeerJ reviewing PDF | (2019:08:40471:2:0:NEW 8 Sep 2020)</ns0:p><ns0:p>Manuscript to be reviewed Individual symbiosis strains were routinely maintained in modified YBC-II medium with no added nitrogen. High densities of Hemiaulus and its symbionts were possible with no N added to the synthetic seawater medium (residual combined inorganic N < 0.1 µM). Maximum cell counts of the host H. hauckii reached ~10,000 cells mL -1 with a maximum chl a concentration of 71 µg L -1 . Typical cell and chl a dynamics are shown in Fig. <ns0:ref type='figure' target='#fig_15'>2</ns0:ref>. High light (200 µmol m -2 s -1 ) chl a concentration reached a maximum approximately 3.5 times greater than the low light (50 µmol m -2 s -1 ) concentrations, although chl a per symbiosis (combined host and symbiont; multiple strains) remained approximately equal over time. In both light conditions, chl a per symbiosis was maximal (~4-5 pg chl a symbiosis -1 ) in early exponential growth and declined over time to ~2-3 pg chl a symbiosis -1 . Extensive chain formation resulted in a high degree of variation in measurements.</ns0:p><ns0:p>Growth rates of H. hauckii in N-deplete medium (Fig. <ns0:ref type='figure'>3</ns0:ref>) followed light saturation kinetics with host and symbiont growth rates highly correlated (r 2 = 0.98, p=0.05, t-test).</ns0:p><ns0:p>Photoinhibition was not observed at the maximum light level used (500 µmol m -2 s -1 ). A modified Jassby-Platt curve fit (Article S1) yielded a realized maximum growth rate µ of 0.74-0.93 div d -1 in replicated experiments (Fig. <ns0:ref type='figure'>3</ns0:ref>, Table <ns0:ref type='table'>1</ns0:ref>). Light-saturated growth occurred with light saturation (E k ) occurring at 84-110 µmol m -2 s -1 and an initial slope () of 0.009 div d -1 (µmol m -2 s -1 ) -1 in both irradiance curves. Compensation light intensity (E c ) calculated from the y-intercept and varied from 7-16 µmol m -2 s -1 . Nitrogen fixation rates estimated by acetylene reduction were tightly linked to the light:dark cycle (Fig. <ns0:ref type='figure'>4</ns0:ref>). The 5-point running average was necessary to smooth the variable Manuscript to be reviewed reduction rate in both H. hauckii and H. membranaceus DDA occurred approximately 4 hours into the photophase (12:12 photoperiod) with a broader maximum acetylene reduction rate extending for 4-6 hours. Acetylene reduction declined over several hours at photophase end to low (1-10% maximum values) but still measurable rates during the scotophase in both H. hauckii (Fig. <ns0:ref type='figure'>4a</ns0:ref>) and H. membranaceus (Fig. <ns0:ref type='figure'>4b</ns0:ref>) DDAs. Unlike the 4-hour discrete incubation diurnal pattern seen in Strain #22, H. hauckii strain #92 rates maintained high values until the end of the photophase (Fig. <ns0:ref type='figure'>4a</ns0:ref>). Hemiaulus membranaceus DDA rates were more symmetrically distributed around the middle of the photoperiod (Fig. <ns0:ref type='figure'>4b</ns0:ref>). In both data sets, the rates reached a maximum in the range of 45-55 fmol N 2 heterocyst -1 h -1 .</ns0:p><ns0:p>Nitrogen fixation-irradiance rates followed a light saturation curve (Fig. <ns0:ref type='figure'>5</ns0:ref>) fit to the hyperbolic tangent function. At the 150 and 200 µmol m -2 s -1 adaptation level (r 2 =0.95 and 0.97, respectively), the curve-fit maximum N 2 -fixation rates was 155 and 144 fmol N 2 heterocyst -1 h -1 , respectively. The maximum rates (light-saturated) at 150 and 200 µmol m -2 s -1 adaptation level were significantly (p<0.01, t-test) greater than the maximum (light-saturated) rate (86 fmol N 2 heterocyst -1 h -1) noted in cultures adapted to 50 µmol m -2 s -1 . The initial slope (light limited portion) of the N 2 fixation curve was approximately 75% higher in the 50 µmol m -2 s -1 adapted culture than the 150 and 200 µmol m -2 s -1 adaptation level. Preliminary experiments in 2010 found that H. hauckii strain #9 did not utilize nitrate (Table <ns0:ref type='table'>S2</ns0:ref>). Subsequent replication experiments found that 40 µM nitrate was not used by a different H. hauckii symbiosis strain (#83) in experiments conducted 1 year later (Fig. <ns0:ref type='figure'>6</ns0:ref>). Ten µM added ammonium declined to ~0.4 µM in 13 days and then remained constant thereafter (Fig. <ns0:ref type='figure'>6</ns0:ref>). Hemiaulus hauckii strain #83 drew down P and Si under all the available N sources at approximately equal rates. The addition of ammonium in an experimental comparison resulted in Manuscript to be reviewed higher percentages (up to 48%) of asymbiotic cells in exponential growth than when either nitrate (10-20%) was added or no N was present in the medium (10-20%) but the strain was not grown free of its symbiont (Fig. <ns0:ref type='figure'>6</ns0:ref>).</ns0:p><ns0:p>A symbiont-free strain of H. hauckii was maintained from March 2017 to August 2017 on a solely nitrate enriched, natural seawater medium (MET-44; Sch. Ammonium concentrations in the aged stock seawater were 0.5 µM or less. When isolated and growing, it was confirmed in April 2017 to be symbiont-free by epifluorescence microscopy and maintained in a seawater-based culture medium (MET-44) that would not support the DDA strains. The strain was lost during Hurricane Harvey in August 2017 and no further information was collected.</ns0:p></ns0:div>
<ns0:div><ns0:head>Discussion</ns0:head><ns0:p>Physiology and rate measurements of Hemiaulus symbioses have previously been limited</ns0:p><ns0:p>to field collected and incubated samples. Using a modification of an artificial seawater medium, we have successfully and reproducibly cultured two species of Hemiaulus with their symbiont.</ns0:p><ns0:p>Caputo et al. ( <ns0:ref type='formula'>2019</ns0:ref>) also reported brief success using an artificial medium; <ns0:ref type='bibr' target='#b34'>Hilton et al. (2013)</ns0:ref> reported genetic sequences from Richelia extracted from Hemiaulus grown using these methods. Greatest isolation success was found when the cells were rapidly removed from the net tow cod-end, suggesting sensitivity to the various exudates found in these concentrated samples.</ns0:p><ns0:p>In addition, the seawater was sterile filtered rather than autoclaved or pasteurized. Sterile filtration leaves the carbonate system and medium pH unaltered compared to heat treatment; however, viruses are not inactivated. Little is known of virus/DDA interactions, but viruses play</ns0:p><ns0:p>PeerJ reviewing PDF | (2019:08:40471:2:0:NEW 8 Sep 2020)</ns0:p><ns0:p>Manuscript to be reviewed a significant role in diatom mortality in general <ns0:ref type='bibr' target='#b43'>(Kranzler et al. 2019)</ns0:ref> and could be a problem for stable cultures.</ns0:p><ns0:p>In addition, culture media designed to support phytoplankton may not support essential phycosphere components or may support difficult to remove lethal bacteria (see van Tol et al.</ns0:p><ns0:p>2017 for an example). The inability to culture Hemiaulus in a seawater-based enrichment medium used for concurrent Rhizosolenia-Richelia cultures suggests that the additional trace metal and chelation in our modified YBCII medium was required for sustained growth or that water quality issues are critical. While we did not perform systematic comparisons, seawater from the Port Aransas pass is heavily influenced by both the inshore bays and coastal Gulf of Mexico. Our modified YBCII medium is free of these influences and we speculate provides a more consistent chemical environment. These difference highlights differing growth needs, sensitivities or tolerances of the Hemiaulus and Rhizosolenia DDAs that remain to be described.</ns0:p><ns0:p>The oscillation between rapidly growing, apparently healthy cultures and less vigorous cultures is clearly an impediment to sustained culture as is the lack of auxospore formation. None of the isolations persisted for more than ~ 3 years making detailed work on model strains problematic at this time. Previous estimates of N 2 fixation tracked 15 N isotope movement from the Richelia symbiont heterocysts to the host Hemiaulus cells using single-cell methods <ns0:ref type='bibr' target='#b20'>(Foster et al. 2011)</ns0:ref> and estimated that it was sufficient to support cell growth with a turnover time of up to 0.59 div <ns0:ref type='bibr'>Foster et al.'s (2011)</ns0:ref> rate measurements for H. hauckii-Richelia (n=17) averaged 20.4 ± 18.5 (std. dev.) fmol N heterocyst -1 h -1 (range 1.15-50.4 fmol heterocyst -1 h -1 ). These heterocyst normalized rates <ns0:ref type='bibr'>(Foster et al.'</ns0:ref>s Table <ns0:ref type='table'>1</ns0:ref> footnote), are lower than the rates observed in the cultures (up to 155 fmol N 2 heterocyst -1 h -1 ). Our culture growth rates (non-limiting Manuscript to be reviewed conditions) suggest some degree of limitation in their field collections. Light/ growth rate adaptation to 150-200 µmol m -2 s -1 PAR was concurrent with a maximum N 2 fixation rate approximately 6 times higher than the maximum rates observed by <ns0:ref type='bibr' target='#b20'>Foster et al. (2011)</ns0:ref>. Data digitized (Plot Digitizer from SourceForge, Slashldot Media, 225 W. Broadway, Suite 1600, San Diego, CA: https://sourceforge.net/) from Carpenter et al.'s (1999) Fig. <ns0:ref type='figure' target='#fig_15'>2</ns0:ref>, allows comparison of our symbiosis culture N 2 fixation rates to those from an Amazon River plume bloom where Hemiaulus DDA abundance reached 1.6 X 10 6 heterocysts L -1 . After extracting their N 2 fixation rates (as mg N m -2 d -1 ) and concurrent heterocyst abundance from Carpenter et al.'s (1999) Fig. <ns0:ref type='figure' target='#fig_15'>2</ns0:ref>, we determined that their rates ranged from 0.6-40.3 fmol N heterocyst -1 h -1 and were ~8 fold lower than the maximum rates seen in cultures (note the unit conversion and comparison: mg N, fmol N or fmol N 2 fixed). These rates were generated from material collected by net and then prescreened to remove Trichodesmium. Based on our isolation attempts, this handling probably adversely affected the rate. <ns0:ref type='bibr' target='#b11'>Carpenter et al. (1999)</ns0:ref> also reported undetectable nitrate uptake in the Hemiaulus DDA bloom, a result consistent with our culture observations that these DDA strains did not utilize nitrate. Many growth characteristics of H. hauckii-R. intracellularis are similar to the Rhizosolenia clevei-R. intracellularis symbiosis. Maximum growth rates are slightly less than 1 div d -1 and are similar between the two DDAs despite their significant size difference. Growth Manuscript to be reviewed expression and acetylene reduction in the Calothrix symbiont of the Chaetoceros DDA <ns0:ref type='bibr' target='#b19'>(Foster et al. 2010)</ns0:ref>.</ns0:p><ns0:formula xml:id='formula_0'>d -1 .</ns0:formula><ns0:p>The differential nitrate use by the Hemiaulus and Rhizosolenia DDAs is a significant difference between the two DDAs. Preferential NO 3utilization drove a higher host growth rate in a strain of the Rhizosolenia DDA, eventually leading to symbiont-free host cultures <ns0:ref type='bibr' target='#b57'>(Villareal 1990</ns0:ref>) growing solely on NO 3 -. In field studies where N 2 appeared to be the primary N source, the Rhizosolenia host and symbiont DDA were tightly coupled <ns0:ref type='bibr' target='#b32'>(Harke et al. 2019)</ns0:ref>. There are at least two mechanisms that could produce this result in the Rhizosolenia DDA: downregulation of Manuscript to be reviewed fixed N 2 even in the presence of combined DIN <ns0:ref type='bibr' target='#b44'>(Mills et al. 2020)</ns0:ref>. Thus, while unusual, the H.</ns0:p><ns0:p>hauckii-Richelia symbiosis is not unique. The truly intracellular location of the Hemiaulus symbiont <ns0:ref type='bibr' target='#b8'>(Caputo et al. 2019)</ns0:ref> clearly limits its contact with the environment and the potential impact of NO 3 -, but our observations also require the host Hemiaulus to be unresponsive to external nitrate. In contrast, Hemiaulus spp. (no information on symbionts) has been reported with growth rates up to 2.2 div d -1 <ns0:ref type='bibr' target='#b26'>(Furnas 1991)</ns0:ref> in field experiments and 3.8 div d -1 in nitrate-based laboratory medium <ns0:ref type='bibr' target='#b2'>(Brand & Guillard 1981)</ns0:ref>. While the symbiont presence is undocumented but seems unlikely given the DDA growth rates reported in our paper of <1.0 div d -1 as well as by modelled symbiont diazotrophy <ns0:ref type='bibr' target='#b38'>(Inomura et al. 2020)</ns0:ref>. The <ns0:ref type='bibr' target='#b26'>Furnas (1991)</ns0:ref> and <ns0:ref type='bibr' target='#b2'>Brand & Guillard (1981)</ns0:ref> reports, as well as our briefly established asymbiotic strain on medium that would not support Hemiaulus DDA growth, all suggest that symbiont-free strains of Hemiaulus are extant in the modern ocean. Hemiaulus DDAs had an ancestral origin 50-100 million years ago <ns0:ref type='bibr' target='#b8'>(Caputo et al. 2019)</ns0:ref>, but asymbiotic H. hauckii strains apparently still persist in the modern ocean. We suggest this data supports, but does not prove, that the Hemiaulus DDA, with its close metabolic coupling of the host-symbiont nitrogen metabolism <ns0:ref type='bibr' target='#b25'>(Foster & Zehr 2019;</ns0:ref><ns0:ref type='bibr' target='#b34'>Hilton et al. 2013)</ns0:ref>, is obligate and that symbiotic host Hemiaulus spp. are distinct from asymbiotic Hemiaulus strains. These asymbiotic strains should provide an invaluable tool for examining evolutionary processes in DDAs.</ns0:p><ns0:p>The growth rate and N 2 fixation results provide useful input to models examining the biogeochemical impact of the Hemiaulus DDA blooms in oceanic regions. The Amazon River plume is particularly noteworthy in that it has an explicit model describing the ecologicalbiogeochemical impacts. Manuscript to be reviewed growth rates > 1 div d -1 with asymbiotic cells growing on ambient N at somewhat greater rates.</ns0:p><ns0:p>Our experimental results are much lower for growth on N 2 (maximum ~0.9 div d -1 ) and indicated no nitrate use. Non-diazotrophic, asymbiotic Hemiaulus growth rates from the literature are much higher than N 2 -based DDA rates. These are significant alterations in the input values available to <ns0:ref type='bibr' target='#b50'>Stukel et al. (2014)</ns0:ref> (2018) depending on the strain used. These higher rates are consistent with the mechanistic model of <ns0:ref type='bibr' target='#b38'>Inomura et al. (2020)</ns0:ref> in that host carbon fixation is substantial enough support to the symbiont N 2 fixation rates required for the unit DDA growth. This host derived carbon is likely to also be the reductant and energy source required to support the lengthy decline of N 2 fixation rates at the beginning of the scotophase noted in the diel experiment (Fig. <ns0:ref type='figure'>4</ns0:ref>). Further experimental verification is required. </ns0:p></ns0:div>
<ns0:div><ns0:head>Table 1(on next page)</ns0:head><ns0:p>Results from the modified hyperbolic tangent function growth rate-irradiance and hyperbolic tangent function N 2 -fixation-irradiance curve fit. Manuscript to be reviewed 0.93 5 7 110 0.99 1 µmol m -2 s -1 2 (fmol N heterocyst -1 h -1 )(µmol m -2 s -1 ) -1 3 (div d -1 )(µmol m -2 s -1 ) -1 4 (fmol N heterocyst -1 h -1 ) 5 div d -1</ns0:p><ns0:p>PeerJ reviewing PDF | (2019:08:40471:2:0:NEW 8 Sep 2020)</ns0:p><ns0:p>Manuscript to be reviewed . See text for details of the methodology for the 4 h and 5 pt average measurements. Error bars are standard deviation.</ns0:p></ns0:div><ns0:figure xml:id='fig_1'><ns0:head /><ns0:label /><ns0:figDesc>approximately 40-50% of rates (144-154 fmol N 2</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_2'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2019:08:40471:2:0:NEW 8 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_3'><ns0:head /><ns0:label /><ns0:figDesc>were attached to tubing to a cotton-plugged mouthpiece (to prevent seawater aspiration) of firepolished glass tubing. Mouth pipetting was used to carefully draw or expel the chain. If mouth PeerJ reviewing PDF | (2019:08:40471:2:0:NEW 8 Sep 2020) Manuscript to be reviewed pipetting is unacceptable, any form of fine control would provide adequate results. Multiple Hemiaulus chains were isolated into one well of a glass depression well plate (16 depressions) containing ~2 ml of medium per depression. Individual chains were then rinsed via serial transfer into other wells containing sterile medium. Extensive rinsing (5-6 rinses) of a single chain before isolating the next chain was much less successful than only 2-3 rinses before placing the chain into a tube of medium. When isolated directly into the N-deplete modified YBCII medium, contaminant growth was minimal even with only 1-2 rinses. These techniques resulted in symbiosis isolation free of other eukaryotes or cyanobacteria in ~30-40% of the attempts. Preliminary experiments used H. hauckii strain #9 isolated during Spring 2010. Subsequent H. hauckii experiments used isolates established during Fall 2010 (strain #22) and Fall 2011 (strain #83, #91, and #92). Hemiaulus membranaceus strain #82 was isolated in the Fall 2011. In all subsequent text, a strain designation indicates a culture of a host diatom containing one or more symbionts. While Hemiaulus hauckii strain #91 was used for most of the experiments, a single strain for the entire suite of experiments was not possible due to loss of the strain or the periodic loss of vitality noted in the results. Strains used are identified in the text and in TableS1.The Hemiaulus DDA could be isolated for short-term growth into MET-44 (Schöne andSchöne 1982) nutrient-amended sterile filtered seawater (0.22 µm filter equipped commercial sterile filtration units) collected at the isolation point. However, the Hemiaulus-Richelia symbioses required re-isolation from the MET-44 medium into the modified YBCII medium for successful maintenance >2-3 weeks. After isolation, cells were placed in a 25 °C incubator under cool white fluorescent illumination of 150-250 µmol m -2 sec -1 on a 12:12 Light:Dark (L:D) cycle. All cultures were grown as batch cultures. Cultures had a high rate of sudden PeerJ reviewing PDF | (2019:08:40471:2:0:NEW 8 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_4'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2019:08:40471:2:0:NEW 8 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_5'><ns0:head /><ns0:label /><ns0:figDesc>cultures were grown at the 7 experimental light levels for 7 days and remained at the assigned light level through the duration of the experiments. Symbiosis growth is used throughout this paper to refer to increases in host diatom numbers containing at least one symbiont.For N 2 fixation, strains were adapted to either 50, 150, or 200 (high light HL) µmol m -2 sec -1 at 25°C and a salinity of 35 under cool white fluorescent lighting for 7 days prior to the acetylene reduction assay. Each adaptation level was then exposed to 7-8 light levels for acetylene reduction assay Diel pattern of N 2 fixation PeerJ reviewing PDF | (2019:08:40471:2:0:NEW 8 Sep 2020) Manuscript to be reviewed H. hauckii strains #22 and #92, and H. membranaceus strain #82 were used for the diel study (12:12 L:D cycles at 200 µmol m -2 sec -1 ) examining the daily rhythm of N 2 fixation on culture medium with no added N. Initial experiments on H. hauckii strain #22 utilized a set of 6 discrete time points between 0600 to 2100. Each incubation lasted 4 hours with initial and final measurements taken in triplicate. Rates were normalized to heterocysts and used the center point of the 4 h incubation period as the time stamp. Subsequent experiments on H. hauckii strain #92</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_6'><ns0:head /><ns0:label /><ns0:figDesc>hauckii) and 78 (H. membranaceus) separate measurements were plotted against time using a 5-point running average (center point plus two on either side) to smooth the data. Rates from different vial series overlapped in time, thus the PeerJ reviewing PDF | (2019:08:40471:2:0:NEW 8 Sep 2020)Manuscript to be reviewed 5-point average has rates from independent time series. Standard deviation was calculated on this 5-point series recognizing this is not a statistically useful value but only a metric for the noise in the data. Experimental cultures were adapted to at 25° C under 200 µmol m -2 s -1 illumination (cool-white fluorescence bulbs) on 12:12 LD cycle. Experimental vials were incubated under these same conditions. Samples during the scotophase were collected/returned to the incubator in a darkened container and shielded from the dimmed laboratory lights during the assay.Nutrient addition experimentsNitrogen source experiments addressed the effect of various inorganic N sources on symbiosis growth and N 2 fixation. In these experiments, H. hauckii strain # 83 was transferred to three 2 L autoclaved glass Erlenmeyer flasks containing the maintenance medium listed above amended with one of the following nitrogen sources: no added nitrogen (control), added nitrate (40µM) or added ammonium (10µM). Samples were maintained at 25 °C and a salinity of 35. Reduced ammonium concentrations were used to avoid toxicity effects; the nitrate concentration duplicated work on the Rhizosolenia-Richelia symbiosis<ns0:ref type='bibr' target='#b56'>(Villareal 1989)</ns0:ref>. Nutrient concentrations and cell abundance were sampled 10 times throughout the duration of the 20-day experiment. Nutrient analyses and cell counts were done in duplicate.Curve-fitting and StatisticsLight-dependent growth was fit to the Jassby-Platt hyperbolic tangent function(Jassby & Platt 1976) with a y-intercept term to permit calculation of compensation light intensity. The yintercept term was omitted for the N 2 fixation rates versus irradiance curves due to timedependent decline in dark N 2 fixation that became evident in the diel measurements. When not omitted, the time-dependent decline in dark N 2 fixation noted in the diel experiment at thePeerJ reviewing PDF | (2019:08:40471:2:0:NEW 8 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_7'><ns0:head /><ns0:label /><ns0:figDesc>point-to-point time series rates into a general diel curve. Two separate experimental treatments (the 4-hour incubations and the 5-point averaging series) indicated the maximum acetylene PeerJ reviewing PDF | (2019:08:40471:2:0:NEW 8 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_8'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2019:08:40471:2:0:NEW 8 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_9'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2019:08:40471:2:0:NEW 8 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_10'><ns0:head /><ns0:label /><ns0:figDesc>rates are not photoinhibited up to 500 µmol photons m -2 s -1 . Rapidly growing cells form extensive chains. Culture agitation, albeit qualitatively measured, negatively affects chain formation and possibly growth rates. The diel pattern of nitrogen fixation in the Hemiaulus DDA cultures parallels the diel nifH nitrogenase gene expression seen in field samples of both Hemiaulus DDAs<ns0:ref type='bibr' target='#b70'>(Zehr et al. 2007</ns0:ref>), the Rhizosolenia DDA<ns0:ref type='bibr' target='#b32'>(Harke et al. 2019)</ns0:ref> and both gene PeerJ reviewing PDF | (2019:08:40471:2:0:NEW 8 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_11'><ns0:head /><ns0:label /><ns0:figDesc>symbiont diazotrophy by exposure to NO 3due to its extra-plasmalemma location and/or induction of host nitrate reductase pathways. The latter would result in diminished carbon flow to the symbiont in order to support nitrate assimilation into protein. Neither of these mechanisms appear to have occurred in the H. hauckii DDA strains we used. These results were replicated in individual experiments four years apart on different strains, excluding the possibility that the results were a laboratory condition artifact. For the Hemiaulus DDA, either nitrate cannot be used or diazotrophic supply exceeded any immediate N demand by the symbiosis and suppressed NO 3uptake. In contrast, ammonium was used and resulted in elevated percentages of symbiontfree hosts, but not a symbiont-free culture. The free-living marine cyanobacterium Trichodesmium can use NO 3either preferentially or concurrently during diazotrophy as an N source<ns0:ref type='bibr' target='#b35'>(Holl & Montoya 2005;</ns0:ref><ns0:ref type='bibr' target='#b42'>Klawonn et al. 2020;</ns0:ref><ns0:ref type='bibr' target='#b45'>Mulholland & Capone 2000)</ns0:ref> and other diazotrophs can simultaneously use N 2 and NO 3 -<ns0:ref type='bibr' target='#b37'>(Inomura et al. 2018)</ns0:ref>. It is unusual for NO 3not to be used at all due to the higher overall energetic cost of nitrogen fixation added to the costs maintaining specialized cellular structures in diazotrophs<ns0:ref type='bibr' target='#b37'>(Inomura et al. 2018)</ns0:ref>. However, in the UCYN-A/haptophyte symbiosis, the host haptophyte only assimilates diazotrophicallyPeerJ reviewing PDF | (2019:08:40471:2:0:NEW 8 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_12'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2019:08:40471:2:0:NEW 8 Sep 2020)Manuscript to be reviewed When comparing rates, the possibility of strain-specific variation betweenFoster et al.'s (2011) Pacific Ocean collections,<ns0:ref type='bibr' target='#b39'>Carpenter et al.'s (1999)</ns0:ref> field collections and our Gulf of Mexico isolations cannot be excluded. Symbionts of the 3 diatom host genera have diverged with strong host specificity within diatom host genera<ns0:ref type='bibr' target='#b24'>(Foster & Zehr 2006;</ns0:ref><ns0:ref type='bibr' target='#b39'>Janson et al. 1999)</ns0:ref>.Bar<ns0:ref type='bibr' target='#b1'>Zeev et al. (2008)</ns0:ref> noted evidence of seasonally varying Hemiaulus-DDA dominated Richelia clades in the Mediterranean but there is little data to assess how physiological characteristics vary with habitat. Rhizosolenia and Hemiaulus DDA symbionts appear limited to vertical transmission during division or possibly transmission during auxosporulation (Foster & Zehr 2019) raising the possibility of genetic drift of various degrees within populations (Bar-Zeev et al. 2008). Conclusions Two symbiotic associations between host diatoms and their intracellular heterocystous cyanobacterium (Hemiaulus hauckii -Richelia intracellularis and Hemiaulus membranaceus-Richelia intracellularis) were successfully cultured for up to 3 years on artificial seawater medium. The N 2 -fixation and growth rate data provided here are, to our knowledge, the first published laboratory-based data for the Hemiaulus DDA. This work provides details on isolation techniques that proved key to successful culturing. The symbioses are sensitive to handling, requiring rapid collection and isolation for successful growth. The cultures did not undergo sexual reproduction, and the lack of auxosporulation and concurrent size increase is a barrier to long-term stable culture. Both symbioses grow without added nitrogen other than dissolved N 2 and are supported at maximum growth rates solely by symbiont nitrogen fixation. Maximum growth rates of the intact diatom-cyanobacterium symbiosis are < 1 div d -1 and are similar to the PeerJ reviewing PDF | (2019:08:40471:2:0:NEW 8 Sep 2020)Manuscript to be reviewed reported rates for another diatom-cyanobacterium symbiosis (Rhizosolenia clevei-Richelia intracellularis). Unlike the Rhizosolenia clevei-Richelia intracellularis symbiosis, the H. hauckii -Richelia intracellularis symbiosis does not assimilate nitrate. Nitrogen fixation by the heterocystous symbiont while within the host diatom has a clear diel pattern with maximum rates occurring during the photophase. The culture nitrogen fixation rates are consistent with field measured rates; however, maximum culture rates are ~6-8 times previously measured field rates. Both growth and nitrogen fixation rates follow light saturation kinetics. These data provide direct input for parameterization of light-dependent growth and nitrogen fixation in biogeochemical models.Both literature reports and our isolation of a nitrate-utilizing, symbiont-free Hemiaulus culture are consistent with distinct symbiont-free and symbiont-containing lines of the diatom Hemiaulus. If correct, these different lineages would be useful models for understanding the evolution of these symbioses in diatoms.</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_13'><ns0:head /><ns0:label /><ns0:figDesc>The three N 2 -fixation experiments were adapted to the given light levels for 7 days. The growth rate experiments were adapted at each of the light levels for 7 days. Strain 91 was used in these experiments PeerJ reviewing PDF | (2019:08:40471:2:0:NEW 8 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_14'><ns0:head>Figure 1</ns0:head><ns0:label>1</ns0:label><ns0:figDesc>Figure 1</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_15'><ns0:head>Figure 2 Typical</ns0:head><ns0:label>2</ns0:label><ns0:figDesc>Figure 2</ns0:figDesc></ns0:figure>
<ns0:figure type='table' xml:id='tab_1'><ns0:head /><ns0:label /><ns0:figDesc>Hemiaulus and Rhizosolenia DDAs are noted. incoming tide) or whole water samples (incoming tide). The net tow sample was collected from a platform under a pier laboratory that both shaded the sample from direct sun the entire time as well as facilitating numerous short tows resulting in dilute samples. Both initial isolations and</ns0:figDesc><ns0:table><ns0:row><ns0:cell>subsequent cultures of symbiont containing Hemiaulus were maintained in sterile filtered, fixed</ns0:cell></ns0:row><ns0:row><ns0:cell>N-free YBCII media with L1 trace metal/ EDTA additions (Ohki et al. 1992; Chen et al. 1996;</ns0:cell></ns0:row><ns0:row><ns0:cell>Guillard & Hargraves 1993), and final concentrations of 1 µM sodium glycerophosphate</ns0:cell></ns0:row><ns0:row><ns0:cell>(C 3 H 7 Na 2 O 6 P), 2.6 µM sodium dihydrogen phosphate monohydrate (NaH 2 PO 4 • H 2 O) and 35.7</ns0:cell></ns0:row><ns0:row><ns0:cell>µM sodium metasilicate (Na 2 SiO 3 • 9H 2 O).</ns0:cell></ns0:row><ns0:row><ns0:cell>Methods and Materials</ns0:cell></ns0:row><ns0:row><ns0:cell>All culturing was conducted at the University of Texas Marine Science Institute (UTMSI)</ns0:cell></ns0:row><ns0:row><ns0:cell>in Port Aransas, Texas. Hemiaulus strains containing symbionts were isolated by micropipette</ns0:cell></ns0:row><ns0:row><ns0:cell>(Andersen & Kawachi 2005) from the Port Aransas ship channel (27° 57' 17.56' N, 90° 03'</ns0:cell></ns0:row><ns0:row><ns0:cell>00.48' W) using material from either net tows (20-35 µm mesh nets, 1-3 minute tows in the</ns0:cell></ns0:row><ns0:row><ns0:cell>PeerJ reviewing PDF | (2019:08:40471:2:0:NEW 8 Sep 2020)</ns0:cell></ns0:row></ns0:table></ns0:figure>
<ns0:figure type='table' xml:id='tab_3'><ns0:head /><ns0:label /><ns0:figDesc>. In addition, our results for H. hauckii DDAs found no evidence of growth rate photoinhibition at the highest light level used (500 µmol m -2 s -1 ). While instantaneous solar PAR may reach ~2,000 µmol m -2 s -1<ns0:ref type='bibr' target='#b3'>(Björkman et al. 2015)</ns0:ref> at Station ALOHA near Hawaii (22° 45' N 158° 00' W), average daily PAR incident at Sta. ALOHA over the diurnal is ~850 µmol m -2 s -1 from June-Aug. (calculated fromLetelier et al. 2017). Vertical mixing rates will both reduce the time averaged PAR exposure exponentially with the depth of mixing as well as being</ns0:figDesc><ns0:table /><ns0:note>rapid enough to preclude general phytoplankton photoacclimation<ns0:ref type='bibr' target='#b53'>(Tomkins et al. 2020</ns0:ref>). Thus, it seems possible that in-situ PAR values would not photoinhibit these DDA strains. However, damaging effects by solar UV wavelengths(Zhu et al. 2020) require further examination. Follet et al. (2018) and Inomura et al. (2020) utilized H. hauckii DDA growth rates extracted from Pyle (2011) for modelling applications. Our report presents the full range of data in Pyle's work and notes that rates can be ~ 0.2 div d -1 higher than the values used by Follet et al.</ns0:note></ns0:figure>
<ns0:note place='foot'>PeerJ reviewing PDF | (2019:08:40471:2:0:NEW 8 Sep 2020)</ns0:note>
</ns0:body>
" | "Dear Dr. Kormas,
I have examined the 3 reviews and made corrections to the manuscript. Only Reviewers 2 and 3 requested changes, so I have truncated the comments from Reviewer 1. My responses are in red italics below the reviewers’ comments. The late review from Reviewer #3 was quite useful and substantially improved the manuscript by noting a variety of typographical and grammatical inconsistencies. My only significant change was to move the analytical methods section earlier in the methods in response to the comments by both Reviewer 2 and 3.
Thank you for your consideration of this manuscript.
Regards, Tracy Villareal
Reviewer 1 (Anonymous)
Comments for the Author
The authors have addressed all of my noted issues in the revision.
Reviewer 2 (Anonymous)
Basic reporting
Description of the experimental design, especially for N2 fixation measurements, is much more clear now. Thank you.
Specific comments:
L17: Hyphenate compound adjectives (“nitrogen-fixing”) but not nouns (“nitrogen fixation” or “nitrogen fixers”).
Done
L55: Insert a space between “fixation” and “(.”
Done
L56: Diazotrophy has been reported in the Arctic Ocean as well. See Harding et al, 2018, PNAS 115 (52) and other publications regarding UCYN-A.
added
L111: N2 abounds, so perhaps you should specify “fixed N.”
Good point. I have made a more global correction in the text specifically noting how the term N-Free is used.
L122: Either use a comma on both sides of “where examined” or leave them out completely.
Done
L140: Insert degree sign after “27” and a comma between lat and lon.
Done
L156: I think rinsing implies discarding, so you don’t need to state it.
Done
L252 - 266: Put the acetylene reduction assay description with the rest of the N2-fixation description.
Reviewer 3 commented on the organization as well, so I have moved the analytical methods section before the experimental design. I prefer not to split the methods up. If I put the ARA assay as part of one of the two sections of N-fixation experiments, then I have to move the chlorophyll and nutrient analysis sections as well. It creates a more cumbersome flow to me as well as forcing a reader to sort through various sections for methods that are used in multiple (but not all) experiments.
L329 - 330: “Light-saturated growth” should be hyphenated while “light saturation” should not be.
corrected
L363: “Symbiont-free”
4 changes made in text
L366: Insert “microscopy.”
done
L393 - 395: Be consistent in using “fmol N2.”
This is one of the unfortunate aspects of the acetylene reduction assay when reporting the units. The 4:1 conversion ratio yields moles N2 reduced, which equals two moles of N reduced. Foster et al. (2011) report their units in fmol N and we must assume they converted from N2 fixed to N fixed. We retained their units in the text when reporting their data for transparency. I have added a parenthetical comment for readers reminding them of that 1 mole of N2 = 2 moles of N. It’s a concept easily overlooked or ascribed to a typographical error. I do thank the reviewer for pointing this out. I doubled checked our calculation based on the Carpenter et al. (1999) data. They reported mg N fixed m-2 d-1 and during the conversion to moles the units were accidently switched to fmol N2 rather than fmol N. This has been corrected in the text along with relevant comparisons. It does not affect the results or conclusions.
Table 1: State the strains used.
done
Figure 2: Put the strain name in context somewhere in a sentence rather than a free-standing phrase. And capitalize “chlorophyll” after “B.”
Done
Figure 3: Be more explicit about the circles. From the graph it looks as if they extend the experiment with the closed squares rather than the open squares.
done
Figure 4: State what the error bars represent.
done
Experimental design
With the clearer description of the methods, I no longer have reservations about the experimental design.
L164: That’s a good success rate! I applaud you.
L171 - 175: Any hypotheses on why YBCII worked better?
I have many guesses, but not enough systematic data to really say much. Once we found something that worked, we focused on collecting data, not testing other media. I have expanded the speculation on water quality since the reviewer asked
Validity of the findings
L319 - 321: I find it interesting that the chl per symbiosis was the same. Is photoacclimation not generally a trait for these diatoms?
Agreed, but there’s not much I can say from this one experiment. I would expect photoacclimation to occur, but we didn’t address it directly in these experiments. The data that I did recover suggest that there is quite a range of chl per cell possible, but I don’t have the concurrent metadata to put it in context (light levels, cell size, etc). There is also an interesting question of two autotrophic organisms responding to the light field in a complex host symbiont relationship.
L346 - 353: Is light or growth affecting N2 fixation rates? If you normalize N2 fixation rates to growth rates instead of volume or biomass, do you see a difference among the treatments?
These N2 fixation experiment incubations were short (only a few hours), so it is unlikely that growth rate changes took place on this time frame. Since growth rate is light-dependent as well, plotting steady state growth rates and N2 fixation would likely yield a strong positive correlation. This is a good question, but it requires parsing the energy flow to N2 fixation and growth in ways we can’t do with the data. We did not measure N2 fixation in cultures acclimated to different light levels like we did for growth rate. We have only two light levels to work with. That’s not enough to build a story around. Perhaps next time…
L396 - 399: Maybe it’s light directly, but maybe all metabolic processes are ramped up. Again, I encourage you to normalize N2 fixation rates to growth rates to compare. Or do you have a proposed mechanism for how light directly affects N2 fixation?
The energy supporting N2 fixation in heterocystous cyanobacteria is well documented to be derived from the light reactions of photosynthesis. The demand is high enough that very little N2 fixation occurs in the dark in these forms as noted in the diel study. There is, of course, a linkage to overall growth rate via synthesis of proteins, enyzmes, etc., as well as how the symbiont grows and divides. The nitrogen fixation light curve is analogous to carbon fixation rates. When a light acclimated culture is exposed to a range of experimental light levels, the P-E curve is follows the Jassby-Platt type of curve. The cultures are all growing at the same rate so normalizing the rates to growth rate is using a fixed number. It is not a useful analysis.
L407 - 408. Goodness, yes!
Comments for the Author
I’m looking forward to seeing more DDA isolates in the future.
Reviewer 3 (Anonymous)
Reviewer follow-up: The authors assert that they included the reference in their original
submission. I can see now that it was indeed in the text; however, it was not in their original list
of references. Here is an excerpt of the original list of references which does not include
Heinbokel.d
Thanks for the copied original text. The reference was included in the resubmission.
Reviewer follow-up: Table S1 is an excel spreadsheet with various pieces of data reported for
at least four different strains (22, 91, 92, 82, 83), and doesn’t provide details of these strains.
The authors make it clear that all strains and some of the data have been lost, but what would
be helpful is a simple summary table listing all the strains for which data are shown in the paper.
The table would include the isolation date and duration in culture, also indicating which strain
was used in which figure, what species each strain represents, and any info about differential
media used in culturing. For a new reader such summary table could save a lot of time and
effort.
This information is in the text, but I have added fields in Table S1 that list the requested information. The records of the specific dates of isolation as well as when it was lost are not available. Unfortunately, our paper notebooks didn’t like 10 days at 100% humidity and 90° F plus temperature in our post-Harvey laboratory without power. Mold does.
Reviewer follow-up: I found the physical address quickly after navigating to the company page.
Ultimately, whether a web address is allowable is up to the editors. Most journals do not allow
them.
I just searched for the software, not the source company! For some reason, it did not have a link to the source company’s address. I thank the reviewer. I’ve included the physical address as well as the link. Software companies don’t have much physical infrastructure to move, so physical site fidelity is not a big thing to them. They are less likely to change a URL .
L17 delete one ‘source’
done
L20 Change wording here for accuracy. ‘Amazon River’ is not a river plume.
done
L27 ‘cyanobacteria symbiont’ seems linguistically incorrect. Cyanobacteria is plural, symbiont is
singular.
Corrected to make symbiont plural
L29-30 Standard unit for reporting growth rate is d-1, not ‘div d-1’.
There is no universally accepted standard for this. DIv d-1 is an intuitive biological number (doublings per day). D-1 is a specific growth rate (base e) that is useful in modelling. They are easily interchanged.
L32 Nitrogen fixation should not have a dash.
Done,
L45 I don’t understand what is meant by ‘with the latter part of an obligate symbiosis’
It was referring to the latter of the asymbiotic, symbiotic Hemiaulus list in the sentence. I’ve changed the text to clarify.
L66-67 Here the discussion is branching into any coccoid cyanobacterial symbionts in
phytoplankton, beyond those in diatoms. In this context statement the ‘little is known about the
characteristics of the coccoid symbionts’ looks odd, because extensive research exists on the
UCYN-A lineage of symbiotic, coccoid cyanobacteria that are hosted by phytoplankton.
Quite valid and shows a bit of tunnel vision on our part. I’ve streamlined the text to stay within the bounds of diatom symbioses.
L81-84 Revise the sentence for clarity.
L89 Since this citation from the 1970s, we have learned a lot about diazotrophs in the ocean,
including the fact that some key oceanic cyanobacterial diazotrophs do not fix CO2 (UCYN-A)
and some oceanic N2 fixation may be heterotrophic bacteria, thereby not CO2 fixers.
I have modified the sentence to be specific to photosynthetic N2 fixation that Eppley and Peterson were referring to.
L100 Tmols > Tmol
Done
L108-109 reword sentence. Currently “growth rates grows up to …” is unclear and
grammatically incorrect.
Done
L109-110 wording is awkward: ‘growth of host and symbiont are uncoupled’.
Done
L112-113 I suggest wording edits here to clarify. Rather than ‘to outgrow the symbiont’, this
observation shows that the host replaces the N source from symbiont’s N2 fixation with nitrate.
The statement misleadingly appears to suggest the host and symbiont are competing of nitrate,
and the host wins, outcompeting the symbiont.
Done
L118-119 Grammar is off here.
Done
L122 add comma after ‘where examined’
done
L145 If EDTA is not part of a typical YBCII, state how much was added. If it is part of typical
YBCII, no reason to single it out from the rest of the media components.
The EDTA addition is part of the L1 formulation. I’ve clarified this by removing the “and” and inserting “/”.
L147 Sodium glycerophosphate chemical formula is not given although it is given for the other
media components.
Corrected
L148 You state ‘additions’: are these final concentrations in the media?
Text has been modified to reflect final concentrations.
L154 How was salinity measured?
With a refractometer. I do not have the make/model.
L156 delete ‘, the rinsed discarded,’
Done
L157-158 ‘rapidly isolated’ How exactly? Were you using a dissecting scope or was this done by
viewing samples by naked eye? Was micromanipulator used?
A stereomicroscope has been specified and more details on the isolation specified.
L158 ‘Depression well’ needs to be clarified. Is this a multiwall plate? If so, what type plastic? Details like these matter in future isolation attempts.
Yes, they do. So much of this is second nature that it is sometimes hard to remember to specify such things. The text has been modified.
L175 The statement: ‘successful maintenance >2-3 weeks’ is unclear. Do you mean: ‘successful
maintenance of 2-3 weeks’ or ‘up to 2-3 weeks’? Still, this appears incorrect as later you state
you kept cultures up to 2 years.
This line was part of the description of how sweater based MET-44 medium was not suitable for long-term maintenance of cultures. I have clarified this.
L182 You say bacteria were rarely visible. How did you check?
Specified in the text now as phase contrast or DIC optics.
L196 You state ‘7 experimental light levels’; however, one of the strains (solid squares in Fig 3)
was grown at 2 light levels.
The text has been modified as per this comment and reviewer 2.
L211 I see that you added description of acetylene reduction assay later in the methods;
however, understanding this section about baseline, ethylene production etc., requires that an
unfamiliar reader has to first jumped ahead to read the AR method details. You may consider
re-organizing.
Since Reviewer 2 commented on this as well, I have moved the Analytical Methods section before the experimental design section.
L213-216 Reword the sentence for clarity and grammar.
Done
L216 ‘The second was’ ? The sentence is missing something. Do you mean: “The second
reason for this approach was…”?
done
L219 I am confused about the labeling. Why are there vials a and b? Were things run in
duplicate? Were 1a and 1b sampled at different times? What do you mean ‘then repeated’. If
the vials are all distinct and serve the same role in the time series, why use a and b
designations?
Vials a and b were started at approximately the same time. They are distinct time series, but separated temporally from the next set by several hours. This is the convention used in both diels by the student and I am reluctant to change it should I ever need to access the dataset again.
What is the difference between the ‘Time’ and ‘Begin time’ columns? The times, if indicated in
both, are always the same, but or many vials ‘Begin time’ is missing.
Begin time is when sampling of series HK 2a, 3a, 1a, etc. began. The time column is the time for the sampling of all the vials. It allows one to know when the series began without having to sort through the entire time series. I’ve modified the column header to reflect this.
L265 ‘injected sampled’?
Text has been clarified
L294 repeated > several
I’ve changed this to multiple. We had 30+ strains going at one point, so several does not quite capture how well this worked.
L312 ‘bacteria were not evident’ – what observation is this based on?
In the Methods, we note now we used phase contrast and DIC optics for this.
L328 ‘Article S1’? Are you referring to Supplementary text S1?
Article S1 is the name of the supplementary text submitted. From my read of the instructions, that’s what I am supposed to call it.
L328 ‘from’ should be followed by ‘to’
I don’t understand why this makes sense. I do see the error in the sentence and have corrected it.
L331 Explain in methods how you calculated Ic. You explain Ec in Supplement S1. Is this Ic?
Yes, a change that didn’t get made in the text. It is corrected now.
L334-335 ‘was necessary to smooth the variability’ - It is good that you state this for
transparency, but know that someone may argue that you are massaging your data by this
averaging. Looking at the data in Table S1 indeed a lot of variability is hidden by this averaging,
although it is shown with the error bars so should be ok. Why do you think there is so much
variability? Is this a sampling or injection issue, culture aggregation, GC standardization, errors
in heterocyst counts, result of repeated subsampling of the same vials, etc.?
It was an annoying amount of variability due to all of the above. I don’t know how to parse this into the components. The general patterns are robust and that’s the point. To discuss it would require multiple paragraphs to no real point, I think. Considering that it is completely accepted to log or square root transform biological data for statistical purposes (which I find to be meaningless), this is (as the reviewer notes) very transparent. I would not do it this way again, though!
L363-368 These statements are anecdotal as no data are shown to support these statements. I
suggest omitting this part.
They are anecdotal only in the sense we don’t have zeros and dates. I have added text that they culture was confirmed to be symbiont-free approximately a month after isolation. We have reported our observations, observations which are supported by the literature we cite in the discussion. Readers can take it or leave it as they see fit, but it is a consistent with the literature on this genus. Should these lineages exist as we suggest, they are a very important tool in understanding symbioses in diatoms. Researchers have to know they can exist to look for them, though.
L373 “two species of Hemialulus with THEIR symbiontS”.
Done
L380 “these concentrated sampleS”
Done
L400 If you navigate to the SourceForge site you will find a physical address. This is up to the
journal, but generally website address as a citation is unacceptable.
https://sourceforge.net/about
done
L471-473 Some issues with sentence structure and/or missing words here. The sentence is
unclear.
I wish the reviewer were more clear. This sentence makes sense to me as an oceanographer.
I have modified the sentence some, but I’m not sure of what is unclear.
L478 rates could presumably also be lower depending on strain used. Or are you saying rates
can be 0.2 d-1 greater than those reported in Pyle?
I have modified the sentence for clarity.
L478-479 with ‘these higher rates’ are you referring to rates from this study?
Sentence has been clarified
L486 Unclear what you mean by ‘three clades’. Do you mean genera, species, or strains, or
something else? Replace ‘clades’ to clarify what you mean here.
I used the language of the paper cited. I have modified the latter part of the sentence some for clarity, but I do not wish to change the language of the authors.
L487 It’s either: “within a GENUS” or “AMONG the genera”. Unclear if you are talking about
Hemiaulus here specifically, or including Rhizosolenia here. Seems you should say “within each
genus”?
this has been clarified
L489 Unclear what you mean by ‘clades’? All clades? Most Richelia were in Hemiaulus? Or
most Richelia found represented clades known to associate with Hemiaulus? Please clarify.
clarified.
L490-491 Do you mean “vertical transmission”? Alternatively, if you are referring to movement
of symbionts among genera or within genus among cells at the population level it would be
“horizontal transmission”.
Yes, vertical transmission. Not sure where exchange came from. Foster and Zehr cited my work for this point, so I do know better!
L496-497 They may be, but I would modify it with “…to our knowledge, the first published
data…”.
done
L499 “… have been successfully cultured”. This statement makes it sound like you have a
culture you can now share. I suggest “…were successfully isolated and kept in culture for up to
3 years.”
done
L510 >”3-4 times previously reported field rates.”
This has been modified with the text and also reflecting the molar units issue mentioned.
The conclusions section is mostly repeating discussion. The section would benefit from addition
of a few overall synthesis statements about importance of this work.
I have slightly modified the conclusions to highlight the significance of the data and observations
There are two versions of Table S1. One of these is included in the ‘Article S1’. The second
one, with different content, is one of the Excel files.
The Table 1 in the Article S1 has been removed.
Figure 2. What was n? Were these run in triplicate? I can’t find the information in Table S1
either.
Duplicates. I’ve added this to the figure legend.
" | Here is a paper. Please give your review comments after reading it. |
9,751 | "<ns0:abstract xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:p>There is growing evidence that herbivory-induced legacy effects permit plants to cope with herbivory. However, herbivory-induced defense strategies in plants against grazing mammals have received little attention. To further understand the grazing-induced legacy effects on plants, we conducted a greenhouse experiment with Leymus chinensis experiencing different grazing histories. We focused on grazing-induced legacy effects on above-ground spatial avoidance and below-ground biomass allocation. Our results showed that L. chinensis collected from the continuous overgrazing plot (OG) exhibited higher performance under simulated grazing in terms of growth and cloning ability than those collected from the 35-year no-grazing plot (NG). The enhanced adaptability of OG was attributed to increased above-ground spatial avoidance, which was mediated by larger leaf angle and shorter height (leaf angle contributed to the above-ground spatial avoidance at a lower herbivory stubble height, while height contributed to above-ground spatial avoidance at a higher herbivory stubble height). Contrary to our prediction, OG preallocated less biomass to the rhizome, which does not benefit the herbivory tolerance and avoidance of L. chinensis; however, this also may reflect a tolerance strategy via shorter rhizome and more tiller ramets .</ns0:p></ns0:div>
</ns0:abstract>
<ns0:body xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:head>INTRODUCTION</ns0:head><ns0:p>Environmental disturbances (e.g., drought, herbivory) can have persistent effects on ecological attributes (e.g., ecological processes, community structure, population dynamics, and plant and soil characteristics) long after they occur (i.e., the legacy effect) <ns0:ref type='bibr' target='#b21'>(Fox et al., 2015;</ns0:ref><ns0:ref type='bibr' target='#b31'>Kafle & Wurst, 2019)</ns0:ref>. Legacy effects are ubiquitous phenomena in nature and have been extensively studied in the context of plant succession, herbivory, invasive plants, ecosystem engineering, and human land use <ns0:ref type='bibr' target='#b11'>(Cuddington, 2011;</ns0:ref><ns0:ref type='bibr' target='#b35'>Kostenko et al., 2012;</ns0:ref><ns0:ref type='bibr' target='#b65'>Wurst & Ohgushi, 2015)</ns0:ref>. In grassland ecosystems, herbivory-induced legacy effects on ecological processes, such as plant succession and biological diversity change, can persist for decades and even millennia <ns0:ref type='bibr' target='#b28'>(Holeski, Jander & Agrawal, 2012;</ns0:ref><ns0:ref type='bibr' target='#b21'>Fox et al., 2015)</ns0:ref>. Some studies have shown that legacies in plant defense strategies can mediate herbivory-induced legacy effects on ecological processes. For instance, defense traits (e.g., chemical defense substances, resource reallocation) can significantly affect the plant-herbivore relationship (generally increasing plant adaptability to herbivores), the interspecific relationship, and soil characteristics <ns0:ref type='bibr' target='#b65'>(Wurst & Ohgushi, 2015)</ns0:ref>. Therefore, clarifying the overgrazing-induced legacy effects in plant defense strategies is critical for understanding processes occurring in grazing ecosystems. However, most previous studies on herbivoryinduced legacy effects on plant defense strategies have focused on short-term insect herbivory, in contrast, studies examining long-term livestock grazing have received less attention by comparison <ns0:ref type='bibr' target='#b28'>(Holeski, Jander & Agrawal, 2012;</ns0:ref><ns0:ref type='bibr' target='#b31'>Kafle & Wurst, 2019)</ns0:ref>.</ns0:p><ns0:p>Strategies by which plants cope with herbivory include resistance, avoidance, and tolerance <ns0:ref type='bibr' target='#b15'>(Didiano et al., 2014)</ns0:ref>. While resistance strategies (e.g., thorns, higher tannin concentrations) play an important role in coping with insect herbivory <ns0:ref type='bibr' target='#b1'>(Agrawal et al., 2002;</ns0:ref><ns0:ref type='bibr' target='#b28'>Holeski, Jander & Agrawal, 2012)</ns0:ref>, they may be less useful in plants experiencing herbivory from livestock or other grazing mammals, as these large animals are unable to selectively graze at such a fine scale <ns0:ref type='bibr' target='#b45'>(Menard et al., 2002;</ns0:ref><ns0:ref type='bibr'>Benot et al., 2013)</ns0:ref>. Empirical evidence indicates that plants under grazing mammal herbivory show adaptive legacy effects via above-ground spatial avoidance traits which resulted in the distribution of more above-ground biomass close to the ground, including larger leaf angles, shorter height, and more prostrate growth forms <ns0:ref type='bibr' target='#b51'>(Polley & Detling, 1988</ns0:ref><ns0:ref type='bibr' target='#b52'>, 1990;</ns0:ref><ns0:ref type='bibr' target='#b49'>Painter, Detling & Steingraeber, 1993;</ns0:ref><ns0:ref type='bibr' target='#b58'>Tomás et al., 2000;</ns0:ref><ns0:ref type='bibr' target='#b38'>Li et al., 2015;</ns0:ref><ns0:ref type='bibr' target='#b55'>Ren et al., 2017)</ns0:ref>. However, studies examining the above-ground biomass vertical distribution have been qualitative-not quantitative. The lack of rigorous quantitative approaches severely limits our understanding of the role of these avoidance traits in the responses of plants to grazing mammal herbivory (e.g., livestock grazing). For example, the specific morphological characters that lead to the near-surface distribution of above-ground biomass to reduce the possibility of defoliation by grazing mammals remain unclear.</ns0:p><ns0:p>Biomass reallocation is a fundamental strategy for plants to cope with herbivory. When subjected to above-ground herbivory, more biomass is mobilized above-ground to facilitate the recovery of growth <ns0:ref type='bibr' target='#b40'>(Liu et al., 2018)</ns0:ref>, and a transient transformation of resources away from herbivores occurs within hours after herbivory <ns0:ref type='bibr' target='#b3'>(Anten & Pierik, 2010;</ns0:ref><ns0:ref type='bibr' target='#b48'>Orians, Thorn & Gomez, 2011)</ns0:ref>. Some studies have indicated that plant tolerance is tightly linked to biomass allocation patterns expressed before herbivory; that is, larger belowground biomass pre-allocation is associated with stronger tolerance of plants to above-ground herbivory <ns0:ref type='bibr' target='#b19'>(Fornoni, 2011;</ns0:ref><ns0:ref type='bibr' target='#b42'>Lurie, Barton & Daehler, 2017)</ns0:ref>. In addition to the importance of tolerance, large belowground biomass pre-allocation potentially helps plants avoid above-ground herbivory. There is growing evidence that long-term grazing induces higher belowground biomass allocation at the community and population levels <ns0:ref type='bibr' target='#b46'>(Milchunas & Lauenroth, 1993;</ns0:ref><ns0:ref type='bibr' target='#b39'>Lindwall et al., 2013;</ns0:ref><ns0:ref type='bibr'>)</ns0:ref>. Furthermore, more biomass allocation to roots for plants collected from grazing areas has been observed in the common garden environments free of herbivory disturbance <ns0:ref type='bibr' target='#b29'>(Jaramillo & Detling, 1988;</ns0:ref><ns0:ref type='bibr' target='#b51'>Polley & Detling, 1988</ns0:ref><ns0:ref type='bibr' target='#b52'>, 1990)</ns0:ref>. However, few studies have examined herbivory-induced legacy effects on plant belowground biomass allocation, especially the allocation of belowground reproductive organs such as rhizomes.</ns0:p><ns0:p>Leymus chinensis, a rhizomatous clonal plant, is the dominant species on the Inner Mongolia typical steppe grasslands. This grass species is highly relished by livestock (e.g., cattle, sheep) and has been subjected to overgrazing for more than 50 years. Similar to plants that have been studied in other regions, L. chinensis exhibits significant grazing-induced legacy effects, such as short height and short leaves <ns0:ref type='bibr' target='#b38'>(Li et al., 2015;</ns0:ref><ns0:ref type='bibr' target='#b55'>Ren et al., 2017)</ns0:ref>. However, no quantitative studies have characterized herbivory-induced legacy effects on the vertical distribution of above-ground biomass and belowground biomass allocation (especially the allocation of rhizome biomass). Although earlier studies <ns0:ref type='bibr' target='#b47'>(Oesterheld & McNaughton, 1988;</ns0:ref><ns0:ref type='bibr' target='#b41'>Loreti, Oesterheld & Sala, 2001)</ns0:ref> have shown that plants subjected to grazing disturbance showed stronger adaptability to herbivory compared with ungrazed plants in a common garden environment, studies reporting legacy effects underlying L. chinensis adaptation to grazing are limited.</ns0:p><ns0:p>To further characterize the grazing-induced legacy effects on plants, we conducted a greenhouse pot experiment with L. chinensis collected from two adjacent plots separated by a pasture fence. The first was a 35-year no-grazing plot and the second is a long-term overgrazing plot. Our study sought to answer three questions. First, does L. chinensis exposed to long-term overgrazing grazing disturbance exhibit enhanced above-ground spatial avoidance (measured by the above-ground biomass vertical distribution)? If so, which individual characteristics contribute to this trait? Second, are there overgrazing-induced legacy effects on L. chinensis in terms of the belowground biomass allocation (i.e., root and rhizome)? Third, does the L. chinensis collected from the grazing plot exhibit stronger adaptation to simulated herbivory compared with those collected from the no-grazing plot?</ns0:p></ns0:div>
<ns0:div><ns0:head>MATERIALS & METHODS</ns0:head></ns0:div>
<ns0:div><ns0:head>Materials</ns0:head></ns0:div>
<ns0:div><ns0:head>Description of studying sites</ns0:head><ns0:p>Samples of L. chinensis were collected from typical steppe grassland located at the Inner Mongolia Grassland Ecosystem Research Station (43 • 38' N, 116 • 42' E). The sampling sites comprise two adjacent plots separated by a pasture fence. The first was a no-grazing plot (600×400 m), which has been fenced since 1983 for long-term ecological observations, and the second was a continuously overgrazing plot (600×100 m) that has been grazed at a stocking rate of ~3 sheep units per hectare for more than 50 years. However, the stocking rate recommended by the local government is ~1.5 sheep units per hectare to achieve a balance between grassland productivity and livestock forage requirements. Thus, the continuously grazed plot has experienced heavy grazing pressure over the last several decades <ns0:ref type='bibr'>(Ma, et al., 2015)</ns0:ref>.</ns0:p></ns0:div>
<ns0:div><ns0:head>Materials Collection at the studying sites</ns0:head><ns0:p>Genotypes showed different phenotypic plasticity to environmental disturbance and different genetic structures may be one of the main mechanisms mediating overgrazing-induced legacies on L. chinensis <ns0:ref type='bibr'>(Josephs, 2018)</ns0:ref>. We could not determine the genotype of each L. chinensis individual collected from both plots. To reduce the possible impacts of genotypes on our experimental results, we conducted our experiment at the population level and used the highreplication sampling.</ns0:p><ns0:p>We sampled 150-segment rhizomes with a root drill at 150 random points in each of the two treatment plots at the beginning of the growing season. Given that the field site from which we collected materials was not free from pseudo-replication, we sampled the entire area for each plot except for the margin and ensured that the distance between every two sampling points was greater than 20 m; this was done both to improve the representativeness of the samples for each plot and to decrease the impacts of pseudo-replication on the experimental results. Each rhizome was approximately 4 cm long and contained at least one sprouting section of the buds. To prevent the sampled rhizome from losing its vitality following removal from the soil environment, we immediately transferred samples into a well-prepared moistened soil container.</ns0:p></ns0:div>
<ns0:div><ns0:head>Materials cultivation in the greenhouse</ns0:head><ns0:p>In the laboratory, each rhizome was dissected into 2-cm long with only one node. All of the rhizomes were cultivated in flowerpots, 20 cm in diameter and 15 cm high, which were kept in the greenhouse. Each flowerpot contained 3 kg of soil collected from points adjacent to the experimental site and was subsequently planted with one rhizome section. There were 150 flowerpots for rhizomes collected from the no grazing plot and the overgrazing plot, respectively. After 20 days, the rhizomes had sprouted in approximately 100 flowerpots from each treatment.</ns0:p></ns0:div>
<ns0:div><ns0:head>Experimental Design & Measurements Experimental design</ns0:head><ns0:p>The experiment was a full factorial design and consisted of two factors. The first was the source of L. chinensis (NG: L. chinensis collected from the no-grazing plot; OG: L. chinensis collected from the overgrazing plot), and the second was simulated grazing (CK: no simulated grazing; H8: simulated moderate grazing; H4: simulated heavy grazing). For NG and OG, we randomly selected 90 flowerpots that had sprouted rhizomes. These flowerpots were randomly arranged in the greenhouse, and we alternated their position every week to exclude the influence of external factors (e.g., light). One-third of the L. chinensis growing in flowerpots ( in both NG and OG separately) were randomly assigned to CK, H8, and H4. However, several L. chinensis died during the experiment and the numbers left for each treatment were NG*CK ( <ns0:ref type='formula'>22</ns0:ref>), NG*H8 (29), NG*H4 (25), OG*CK ( <ns0:ref type='formula'>22</ns0:ref>), OG*H8 (26), and OG*H4 (28). We clipped plants with scissors to conduct the simulated grazing as per <ns0:ref type='bibr' target='#b59'>Turley (2013)</ns0:ref> and <ns0:ref type='bibr' target='#b15'>Didiano et al (2014)</ns0:ref>. We simulated moderate and heavy grazing by clipping the above-ground part of plants 8 cm and 4 cm above the soil surface. The stubble height in the simulated grazing treatment was per <ns0:ref type='bibr' target='#b22'>Gao (2008)</ns0:ref>. Sloping parts (e.g., the sloping leaves and stems) were not straightened in the simulated grazing treatments. We only removed plant parts that were distributed above 8 cm or 4 cm in its natural state. The herbivory simulation was carried out three times at 45, 60, and 75 days after the rhizomes had sprouted to imitate repeated grazing in natural environments.</ns0:p></ns0:div>
<ns0:div><ns0:head>Measurements</ns0:head><ns0:p>L. chinensis genet consists of tiller ramets and rhizome ramets. Tiller ramets sprouted from the tiller sections (i.e., the initial rhizome section that transplanted to flowerpots) while rhizome ramets sprouted from the new rhizome section (Fig. <ns0:ref type='figure' target='#fig_0'>S1</ns0:ref>). The clipped biomass from the simulated grazing treatment was oven-dried (60 ºC for 48 h, the same below) and weighed. L. chinensis stopped growing 90 days after the rhizomes had sprouted and the experiment was terminated.</ns0:p><ns0:p>First, we measured the morphological characters of the ramet from the CK treatment, including natural height, vertical height, stem height, leaf length, leaf width, leaf number, and leaf angle. There were about 10 ramets in each flowerpot and about five leaves on each ramet. Generally, the tiller ramet that sprouted first was taller than others in the same flowerpot, and the second leaf from the ground was the longest. Thus, we selected the tiller ramet that sprouted first in each flowerpot and chose the second leaf on the selected ramet for measurement. Leaf angle was measured using a protractor as degrees (0-90°) from the ramet stem to the measured leaf. Next, we clipped the selected ramet biomass above 8 cm in height, those between 4-8 cm in height, and those below 4 cm in height (Note: the sloping parts of selected ramets were not straightened when clipped, and biomass data of each of the three parts were collected in their natural state), followed by oven-dring and weighing. Second, after counting the tiller ramets and rhizome ramets, we harvested the genet above-ground biomass at different vertical distributions (i.e., above 8 cm in height, between 4-8 cm in height and below 4 cm in height for CK and H8; above and below 4 cm in height for H4), followed by oven-dring and weighing. Third, all soil with roots and rhizomes in the flowerpot was transferred into the mesh bag and rinsed until only the clean roots and rhizomes remained. The stem below the soil surface (ca., 2 cm deep) was treated as the genet above-ground biomass below 4 cm. We also measured the morphological characteristics of rhizomes, including length and internode number.</ns0:p></ns0:div>
<ns0:div><ns0:head>Statistical Analysis</ns0:head><ns0:p>Root biomass allocation, rhizome biomass allocation, and below-ground biomass allocation were estimated using root biomass divided by total biomass, rhizome biomass divided by total biomass, and below-ground biomass divided by total biomass, respectively. The above-ground biomass vertical distribution parameters, including above-ground biomass distribution below 4 cm and 8 cm, were estimated using below 4-cm biomass divided by above-ground biomass, and below 8-cm biomass divided by total above-ground biomass, respectively. The above-ground biomass for each layer included the clipping biomass during the simulated grazing treatment.</ns0:p><ns0:p>The genet biomass allocation parameters and the above-ground biomass vertical distribution parameters were used to evaluate spatial grazing avoidance and the induced spatial grazing avoidance using a two-way analysis of variance (two-way ANOVA). Similarly, two-way ANOVA was used to analyze the influence of long-term overgrazing-induced legacy effects on the adaptation of L. chinensis to grazing with respected to its growth ability (biomass accumulation) and cloning ability (ramet number, rhizome length and rhizome section number). A significant interaction between the two factors under study indicates an effect of long-term overgrazing on the response of L. chinensis to simulated grazing. If the data for a trait were not normally distributed or homogeneous, such data were transformed using various methods (e.g., logarithmic, square root, square, reciprocal, or square root inverse rotation conversion) to attain normality and homoscedasticity. However, if data transformation could not make the data normally distributed and variances homogeneous, we conducted a one-way analysis of variance (one-way ANOVA) or Kruskal-Wallis test. If only NG or GZ had a significant response to simulated grazing, this implied that responses to simulated grazing between NG and OG were disparate. In contrast, if significant differences were observed in the simulated grazing treatments in both NG and OG, we calculated the plasticity index to simulated herbivory of these traits (PI) using the following formula (the content in parentheses in the formula refers to the trait for calculating the plasticity index):</ns0:p><ns0:p>'PI( ) = (CK( ) -H4( )) / CK( )'.</ns0:p><ns0:p>Student's t-test or Kruskal-Wallis test was used to compare differences in the phenotypic traits of L. chinensis ramets collected from the different plots (i.e., NG vs OG). The relationship between the traits was explored using the Pearson correlation method.</ns0:p><ns0:p>Structural equation modeling (SEM) was used to evaluate the importance of individual morphological traits for the above-ground spatial grazing avoidance of L. chinensis. The model assumed that the above-ground spatial grazing avoidance of L. chinensis, which was measured by the above-ground biomass vertical distribution, was attributed to individual morphological characters. The Pearson correlation analysis was conducted between all parameters included in the model. The initial model was developed according to the results of the correlation analysis and basic knowledge of plant science. Furthermore, the model was modified by deleting nonsignificant pathways and increasing pathways with covarying parameters. χ 2 statistics with the associated probability, the root mean square errors of approximation with the associated probability, and the Bentler-Bonett Index or Normed Fit Index were used to evaluate the overall fit of the model.</ns0:p><ns0:p>All the analyses were completed using IBM SPSS Statistics 19.0 and the means were compared using Tukey's HSD test (P < 0.05). The figures were generated in Oringin2019b.</ns0:p></ns0:div>
<ns0:div><ns0:head>RESULTS</ns0:head></ns0:div>
<ns0:div><ns0:head>Genet biomass allocation and above-ground biomass vertical distribution</ns0:head><ns0:p>There were significant legacy effects induced by long-term overgrazing in the biomass allocation of L. chinensis genet. In CK, the rhizome and belowground biomass allocation of OG decreased significantly compared with that observed from NG (P< 0.001 ), and root biomass allocation was not sensitive to overgrazing-induced legacies (P=0.057). Under simulated heavy grazing, OG had greater belowground biomass allocation (P=0.004) and root biomass allocation (P<0.05) than NG. There were significant decreases in root and rhizome biomass allocation under the simulated grazing treatment (P<0.05). Compared with NG, OG showed a smaller plastic index (PI) under simulated grazing treatment in terms of root and rhizome biomass allocation (Fig. <ns0:ref type='figure' target='#fig_0'>1</ns0:ref>).</ns0:p><ns0:p>There was a significant difference in the genet above-ground biomass vertical distribution between NG and OG. OG tended to allocate more biomass close to the ground with a larger above-ground biomass distribution below 4 cm and 8 cm than NG (P<0.001). The simulated grazing treatment did not alter the genet vertical distribution of biomass of NG but reduced the above-ground biomass distribution of OG close to the ground (P<0.05) (Fig. <ns0:ref type='figure' target='#fig_1'>2</ns0:ref>).</ns0:p></ns0:div>
<ns0:div><ns0:head>Ramet above-ground biomass vertical distribution</ns0:head><ns0:p>There was a significant correlation between the ramet and genet above-ground biomass vertical distribution (P<0.001) (Fig. <ns0:ref type='figure' target='#fig_1'>S2</ns0:ref>). The ramet distribution below 4 cm and 8 cm of OG were 89.16% and 69.29% larger than those of NG, respectively (P<0.001). L. chinensis ramet morphological traits (e.g., natural height, vertical height) showed significant grazing-induced legacies when growing in the homogenous environment without grazing (Table <ns0:ref type='table'>1</ns0:ref>). Leaf angle showed the most pronounced change, increasing by 129.91%, while the leaf number did not respond significantly to grazing legacies (Fig. <ns0:ref type='figure' target='#fig_2'>S3</ns0:ref>). Although OG ramets accumulated fewer photosynthetic products above 4 cm and 8 cm, it had larger accumulated biomass below 4 cm and 8 cm (P<0.001) (Fig. <ns0:ref type='figure' target='#fig_3'>S4</ns0:ref>).</ns0:p><ns0:p>SEM explained 93% and 94% of the variation in the ramet above-ground biomass vertical distribution below 4 cm and 8 cm, respectively. The larger near-surface distribution of aboveground biomass of OG ramets stemmed from the larger near-surface biomass accumulation and smaller biomass accumulation farther from the ground, and this above-ground spacial avoidance was induced by individual morphological characteristics (Fig. <ns0:ref type='figure' target='#fig_2'>3</ns0:ref>). According to the 'Standardized Total Effects, ' leaf angle played the most important role in inducing the above-ground spatial avoidance below 4 cm, while natural height made the highest contribution to above-ground spatial avoidance below 8 cm. The standardized total effects of grazing-induced legacies on the below 4-cm and 8-cm ramet above-ground biomass vertical distributions were -0.73 and -0.77, respectively (Table <ns0:ref type='table'>S1, S2</ns0:ref>).</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:05:49307:1:1:NEW 13 Aug 2020)</ns0:p><ns0:p>Manuscript to be reviewed</ns0:p></ns0:div>
<ns0:div><ns0:head>L. chinensis genet performance</ns0:head><ns0:p>Simulated grazing significantly reduced the below-ground, rhizome, root, and total biomass of L. chinensis genet, but did not affect the above-ground biomass. NG experienced more serious disturbance compared with OG and had a larger 'PI' based on the simulated herbivory in terms of the below-ground, rhizome, root, and total biomass accumulation. Besides, there were significant interactions between the material source and simulated herbivory treatment in terms of the below-ground, root, and total biomass. In the control treatment, there were no differences in the total, root, and below-ground biomass accumulation between NG and OG (P>0.05). In contrast, OG accumulated higher total, root, and belowground biomass than NG under simulated herbivory (P<0.01). Furthermore, NG accumulated more rhizome biomass than OG under CK treatment, while there were no significant differences were observed between NG and OG under simulated grazing treatments (Fig. <ns0:ref type='figure' target='#fig_3'>4</ns0:ref>).</ns0:p><ns0:p>Simulated herbivory had no effects on rhizome ramet number, tiller ramet number, and total ramet number (P>0.05). Compared with NG, OG showed more rhizome ramets, tiller ramets, and the total. Simulated herbivory significantly reduced the rhizome length and rhizome internode number of NG but did not affect that of OG. Under CK, the rhizome length and the number of rhizome sections of NG were larger than those of OG (P<0.001), and no differences in these variables were detected under H4 (Fig. <ns0:ref type='figure' target='#fig_4'>5</ns0:ref>).</ns0:p></ns0:div>
<ns0:div><ns0:head>DISCUSSION</ns0:head><ns0:p>The vertical distribution of above-ground biomass Compared with species that decrease in abundance under intensive grazing, plant species that exhibit an increasing trend in abundance are generally shorter and more prostrate <ns0:ref type='bibr'>(DÍAZ et al., 2007;</ns0:ref><ns0:ref type='bibr' target='#b20'>Forrestel, Donoghue & Smith, 2015)</ns0:ref>. Besides, previous work has shown that one species population in over-grazing fields are more prostrate and shorter compared with that in fields where grazing is excluded, and this phenomenon could persist for several generations after transplanting plants into the common garden environments <ns0:ref type='bibr' target='#b8'>(Carman, 1985;</ns0:ref><ns0:ref type='bibr' target='#b52'>Polley & Detling, 1990;</ns0:ref><ns0:ref type='bibr' target='#b49'>Painter, Detling & Steingraeber, 1993;</ns0:ref><ns0:ref type='bibr' target='#b56'>Rotundo & Aguiar, 2008;</ns0:ref><ns0:ref type='bibr' target='#b15'>Didiano et al., 2014)</ns0:ref>. Consistent with these reports, our direct measurements of the vertical distribution of aboveground biomass showed that OG allocated more above-ground biomass close to the ground. SEM analysis indicated that the smaller plant natural height and larger leaf angles of OG drove this pattern. This study confirms the findings of previous studies by clarifying changes in plant phenotypes under intensive grazing, specifically, more biomass is distributed close to the ground via shorter and more prostrate growth forms <ns0:ref type='bibr' target='#b15'>(Didiano et al., 2014)</ns0:ref>.</ns0:p><ns0:p>There was no marked plasticity in the above-ground biomass vertical distribution of L. chinensis to simulated grazing; however, we expected that simulated grazing should significantly increase plant biomass allocation toward the ground. This might be expected given that the ramet number did not respond significantly to simulated grazing in our experiment. A large number of new shorter ramets could contribute to the near-surface allocation of plant biomass via the shorter natural height as mentioned above. On the other hand, we speculate that leaf angle, an important morphological trait influencing the above-ground biomass vertical distribution, may also do not exhibit significant responses to simulated grazing without saliva and trampling.</ns0:p><ns0:p>Furthermore, the lack of significant responses of the vertical distribution of L. chinensis aboveground biomass to simulated grazing indicates that overgrazing-induced legacies in terms of the vertical distribution of above-ground biomass cannot be induced by short-term defoliation, and may be attributed instead to long-term defoliation or other pathways. In addition to defoliation, livestock can also influence plants by trampling, saliva, and indirect effects (e.g., increasing soil density, changing the rate of light interception by plants, etc.) <ns0:ref type='bibr' target='#b25'>(Heggenes, Odland & Bjerketvedt, 2018)</ns0:ref>. Trampling may play an important role in overgrazing-induced variation in the vertical distribution of above-ground biomass. Generally, short and prostrate plants have a higher resistance to trampling than those with high and erect growth forms <ns0:ref type='bibr' target='#b64'>(Warwick, 1980;</ns0:ref><ns0:ref type='bibr' target='#b57'>Sun & Liddle, 1993;</ns0:ref><ns0:ref type='bibr' target='#b34'>Kobayashi, Ikeda & Hori, 1999)</ns0:ref>. Therefore, the procumbent growth form exhibited by OG may be related to livestock trampling. Aside from this consideration, overgrazing-resulted more severe drought may be another contributor to the observed overgrazing-induced legacy effect on L. chinensis. Drought-adaptive morphological characteristics, such as small stature, have been reported to be advantageous for avoiding and recovering from herbivory <ns0:ref type='bibr' target='#b0'>(Adler et al., 2004;</ns0:ref><ns0:ref type='bibr' target='#b50'>Patty et al., 2010)</ns0:ref>. However, the pathways by which more above-ground biomass is allocated close to the ground as a result of overgrazing require additional research.</ns0:p></ns0:div>
<ns0:div><ns0:head>Biomass allocation above-and below-ground</ns0:head><ns0:p>Contrary to our expectation, OG allocated less biomass belowground under CK than NG; this was attributed to the lower investment in rhizomes by OG. The lower pre-allocating biomass to the rhizome of OG does not benefit the herbivory tolerance and avoidance of L. chinensis <ns0:ref type='bibr' target='#b19'>(Fornoni, 2011;</ns0:ref><ns0:ref type='bibr' target='#b42'>Lurie, Barton & Daehler, 2017)</ns0:ref> and can be considered a cost of herbivory tolerance in the absence of grazing <ns0:ref type='bibr' target='#b37'>(Lennartsson, Ramula & Tuomi, 2018)</ns0:ref>. However, the reduced rhizome allocation of OG may also be the result of the trade-off between colonization and viability, as we found a significant negative correlation between rhizome length and ramet number (Fig. <ns0:ref type='figure' target='#fig_4'>S5B</ns0:ref>). Long-dispersing ramets receive less support from the mother plant <ns0:ref type='bibr' target='#b66'>(Zobel, Moora & Herben, 2010)</ns0:ref>; hence, shorter rhizomes induced by the lower investment of L. chinensis in this organ could prevent the ramets from overgrazing-induced death. On the other hand, the shorter rhizome implies more tiller ramets, and this could reduce the risk of extinction of the ramets that emerge from the same rhizome section under overgrazing.</ns0:p><ns0:p>Colonization-competition trade-off <ns0:ref type='bibr' target='#b26'>(Herben et al., 1997;</ns0:ref><ns0:ref type='bibr' target='#b66'>Zobel, Moora & Herben, 2010;</ns0:ref><ns0:ref type='bibr' target='#b24'>Gough et al., 2012)</ns0:ref> may provide another explanation for the larger investment in rhizomes observed in NG. Extreme droughts that occur every few years resulted in many open patches in both overgrazing and no-grazing plots <ns0:ref type='bibr' target='#b62'>(Wang, Liu & Guo, 2019)</ns0:ref>. In the absence of grazing, plants with the most rapid colonizing ability in the open patches become dominant components of the vegetation <ns0:ref type='bibr' target='#b17'>( Fahrig et al., 1994;</ns0:ref><ns0:ref type='bibr'>Benot et al., 2013)</ns0:ref>. Therefore, intermittent extreme droughts may have contributed to the longer observed rhizomes of NG. Another possible mechanism driving this phenomenon is the fact that NG plants reserved resources for ensuring the next generation of new ramets through a dense canopy and litter layer, which seriously hinders the growth and development of new ramets <ns0:ref type='bibr' target='#b10'>(Craine et al., 2001;</ns0:ref><ns0:ref type='bibr'>Benot et al., 2013)</ns0:ref>.</ns0:p><ns0:p>The immediate responses of above-and below-ground allocation to simulated herbivory showed that L. chinensis increases the allocation of above-ground biomass, and this result supports the functional equilibrium theory <ns0:ref type='bibr' target='#b53'>(Poorter et al., 2012;</ns0:ref><ns0:ref type='bibr' target='#b23'>Gong et al., 2015)</ns0:ref>. According to this theory, more resources stored in the roots and rhizomes are mobilized to stimulate the growth of newly emerged leaves and new tiller ramets <ns0:ref type='bibr' target='#b16'>(Donaghy & Fulkerson, 1998)</ns0:ref>. Numerous studies have shown that this process leads to a reduced allocation of resources belowground for plants that experience leaf damage <ns0:ref type='bibr' target='#b23'>(Gong et al., 2015;</ns0:ref><ns0:ref type='bibr' target='#b4'>Barton, 2016;</ns0:ref><ns0:ref type='bibr' target='#b40'>Liu et al., 2018)</ns0:ref>. However, biomass allocation of OG was less affected by simulated herbivory; this was attributed to the enhanced above-ground spatial avoidance displayed by OG, which reduced the degree of herbivory damage (Fig. <ns0:ref type='figure'>S6</ns0:ref>).</ns0:p></ns0:div>
<ns0:div><ns0:head>Ramet number</ns0:head><ns0:p>The stimulation of plant density through increased grazing disturbance has been reported by numerous previous studies <ns0:ref type='bibr' target='#b55'>(Wang et al., 2017)</ns0:ref>. Concurrent to this observation, we found that there were more ramets (both tiller ramets and rhizome ramets) for OG than NG, and this is consistent with previous studies <ns0:ref type='bibr' target='#b12'>(Detling & Painter, 1983;</ns0:ref><ns0:ref type='bibr' target='#b56'>Rotundo & Aguiar, 2008)</ns0:ref>. Thus, while many scientists have focused on the disruption of plant apical dominance caused by livestock defoliation <ns0:ref type='bibr' target='#b54'>(Rautio et al., 2005;</ns0:ref><ns0:ref type='bibr' target='#b37'>Lennartsson, Ramula & Tuomi, 2018)</ns0:ref>, overgrazinginduced legacy effects on ramet number could partially explain the relatively larger plant density in grazing ecosystems.</ns0:p><ns0:p>Based on the negative relationship observed between the ramet number and ramet vertical height (Fig. <ns0:ref type='figure' target='#fig_4'>S5A</ns0:ref>), we speculate that the observed significant increase in the number of ramets of OG may be stem from decreased apical dominance. Weak apical dominance is a coping strategy for avoiding browsing damage caused by livestock. Shorter height in plants results in superior grazing avoidance, which reduces the probability of being defoliated because more biomass is allocated close to the ground. Furthermore, a higher number of tiller ramets could enhance tolerance to grazing by reducing the risk of extinction of ramets that emerge from the same rhizome section under overgrazing. Considering this phenomenon, both grazing tolerance and grazing avoidance could thus persist simultaneously, although some studies have suggested that there is a trade-off between these two strategies <ns0:ref type='bibr' target='#b18'>(Fineblum & Rausher, 1995;</ns0:ref><ns0:ref type='bibr' target='#b44'>Mauricio, 2000;</ns0:ref><ns0:ref type='bibr' target='#b36'>Krimmel & Pearse, 2016)</ns0:ref>. On the other hand, the decreased apical dominance induced by overgrazing-induced legacy effects may be attributed to an improvement in light penetration into the environment under grazing, and weaker apical dominance is most likely in the absence of competition for light <ns0:ref type='bibr' target='#b37'>( Lennartsson, Ramula & Tuomi, 2018)</ns0:ref>.</ns0:p><ns0:p>In Contrast to our prediction, we did not observe significant differences in L. chinensis ramet numbers between the simulated herbivory treatments in our experiment. Although many studies have indicated that plants can produce more branches when the apically dominant shoot is subjected to physical damage or herbivorous defoliation <ns0:ref type='bibr' target='#b33'>(Klimešová & Klimeš, 2003;</ns0:ref><ns0:ref type='bibr' target='#b54'>Rautio et al., 2005;</ns0:ref><ns0:ref type='bibr' target='#b55'>Wang et al., 2017)</ns0:ref>, many studies have also suggested that herbivory or cutting cannot increase the ramet number <ns0:ref type='bibr' target='#b63'>(Wang et al, 2004;</ns0:ref><ns0:ref type='bibr' target='#b6'>Benot et al., 2009)</ns0:ref> and might even reduce it <ns0:ref type='bibr' target='#b27'>(Hicks & Turkington, 2000)</ns0:ref>. Two preconditions are required for broken apical dominance to induce an increase in ramet numbers: a sufficient number of resources and meristems for regrowth and sufficient apical suppression of basal meristems <ns0:ref type='bibr' target='#b32'>(Klimesova et al., 2014;</ns0:ref><ns0:ref type='bibr' target='#b37'>Lennartsson, Ramula & Tuomi, 2018)</ns0:ref>. In this study, sufficient apical suppression of the basal meristems was achieved during the simulated grazing experiment. Specifically, most of the leaves were removed and only a small section of the stem was left in the 'H4' treatment. Hence, there may have been limited resources and meristems for the regrowth of L. chinensis because of its short growth times from tiller emergence to the first simulated grazing treatment and from the last treatment to harvest.</ns0:p></ns0:div>
<ns0:div><ns0:head>CONCLUSIONS</ns0:head><ns0:p>Consistent with studies on contemporary evolution and stress memory <ns0:ref type='bibr' target='#b12'>( Detling & Painter, 1983;</ns0:ref><ns0:ref type='bibr' target='#b47'>Oesterheld & McNaughton, 1988;</ns0:ref><ns0:ref type='bibr' target='#b2'>Agrawal et al., 2012;</ns0:ref><ns0:ref type='bibr' target='#b67'>Züst et al., 2012;</ns0:ref><ns0:ref type='bibr' target='#b15'>Didiano et al., 2014)</ns0:ref>, our study showed that OG exhibited higher adaptation to simulated grazing in terms of the growth and cloning ability than NG. This stronger adaptation was attributed to enhanced above-ground spatial avoidance. Contrary to our prediction, OG pre-allocated less biomass to the rhizome, which does not benefit the herbivory tolerance and avoidance of L. chinensis; however, this also may reflect a tolerance strategy via shorter rhizome and more tiller ramets. Here, we quantitatively studied the grazing-induced spatial avoidance of plants for the first time and found that enhanced above-ground spatial avoidance was induced by a larger leaf angle and shorter natural height. However, because of the pseudo-replication and because the sample areas only consisted of two small adjacent pastures in this study, the findings from our research may be limited to our study site. </ns0:p></ns0:div><ns0:figure xml:id='fig_0'><ns0:head>Figure 1</ns0:head><ns0:label>1</ns0:label><ns0:figDesc>Figure 1</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_1'><ns0:head>Figure 2</ns0:head><ns0:label>2</ns0:label><ns0:figDesc>Figure 2</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_2'><ns0:head>Figure 3 The</ns0:head><ns0:label>3</ns0:label><ns0:figDesc>Figure 3</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_3'><ns0:head>Figure 4</ns0:head><ns0:label>4</ns0:label><ns0:figDesc>Figure 4</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_4'><ns0:head>Figure 5</ns0:head><ns0:label>5</ns0:label><ns0:figDesc>Figure 5</ns0:figDesc></ns0:figure>
<ns0:note place='foot'>PeerJ reviewing PDF | (2020:05:49307:1:1:NEW 13 Aug 2020)</ns0:note>
</ns0:body>
" | "Pratacultural College
Gansu Agricultural University
Lanzhou, Gansu Province, PR China
August 5, 2020
Dear Editor and Reviewers:
We thank the editor and reviewers for their valuable scholarly comments and suggestions on our manuscript entitled “Overgrazing-induced legacy effects on phenotypes prepare Leymus chinensis to cope with herbivory” (#49307). We have read the comments carefully and revised the manuscript to address the concerns raised by the reviewers. The major changes to the manuscript are described in detail in the attached point-to-point response below. The comments from Editor and Reviewers are listed in italicized font and our responses are provided without italics.
Sincerely,
Fenghui Guo
Reply to Editor’s comments：
Issue 1: Pseudo replication – your inferences are limited to your 2 specific study plots. This should be acknowledged.
Response: Agreed. As suggested by the editor, we have included a conditional statement indicating that the findings from our research may be limited to our study area due to the adoption of pseudoreplication in our experiment. Please see the section of “Conclusion” (Line 395~400).
Issue 2: Why did you select these clipping treatments (4 cm and 8 cm). How do they relate to the ecology of your study system and what is the hypothesis regarding the two different clipping heights?
Response: The clipping height of 4 cm was selected to simulate heavy grazing while 8 cm was selected to simulate moderate grazing. This is based on Gao (2008) on the simulation of grazing (Line 150~152).
Reference:
Gao, Y. 2008. The study on tolerance to simulated herbivory in Leymus chinensis on Songnen plain. Northeast Normal University.
Issue 3: Does clipping with scissors yield plant responses similar to sheep grazing? It may be helpful to cite other recent studies supporting this. Commonly, jasmonic acid is applied to mechanical wounds to initiate a plant response that is more similar to natural herbivory.
Response: Thanks for this comment. We acknowledge the use of jasmonic acid to initiate plant response to herbivory under mechanical simulation of grazing. Please see some recent studies that support the method of clipping adopted in our study (Line 150).
Reference:
Didiano, T. J., Turley, N. E., Everwand, G., Schaefer, H., Crawley, M. J., & Johnson, M. T. (2014). Experimental test of plant defence evolution in four species using long‐term rabbit exclosures. Journal of Ecology, 102(3), 584-594.
Turley, N. E., Odell, W. C., Schaefer, H., Everwand, G., Crawley, M. J., & Johnson, M. T. (2013). Contemporary evolution of plant growth rate following experimental removal of herbivores. The American Naturalist, 181(S1), S21-S34.
Methods clarifications:
Issue 4:Reviewer 1 point out various clarifications that are needed in the Methods. Additionally, your definition of “ramet” should be provided. Furthermore, even after viewing Fig S2, I have no idea what is meant by the terms “tiller ramet” and “rhizome ramet” (L 151). Both of them seem to be composed of tillers in Fig S2.
Response: We have revised the Method section of the paper as suggested by the reviewer. In the current version of the paper, “tiller ramet” and “rhizome ramet” have been defined in section “Measurements” (Line 158~160). Furthermore, Fig. S1 has been modified to capture the explanation of “tiller ramet” and “rhizome ramet”.
Issue 5: Why were plant parts below the soil surface treated as aboveground biomass? Cite a reference to support this.
Response: Since this part belongs to the stem, we treated them as aboveground biomass. As far as we know, taking the Leymus chinensis stem under the soil surface (about 2 cm) as aboveground biomass is common; however, no studies have stated that. Furthermore, because of the rain-erosion and livestock-trampling, we considered that this part cannot contribute to spatial avoidance for plants under grazing. More importantly, this part accounts for a small part of the aboveground and total biomass, we believe it does not affect the final results obtained.
Issue 6: What is the reasoning for separating biomass layers into strata 0-4 cm, 4-8 cm, and 8+ cm? How are they predicted to differ? I understand that 0-4 cm could refer to “close to the ground” (escape from grazers?) but how are the other two strata expected to differ? Aren’t those strata both in easy reach of grazers? Assuming the sheep browse typically to ~4 cm (?), then it seems clearer to me to compare plant properties/biomass proportions < 4cm versus > 4 cm.
Response: The biomass distributing under 4 cm could escape from heavy grazing while the biomass distributing under 8 cm could escape from moderate grazing. To calculate the biomass under 4 cm distribution and 8 cm distribution, we separated biomass layers into 0-4 cm, 4-8 cm, and 8+ cm. This is per Gao (2008) on the simulation of grazing, and please see the responses for Issue 2.
Issue 7: Manuscript Structure – All sections need attention as pointed out by Reviewer 1. Some material may be better shifted to supplemental material or cut entirely. Careful English editing is required.
Response: Agreed. We have revised all sections of the paper as suggested by the reviewer. Some figures have been removed from the manuscript while others have now been moved to the supplemental material. The English language has been improved by a native English speaker.
Issue 8: Please avoid the term 'macro mammals'; I think it can be replaced with 'grazing mammals'.
Response: We have replaced 'macro mammals' with 'grazing mammals' as suggested by the editor.
Further comments:
Issue 9: Discussion –The Discussion has not raised the ecological concept of (over)compensation at all, although it has been suggested for other grasses subjected to grazing. Did you find any evidence in terms of biomass?
Response: Our study did not find evidence for (over)compensation in the Leymus chinensis collected from both the no grazing and heavy grazing area. This may be related to the frequent simulated grazing in this study. Notably, the objective of this study was to compare the responses of Leymus chinensis collected from the no grazing and heavy grazing area to simulated grazing, hence, the concept of (over)compensation was not observed in our study.
Issue 10: Conclusion L 351 -- What is meant by “cloning ability”? Do you mean ramet production or perhaps tiller production? Biomass production is not the same as cloning ability, and you have not presented ramet counts, so this conclusion may need to be re-thought.
Response: Leymus Chinensis clones through rhizomes to form clonal ramets. We used ramet number, rhizome length, and the number of rhizome sections to estimate the cloning ability of L. chinensis. We have added an explanation for cloning ability in section “Statistical Analysis” (Line 193~194).
Issue 11: Figure 2B – The y-axis should be back-transformed to match 2A.
Response: Agreed. We have made the suggested correction.
Issue 12: Figure 3 – since ramet and genet distributions are correlated as shown here, it could make sense to only present one of the results in the main manuscript and move the other to supplemental materials.
Response: Agreed.We have moved the figure to supplementary material as suggested by the reviewer.
Issue 13: Figure 6 – What is the ecological relevance of “below 8 cm”? This needs to be clarified in the text otherwise it should not be part of SEM modeling and I would also question why any data or analyses are presented on 8 cm (versus 4 cm). This study could be easier to understand if one ecologically meaningful height were chosen for analyses, corresponding to the height that is often protected from sheep, if such a height exists.
Response: As pointed out in Issue 2 and 6, biomass distribution below 8 cm could escape from moderate grazing while biomass below 4 cm could escape from heavy grazing. The clipping height of 4 cm was selected to simulate heavy grazing while 8 cm was selected to simulate moderate grazing. This is per Gao (2008) on the simulation of grazing, and please see the responses for Issue 2.
Issue 14:Table 1 What is “Nature height”? This same term appears on L 351. Did any of the culms flower? It would be valuable to compare the height of flowering culms between grazing histories.
Response: The use of the term ‘Nature height’ was an oversight. We have replaced it ‘Natural height’. There were no flowers on Leymus chinensis culms.
Reply to Reviewer 1:
Basic reporting
Review of Guo et al. “Overgrazing-induced legacy effects on phenotypes prepare Leymus chinensis to cope with herbivory” for PeerJ MS# 49307 June 2020
In this manuscript, Guo et al. seek to demonstrate if evolutionary adaptations/herbivory-induced legacy effects alter the phenotypes of populations of a highly palatable rhizomatous clonal grass species, Leymus chinensis, that have experienced significantly different grazing histories. Collecting samples from two pastures located in Inner Mongolian steppe grasslands that have experienced either high intensity, long-term livestock grazing or been protected from mammalian grazing they grew potted transplants in a greenhouse and subjected them to different simulated herbivory/clipping “grazing” intensities. The plants were then assessed after 90 days for differential demographic population responses based on the plants origin (legacy) and herbivory intensity treatments.
Issue 1: Utilizing plant material from only two different small pastures and comparing them could be considered problematic, both ecologically and statistically (pseudoreplication). Concerns regarding the latter are somewhat mitigated since the authors are primarily interested in how plant populations from these different “legacies” differentially respond to the different “grazing” intensity treatments that they experimentally implement in controlled and randomized conditions, but the former concern regarding the broader generality and applicability of these sampling areas remains.
Response: Thanks for these comments. We acknowledged that there are some ecological and statistical problems in this study design. To decrease the impacts of pseudo-replication on experimental results, we sampled the entire area for each plot except for the margin and ensured that the distance between every two sampling points was greater than 20 m.
A similar sampling area that has been enclosed for more than 30 years with Leymus chinensis as dominant species is practically rare in China. This limited us to improve the broader generality and applicability of the sampling areas. However, some studies which have conducted on other grasslands got similar results with our study (please see below for these studies).
Furthermore, we have included a conditional statement section indicating that the findings from our research may be limited to our study area due to the pseudo-replication and limiting generality (Line 395~400).
Reference:
Polley H W, Detling J K. 1988. Herbivory tolerance of Agropyron smithii populations with different grazing histories. Oecologia, 77(2): 261-267.
Polley H W, Detling J K. 1990. Grazing-mediated differentiation in Agropyron smithii: evidence from populations with different grazing histories.[J]. Oikos, 57(3): 326-332.
Painter E L, Detling J K, Steingraeber D A. 1993. Plant morphology and grazing history: relationships between native grasses and herbivores.[J]. Vegetatio, 106(1): 37-62.
Turley, N. E., Odell, W. C., Schaefer, H., Everwand, G., Crawley, M. J., & Johnson, M. T. (2013). Contemporary evolution of plant growth rate following experimental removal of herbivores. The American Naturalist, 181(S1), S21-S34.
Didiano, T. J., Turley, N. E., Everwand, G., Schaefer, H., Crawley, M. J., & Johnson, M. T. (2014). Experimental test of plant defence evolution in four species using long‐term rabbit exclosures. Journal of Ecology, 102(3), 584-594.
Issue 2: There are a number of interesting demographic/population results reported herein, but unfortunately, the manuscript suffers from numerous shortcomings including significant weaknesses in the study design, awkward structural organization of the text, confusing conceptual development of the central themes, and issues with the interpretation and presentation of the data.
Response: Thanks for these comments. We have substantially improved all parts of the article in the revised version.
Issue 3: Additionally, I encourage the authors to seek language editing assistance as the current grammatical state of their manuscript exceeds the reasonable editing expectations of external referees. I have made a few suggestions below throughout the manuscript, but the overall revisions needed are extensive and will require a committed editorial service.
Response: Thanks for this comment. The English language has been improved by a native English speaker and I also have revised the language as Reviewer 1 suggested.
Issue 4: Finally, simulated herbivory studies in controlled greenhouse conditions are not novel in the ecological literature. The authors claim they are contributing information that fills a critical knowledge gap, but I, at least as presented here (and they, to an extent), am not convinced (see below). As such, I do not believe most of these concerns could be easily addressed without the revision constituting an entirely new submission.
Response: Agreed. We have replaced “to close the knowledge gap” with “To further understand grazing-induced legacy effects on plants” in the Abstract and Introduction sections of the paper (Line 17, 93).
My major concerns include:
Issue 5: Introduction - The introduction is poorly organized. It fails to both identify and justify the central thesis the authors seek to examine and adequately support it with previously published literature. The large paragraph describing resistance, avoidance, and tolerance is overly general and can be shortened considerably.
Response: Agreed. We have substantially revised the introduction. Also, we have shortened the description of plant strategies to cope with herbivory (Line 50~65).
Issue 6: The Methods sections suffer from numerous inadequacies and insufficient descriptions of critical procedures. Section 2.1.2. entitled “The materials sampling and cultivating” is particularly confusing and unclear.
Response: We have improved the descriptions in the Method section and replaced “The materials sampling and cultivating” with “Materials Collection at the studying sites” (Line 115) and ‘‘Materials cultivation in the greenhouse” (Line 131).
Issue 7: Methods - There is little to provide confidence that the arbitrary heights of 4cm and 8cm aboveground selected to differentiate morphological adaptations to grazing have a sound ecological justification. Are there specific sheep grazing behaviors that affect suitable forage heights? Would 3cm and 7cm or 5cm and 9cm yield different results?
Response: We simulated moderate and heavy grazing by clipping the above-ground part of the plants at 8 cm and 4 cm above the soil surface, respectively. The stubble height used in the simulated grazing treatment is consistent with the method reported by Gao (2008) (Line 150~152).
Reference:
Gao, Y. 2008. The study on tolerance to simulated herbivory in Leymus chinensis on Songnen plain. Northeast Normal University.
Issue 8: And given that the authors merely measured the heights based on their growth in pots in artificial greenhouse conditions how would these be affected by exposure to the real-world conditions of climatic elements and animal trampling? Both here and in the Introduction, the authors fail to develop a convincing case why these results will potentially translate into ecologically meaningful patterns. Moreover, since these leaf angle and height data end up being some of the most significant results reported and strongly influence the SEM interpretation I remain underwhelmed and skeptical. There are some interesting population demography data reported, particularly the root biomass responses, but these fail to rescue the overall deficiencies noted above.
Response: We appreciate these comments. We acknowledge that the effect of grazing disturbance on plants is a very complicated process – including defoliation, trampling, saliva effect, changing the soil environment, etc. However, to study overgrazing-induced legacy effects on plants, there is a need for an experiment in the greenhouse to eliminate the differences in plant living environmental conditions under different grazing histories. Furthermore, simulated grazing experiments in a greenhouse is a common approach in grassland ecological studies. Please see some similar articles in this regard we have cited (Line 149~150).
Reference:
Turley, N. E., Odell, W. C., Schaefer, H., Everwand, G., Crawley, M. J., & Johnson, M. T. (2013). Contemporary evolution of plant growth rate following experimental removal of herbivores. The American Naturalist, 181(S1), S21-S34.
Didiano, T. J., Turley, N. E., Everwand, G., Schaefer, H., Crawley, M. J., & Johnson, M. T. (2014). Experimental test of plant defence evolution in four species using long‐term rabbit exclosures. Journal of Ecology, 102(3), 584-594.
Issue 9: Results – The results are very dense and could be improved by adding more description with less dependence on referencing figures and tables. Indeed, based on my downstream suggestion to eliminate and/or reduce some of the numerous figures and tables, converting some of the key findings into more comprehensive or detailed text may improve the readability of this section.
Response: Agreed. We have improved the entire result section as suggested by the reviewer. Also, we have moved to Fig.3, Fig,4, Fig.5, and Table.2 to supplementary material.
Issue 10: Discussion – The most parsimonious explanation for L. chinensis in the OG plot pre-allocating less biomass to rhizomes, is not so much a tolerance strategy where the plant benefits from shorter rhizomes and more tiller ramets, but a cost or tradeoff associated with biomass losses and altered source-sink relationships within the plant. The authors describe this type of biomass reallocation in their Introduction and somewhat revisit it in the Discussion, but the text takes on a meandering explanation that merely leaves the reader confused. Indeed, I believe these are some of the strongest and most interesting results in the study and they merit a thoughtful and thorough interpretation in the Discussion.
Response: Thanks for this comment. We have improved the explanation for L. chinensis in the OG plot pre-allocating less biomass to rhizomes in section “biomass allocation between above- and below-ground” (Line 319~310).
We agreed with reviewer 1 that pre-allocating less biomass to rhizomes for L. chinensis in the OG plot did not benefit the herbivory tolerance and maybe a cost of herbivory tolerance for OG under no-grazing. We added this viewpoint in section “Biomass allocation above- and below-ground” (Line 320~323).
However, based on our result in terms of “negative correlation between rhizome length and ramet number” and previous studies on the relationship between the mother plant and rhizome ramets, we retained our initial speculation (OG pre-allocated less biomass to rhizomes may reflect a tolerance strategy where plants benefit from shorter rhizomes and more tiller ramets) (Line 324~330).
Issue 11: Lines 254-279 – This entire section of the Discussion needs to be restructured, better clarified, and edited with improved grammar. I reread it multiple times and am still not entirely certain what takeaway message the authors are attempting to convey.
Response: As suggested by the reviewer, this section has been restructured and edited with improved grammar. Part of the information provided in this section was that our results regarding the allocation of more biomass distribution close to the ground by L. chinensis in the OG plot are similar to the findings from other studies. Further, we gave a possible explanation for why the aboveground biomass vertical distribution did not respond significantly to the simulated grazing. We also dwell on through which pathway (trampling, saliva, or other mechanisms), overgrazing resulted in the observed legacies in terms of more allocation of above-ground biomass close to the ground through. (Line 280~300)
Issue 12: Section beginning at line 280 – there is some interesting speculation here that would be of interest to a broader base of ecological readers. Unfortunately, the authors really need field data, which they do not have, on the frequency and abundance of asexual vs sexual reproduction of the different genotypes in the different pastures to answer these questions in a satisfactory manner.
Response: We appreciate this comment. Although we do not have the suggested field data at hand to provide further information in this regard. However, the observation raised by the reviewer has provided a very good insight into future research. We hope to consider filling this knowledge by conducting field experiments regarding these speculations in the future.
Issue 13: It is discouraging to read the authors state “more detailed research is needed to understand the responses of plant ramet numbers to herbivore defoliation, especially in combination with grazing-induced legacy effects on plants” at the end of their manuscript when the primary intention of their study was to make exactly this contribution! I appreciate that no study is wholly comprehensive and there are always moving goalposts on opportunities to gain more knowledge and insights but regret that (and as the authors confess above) the current iteration of this submission falls considerably short of adding sufficiently to existing published ecological literature.
Response: Thanks for this comment. We have removed the part in question due to the observed negativity it confers on the objective and outcome of our research. Our study found that Leymus chinensis from the overgrazing plot had more ramets than that from the no grazing plot. However, the Leymus chinensis from the two different plots did not respond significantly to the simulated grazing, while many studies have reported that increasing ramet or branch number plays an important role in how plants cope with herbivore defoliation, and we offered a possible explanation for this. (Line 374~387)
Issue 14: Figures & Tables - While I appreciate that supplemental electronic figures and data eliminate physical journal space restrictions and provide authors an opportunity to be more comprehensive, the cartoon drawings for Supplemental Figures 1-4 are overly simplistic and unnecessary. In their place, I recommend that several of the main figures and tables be moved to the supplemental category. Eight multi-panel figures and three data intensive tables are excessive for a study as simple and straightforward as presented here. I recommend considerably reducing the primary visual data presented to only the essential results that best tell the story the authors are crafting from their study.
Response: We have made corrections as suggested by the reviewer and the result section has also been improved.
A non-comprehensive sampling of other grammatical issues:
Issue 15: Line 12 – conditions
Response: We have made corrections as suggested by the reviewer.
Issue 16: Line 21 - “against our prospection” is unclear. Do the authors mean “Contrary to our prediction…”?
Response: We have made corrections as suggested by the reviewer. (Line 26)
Issue 17: Lines 26-27 - Awkwardly phrased sentence. Requires grammatical clarity. I also submit that the definition is either incomplete or insufficient.
Response: We have made corrections as suggested by the reviewer as regards the definition of legacy effects (Line 32~34).
Issue 18: Line 31 - extra space after the bracket.
Response: We have made corrections as suggested by the reviewer.
Issue 19: Line 32 – above the authors define legacy effects as effects on ecosystem structure and/or function, but here they refer to plant defense strategies. Plant defense strategies are conventionally thought of as plant population processes, not ecosystem process. This may seem semantic, but it is a rather critical point and influences not only the accurate usage of the legacy effect terminology, but the entire context and interpretation of it both here and as applied to other species in other ecosystems.
Response: We have redefined the legacy effects as suggested by the reviewer (Line 32~34).
Issue 20: Lines 37-39 - You emphasize the majority of published herbivory induced legacy effects research over recent decades has focused on short-term insect damage but then fail to provide any references to support this assertion.
Response: Agreed. We have corrected by providing two references (Holeski et al. 2012, Kafle et al. 2019) to support the concern raised above (Line 46~49).
Reference:
Holeski, L.M., Jander, G., and Agrawal, A.A. 2012. Transgenerational defense induction and epigenetic inheritance in plants. TRENDS ECOL EVOL 27(11): 618-626. doi: 10.1016/j.tree.2012. 07.011.
Kafle, D., Wurst, S. 2019. Legacy effects of herbivory enhance performance and resistance of progeny plants. Jouanal of Ecology 107(1):58-68. doi: 10.1111/1365‐2745.13038.
Issue 21: Line 84 – precariously?
Response: We have made corrections as suggested by the reviewer.
Issue 22: The term “past decades” is overused in the Introduction.
Response: We have made corrections as suggested by the reviewer.
Issue 23: Line 140 – replace “weighted with a weighing balance” with “weighed”
Response: We have made corrections as suggested by the reviewer. (Line 161)
Issue 24: Line 154 – mesh bag
Response: We have made corrections as suggested by the reviewer.
Issue 25: Line 255 – increasing trend of what?
Response: We have replaced the “increasing trend” with “increasing trend in abundance”. (Line 281)
Issue 26: Figure 5 legend – light
Response: We have made corrections as suggested by the reviewer.
Experimental design
see above
Validity of the findings
see above
Comments for the author
see above
Reply to Reviewer 2:
Basic reporting
no comment
Experimental design
no comment
Validity of the findings
no comment
Comments for the author
The manuscript reported a novel experiment aiming to declare how the performance of Leymus chinensis in response to simulated grazing is influenced by the overgrazing-induced legacy, and what phenotypic traits contribute to the changes in the performance. The novelty of the study lies in that it not only declared the legacy effect of overgrazing by livestock rather than by insects but declared its consequences for future coping strategies in responses to grazing. In my view, the experiment was nicely designed, the introduction section has given sufficient context of the study, the results presented are necessary and clearly stated, and the discussion section is cohesive and in-depth. However, I still have some concerns as follows.
Major concerns:
Issue 1: L217, when you calculated the aboveground, either for every vertical layer or the total, did you pool together all the three harvest?
Response: Yes, we pool all the three harvests together when calculating the aboveground biomass to better reflect the plant growth. We added the station in section “Statistical Analysis” (Line 188).
Issue 2: L214-215, Why simulated grazing reduce, but not increase the aboveground biomass distribution close to the ground? As expected, the aboveground biomass distribution should be more towards the ground surface in response to grazing. I cannot see the explanation for this in the Discussion section.
Response: Thank you for this comment. We have added a paragraph in section ‘‘The vertical distribution of above-ground biomass' to explain the observation raised above (Line 293~300).
Issue 3: L243-245, I think this result is very interesting and worth more deeply exploited in the Discussion section. For example, the OG plants are more conservative compared to NG plants, which are more exploitative, as shown by its greater rhizome biomass than OG plants when cultivated in CK.
Response: Thanks for this comment. We have improved the discussion about the observed higher cloning ability of NG under the CK in the second paragraph in section “Biomass allocation above- and below-ground” (Line 331~340).
Minor concerns:
Issue 4:L179, “not both responded differently…” is grammatically wrong.
Response: Thanks for this comment. We have made the necessary correction.
Issue 5: L183, there is nothing in the parentheses in the formula.
Response: We have explained what the empty parentheses mean in the section “Statistical Analysis” (Line 205~206).
Issue 6:L207, “NG had smaller….”, can this sentence be written otherwise like: “OG had greater…..”, because intuitively, the readers will take NG as a control to OG, so you’d better use OG as the subject of the sentence, and compare with NG. The same problem with the sentence in L209-210.
Response: Agreed. We have made corrections as suggested by the reviewer.
Issue 7: L218, here the authors mentioned Fig.3, but in my view, this figure is of little significance, because genet is composed of ramets, genet and ramet should have the same biomass distribution.
Response: Agreed. We have moved Fig.3 to supplementary material.
Issue 8: L269, “……may not be ascribed to defoliation by livestock.” I don’t agree. Over-grazing induced legacy is formed during a long-term period through many generations. It is a bit arbitrary to conclude that “the over-grazing induced legacies distribution may not be ascribed to defoliation by livestock” only based on one-off and short-term tests.
Response: We have made corrections as suggested by the reviewer (Line 301~317).
Issue 9: L298-301, this explanation is a bit far-fetching. Besides, I am afraid that the vegetative tillers sprouted from the rhizome nodes cannot be termed as seedlings, as they were not originated from seeds.
Response: We have replaced “next-generation seedlings” with the “next generation of new ramets” (Line 338~340).
Issue 10: Since you have defined the codes like OG and NG, you should use them when necessary, to avoid redundancy. See Line 308-310 in the annotated manuscript. Please make similar changes where else when necessary throughout the entire manuscript.
Response: Thanks for this comment. We have made corrections as suggested by the reviewer throughout the paper.
Issue 11: I would term the OG plot overgrazed plot instead of overgrazing plot, and term the NG plot ungrazed plot instead of no grazing plot.
Response: Agreed. We have made corrections as suggested by the reviewer throughout the paper.
Issue 12: Figure 2, I would like the two panels to combine into one panel (of course you should first do the same data transformation for below 4 cm distribution as for below 8 cm distribution).
Response: We have tried to follow the reviewer’s suggestion regarding the figure. However, we found that the revised figure is more confusing than the initial one. Therefore, we retained the original figure.
Issue 13: Figure 7, GZ in the parentheses should be OG?
Response: Thanks for this comment. We have corrected as suggested by the reviewer.
" | Here is a paper. Please give your review comments after reading it. |
9,752 | "<ns0:abstract xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:p>There is growing evidence that herbivory-induced legacy effects permit plants to cope with herbivory. However, herbivory-induced defense strategies in plants against grazing mammals have received little attention. To further understand the grazing-induced legacy effects on plants, we conducted a greenhouse experiment with Leymus chinensis experiencing different grazing histories. We focused on grazing-induced legacy effects on above-ground spatial avoidance and below-ground biomass allocation. Our results showed that L. chinensis collected from the continuous overgrazing plot (OG) exhibited higher performance under simulated grazing in terms of growth, cloning and colonizing ability than those collected from the 35-year no-grazing plot (NG). The enhanced adaptability of OG was attributed to increased above-ground spatial avoidance, which was mediated by larger leaf angle and shorter height (reduced vertical height and increased leaf angle contributed to the above-ground spatial avoidance at a lower herbivory stubble height, while reduced tiller natural height contributed to above-ground spatial avoidance at a higher herbivory stubble height). Contrary to our prediction, OG pre-allocated less biomass to the rhizome, which does not benefit the herbivory tolerance and avoidance of L. chinensis; however, this also may reflect a tolerance strategy where reduced allocation to rhizomes is associated with increased production of ramets.</ns0:p></ns0:div>
</ns0:abstract>
<ns0:body xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:head>INTRODUCTION</ns0:head><ns0:p>Environmental disturbances (e.g., drought, herbivory) can have persistent effects on ecological attributes (e.g., ecological processes, community structure, population dynamics, and plant and soil characteristics) long after they occur (i.e., the legacy effect) <ns0:ref type='bibr' target='#b21'>(Fox et al., 2015;</ns0:ref><ns0:ref type='bibr' target='#b31'>Kafle & Wurst, 2019)</ns0:ref>. Legacy effects are ubiquitous phenomena in nature and have been extensively studied in the context of plant succession, herbivory, invasive plants, ecosystem engineering, and human land use <ns0:ref type='bibr' target='#b11'>(Cuddington, 2011;</ns0:ref><ns0:ref type='bibr' target='#b35'>Kostenko et al., 2012;</ns0:ref><ns0:ref type='bibr' target='#b65'>Wurst & Ohgushi, 2015)</ns0:ref>. In grassland ecosystems, herbivory-induced legacy effects on ecological processes, such as plant succession and biological diversity change, can persist for decades and even millennia <ns0:ref type='bibr' target='#b28'>(Holeski, Jander & Agrawal, 2012;</ns0:ref><ns0:ref type='bibr' target='#b21'>Fox et al., 2015)</ns0:ref>. Some studies have shown that legacies in plant defense strategies can mediate herbivory-induced legacy effects on ecological processes. For instance, defense traits (e.g., chemical defense substances, resource reallocation) can significantly affect plant-herbivore relationships (generally increasing plant adaptations to herbivores), as well as interspecific relationships and soil characteristics <ns0:ref type='bibr' target='#b65'>(Wurst & Ohgushi, 2015)</ns0:ref>. Therefore, clarifying the overgrazing-induced legacy effects in plant defense strategies is critical for understanding processes occurring in grazing ecosystems. However, most previous studies on herbivory-induced legacy effects on plant defense strategies have focused on short-term insect herbivory, in contrast, studies examining long-term livestock grazing have received less attention by comparison <ns0:ref type='bibr' target='#b28'>(Holeski, Jander & Agrawal, 2012;</ns0:ref><ns0:ref type='bibr' target='#b31'>Kafle & Wurst, 2019)</ns0:ref>.</ns0:p><ns0:p>Strategies by which plants cope with herbivory include resistance, avoidance, and tolerance <ns0:ref type='bibr' target='#b15'>(Didiano et al., 2014)</ns0:ref>. While resistance strategies (e.g., thorns, higher tannin concentrations) play an important role in coping with insect herbivory <ns0:ref type='bibr' target='#b1'>(Agrawal et al., 2002;</ns0:ref><ns0:ref type='bibr' target='#b28'>Holeski, Jander & Agrawal, 2012)</ns0:ref>, they may be less useful in plants experiencing herbivory from livestock or other grazing mammals, as these large animals are unable to selectively graze at such a fine scale <ns0:ref type='bibr' target='#b45'>(Menard et al., 2002;</ns0:ref><ns0:ref type='bibr'>Benot et al., 2013)</ns0:ref>. Empirical evidence indicates that plants under grazing mammal herbivory show adaptive legacy effects via above-ground spatial avoidance traits which resulted in the distribution of more above-ground biomass close to the ground, including larger leaf angles, shorter height, and more prostrate growth forms <ns0:ref type='bibr' target='#b51'>(Polley & Detling, 1988</ns0:ref><ns0:ref type='bibr' target='#b52'>, 1990;</ns0:ref><ns0:ref type='bibr' target='#b49'>Painter, Detling & Steingraeber, 1993;</ns0:ref><ns0:ref type='bibr' target='#b58'>Tomás et al., 2000;</ns0:ref><ns0:ref type='bibr' target='#b38'>Li et al., 2015;</ns0:ref><ns0:ref type='bibr' target='#b55'>Ren et al., 2017)</ns0:ref>. However, studies examining the above-ground biomass vertical distribution have been qualitative-not quantitative. The lack of rigorous quantitative approaches severely limits our understanding of the role of these avoidance traits in the responses of plants to grazing mammal herbivory (e.g., livestock grazing). For example, the specific morphological characters that lead to the near-surface distribution of above-ground biomass to reduce the possibility of defoliation by grazing mammals remain unclear.</ns0:p><ns0:p>Biomass reallocation is a fundamental strategy for plants to cope with herbivory. When subjected to above-ground herbivory, more biomass is mobilized above-ground to facilitate the recovery of growth <ns0:ref type='bibr' target='#b40'>(Liu et al., 2018)</ns0:ref>, and a transient transformation of resources away from herbivores occurs within hours after herbivory <ns0:ref type='bibr' target='#b3'>(Anten & Pierik, 2010;</ns0:ref><ns0:ref type='bibr' target='#b48'>Orians, Thorn & Gomez, 2011)</ns0:ref>. Some studies have indicated that plant tolerance is tightly linked to biomass allocation patterns expressed before herbivory; that is, larger belowground biomass pre-allocation is associated with stronger tolerance of plants to above-ground herbivory <ns0:ref type='bibr' target='#b19'>(Fornoni, 2011;</ns0:ref><ns0:ref type='bibr' target='#b42'>Lurie, Barton & Daehler, 2017)</ns0:ref>. In addition to the importance of tolerance, large belowground biomass pre-allocation potentially helps plants avoid above-ground herbivory. There is growing evidence that long-term grazing induces higher belowground biomass allocation at the community and population levels <ns0:ref type='bibr' target='#b46'>(Milchunas & Lauenroth, 1993;</ns0:ref><ns0:ref type='bibr' target='#b39'>Lindwall et al., 2013;</ns0:ref><ns0:ref type='bibr'>)</ns0:ref>. Furthermore, more biomass allocation to roots for plants collected from grazing areas has been observed in common garden environments free of herbivory disturbance <ns0:ref type='bibr' target='#b29'>(Jaramillo & Detling, 1988;</ns0:ref><ns0:ref type='bibr' target='#b51'>Polley & Detling, 1988</ns0:ref><ns0:ref type='bibr' target='#b52'>, 1990)</ns0:ref>. However, few studies have examined herbivory-induced legacy effects on plant belowground biomass allocation, especially the allocation of belowground reproductive organs such as rhizomes.</ns0:p><ns0:p>Leymus chinensis, a rhizomatous clonal plant, is the dominant species on the Inner Mongolia typical steppe grasslands. This grass species is highly relished by livestock (e.g., cattle, sheep) and has been subjected to overgrazing for more than 50 years. Similar to plants that have been studied in other regions, L. chinensis exhibits significant grazing-induced legacy effects, such as short height and short leaves <ns0:ref type='bibr' target='#b38'>(Li et al., 2015;</ns0:ref><ns0:ref type='bibr' target='#b55'>Ren et al., 2017)</ns0:ref>. However, no quantitative studies have characterized herbivory-induced legacy effects on the vertical distribution of above-ground biomass and belowground biomass allocation (especially the allocation of rhizome biomass). Although earlier studies <ns0:ref type='bibr' target='#b47'>(Oesterheld & McNaughton, 1988;</ns0:ref><ns0:ref type='bibr' target='#b41'>Loreti, Oesterheld & Sala, 2001)</ns0:ref> have shown that plants subjected to grazing disturbance showed stronger adaptability to herbivory compared with ungrazed plants in a common garden environment, studies reporting legacy effects underlying L. chinensis adaptation to grazing are limited.</ns0:p><ns0:p>To further characterize the grazing-induced legacy effects on plants, we conducted a greenhouse pot experiment with L. chinensis collected from two adjacent plots separated by a pasture fence. The first was a 35-year no-grazing plot and the second is a long-term overgrazing plot. Our study sought to answer three questions. First, does L. chinensis exposed to long-term overgrazing disturbance exhibit enhanced above-ground spatial avoidance (measured by the above-ground biomass vertical distribution)? If so, which individual characteristics contribute to this trait? Second, are there overgrazing-induced legacy effects on L. chinensis in terms of the belowground biomass allocation (i.e., root and rhizome)? Third, does the L. chinensis collected from the grazing plot exhibit stronger adaptation to simulated herbivory compared with those collected from the no-grazing plot?</ns0:p></ns0:div>
<ns0:div><ns0:head>MATERIALS & METHODS</ns0:head></ns0:div>
<ns0:div><ns0:head>Materials</ns0:head></ns0:div>
<ns0:div><ns0:head>Description of studying sites</ns0:head><ns0:p>Samples of L. chinensis were collected from typical steppe grassland located at the Inner Mongolia Grassland Ecosystem Research Station (43 • 38' N, 116 • 42' E). The sampling sites comprise two adjacent plots separated by a pasture fence. The first was a no-grazing plot (600×400 m), which has been fenced since 1983 for long-term ecological observations, and the second was a continuously overgrazing plot (600×100 m) that has been grazed at a stocking rate of ~3 sheep units per hectare for more than 50 years. However, the stocking rate recommended by the local government is ~1.5 sheep units per hectare to achieve a balance between grassland productivity and livestock forage requirements. Thus, the continuously grazed plot has experienced heavy grazing pressure over the last several decades <ns0:ref type='bibr'>(Ma, et al., 2015)</ns0:ref>.</ns0:p></ns0:div>
<ns0:div><ns0:head>Materials Collection at the studying sites</ns0:head><ns0:p>Genotypes showed different phenotypic plasticity to environmental disturbance and different genetic structures may be one of the main mechanisms mediating overgrazing-induced legacies on L. chinensis <ns0:ref type='bibr'>(Josephs, 2018)</ns0:ref>. We could not determine the genotype of each L. chinensis individual collected from both plots. To reduce the possible impacts of genotypes on our experimental results, we conducted our experiment at the population level and used the highreplication sampling.</ns0:p><ns0:p>We sampled 150-segment rhizomes with a root drill at 150 random points in each of the two treatment plots at the beginning of the growing season. Although we acknowledge pseudoreplication in the experimental design, as each treatment consisted of one large plot with subsamples as replicates, we sampled the entire area for each plot except for the margin and ensured that the distance between every two sampling points was greater than 20 m; this was done both to improve the representativeness of the samples for each plot across each of the large neighboring plots. Each rhizome was approximately 4 cm long and contained at least one sprouting section of the buds. To prevent the sampled rhizome from losing its vitality following removal from the soil environment, we immediately transferred samples into a container with moistened soil.</ns0:p></ns0:div>
<ns0:div><ns0:head>Cultivation in the greenhouse</ns0:head><ns0:p>In the laboratory, each rhizome was dissected into a 2-cm long section with only one node. All of the rhizomes were cultivated in flowerpots, 20 cm in diameter and 15 cm high, which were kept in the greenhouse. Each flowerpot contained 3 kg of soil collected from points adjacent to the experimental site and was subsequently planted with one rhizome section. There were 150 flowerpots for rhizomes collected from the no grazing plot and the overgrazing plot, respectively. After 20 days, the rhizomes had sprouted in approximately 100 flowerpots from each treatment.</ns0:p></ns0:div>
<ns0:div><ns0:head>Experimental Design & Measurements</ns0:head></ns0:div>
<ns0:div><ns0:head>Experimental design</ns0:head><ns0:p>The experiment was a full factorial design and consisted of two factors. The first was the source of L. chinensis (NG: L. chinensis collected from the no-grazing plot; OG: L. chinensis collected from the overgrazing plot), and the second was simulated grazing (CK: no simulated grazing; H8: simulated moderate grazing; H4: simulated heavy grazing). For NG and OG, we randomly selected 90 flowerpots that had sprouted rhizomes. These flowerpots were randomly arranged in the greenhouse, and we alternated their position every week to exclude the influence of external factors (e.g., light). One-third of the L. chinensis growing in flowerpots ( in both NG and OG separately) were randomly assigned to CK, H8, and H4. However, several L. chinensis died during the experiment and the numbers left for each treatment were NG*CK ( <ns0:ref type='formula'>22</ns0:ref>), NG*H8 (29), NG*H4 (25), OG*CK ( <ns0:ref type='formula'>22</ns0:ref>), OG*H8 (26), and OG*H4 (28). We clipped plants with scissors to conduct the simulated grazing as per <ns0:ref type='bibr' target='#b59'>Turley (2013)</ns0:ref> and <ns0:ref type='bibr' target='#b15'>Didiano et al (2014)</ns0:ref>. We simulated moderate and heavy grazing by clipping the above-ground part of plants 8 cm and 4 cm above the soil surface, respectively. The stubble height in the simulated grazing treatment was per <ns0:ref type='bibr' target='#b22'>Gao (2008)</ns0:ref>. Sloping parts (e.g., the sloping leaves and stems) were not straightened in the simulated grazing treatments. We only removed plant parts that were distributed above 8 cm or 4 cm in their natural state. The herbivory simulation was carried out three times at 45, 60, and 75 days after the rhizomes had sprouted to simulate repeated grazing in natural environments.</ns0:p></ns0:div>
<ns0:div><ns0:head>Measurements</ns0:head><ns0:p>Although we did not assess genotypic differences among plants in our replicate pots, we nevertheless refer to each replicate potted plant as a genet. We use the term ramet to refer to a tiller that has sprouted from a rhizome bud (i.e. an individual that is a physiologically integrated component of the genet). The clipped biomass from the simulated grazing treatment was oven-dried (60 ºC for 48 h, the same below) and weighed. L. chinensis stopped growing 90 days after the rhizomes had sprouted and the experiment was terminated. First, we measured the morphological characters of the ramet from the CK treatment, including natural height, vertical height, stem height, leaf length, leaf width, leaf number, and leaf angle. There were about 10 ramets in each flowerpot and about five leaves on each ramet. Generally, the ramet that sprouted first was the tallest and the second leaf from the ground was the longest. Thus, we selected the ramet that sprouted first in each flowerpot and chose the second leaf on the selected ramet for measurement. Leaf angle was measured using a protractor as degrees (0-90°) from the ramet stem to the measured leaf. Next, we clipped the selected ramet biomass above 8 cm in height, those between 4-8 cm in height, and those below 4 cm in height (Note: the sloping parts of selected ramets were not straightened when clipped, and biomass data of each of the three parts were collected in their natural state), followed by oven-drying and weighing. Second, after counting the ramets, we harvested the genet above-ground biomass at different vertical distributions (i.e., above 8 cm in height, between 4-8 cm in height and below 4 cm in height for CK and H8; above and below 4 cm in height for H4), followed by oven-drying and weighing. Third, all soil with roots and rhizomes in the flowerpot was transferred into mesh bags and rinsed until only the clean roots and rhizomes remained. The stem below the soil surface (ca., 2 cm deep) was treated as the genet above-ground biomass below 4 cm. We also measured the morphological characteristics of rhizomes, including length and internode number.</ns0:p></ns0:div>
<ns0:div><ns0:head>Statistical Analysis</ns0:head><ns0:p>Root biomass allocation, rhizome biomass allocation, and below-ground biomass allocation were estimated using root biomass divided by total biomass, rhizome biomass divided by total biomass, and below-ground biomass divided by total biomass, respectively. The above-ground biomass vertical distribution parameters, including above-ground biomass distribution below 4 cm and 8 cm, were estimated using below 4-cm biomass divided by above-ground biomass, and below 8-cm biomass divided by total above-ground biomass, respectively. The above-ground biomass for each layer included the clipping biomass during the simulated grazing treatment.</ns0:p><ns0:p>The genet biomass allocation parameters and the above-ground biomass vertical distribution parameters were used to evaluate spatial grazing avoidance and the induced spatial grazing avoidance using a two-way analysis of variance (two-way ANOVA). Similarly, two-way ANOVA was used to analyze the influence of long-term overgrazing-induced legacy effects on the adaptation of L. chinensis to grazing with respect to its growth ability (biomass accumulation), cloning ability (ramet number) and colonizing ability (rhizome length and rhizome internode number). A significant interaction between the two factors under study indicates an effect of long-term overgrazing on the response of L. chinensis to simulated grazing. If the data for a trait were not normally distributed or homogeneous, such data were transformed using various methods (e.g., logarithmic, square root, square, reciprocal, or square root inverse rotation conversion) to attain normality and homoscedasticity. However, if data transformation could not make the data normally distributed and variances homogeneous, we conducted a oneway analysis of variance (one-way ANOVA) or Kruskal-Wallis test. If only NG or OG had a significant response to simulated grazing, this implied that responses to simulated grazing between NG and OG were disparate. In contrast, if significant differences were observed in the simulated grazing treatments in both NG and OG, we calculated the plasticity index (PI) to simulated herbivory of these traits using the following formula:</ns0:p><ns0:p>'PI = (CK -H4) / CK'.</ns0:p><ns0:p>Student's t-test or Kruskal-Wallis test was used to compare differences in the phenotypic traits of L. chinensis ramets collected from the different plots (i.e., NG vs OG). The relationship between the traits was explored using the Pearson correlation method.</ns0:p><ns0:p>Structural equation modeling (SEM) was used to evaluate the importance of individual morphological traits for the above-ground spatial grazing avoidance of L. chinensis. The model assumed that the above-ground spatial grazing avoidance of L. chinensis, which was measured by the above-ground biomass vertical distribution, was attributed to individual morphological characters. The Pearson correlation analysis was conducted between all parameters included in the model. The initial model was developed according to the results of the correlation analysis and basic knowledge of plant science. Furthermore, the model was modified by deleting nonsignificant pathways and by increasing pathways between residual variables. χ 2 statistics with the associated probability, the root mean square errors of approximation with the associated probability, and the Bentler-Bonett Index or Normed Fit Index were used to evaluate the overall fit of the model.</ns0:p><ns0:p>All the analyses were completed using IBM SPSS Statistics 19.0 and the means were compared using Tukey's HSD test (P < 0.05). The figures were generated in Oringin2019b.</ns0:p></ns0:div>
<ns0:div><ns0:head>RESULTS</ns0:head></ns0:div>
<ns0:div><ns0:head>Genet biomass allocation and above-ground biomass vertical distribution</ns0:head><ns0:p>There were significant legacy effects in the biomass allocation of L. chinensis genets sampled from the over-grazing plot. In CK, the rhizome and belowground biomass allocation of OG decreased significantly compared with that observed from NG (P< 0.001 ), while root biomass allocation was not sensitive to overgrazing-induced legacies (P=0.057). Under simulated heavy grazing, OG had greater belowground biomass allocation (P=0.004) and root biomass allocation (P<0.05) than NG. There were significant decreases in root and rhizome biomass allocation under the simulated grazing treatment (P<0.05). Compared with NG, OG showed a smaller plastic index (PI) under simulated grazing treatment in terms of root and rhizome biomass allocation (Fig. <ns0:ref type='figure' target='#fig_0'>1</ns0:ref>).</ns0:p><ns0:p>There was a significant difference in the genet above-ground biomass vertical distribution between NG and OG. OG tended to allocate more biomass close to the ground with a larger above-ground biomass distribution below 4 cm and 8 cm than NG (P<0.001). The simulated grazing treatment did not alter the genet vertical distribution of biomass of NG but reduced the above-ground biomass distribution of OG close to the ground (P<0.05) (Fig. <ns0:ref type='figure' target='#fig_1'>2</ns0:ref>).</ns0:p></ns0:div>
<ns0:div><ns0:head>Ramet above-ground biomass vertical distribution</ns0:head><ns0:p>There was a significant correlation between the ramet and genet above-ground biomass vertical distribution (P<0.001) (Fig. <ns0:ref type='figure' target='#fig_0'>S1</ns0:ref>). The ramet distribution below 4 cm and 8 cm of OG were 89% and 69% larger than those of NG, respectively (P<0.001). L. chinensis ramet morphological traits (e.g., natural height, vertical height) showed significant grazing-induced legacies under CK treatment (Table <ns0:ref type='table'>1</ns0:ref>). Leaf angle showed the most pronounced change, increasing by 130%, while the leaf number did not respond significantly to grazing legacies (Fig. <ns0:ref type='figure' target='#fig_1'>S2</ns0:ref>). Although OG ramets accumulated fewer photosynthetic products above 4 cm and 8 cm, they had larger accumulated biomass below 4 cm and 8 cm (P<0.001) (Table <ns0:ref type='table'>1</ns0:ref>).</ns0:p><ns0:p>SEM explained 93% and 94% of the variation in the ramet above-ground biomass vertical distribution below 4 cm and 8 cm, respectively. The larger near-surface distribution of aboveground biomass of OG ramets stemmed from the larger near-surface biomass accumulation and smaller biomass accumulation farther from the ground, and this above-ground spacial avoidance was induced by individual morphological characteristics (Fig. <ns0:ref type='figure' target='#fig_2'>3</ns0:ref>). According to the 'Standardized Total Effects, ' vertical height and leaf angle played a more important role than other traits in inducing the above-ground spatial avoidance below 4 cm, while natural height made the highest contribution to above-ground spatial avoidance below 8 cm. The standardized total effects of grazing-induced legacies on the below 4-cm and 8-cm ramet above-ground biomass vertical distributions were -0.73 and -0.77, respectively (Table <ns0:ref type='table'>S1, S2</ns0:ref>).</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:05:49307:2:0:NEW 27 Aug 2020)</ns0:p><ns0:p>Manuscript to be reviewed</ns0:p></ns0:div>
<ns0:div><ns0:head>L. chinensis genet performance</ns0:head><ns0:p>Simulated grazing significantly reduced the below-ground, rhizome, root, and total biomass of L. chinensis genets, but did not affect the above-ground biomass. NG experienced more serious biomass loss compared with OG and had a larger 'PI' in response to the simulated herbivory in terms of the below-ground, rhizome, root, and total biomass accumulation. Additionally, there were significant interactions between plant source (NG or OG) and simulated herbivory treatment in terms of the below-ground, root, and total biomass. In the control treatment, there were no differences in the total, root, and below-ground biomass accumulation between NG and OG (P>0.05). In contrast, OG accumulated higher total, root, and belowground biomass than NG under simulated herbivory (P<0.01). Furthermore, NG accumulated more rhizome biomass than OG under CK treatment, while there were no significant differences were observed between NG and OG under simulated grazing treatments (Fig. <ns0:ref type='figure' target='#fig_3'>4</ns0:ref>).</ns0:p><ns0:p>Simulated herbivory had no effects on ramet number (P>0.05) and OG showed more ramets than NG. Simulated herbivory significantly reduced the rhizome length and rhizome internode number of NG but did not affect that of OG. Under CK, the rhizome length and the number of rhizome sections of NG were larger than those of OG (P<0.001), and no differences in these variables were detected in H4 (Fig. <ns0:ref type='figure' target='#fig_4'>5</ns0:ref>).</ns0:p></ns0:div>
<ns0:div><ns0:head>DISCUSSION</ns0:head></ns0:div>
<ns0:div><ns0:head>The vertical distribution of above-ground biomass</ns0:head><ns0:p>In comparison to species that decrease in abundance under intensive grazing, plant species that exhibit increase in abundance are generally shorter and more prostrate <ns0:ref type='bibr'>(DÍAZ et al., 2007;</ns0:ref><ns0:ref type='bibr' target='#b20'>Forrestel, Donoghue & Smith, 2015)</ns0:ref>. Besides, previous work has shown that populations of a single species may vary, with populations in over-grazing fields are more prostrate and shorter compared with that in fields where grazing is excluded, and this phenomenon could persist for several generations after transplanting plants into common garden environments <ns0:ref type='bibr' target='#b8'>(Carman, 1985;</ns0:ref><ns0:ref type='bibr' target='#b52'>Polley & Detling, 1990;</ns0:ref><ns0:ref type='bibr' target='#b49'>Painter, Detling & Steingraeber, 1993;</ns0:ref><ns0:ref type='bibr' target='#b56'>Rotundo & Aguiar, 2008;</ns0:ref><ns0:ref type='bibr' target='#b15'>Didiano et al., 2014)</ns0:ref>. Consistent with these reports, our direct measurements of the vertical distribution of above-ground biomass showed that OG allocated more above-ground biomass close to the ground. SEM analysis indicated that the shorter plant height and larger leaf angles of OG drove this pattern. This study confirms the findings of previous studies by clarifying changes in plant phenotypes under intensive grazing, specifically, more biomass is distributed close to the ground via shorter and more prostrate growth forms <ns0:ref type='bibr' target='#b15'>(Didiano et al., 2014)</ns0:ref>.</ns0:p><ns0:p>There was no marked plasticity in the above-ground biomass vertical distribution of L. chinensis to simulated grazing; however, we expected that simulated grazing should significantly increase plant biomass allocation toward the ground. This might be expected given that the ramet number did not respond significantly to simulated grazing in our experiment. A large number of new shorter ramets could contribute to the near-surface allocation of plant biomass via the shorter natural height as mentioned above. On the other hand, we speculate that leaf angle, an important morphological trait influencing the above-ground biomass vertical distribution, may not exhibit significant responses to simulated grazing without grazer saliva and trampling.</ns0:p><ns0:p>Furthermore, the lack of significant responses of the vertical distribution of L. chinensis aboveground biomass to simulated grazing indicates that overgrazing-induced legacies in terms of the vertical distribution of above-ground biomass cannot be induced by short-term defoliation, and may be attributed instead to long-term defoliation or other pathways. In addition to defoliation, livestock can also influence plants by trampling, saliva, and indirect effects (e.g., increasing soil density, changing the rate of light interception by plants, etc.) <ns0:ref type='bibr' target='#b25'>(Heggenes, Odland & Bjerketvedt, 2018)</ns0:ref>. Trampling may play an important role in overgrazing-induced variation in the vertical distribution of above-ground biomass. Generally, short and prostrate plants have a higher resistance to trampling than those with taller and erect growth forms <ns0:ref type='bibr' target='#b64'>(Warwick, 1980;</ns0:ref><ns0:ref type='bibr' target='#b57'>Sun & Liddle, 1993;</ns0:ref><ns0:ref type='bibr' target='#b34'>Kobayashi, Ikeda & Hori, 1999)</ns0:ref>. Therefore, the procumbent growth form exhibited by OG may be related to livestock trampling. Aside from this consideration, overgrazing may promote increases in localized drought, which may be another contributor to the observed overgrazing-induced legacy effect on L. chinensis. Drought-adaptive morphological characteristics, such as small stature, have been reported to be advantageous for avoiding and recovering from herbivory <ns0:ref type='bibr' target='#b0'>(Adler et al., 2004;</ns0:ref><ns0:ref type='bibr' target='#b50'>Patty et al., 2010)</ns0:ref>. However, the pathways by which more above-ground biomass is allocated close to the ground as a result of overgrazing require additional research.</ns0:p></ns0:div>
<ns0:div><ns0:head>Biomass allocation above-and below-ground</ns0:head><ns0:p>Contrary to our expectation, OG allocated less biomass belowground under CK than NG; this was attributed to the lower investment in rhizomes by OG. The lower pre-allocating biomass to the rhizome of OG does not benefit the herbivory tolerance and avoidance of L. chinensis <ns0:ref type='bibr' target='#b19'>(Fornoni, 2011;</ns0:ref><ns0:ref type='bibr' target='#b42'>Lurie, Barton & Daehler, 2017)</ns0:ref> and can be considered a cost of herbivory tolerance in the absence of grazing <ns0:ref type='bibr' target='#b37'>(Lennartsson, Ramula & Tuomi, 2018)</ns0:ref>. However, the reduced rhizome allocation of OG may also be the result of the trade-off between colonization and grazing tolerance, as we found a significant negative correlation between rhizome length and ramet number (Fig. <ns0:ref type='figure' target='#fig_2'>S3B</ns0:ref>). Long-dispersing ramets receive less support from the mother plant <ns0:ref type='bibr' target='#b66'>(Zobel, Moora & Herben, 2010)</ns0:ref>; hence, shorter rhizomes which are induced by the lower investment of L. chinensis in this organ could prevent the ramets from overgrazing-induced death. On the other hand, the shorter rhizome implies more ramets, and this could reduce the risk of ramets extinction under overgrazing.</ns0:p><ns0:p>A colonization-competition trade-off <ns0:ref type='bibr' target='#b26'>(Herben et al., 1997;</ns0:ref><ns0:ref type='bibr' target='#b66'>Zobel, Moora & Herben, 2010;</ns0:ref><ns0:ref type='bibr' target='#b24'>Gough et al., 2012)</ns0:ref> may provide another explanation for the larger investment in rhizomes observed in NG. Extreme droughts occur every few years which result in many open patches in both overgrazing and no-grazing plots <ns0:ref type='bibr' target='#b62'>(Wang, Liu & Guo, 2019)</ns0:ref>. In the absence of grazing, plants with the most rapid colonizing ability in the open patches become dominant components of the vegetation <ns0:ref type='bibr' target='#b17'>( Fahrig et al., 1994;</ns0:ref><ns0:ref type='bibr'>Benot et al., 2013)</ns0:ref>. Therefore, intermittent extreme droughts may have contributed to the longer observed rhizomes of NG. Another possible mechanism driving this phenomenon is the fact that NG plants reserved resources for ensuring the next generation of new ramets through a dense canopy and litter layer, which seriously hinders the growth and development of new ramets <ns0:ref type='bibr' target='#b10'>(Craine et al., 2001;</ns0:ref><ns0:ref type='bibr'>Benot et al., 2013)</ns0:ref>.</ns0:p><ns0:p>The immediate responses of above-and below-ground allocation to simulated herbivory showed that L. chinensis increases the allocation of above-ground biomass, and this result supports the functional equilibrium theory <ns0:ref type='bibr' target='#b53'>(Poorter et al., 2012;</ns0:ref><ns0:ref type='bibr' target='#b23'>Gong et al., 2015)</ns0:ref>. According to this theory, more resources stored in the roots and rhizomes are mobilized to stimulate the growth of newly emerged leaves and new ramets <ns0:ref type='bibr' target='#b16'>(Donaghy & Fulkerson, 1998)</ns0:ref>. Numerous studies have shown that this process leads to a reduced allocation of resources belowground for plants that experience leaf damage <ns0:ref type='bibr' target='#b23'>(Gong et al., 2015;</ns0:ref><ns0:ref type='bibr' target='#b4'>Barton, 2016;</ns0:ref><ns0:ref type='bibr' target='#b40'>Liu et al., 2018)</ns0:ref>. However, biomass allocation of OG was less affected by simulated herbivory; this was attributed to the enhanced above-ground spatial avoidance displayed by OG, which reduced the degree of herbivory damage (Fig. <ns0:ref type='figure' target='#fig_3'>S4</ns0:ref>).</ns0:p></ns0:div>
<ns0:div><ns0:head>Ramet number</ns0:head><ns0:p>The stimulation of plant density through increased grazing disturbance has been reported by numerous previous studies <ns0:ref type='bibr' target='#b55'>(Wang et al., 2017)</ns0:ref>. Concurrent to this observation, we found that there were more ramets for OG than NG, and this is consistent with previous studies <ns0:ref type='bibr' target='#b12'>(Detling & Painter, 1983;</ns0:ref><ns0:ref type='bibr' target='#b56'>Rotundo & Aguiar, 2008)</ns0:ref>. Thus, while many scientists have focused on the disruption of plant apical dominance caused by livestock defoliation <ns0:ref type='bibr' target='#b54'>(Rautio et al., 2005;</ns0:ref><ns0:ref type='bibr' target='#b37'>Lennartsson, Ramula & Tuomi, 2018)</ns0:ref>, overgrazing-induced legacy effects on ramet number could partially explain the relatively larger plant density in grazing ecosystems.</ns0:p><ns0:p>Based on the negative relationship observed between the ramet number and ramet vertical height (Fig. <ns0:ref type='figure' target='#fig_2'>S3A</ns0:ref>), we speculate that the observed significant increase in the number of ramets of OG may stem from decreased apical dominance. Weak apical dominance is a coping strategy for avoiding browsing damage caused by livestock. Shorter height in plants results in superior grazing avoidance, which reduces the probability of being defoliated because more biomass is allocated close to the ground. Furthermore, a higher number of ramets could enhance tolerance to grazing by reducing the risk of ramets extinction under overgrazing. Considering this phenomenon, both grazing tolerance and grazing avoidance could thus persist simultaneously, although some studies have suggested that there is a trade-off between these two strategies <ns0:ref type='bibr' target='#b18'>(Fineblum & Rausher, 1995;</ns0:ref><ns0:ref type='bibr' target='#b44'>Mauricio, 2000;</ns0:ref><ns0:ref type='bibr' target='#b36'>Krimmel & Pearse, 2016)</ns0:ref>. On the other hand, the decreased apical dominance induced by overgrazing-induced legacy effects may be attributed to an improvement in light penetration into the environment under grazing, and weaker apical dominance is most likely in the absence of competition for light <ns0:ref type='bibr' target='#b37'>( Lennartsson, Ramula & Tuomi, 2018)</ns0:ref>.</ns0:p><ns0:p>In contrast to our prediction, we did not observe significant differences in L. chinensis ramet numbers between the simulated herbivory treatments in our experiment. Although many studies have indicated that plants can produce more branches when the apically dominant shoot is subjected to physical damage or herbivorous defoliation <ns0:ref type='bibr' target='#b33'>(Klimešová & Klimeš, 2003;</ns0:ref><ns0:ref type='bibr' target='#b54'>Rautio et al., 2005;</ns0:ref><ns0:ref type='bibr' target='#b55'>Wang et al., 2017)</ns0:ref>, many studies have also suggested that herbivory or cutting cannot increase the ramet number <ns0:ref type='bibr' target='#b63'>(Wang et al, 2004;</ns0:ref><ns0:ref type='bibr' target='#b6'>Benot et al., 2009)</ns0:ref> and might even reduce it <ns0:ref type='bibr' target='#b27'>(Hicks & Turkington, 2000)</ns0:ref>. Two preconditions are required for broken apical dominance to induce an increase in ramet numbers: a sufficient number of resources and meristems for regrowth and sufficient apical suppression of basal meristems <ns0:ref type='bibr' target='#b32'>(Klimesova et al., 2014;</ns0:ref><ns0:ref type='bibr' target='#b37'>Lennartsson, Ramula & Tuomi, 2018)</ns0:ref>. In this study, sufficient apical suppression of the basal meristems was achieved during the simulated grazing experiment. Specifically, most of the leaves were removed and only a small section of the stem was left in the 'H4' treatment. Hence, there may have been limited resources and meristems for the regrowth of L. chinensis because of its short growth times from tiller emergence to the first simulated grazing treatment and from the last treatment to harvest.</ns0:p></ns0:div>
<ns0:div><ns0:head>CONCLUSIONS</ns0:head><ns0:p>Consistent with studies on contemporary evolution and stress memory <ns0:ref type='bibr' target='#b12'>( Detling & Painter, 1983;</ns0:ref><ns0:ref type='bibr' target='#b47'>Oesterheld & McNaughton, 1988;</ns0:ref><ns0:ref type='bibr' target='#b2'>Agrawal et al., 2012;</ns0:ref><ns0:ref type='bibr' target='#b67'>Züst et al., 2012;</ns0:ref><ns0:ref type='bibr' target='#b15'>Didiano et al., 2014)</ns0:ref>, our study showed that OG exhibited higher adaptation to simulated grazing in terms of the growth, cloning and colonizing ability than NG. This stronger adaptation was attributed to enhanced above-ground spatial avoidance. Contrary to our prediction, OG pre-allocated less biomass to the rhizome, which does not seem to promote herbivory tolerance and avoidance in L. chinensis; however, this also may reflect a tolerance strategy via shorter rhizomes and more ramets. Here, we quantitatively studied the grazing-induced spatial avoidance of plants for the first time and found that enhanced above-ground spatial avoidance was induced by a larger leaf angle and shorter height. However, because of the pseudo-replication and because the sample areas only consisted of two adjacent pastures in this study, the findings from our research may be limited to our study site. </ns0:p></ns0:div><ns0:figure xml:id='fig_0'><ns0:head>Figure 1</ns0:head><ns0:label>1</ns0:label><ns0:figDesc>Figure 1</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_1'><ns0:head>Figure 2</ns0:head><ns0:label>2</ns0:label><ns0:figDesc>Figure 2</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_2'><ns0:head>Figure 3 The</ns0:head><ns0:label>3</ns0:label><ns0:figDesc>Figure 3</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_3'><ns0:head>Figure 4</ns0:head><ns0:label>4</ns0:label><ns0:figDesc>Figure 4</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_4'><ns0:head>Figure 5</ns0:head><ns0:label>5</ns0:label><ns0:figDesc>Figure 5</ns0:figDesc></ns0:figure>
<ns0:note place='foot'>PeerJ reviewing PDF | (2020:05:49307:2:0:NEW 27 Aug 2020)</ns0:note>
</ns0:body>
" | "Pratacultural College
Gansu Agricultural University
Lanzhou, Gansu Province, PR China
August 5, 2020
Dear Mr. Curtis:
We thank you for the valuable scholarly comments and suggestions on our manuscript entitled “Overgrazing-induced legacy effects permit Leymus chinensis to cope with herbivory” (#49307). We have read the comments carefully and revised the manuscript to address your raised concerns. The major changes to the manuscript are described in detail in the attached point-to-point response below. The comments are listed in italicized font and our responses are provided without italics.
Sincerely,
Fenghui Guo
Reply to Editor’s comments：
This manuscript has been substantially improved. However, I have identified additional necessary as follows:
Issue1: Title – I don’t think the current wording is supported by the available data. Replacing “permit” with “may permit” would be improve accuracy.
Response: Thank you for this comment. We have revised it. (Line 1)
Issue2: L 22 “cloning ability” – the term needs to be clearly defined or avoided (see further comments below)
Response: Thank you for this comment. We have redefined the cloning ability which relates to ramet number and redefined colonizing ability which relates to rhizome length and rhizome internode number. (Line 22, 195-196, 390)
Issue3: L 25 “while height” to “while reduced tiller natural height”
Response: We have revised it. (Line 26)
Issue4: L 28 “via shorter rhizome and more tiller ramets.” to “where reduced allocation to rhizomes is associated with increased production of ramets”.
Response: We have revised it. (Line 29-30)
Issue5: L 43 “the plant-herbivore relationship” to “plant-herbivore relationships”
Response: We have revised it. (Line 43)
Issue6: L 43 “adaptability” to “adaptations”
Response: We have revised it. (Line 43)
Issue7: L 44 “the interspecific relationship, and” to “as well as interspecific relationships and”
Response: We have revised it. (Line 44)
Issue8: L 77 “the common garden” to “common garden”
Response: We have revised it. (Line 77)
Issue9: L 97 “grazing disturbance” to “disturbance”
Response: We have revised it. (Line 97)
Issue10: L 123 “Given that the field site from which we collected materials was not free from pseudo-replication,” to “Although we acknowledge pseudoreplication in the experimental design, as each treatment consisted of one large plot with subsamples as replicates,”
Response: We have revised it. (Line 123-124)
Issue11: L 127 “and to decrease the impacts of pseudo-replication on the experimental results.” to “across each of the large neighboring plots.”
Response: We have revised it. (Line 127)
Issue12: L 130 what does “well-prepared” mean?
Response: we replaced “into a well-prepared moistened soil container” with “into a container with moistened soil”. (Line 130)
Issue13: L 131 “Materials cultivation in the greenhouse” to “Cultivation in the greenhouse”
Response: Thank you for this comment. We have revised it. (Line 131)
Issue14: L 132 “into 2-cm long” to “ into a 2-cm long section”
Response: We have revised it. (Line 133)
Issue15: L 152 “the soil surface.” to “the soil surface, respectively”.
Response: We have revised it. (Line 153)
Issue16: L 154 “in its” to “in their”
Response: We have revised it. (Line 156)
Issue17: L 156 “imitate” to “simulate”
Response: We have revised it. (Line 157)
Issue18: L 158 I think more explanation / justification is needed for these terms. To start, I suggest something like “Although we did not assess genotypic differences among plants in our replicate pots, we nevertheless refer to each replicate potted plant as a genet. We use the term ramet to refer to a tiller that has sprouted from a rhizome bud (i.e. an individual that is a physiologically integrated component of the genet).” I find the following sentence (L 159) very confusing . The problem is that a tiller is simply defined as a grass stem. Therefore, what you have drawn on Fig S1 as “rhizome ramet” is actually a tiller (individual grass stem). I think new terms may be needed here, and just as importantly, please explain to readers why it is important or interesting to separate these two types of tillers. I suppose one type contributes to spatial spread while the other favors increasing biomass density. Is this difference relevant to your study on grazing effects? After reading the entire manuscript, I see that you have not described differences between ramet types in Results or discussed the two ramet types in Discussion (except brief mention of “tiller ramets' on L 365), so I wonder if confusion could be avoided by combining all ramets rather than defining two types? See also comments on Fig 5 below.
Response: Thank you for these comments. We have combined all ramets to avoid the confusion. We have revised that in MATERIALS & METHODS, Results, Discussion, and Fig 5. Please see Line 159-162, 175, 272, 353, and Fig 5.
Issue19: L 173 oven-dring to oven-drying
Response: We have revised it. (Line 174)
Issue20: L 177 “the mesh bag” to “mesh bags”
Response: We have revised it. (Line 178)
Issue21: L 193 “respected to” to “respect to”
Response: We have revised it. (Line 194)
Issue22: L 194 “rhizome section number” – what does it mean?
Response: We have replaced the “rhizome section number” with “rhizome internode number”.
Issue23: L 194 Cloning ability if used here, needs to be clearly defined. Rhizome length doesn’t seem to relate to cloning ability, rather it relates to physical spread.
Response: We have revised it. Please see the Responses of Issue 2.
Issue24: L 201 What is GZ?
Response: We are sorry that it was an error. We have revised it. (Line 202)
Issue25: L 204 “plasticity index to simulated herbivory of these traits (PI)” to “plasticity index (PI) to simulated herbivory of these traits”
Response: We have revised it. (Line 205)
Issue26: L 207 Why does this formula have “( )” to the right of each variable?
Response: Thank you for this comment. The “( )” indicate phenotypic traits which is used to calculate PI, such as PI (root biomass). However, this may be not necessary in the formula. We have delete the “( )” in the formula. (Line 207)
Issue27: L 207 Why was H4 selected for the PI?
Response: The PI was used to compare the different responses to simulated grazing between NG and OG. The H4 had a greater effect on L. chinensis than H8, and the changing trend for indicators under different treatment was CK-H8-H4, so H4 was selected for the PI.
Issue28: L 218 What does “increasing pathways” mean?
Response: The initial model could be modified with the Modification Index in the AMOS. According to the Modification Index, we could increase pathways between residual variables to get a better model. We have revised this description. (Line 218)
Issue29: L 226 Due to the pseudoreplicated design, this line should be worded something like “There were significant legacy effects in the biomass allocation of L. chinensis genets sampled from the over-grazing plot.”
Response: Thank you for this comment. We have revised it. (Line 226)
Issue30: L 228 “and root biomass” to “while root biomass”
Response: We have revised it. (Line 228)
Issue31: L 241-242 round off percent values here to whole numbers (89%, 69%, 130%)
Response: We have revised it. (Line 243-245)
Issue32: L 244 What does “homogenous environment” mean here?
Response: We emphasized that OG and NG grew in the same environment without grazing. We have replaced the “in the homogenous environment without grazing” with “under CK treatment” to get a more concise description. (Line 244-245)
Issue33: L 245 It is impossible for any value to decrease by more than 100%. If a value decreases by 100% it means there is nothing left.
Response: We calculated this with the formula “(NG-OG)/NG”. Because the leaf angle of OG is lager than NG, so the value is negative, and we described that with “increasing by 130%”.
Issue34: L 247 “it had” to “they had”
Response: We have revised it. (Line 247)
Issue35: L 262 “genet” to “genets”
Response: We have revised it. (Line 262)
Issue36: L 263 What is meant by “disturbances”? Biomass loss?
Response: Yeah, “disturbances” indicated “Biomass loss”. We have revised it. (Line 263)
Issue37: L 263 “based on” to “in response to”
Response: We have revised it. (Line 263)
Issue38: L 264 “Besides” to “Additionally”
Response: We have revised it. (Line 264)
Issue39: L 277 “under H4” do your mean “for H4” or “in H4”?
Response: Year, we have replaced “under H4” with “in H4”. (Line 276)
Issue40: L 280 “Compared with” to “In comparison to”
Response: We have revised it. (Line 279)
Issue41: L 281 “an increasing trend in” to “increase in”
Response: We have revised it. (Line 280)
Issue42: L 282 “one species population” to “populations of a single species may vary, with populations”
Response: We have revised it. (Line 282)
Issue43: L 285 “the common garden” to “common garden”
Response: We have revised it. (Line 284)
Issue44: L 289 “smaller” to “shorter”
Response: We have revised it. (Line 288)
Issue45: L 299 “may also do not” to “may not”
Response: We have revised it. (Line 299)
Issue46: L 300 “saliva” to “grazer saliva”
Response: We have revised it. (Line 299)
Issue47: L 309 “high” to “taller”
Response: We have revised it. (Line 308)
Issue48: L 322 “overgrazing-resulted more severe drought” to “overgrazing may promote increases in localized drought, which” [I think this is what you mean?]
Response: Yeah, that is what we mean, and we have revised it. (Line 312)
Issue49: L 325 I don’t think “viability” is the correct word. Do you mean “aboveground growth and grazing tolerance” or something similar?
Response: Thank you for this comment. We have replaced the “viability” with “grazing tolerance” (Line 324)
Issue50: L 331 “ Colonization-competition” “ A colonization-competition”
Response: We have revised it. (Line 330)
Issue51: L 333 “resulted in” to “which result in”
Response: We have revised it. (Line 332)
Issue52: L 362 “be stem” to “stem”
Response: We have revised it. (Line 360)
Issue53: L 374 “In Contrast” to “In contrast”
Response: We have revised it. (Line 372)
Issue54: L 394 “benefit the” to “seem to promote”
Response: We have revised it. (Line 392)
Issue55: L 394 “of L. chinensis” to “in L. chinensis”
Response: We have revised it. (Line 392)
Issue56: L 395 “shorter rhizome” to “shorter rhizomes”
Response: We have revised it. (Line 393)
Issue57: L 399 delete “small”
Response: We have revised it. (Line 37)
Issue58: Fig 1 caption – delete “light-dark”
“the L. chinensis genet collected from grazing” to “L. chinensis genets collected from the grazing”
“the L. chinensis genet collected from” to “L. chinensis genets collected from”
Response: Thank you for this comment. We have revised it. (Please see Fig.1)
Issue59: Fig 3 – For “material source” I suggest “plant source (NG or OG)” to increase clarity.
Response: Thank you for this comment. We have revised it. (Please see Fig.3)
Issue60: Fig 5 It is not clear how these 5 graphs relate to “cloning ability”. Perhaps production each new ramet could be considered a new clone that is integrated as one plant; however, how does rhizome length indicate cloning ability? The “total ramet production” graph could be presented in place of the so called “tiller ramet” and “rhizome ramet” graphs as the same statistical findings and trends are the same.
Response: Thank you for this comment. We have revised it. Please see the responses for Issue 2 and Issue 18.
Issue61: Table 1 caption “while oter phenotypic plastics” what does this mean? Do you mean PI? I don’t see PI in the table.
Response: Thank you for this comment. That is an error, and we have replaced the “while oter phenotypic plastics” with “while other phenotypic traits”. (Please see Fig.3)
Issue62: Define “Lg” in the table.
Response: We have revised it.
" | Here is a paper. Please give your review comments after reading it. |
9,753 | "<ns0:abstract xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:p>There is growing evidence that herbivory-induced legacy effects permit plants to cope with herbivory. However, herbivory-induced defense strategies in plants against grazing mammals have received little attention. To further understand the grazing-induced legacy effects on plants, we conducted a greenhouse experiment with Leymus chinensis experiencing different grazing histories. We focused on grazing-induced legacy effects on above-ground spatial avoidance and below-ground biomass allocation. Our results showed that L. chinensis collected from the continuous overgrazing plot (OG) exhibited higher performance under simulated grazing in terms of growth, cloning and colonizing ability than those collected from the 35-year no-grazing plot (NG). The enhanced adaptability of OG was attributed to increased above-ground spatial avoidance, which was mediated by larger leaf angle and shorter height (reduced vertical height and increased leaf angle contributed to the above-ground spatial avoidance at a lower herbivory stubble height, while reduced tiller natural height contributed to above-ground spatial avoidance at a higher herbivory stubble height). Contrary to our prediction, OG pre-allocated less biomass to the rhizome, which does not benefit the herbivory tolerance and avoidance of L. chinensis; however, this also may reflect a tolerance strategy where reduced allocation to rhizomes is associated with increased production of ramets.</ns0:p></ns0:div>
</ns0:abstract>
<ns0:body xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:head>INTRODUCTION</ns0:head><ns0:p>Environmental disturbances (e.g., drought, herbivory) can have persistent effects on ecological attributes (e.g., ecological processes, community structure, population dynamics, and plant and soil characteristics) long after they occur (i.e., the legacy effect) <ns0:ref type='bibr' target='#b21'>(Fox et al., 2015;</ns0:ref><ns0:ref type='bibr' target='#b31'>Kafle & Wurst, 2019)</ns0:ref>. Legacy effects are ubiquitous phenomena in nature and have been extensively studied in the context of plant succession, herbivory, invasive plants, ecosystem engineering, and human land use <ns0:ref type='bibr' target='#b11'>(Cuddington, 2011;</ns0:ref><ns0:ref type='bibr' target='#b35'>Kostenko et al., 2012;</ns0:ref><ns0:ref type='bibr' target='#b65'>Wurst & Ohgushi, 2015)</ns0:ref>. In grassland ecosystems, herbivory-induced legacy effects on ecological processes, such as plant succession and biological diversity change, can persist for decades and even millennia <ns0:ref type='bibr' target='#b28'>(Holeski, Jander & Agrawal, 2012;</ns0:ref><ns0:ref type='bibr' target='#b21'>Fox et al., 2015)</ns0:ref>. Some studies have shown that legacies in plant defense strategies can mediate herbivory-induced legacy effects on ecological processes. For instance, defense traits (e.g., chemical defense substances, resource reallocation) can significantly affect plant-herbivore relationships (generally increasing plant adaptations to herbivores), as well as interspecific relationships and soil characteristics <ns0:ref type='bibr' target='#b65'>(Wurst & Ohgushi, 2015)</ns0:ref>. Therefore, clarifying the overgrazing-induced legacy effects in plant defense strategies is critical for understanding processes occurring in grazing ecosystems. However, most previous studies on herbivory-induced legacy effects on plant defense strategies have focused on short-term insect herbivory, in contrast, studies examining long-term livestock grazing have received less attention by comparison <ns0:ref type='bibr' target='#b28'>(Holeski, Jander & Agrawal, 2012;</ns0:ref><ns0:ref type='bibr' target='#b31'>Kafle & Wurst, 2019)</ns0:ref>.</ns0:p><ns0:p>Strategies by which plants cope with herbivory include resistance, avoidance, and tolerance <ns0:ref type='bibr' target='#b15'>(Didiano et al., 2014)</ns0:ref>. While resistance strategies (e.g., thorns, higher tannin concentrations) play an important role in coping with insect herbivory <ns0:ref type='bibr' target='#b1'>(Agrawal et al., 2002;</ns0:ref><ns0:ref type='bibr' target='#b28'>Holeski, Jander & Agrawal, 2012)</ns0:ref>, they may be less useful in plants experiencing herbivory from livestock or other grazing mammals, as these large animals are unable to selectively graze at such a fine scale <ns0:ref type='bibr' target='#b45'>(Menard et al., 2002;</ns0:ref><ns0:ref type='bibr'>Benot et al., 2013)</ns0:ref>. Empirical evidence indicates that plants under grazing mammal herbivory show adaptive legacy effects via above-ground spatial avoidance traits which resulted in the distribution of more above-ground biomass close to the ground, including larger leaf angles, shorter height, and more prostrate growth forms <ns0:ref type='bibr' target='#b51'>(Polley & Detling, 1988</ns0:ref><ns0:ref type='bibr' target='#b52'>, 1990;</ns0:ref><ns0:ref type='bibr' target='#b49'>Painter, Detling & Steingraeber, 1993;</ns0:ref><ns0:ref type='bibr' target='#b58'>Tomás et al., 2000;</ns0:ref><ns0:ref type='bibr' target='#b38'>Li et al., 2015;</ns0:ref><ns0:ref type='bibr' target='#b55'>Ren et al., 2017)</ns0:ref>. However, studies examining the above-ground biomass vertical distribution have been qualitative-not quantitative. The lack of rigorous quantitative approaches severely limits our understanding of the role of these avoidance traits in the responses of plants to grazing mammal herbivory (e.g., livestock grazing). For example, the specific morphological characters that lead to the near-surface distribution of above-ground biomass to reduce the possibility of defoliation by grazing mammals remain unclear.</ns0:p><ns0:p>Biomass reallocation is a fundamental strategy for plants to cope with herbivory. When subjected to above-ground herbivory, more biomass is mobilized above-ground to facilitate the recovery of growth <ns0:ref type='bibr' target='#b40'>(Liu et al., 2018)</ns0:ref>, and a transient transformation of resources away from herbivores occurs within hours after herbivory <ns0:ref type='bibr' target='#b3'>(Anten & Pierik, 2010;</ns0:ref><ns0:ref type='bibr' target='#b48'>Orians, Thorn & Gomez, 2011)</ns0:ref>. Some studies have indicated that plant tolerance is tightly linked to biomass allocation patterns expressed before herbivory; that is, larger belowground biomass pre-allocation is associated with stronger tolerance of plants to above-ground herbivory <ns0:ref type='bibr' target='#b19'>(Fornoni, 2011;</ns0:ref><ns0:ref type='bibr' target='#b42'>Lurie, Barton & Daehler, 2017)</ns0:ref>. In addition to the importance of tolerance, large belowground biomass pre-allocation potentially helps plants avoid above-ground herbivory. There is growing evidence that long-term grazing induces higher belowground biomass allocation at the community and population levels <ns0:ref type='bibr' target='#b46'>(Milchunas & Lauenroth, 1993;</ns0:ref><ns0:ref type='bibr' target='#b39'>Lindwall et al., 2013;</ns0:ref><ns0:ref type='bibr'>)</ns0:ref>. Furthermore, more biomass allocation to roots for plants collected from grazing areas has been observed in common garden environments free of herbivory disturbance <ns0:ref type='bibr' target='#b29'>(Jaramillo & Detling, 1988;</ns0:ref><ns0:ref type='bibr' target='#b51'>Polley & Detling, 1988</ns0:ref><ns0:ref type='bibr' target='#b52'>, 1990)</ns0:ref>. However, few studies have examined herbivory-induced legacy effects on plant belowground biomass allocation, especially the allocation of belowground reproductive organs such as rhizomes.</ns0:p><ns0:p>Leymus chinensis, a rhizomatous clonal plant, is the dominant species on the Inner Mongolia typical steppe grasslands. This grass species is relished by livestock (e.g., cattle, sheep) and has been subjected to overgrazing for more than 50 years. Similar to plants that have been studied in other regions, L. chinensis exhibits significant grazing-induced legacy effects, such as short height and short leaves <ns0:ref type='bibr' target='#b38'>(Li et al., 2015;</ns0:ref><ns0:ref type='bibr' target='#b55'>Ren et al., 2017)</ns0:ref>. However, no quantitative studies have characterized herbivory-induced legacy effects on the vertical distribution of above-ground biomass and belowground biomass allocation (especially the allocation of rhizome biomass). Although earlier studies <ns0:ref type='bibr' target='#b47'>(Oesterheld & McNaughton, 1988;</ns0:ref><ns0:ref type='bibr' target='#b41'>Loreti, Oesterheld & Sala, 2001)</ns0:ref> have shown that plants subjected to grazing disturbance showed stronger adaptability to herbivory compared with ungrazed plants in a common garden environment, studies reporting legacy effects underlying L. chinensis adaptation to grazing are limited.</ns0:p><ns0:p>To further characterize the grazing-induced legacy effects on plants, we conducted a greenhouse pot experiment with L. chinensis collected from two adjacent plots separated by a pasture fence. The first was a 35-year no-grazing plot and the second is a long-term overgrazing plot. Our study sought to answer three questions. First, does L. chinensis exposed to long-term overgrazing disturbance exhibit enhanced above-ground spatial avoidance (measured by the above-ground biomass vertical distribution)? If so, which individual characteristics contribute to this trait? Second, are there overgrazing-induced legacy effects on L. chinensis in terms of the belowground biomass allocation (i.e., root and rhizome)? Third, does the L. chinensis collected from the grazing plot exhibit stronger adaptation to simulated herbivory compared with those collected from the no-grazing plot?</ns0:p></ns0:div>
<ns0:div><ns0:head>MATERIALS & METHODS</ns0:head></ns0:div>
<ns0:div><ns0:head>Materials</ns0:head></ns0:div>
<ns0:div><ns0:head>Description of studying sites</ns0:head><ns0:p>Samples of L. chinensis were collected from typical steppe grassland located at the Inner Mongolia Grassland Ecosystem Research Station (43 • 38' N, 116 • 42' E). The sampling sites comprise two adjacent plots separated by a pasture fence. The first was a no-grazing plot (600×400 m), which has been fenced since 1983 for long-term ecological observations, and the second was a continuously overgrazing plot (600×100 m) that has been grazed at a stocking rate of ~3 sheep units per hectare for more than 50 years. However, the stocking rate recommended by the local government is ~1.5 sheep units per hectare to achieve a balance between grassland productivity and livestock forage requirements. Thus, the continuously grazed plot has experienced heavy grazing pressure over the last several decades <ns0:ref type='bibr'>(Ma, et al., 2015)</ns0:ref>.</ns0:p></ns0:div>
<ns0:div><ns0:head>Materials Collection at the studying sites</ns0:head><ns0:p>Genotypes showed different phenotypic plasticity to environmental disturbance and different genetic structures may be one of the main mechanisms mediating overgrazing-induced legacies on L. chinensis <ns0:ref type='bibr'>(Josephs, 2018)</ns0:ref>. We could not determine the genotype of each L. chinensis individual collected from both plots. To reduce the possible impacts of genotypes on our experimental results, we conducted our experiment at the population level and used the highreplication sampling.</ns0:p><ns0:p>We sampled 150-segment rhizomes with a root drill at 150 random points in each of the two treatment plots at the beginning of the growing season. Although we acknowledge pseudoreplication in the experimental design, as each treatment consisted of one large plot with subsamples as replicates, we sampled the entire area for each plot except for the margin and ensured that the distance between every two sampling points was greater than 20 m; this was done both to improve the representativeness of the samples for each plot across each of the large neighboring plots. Each rhizome was approximately 4 cm long and contained at least one sprouting section containing buds. To prevent the sampled rhizome from losing its vitality following removal from the soil environment, we immediately transferred samples into a container with moistened soil.</ns0:p></ns0:div>
<ns0:div><ns0:head>Cultivation in the greenhouse</ns0:head><ns0:p>In the laboratory, each rhizome was dissected into a 2-cm long section with only one node. All of the rhizomes were cultivated in flowerpots, 20 cm in diameter and 15 cm high, which were kept in the greenhouse. Each flowerpot contained 3 kg of soil collected from points adjacent to the experimental site and was subsequently planted with one rhizome section. There were 150 flowerpots for rhizomes collected from the no grazing plot and the overgrazing plot, respectively. After 20 days, the rhizomes had sprouted in approximately 100 flowerpots from each treatment.</ns0:p></ns0:div>
<ns0:div><ns0:head>Experimental Design & Measurements</ns0:head></ns0:div>
<ns0:div><ns0:head>Experimental design</ns0:head><ns0:p>The experiment was a full factorial design and consisted of two factors. The first was the source of L. chinensis (NG: L. chinensis collected from the no-grazing plot; OG: L. chinensis collected from the overgrazing plot), and the second was simulated grazing (CK: no simulated grazing; H8: simulated moderate grazing; H4: simulated heavy grazing). For NG and OG, we randomly selected 90 flowerpots that had sprouted rhizomes. These flowerpots were randomly arranged in the greenhouse, and we alternated their position every week to exclude the influence of external factors (e.g., light). One-third of the L. chinensis growing in flowerpots ( in both NG and OG separately) were randomly assigned to CK, H8, and H4. However, several L. chinensis died during the experiment and the numbers left for each treatment were NG*CK ( <ns0:ref type='formula'>22</ns0:ref>), NG*H8 (29), NG*H4 (25), OG*CK ( <ns0:ref type='formula'>22</ns0:ref>), OG*H8 (26), and OG*H4 (28). We clipped plants with scissors to conduct the simulated grazing as per <ns0:ref type='bibr' target='#b59'>Turley (2013)</ns0:ref> and <ns0:ref type='bibr' target='#b15'>Didiano et al (2014)</ns0:ref>. We simulated moderate and heavy grazing by clipping the above-ground part of plants 8 cm and 4 cm above the soil surface, respectively. The stubble height in the simulated grazing treatment was per <ns0:ref type='bibr' target='#b22'>Gao (2008)</ns0:ref>. Sloping parts (e.g., the sloping leaves and stems) were not straightened in the simulated grazing treatments. We only removed plant parts that were distributed above 8 cm or 4 cm in their natural state. The herbivory simulation was carried out three times at 45, 60, and 75 days after the rhizomes had sprouted to simulate repeated grazing in natural environments.</ns0:p></ns0:div>
<ns0:div><ns0:head>Measurements</ns0:head><ns0:p>Although we did not assess genotypic differences among plants in our replicate pots, we nevertheless refer to each replicate potted plant as a genet. We use the term ramet to refer to a tiller that has sprouted from a rhizome bud (i.e. an individual that is a physiologically integrated component of the genet). The clipped biomass from the simulated grazing treatment was oven-dried (60 ºC for 48 h, the same below) and weighed. L. chinensis stopped growing 90 days after the rhizomes had sprouted and the experiment was terminated. First, we measured the morphological characters of the ramets from the CK treatment, including natural height, vertical height, stem height, leaf length, leaf width, leaf number, and leaf angle. There were about 10 ramets in each flowerpot and about five leaves on each ramet. Generally, the ramet that sprouted first was the tallest and the second leaf from the ground was the longest. Thus, we selected the ramet that sprouted first in each flowerpot and chose the second leaf on the selected ramet for measurement. Leaf angle was measured using a protractor as degrees (0-90°) from the ramet stem to the measured leaf. Next, we clipped the selected ramet biomass above 8 cm in height, those between 4-8 cm in height, and those below 4 cm in height (Note: the sloping parts of selected ramets were not straightened when clipped, and biomass data of each of the three parts were collected in their natural state), followed by oven-drying and weighing. Second, after counting the ramets, we harvested the genet above-ground biomass at different vertical distributions (i.e., above 8 cm in height, between 4-8 cm in height and below 4 cm in height for CK and H8; above and below 4 cm in height for H4), followed by oven-drying and weighing. Third, all soil with roots and rhizomes in the flowerpot was transferred into mesh bags and rinsed until only the clean roots and rhizomes remained. The stem below the soil surface (ca., 2 cm deep) was treated as the genet above-ground biomass below 4 cm. We also measured the morphological characteristics of rhizomes, including length and internode number.</ns0:p></ns0:div>
<ns0:div><ns0:head>Statistical Analysis</ns0:head><ns0:p>Root biomass allocation, rhizome biomass allocation, and below-ground biomass allocation were estimated using root biomass divided by total biomass, rhizome biomass divided by total biomass, and below-ground biomass divided by total biomass, respectively. The above-ground biomass vertical distribution parameters, including above-ground biomass distribution below 4 cm and 8 cm, were estimated using below 4-cm biomass divided by above-ground biomass, and below 8-cm biomass divided by total above-ground biomass, respectively. The above-ground biomass for each layer included the clipping biomass during the simulated grazing treatment.</ns0:p><ns0:p>The genet biomass allocation parameters and the above-ground biomass vertical distribution parameters were used to evaluate spatial grazing avoidance and the induced spatial grazing avoidance using a two-way analysis of variance (two-way ANOVA). Similarly, two-way ANOVA was used to analyze the influence of long-term overgrazing-induced legacy effects on the adaptation of L. chinensis to grazing with respect to its growth ability (biomass accumulation), cloning ability (ramet number) and colonizing ability (rhizome length and rhizome internode number). A significant interaction between the two factors under study indicates an effect of long-term overgrazing on the response of L. chinensis to simulated grazing. If the data for a trait were not normally distributed or homogeneous in variance, such data were transformed using various methods (e.g., logarithmic, square root, square, reciprocal, or square root inverse rotation conversion) to attain normality and homoscedasticity. However, if data transformation could not make the data normally distributed and variances homogeneous, we conducted a one-way analysis of variance (one-way ANOVA) or Kruskal-Wallis test. If only NG or OG had a significant response to simulated grazing, this implied that responses to simulated grazing between NG and OG were disparate. In contrast, if significant differences were observed in the simulated grazing treatments in both NG and OG, we calculated the plasticity index (PI) to simulated herbivory of these traits using the following formula:</ns0:p><ns0:p>'PI = (CK -H4) / CK'.</ns0:p><ns0:p>H4 was chosen for the PI estimate because H4 had a greater effect on L. chinensis than H8. Student's t-test or Kruskal-Wallis test was used to compare differences in the phenotypic traits of L. chinensis ramets collected from the different plots (i.e., NG vs OG). The relationship between the traits was explored using the Pearson correlation method.</ns0:p><ns0:p>Structural equation modeling (SEM) was used to evaluate the importance of individual morphological traits for the above-ground spatial grazing avoidance of L. chinensis. The model assumed that the above-ground spatial grazing avoidance of L. chinensis, which was measured by the above-ground biomass vertical distribution, was attributed to individual morphological characters. The Pearson correlation analysis was conducted between all parameters included in the model. The initial model was developed according to the results of the correlation analysis and basic knowledge of plant science. Furthermore, the model was modified by deleting nonsignificant pathways and by increasing pathways between residual variables. χ 2 statistics with the associated probability, the root mean square errors of approximation with the associated probability, and the Bentler-Bonett Index or Normed Fit Index were used to evaluate the overall fit of the model.</ns0:p><ns0:p>All the analyses were completed using IBM SPSS Statistics 19.0 and the means were compared using Tukey's HSD test (P < 0.05). The figures were generated in Oringin2019b.</ns0:p></ns0:div>
<ns0:div><ns0:head>RESULTS</ns0:head></ns0:div>
<ns0:div><ns0:head>Genet biomass allocation and above-ground biomass vertical distribution</ns0:head><ns0:p>There were significant legacy effects in the biomass allocation of L. chinensis genets sampled from the over-grazing plot. In CK, the rhizome and belowground biomass allocation of OG decreased significantly compared with that observed from NG (P< 0.001 ), while root biomass allocation was not sensitive to overgrazing-induced legacies (P=0.057). Under simulated heavy grazing, OG had greater belowground biomass allocation (P=0.004) and root biomass allocation (P<0.05) than NG. There were significant decreases in root and rhizome biomass allocation under the simulated grazing treatment (P<0.05). Compared with NG, OG showed a smaller plastic index (PI) under simulated grazing treatment in terms of root and rhizome biomass allocation (Fig. <ns0:ref type='figure' target='#fig_1'>1</ns0:ref>).</ns0:p><ns0:p>There was a significant difference in the genet above-ground biomass vertical distribution between NG and OG. OG tended to allocate more biomass close to the ground with a larger above-ground biomass distribution below 4 cm and 8 cm than NG (P<0.001). The simulated grazing treatment did not alter the genet vertical distribution of biomass of NG but reduced the above-ground biomass distribution of OG close to the ground (P<0.05) (Fig. <ns0:ref type='figure' target='#fig_2'>2</ns0:ref>).</ns0:p></ns0:div>
<ns0:div><ns0:head>Ramet above-ground biomass vertical distribution</ns0:head><ns0:p>There was a significant correlation between the ramet and genet above-ground biomass vertical distribution (P<0.001) (Fig. <ns0:ref type='figure' target='#fig_1'>S1</ns0:ref>). The ramet distribution below 4 cm and 8 cm of OG were 89% and 69% larger than those of NG, respectively (P<0.001). L. chinensis ramet morphological traits (e.g., natural height, vertical height) showed significant grazing-induced legacies under CK treatment (Table <ns0:ref type='table'>1</ns0:ref>). Leaf angle showed the most pronounced change, increasing by 130%, while the leaf number did not respond significantly to grazing legacies (Fig. <ns0:ref type='figure' target='#fig_2'>S2</ns0:ref>). Although OG ramets accumulated fewer photosynthetic products above 4 cm and 8 cm, they had larger accumulated biomass below 4 cm and 8 cm (P<0.001) (Table <ns0:ref type='table'>1</ns0:ref>). SEM explained 93% and 94% of the variation in the ramet above-ground biomass vertical distribution below 4 cm and 8 cm, respectively. The larger near-surface distribution of aboveground biomass of OG ramets stemmed from the larger near-surface biomass accumulation and smaller biomass accumulation farther from the ground, and this above-ground spacial avoidance was induced by individual morphological characteristics (Fig. <ns0:ref type='figure' target='#fig_3'>3</ns0:ref>). According to the 'Standardized Total Effects, ' vertical height and leaf angle played a more important role than other traits in inducing the above-ground spatial avoidance below 4 cm, while natural height made the highest contribution to above-ground spatial avoidance below 8 cm. The standardized total effects of grazing-induced legacies on the below 4-cm and 8-cm ramet above-ground biomass vertical distributions were -0.73 and -0.77, respectively (Table <ns0:ref type='table'>S1, S2</ns0:ref>). Manuscript to be reviewed</ns0:p></ns0:div>
<ns0:div><ns0:head>L. chinensis genet performance</ns0:head><ns0:p>Simulated grazing significantly reduced the below-ground, rhizome, root, and total biomass of L. chinensis genets, but did not affect the above-ground biomass. NG experienced more serious biomass loss compared with OG and had a larger 'PI' in response to the simulated herbivory in terms of the below-ground, rhizome, root, and total biomass accumulation. Additionally, there were significant interactions between plant source (NG or OG) and simulated herbivory treatment in terms of the below-ground, root, and total biomass. In the control treatment, there were no differences in the total, root, and below-ground biomass accumulation between NG and OG (P>0.05). In contrast, OG accumulated higher total, root, and belowground biomass than NG under simulated herbivory (P<0.01). Furthermore, NG accumulated more rhizome biomass than OG under CK treatment, while there were no significant differences were observed between NG and OG under simulated grazing treatments (Fig. <ns0:ref type='figure' target='#fig_4'>4</ns0:ref>).</ns0:p><ns0:p>Simulated herbivory had no effects on ramet number (P>0.05) and OG showed more ramets than NG. Simulated herbivory significantly reduced the rhizome length and rhizome internode number of NG but did not affect that of OG. Under CK, the rhizome length and the number of rhizome sections of NG were larger than those of OG (P<0.001), and no differences in these variables were detected in H4 (Fig. <ns0:ref type='figure' target='#fig_5'>5</ns0:ref>).</ns0:p></ns0:div>
<ns0:div><ns0:head>DISCUSSION</ns0:head></ns0:div>
<ns0:div><ns0:head>The vertical distribution of above-ground biomass</ns0:head><ns0:p>In comparison to species that decrease in abundance under intensive grazing, plant species that exhibit increase in abundance are generally shorter and more prostrate <ns0:ref type='bibr'>(DÍAZ et al., 2007;</ns0:ref><ns0:ref type='bibr' target='#b20'>Forrestel, Donoghue & Smith, 2015)</ns0:ref>. Previous work has also shown that populations of a single species may vary, with populations in over-grazed fields are characterized by more prostrate and shorter growth compared with populations in fields where grazing is excluded, and this phenomenon could persist for several generations after transplanting plants into common garden environments <ns0:ref type='bibr' target='#b8'>(Carman, 1985;</ns0:ref><ns0:ref type='bibr' target='#b52'>Polley & Detling, 1990;</ns0:ref><ns0:ref type='bibr' target='#b49'>Painter, Detling & Steingraeber, 1993;</ns0:ref><ns0:ref type='bibr' target='#b56'>Rotundo & Aguiar, 2008;</ns0:ref><ns0:ref type='bibr' target='#b15'>Didiano et al., 2014)</ns0:ref>. Consistent with these reports, our direct measurements of the vertical distribution of above-ground biomass showed that OG allocated more above-ground biomass close to the ground. SEM analysis indicated that the shorter plant height and larger leaf angles of OG drove this pattern. This study confirms the findings of previous studies by clarifying changes in plant phenotypes under intensive grazing, specifically, more biomass is distributed close to the ground via shorter and more prostrate growth forms <ns0:ref type='bibr' target='#b15'>(Didiano et al., 2014)</ns0:ref>.</ns0:p><ns0:p>There was no marked plasticity in the above-ground biomass vertical distribution of L. chinensis to simulated grazing; however, we expected that simulated grazing should significantly increase plant biomass allocation toward the ground. This might be expected given that the ramet number did not respond significantly to simulated grazing in our experiment. A large number of new shorter ramets could contribute to the near-surface allocation of plant biomass via the shorter natural height as mentioned above. On the other hand, we speculate that leaf angle, an important morphological trait influencing the above-ground biomass vertical distribution, may not exhibit significant responses to simulated grazing without grazer saliva and trampling.</ns0:p><ns0:p>Furthermore, the lack of significant responses of the vertical distribution of L. chinensis aboveground biomass to simulated grazing indicates that overgrazing-induced legacies in terms of the vertical distribution of above-ground biomass cannot be induced by short-term defoliation, and may be attributed instead to long-term defoliation or other pathways. In addition to defoliation, livestock can also influence plants by trampling, saliva, and indirect effects (e.g., increasing soil density, changing the rate of light interception by plants, etc.) <ns0:ref type='bibr' target='#b25'>(Heggenes, Odland & Bjerketvedt, 2018)</ns0:ref>. Trampling may play an important role in overgrazing-induced variation in the vertical distribution of above-ground biomass. Generally, short and prostrate plants have a higher resistance to trampling than those with taller and erect growth forms <ns0:ref type='bibr' target='#b64'>(Warwick, 1980;</ns0:ref><ns0:ref type='bibr' target='#b57'>Sun & Liddle, 1993;</ns0:ref><ns0:ref type='bibr' target='#b34'>Kobayashi, Ikeda & Hori, 1999)</ns0:ref>. Therefore, the procumbent growth form exhibited by OG may be related to livestock trampling. Aside from this consideration, overgrazing may promote increases in localized drought, which may be another contributor to the observed overgrazing-induced legacy effect on L. chinensis. Drought-adaptive morphological characteristics, such as small stature, have been reported to be advantageous for avoiding and recovering from herbivory <ns0:ref type='bibr' target='#b0'>(Adler et al., 2004;</ns0:ref><ns0:ref type='bibr' target='#b50'>Patty et al., 2010)</ns0:ref>. However, the pathways by which more above-ground biomass is allocated close to the ground as a result of overgrazing require additional research.</ns0:p></ns0:div>
<ns0:div><ns0:head>Biomass allocation above-and below-ground</ns0:head><ns0:p>Contrary to our expectation, OG allocated less biomass belowground under CK than NG; this was attributed to the lower investment in rhizomes by OG. The lower biomass pre-allocation to the rhizome of OG does not benefit the herbivory tolerance and avoidance of L. chinensis <ns0:ref type='bibr' target='#b19'>(Fornoni, 2011;</ns0:ref><ns0:ref type='bibr' target='#b42'>Lurie, Barton & Daehler, 2017)</ns0:ref> and can be considered a cost of herbivory tolerance in the absence of grazing <ns0:ref type='bibr' target='#b37'>(Lennartsson, Ramula & Tuomi, 2018)</ns0:ref>. However, the reduced rhizome allocation of OG may also be the result of the trade-off between colonization and grazing tolerance, as we found a significant negative correlation between rhizome length and ramet number (Fig. <ns0:ref type='figure' target='#fig_3'>S3B</ns0:ref>). Long-dispersing ramets receive less support from the mother plant <ns0:ref type='bibr' target='#b66'>(Zobel, Moora & Herben, 2010)</ns0:ref>; hence, shorter rhizomes which are induced by the lower investment of L. chinensis in this organ could prevent the ramets from overgrazing-induced death. On the other hand, the shorter rhizome implies more ramets, and this could reduce the risk of ramets extinction under overgrazing.</ns0:p><ns0:p>A colonization-competition trade-off <ns0:ref type='bibr' target='#b26'>(Herben et al., 1997;</ns0:ref><ns0:ref type='bibr' target='#b66'>Zobel, Moora & Herben, 2010;</ns0:ref><ns0:ref type='bibr' target='#b24'>Gough et al., 2012)</ns0:ref> may provide another explanation for the larger investment in rhizomes observed in NG. Extreme droughts occur every few years which result in many open patches in both overgrazing and no-grazing plots <ns0:ref type='bibr' target='#b62'>(Wang, Liu & Guo, 2019)</ns0:ref>. In the absence of grazing, plants with the most rapid colonizing ability in the open patches become dominant components of the vegetation <ns0:ref type='bibr' target='#b17'>( Fahrig et al., 1994;</ns0:ref><ns0:ref type='bibr'>Benot et al., 2013)</ns0:ref>. Therefore, intermittent extreme droughts may have contributed to the longer observed rhizomes of NG. Another possible mechanism driving this phenomenon is the fact that NG plants reserved resources for ensuring the next generation of new ramets through a dense canopy and litter layer, which seriously hinders the growth and development of new ramets <ns0:ref type='bibr' target='#b10'>(Craine et al., 2001;</ns0:ref><ns0:ref type='bibr'>Benot et al., 2013)</ns0:ref>.</ns0:p><ns0:p>The immediate responses of above-and below-ground allocation to simulated herbivory showed that L. chinensis increases the allocation of above-ground biomass, and this result supports the functional equilibrium theory <ns0:ref type='bibr' target='#b53'>(Poorter et al., 2012;</ns0:ref><ns0:ref type='bibr' target='#b23'>Gong et al., 2015)</ns0:ref>. According to this theory, more resources stored in the roots and rhizomes are mobilized to stimulate the growth of newly emerged leaves and new ramets <ns0:ref type='bibr' target='#b16'>(Donaghy & Fulkerson, 1998)</ns0:ref>. Numerous studies have shown that this process leads to a reduced allocation of resources belowground for plants that experience leaf damage <ns0:ref type='bibr' target='#b23'>(Gong et al., 2015;</ns0:ref><ns0:ref type='bibr' target='#b4'>Barton, 2016;</ns0:ref><ns0:ref type='bibr' target='#b40'>Liu et al., 2018)</ns0:ref>. However, biomass allocation of OG was less affected by simulated herbivory; this was attributed to the enhanced above-ground spatial avoidance displayed by OG, which reduced the degree of herbivory damage (Fig. <ns0:ref type='figure' target='#fig_4'>S4</ns0:ref>).</ns0:p></ns0:div>
<ns0:div><ns0:head>Ramet number</ns0:head><ns0:p>The stimulation of plant density through increased grazing disturbance has been reported by numerous previous studies <ns0:ref type='bibr' target='#b55'>(Wang et al., 2017)</ns0:ref>. Concurrent to this observation, we found that there were more ramets for OG than NG, and this is consistent with previous studies <ns0:ref type='bibr' target='#b12'>(Detling & Painter, 1983;</ns0:ref><ns0:ref type='bibr' target='#b56'>Rotundo & Aguiar, 2008)</ns0:ref>. Thus, while many scientists have focused on the disruption of plant apical dominance caused by livestock defoliation <ns0:ref type='bibr' target='#b54'>(Rautio et al., 2005;</ns0:ref><ns0:ref type='bibr' target='#b37'>Lennartsson, Ramula & Tuomi, 2018)</ns0:ref>, overgrazing-induced legacy effects on ramet number could partially explain the relatively larger plant density in grazing ecosystems.</ns0:p><ns0:p>Based on the negative relationship observed between the ramet number and ramet vertical height (Fig. <ns0:ref type='figure' target='#fig_3'>S3A</ns0:ref>), we speculate that the observed significant increase in the number of ramets of OG may stem from decreased apical dominance. Weak apical dominance is a coping strategy for avoiding browsing damage caused by livestock. Shorter height in plants results in superior grazing avoidance, which reduces the probability of being defoliated because more biomass is allocated close to the ground. Furthermore, a higher number of ramets could enhance tolerance to grazing by reducing the risk of ramets extinction under overgrazing. Considering this phenomenon, both grazing tolerance and grazing avoidance could thus persist simultaneously, although some studies have suggested that there is a trade-off between these two strategies <ns0:ref type='bibr' target='#b18'>(Fineblum & Rausher, 1995;</ns0:ref><ns0:ref type='bibr' target='#b44'>Mauricio, 2000;</ns0:ref><ns0:ref type='bibr' target='#b36'>Krimmel & Pearse, 2016)</ns0:ref>. On the other hand, the decreased apical dominance induced by overgrazing-induced legacy effects may be attributed to an improvement in light penetration into the environment under grazing, and weaker apical dominance is most likely in the absence of competition for light <ns0:ref type='bibr' target='#b37'>( Lennartsson, Ramula & Tuomi, 2018)</ns0:ref>.</ns0:p><ns0:p>In contrast to our prediction, we did not observe significant differences in L. chinensis ramet numbers between the simulated herbivory treatments in our experiment. Although many studies have indicated that plants can produce more branches when the apically dominant shoot is subjected to physical damage or herbivorous defoliation <ns0:ref type='bibr' target='#b33'>(Klimešová & Klimeš, 2003;</ns0:ref><ns0:ref type='bibr' target='#b54'>Rautio et al., 2005;</ns0:ref><ns0:ref type='bibr' target='#b55'>Wang et al., 2017)</ns0:ref>, many studies have also suggested that herbivory or cutting cannot increase the ramet number <ns0:ref type='bibr' target='#b63'>(Wang et al, 2004;</ns0:ref><ns0:ref type='bibr' target='#b6'>Benot et al., 2009)</ns0:ref> and might even reduce it <ns0:ref type='bibr' target='#b27'>(Hicks & Turkington, 2000)</ns0:ref>. Two preconditions are required for broken apical dominance to induce an increase in ramet numbers: a sufficient number of resources and meristems for regrowth and sufficient apical suppression of basal meristems <ns0:ref type='bibr' target='#b32'>(Klimesova et al., 2014;</ns0:ref><ns0:ref type='bibr' target='#b37'>Lennartsson, Ramula & Tuomi, 2018)</ns0:ref>. In this study, sufficient apical suppression of the basal meristems was achieved during the simulated grazing experiment. Specifically, most of the leaves were removed and only a small section of the stem was left in the 'H4' treatment. Hence, there may have been limited resources and meristems for the regrowth of L. chinensis because of its short growth times from tiller emergence to the first simulated grazing treatment and from the last treatment to harvest.</ns0:p></ns0:div>
<ns0:div><ns0:head>CONCLUSIONS</ns0:head><ns0:p>Consistent with studies on contemporary evolution and stress memory <ns0:ref type='bibr' target='#b12'>( Detling & Painter, 1983;</ns0:ref><ns0:ref type='bibr' target='#b47'>Oesterheld & McNaughton, 1988;</ns0:ref><ns0:ref type='bibr' target='#b2'>Agrawal et al., 2012;</ns0:ref><ns0:ref type='bibr' target='#b67'>Züst et al., 2012;</ns0:ref><ns0:ref type='bibr' target='#b15'>Didiano et al., 2014)</ns0:ref>, our study showed that OG exhibited higher adaptation to simulated grazing in terms of the growth, cloning and colonizing ability than NG. This stronger adaptation was attributed to enhanced above-ground spatial avoidance. Contrary to our prediction, OG pre-allocated less biomass to the rhizome, which does not seem to promote herbivory tolerance and avoidance in L. chinensis; however, this also may reflect a tolerance strategy via shorter rhizomes and more ramets. Here, we quantitatively studied the grazing-induced spatial avoidance of plants for the first time and found that enhanced above-ground spatial avoidance was induced by a larger leaf angle and shorter height. However, because of the pseudo-replication and because the sample areas only consisted of two adjacent pastures in this study, the findings from our research may be limited to our study site. Manuscript to be reviewed Manuscript to be reviewed </ns0:p></ns0:div><ns0:figure xml:id='fig_0'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:05:49307:3:0:NEW 8 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_1'><ns0:head>Figure 1</ns0:head><ns0:label>1</ns0:label><ns0:figDesc>Figure 1</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_2'><ns0:head>Figure 2</ns0:head><ns0:label>2</ns0:label><ns0:figDesc>Figure 2</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_3'><ns0:head>Figure 3 The</ns0:head><ns0:label>3</ns0:label><ns0:figDesc>Figure 3</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_4'><ns0:head>Figure 4</ns0:head><ns0:label>4</ns0:label><ns0:figDesc>Figure 4</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_5'><ns0:head>Figure 5</ns0:head><ns0:label>5</ns0:label><ns0:figDesc>Figure 5</ns0:figDesc></ns0:figure>
<ns0:note place='foot'>PeerJ reviewing PDF | (2020:05:49307:3:0:NEW 8 Sep 2020)</ns0:note>
</ns0:body>
" | "Pratacultural College
Gansu Agricultural University
Lanzhou, Gansu Province, PR China
August 8, 2020
Dear Mr. Curtis:
Thanks for your valuable scholarly comments and suggestions on our manuscript entitled “Overgrazing-induced legacy effects may permit Leymus chinensis to cope with herbivory” (#49307). We have read the comments carefully and revised the manuscript to address your raised concerns. The major changes to the manuscript are described in detail in the attached point-to-point response below. The comments are listed in italicized font and our responses are provided without italics.
Sincerely,
Fenghui Guo
The revised text is generally clear and assertions are supported by the data. I identified a few minor edits as follows:
Issue1: L 83 delete “highly” (relished already implies this so “highly” is redundant)
Response: Thank you for this comment. We have revised it. (Line 83)
Issue2: L 129 “of the buds” to “containing buds”
Response: We have revised it. (Line 129)
Issue3: L 165 “ramet” to “ramets”
Response: We have revised it. (Line 165)
Issue4: L 198 “homogeneous” to “homogeneous in variance”
Response: We have revised it. (Line 198)
Issue5: L 208 can you add a sentence “H4 was chosen for the PI estimate because H4 had a greater effect on L. chinensis than H8.”
Response: We have revised it. (Line 208)
Issue6: L 281 “Besides, previous work has shown” to “Previous work has also shown”
Response: We have revised it. (Line 282)
Issue7: L 282 “over-grazing fields are more prostrate and shorter” to “over-grazed fields are characterized by more prostrate and shorter growth”
Response: We have revised it. (Line 283)
Issue8: L 283 “with that in fields” to “with populations in fields”
Response: We have revised it. (Line 284)
Issue9: L 319 “pre-allocating biomass” to “biomass pre-allocation”
Response: We have revised it. (Line 321)
" | Here is a paper. Please give your review comments after reading it. |
9,754 | "<ns0:abstract xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:p>The widespread occurrence of pathogenic bacteria resistant to last-line antibiotics has resulted in significant challenges in human and veterinary medicine. There is an urgent need for new antimicrobial agents that can be used to control these life threating pathogens. We report the identification of antimicrobial activities, against a broad range of bacterial pathogens, from a collection of marine-derived spore-forming bacteria. Although marine environments have been previously investigated as sources of novel antibiotics, studies on such environments are still limited and there remain opportunities for further discoveries and this study has used resources derived from an under-exploited region, the Vietnam Sea. Antimicrobial activity was assessed against a panel of Gram-positive and Gram-negative bacteria, including several multi-drug resistant pathogens. From a total of 489 isolates, 16.4% had antimicrobial activity. Of 23 shortlisted isolates with the greatest antimicrobial activity, 22 were Bacillus spp. isolates and one was a Paenibacillus polymyxa isolate. Most of the antimicrobial compounds were sensitive to proteases, indicating that they were proteins rather than secondary metabolites. The study demonstrated that marine bacteria derived from the Vietnam Sea represent a rich resource, producing antimicrobial compounds with activity against a broad range of clinically relevant bacterial pathogens, including important antibiotic resistant pathogens. Several isolates were identified that have particularly broad range activities and produce antimicrobial compounds that may have value for future drug development.</ns0:p></ns0:div>
</ns0:abstract>
<ns0:body xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:head>Introduction</ns0:head><ns0:p>The emergence and spread of bacterial pathogens that are resistant to last-line antibiotics, for example carbapenem resistant Gram-negative pathogens, methicillin resistant Staphylococcus aureus (MRSA), and vancomycin resistant Enterococci (VRE), is of great concern in both human and veterinary medicine <ns0:ref type='bibr' target='#b3'>(Datta & Huang, 2008;</ns0:ref><ns0:ref type='bibr' target='#b14'>Loomba, Taneja & Mishra, 2010;</ns0:ref><ns0:ref type='bibr' target='#b27'>Raghunath, 2010;</ns0:ref><ns0:ref type='bibr' target='#b22'>O'Driscoll & Crank, 2015;</ns0:ref><ns0:ref type='bibr' target='#b19'>Meletis, 2016;</ns0:ref><ns0:ref type='bibr' target='#b38'>Zaman et al., 2017)</ns0:ref>. Antibiotic resistance genes are frequently located on mobile elements such as conjugative plasmids and transposons which facilitate horizontal and vertical transmission, leading to increasing numbers of multi-resistant bacteria worldwide <ns0:ref type='bibr' target='#b6'>(Devaud, Kayser & Bächi, 1982;</ns0:ref><ns0:ref type='bibr' target='#b35'>Turner et al., 2014)</ns0:ref>. This high incidence of antibiotic resistant bacteria has serious implications for pathogen control and there is an urgent need for alternatives to the currently available antibiotics to re-control these life-threating antibiotic-resistant pathogens. The discovery of novel antibiotics from terrestrial environments over the last few decades has been challenging, due to an exhaustion of traditional antibiotic sources <ns0:ref type='bibr' target='#b38'>(Zaman et al., 2017)</ns0:ref>. According to the FDA, approval of new medically important antibiotics had decreased by 56% over the last few decades <ns0:ref type='bibr' target='#b30'>(Spellberg et al., 2004)</ns0:ref> and there has been no evidence for an increase in discovery rate since that study.</ns0:p><ns0:p>The marine environment has been identified as a promising alternative source for antibiotic discovery, due to the apparently high abundance of antibiotics produced by various members of diverse marine microbial communities <ns0:ref type='bibr' target='#b18'>(Mayer et al., 2011;</ns0:ref><ns0:ref type='bibr' target='#b16'>Malve, 2016)</ns0:ref>. It has been hypothesized that marine bacteria are under unusually rigorous selection pressures because of the environmental conditions with which they must contend. They are typically exposed to low levels of nutrition and rapid changes of nutrition and physical conditions due to wave and tidal action. These harsh chemo-physical conditions have selected bacteria that deploy various mechanisms to out-compete other bacteria. Diversification of antimicrobial production is one such adaptation that can be harnessed for occupying and defending an ecological niche <ns0:ref type='bibr' target='#b10'>(Jensen & Fenical, 1996;</ns0:ref><ns0:ref type='bibr' target='#b36'>Valentine, 2007;</ns0:ref><ns0:ref type='bibr' target='#b5'>Desriac et al., 2010)</ns0:ref>. It is hypothesized that marine environments may harbor novel antimicrobial producing species, which could be effective against a range of bacteria, including antibiotic resistant pathogenic bacteria. Among the diverse marine bacterial communities, members of the Bacillus genus, characterized as spore formers, have been shown to produce an array of structurally diverse antibiotics, including ribosomally synthesized peptides (bacteriocins), and non-ribosomal secondary metabolites such as polyketide, lipopeptide and bacilysin <ns0:ref type='bibr' target='#b32'>(Sumi et al., 2014)</ns0:ref>. Bacteriocins are of particular interest. These small cationic peptides commonly have a narrow spectrum of activity, affecting a limited range of bacteria, and a low resistance rate, making them attractive antimicrobials for application against bacteria that have acquired resistance to the current range of available antibiotics <ns0:ref type='bibr' target='#b1'>(Cotter, Ross & Hill, 2013;</ns0:ref><ns0:ref type='bibr' target='#b17'>Mathur et al., 2017)</ns0:ref>. Many terrestrially derived bacteriocins have been characterized, but little is known of marine-derived bacteriocins. Bacteria that produce antimicrobial compounds have the potential to be used directly, for example as probiotics, to combat pathogenic bacteria. Spore-forming bacteria have particular advantages for such applications as the spores are robust and resistant to processing and storage conditions. Hence, spore-forming antimicrobial producing bacteria such as Bacillus and Paenibacillus species can be more easily incorporated into processed food products than more vulnerable species such as Lactobacillus and Lactococcus <ns0:ref type='bibr' target='#b7'>(Elshaghabee et al., 2017)</ns0:ref>. Vietnam has more than 3,400 km of coastline, incorporating a variety of tropical marine ecosystems with abundant marine species and little previous study of bacteria derived from its marine environments. These characteristics make the Vietnam Sea a promising source to explore for novel antimicrobials. The objective of this study was to test the hypothesis that bacteria that produce potentially useful antimicrobials could be found in the Vietnam Sea. A diverse collection of spore-forming bacteria was assembled and evaluated for antimicrobial molecules with activity against a range of important pathogens, including multi-antibiotic resistant pathogens.</ns0:p></ns0:div>
<ns0:div><ns0:head>Materials & Methods</ns0:head></ns0:div>
<ns0:div><ns0:head>Collection of marine samples and bacterial isolation</ns0:head><ns0:p>Fifty marine samples, from sponges, seaweeds, sediments, and seawaters, were collected by scuba diving at Nha Trang <ns0:ref type='bibr'>Bay, around Hon Mieu Island (12.191837, 109.235086), and around Hon Mot Island (12.173368, 109.271591) at depth of 5-10 meters, and Hon Rua Island (12.289501, 109.242413</ns0:ref>) at depth of 1-3 meters. These locations were chosen because of their known high diversity of marine habitats. The collection of environmental samples was approved by the Nha Trang Institute of Technology Research and Application. Sponge and seaweed samples were washed three times with sterile sea water (SSW) to remove loosely attached external microbes. Ten grams of each sponge and seaweed sample were homogenized with a sterile mortar and pestle in 90 mL of sterile sea water to release microbes into the seawater. Sediments were air dried and 10 g vortexed in 90 mL of sterile seawater to detach microbes. Fifty mL of sea waters were centrifuged at 2,500g for 10 minutes and the 1 mL at the bottom of the tube was used to resuspend pelleted material. Aliquots of all samples were heated at 80°C for 20 minutes to kill non-spore-forming bacteria, followed by serial dilution for plating and isolation. From each dilution 100 µL was plated onto laboratory-prepared marine agar (LPMA) (2.5 g/L yeast extract, 5.0 g/L peptone, 1.0 g/L dextrose, 0.2 g/L K 2 HPO 4 , 0.05g/L MgSO 4 .7H 2 O, 750 mL/L aged sea water, 250 mL/L tap water, pH= 7.5). The LPMA was Youshimizu and Kimura medium as modified by Mikhailov et al. <ns0:ref type='bibr' target='#b20'>(Mikhailov, Romanenko & Ivanova, 2006)</ns0:ref>. Aged sea water was prepared by storing fresh sea water in the dark for 2-4 weeks to stabilize or neutralize the heavy metals or toxic compounds which may affect bacterial recovery. The inoculated plates were incubated at 30°C for up to 4 days. Colonies were re-streaked until only colonies with similar morphologies were observed on the plates. Pure colonies were scrapped off plates to mix in laboratory-prepared marine broth (LPMB) supplemented with 20% glycerol and stored at -80°C.</ns0:p></ns0:div>
<ns0:div><ns0:head>Primary screening for antimicrobial activity by cross-streak assay</ns0:head><ns0:p>Each bacterial isolate was streaked vertically onto an LPMA plate and incubated overnight at 30 o C. The six indicator strains used in the initial screening were, S. aureus, B. cereus, C. albicans, E. coli, P. aeruginosa, and a methicillin resistant S. aureus. Each indicator strain was streaked, horizontally, from the edge of the plate to the pre-grown isolate streak. The cross-streak assay plates were then incubated in the growth conditions appropriate for each indicator strain, as detailed in Table <ns0:ref type='table' target='#tab_0'>1</ns0:ref>. The gap between indicator strain and the test isolates' growth indicated the presence or absence of antimicrobial activity. After inspection of the strength and spectrum of antimicrobial activity, a short-list of isolates with the strongest antimicrobial activity was chosen for more detailed analysis. Their antagonistic activities against 14 pathogens were evaluated. The cross-streak assay was also used to detect antimicrobial activities produced by the 23 shortlisted isolates against the other members of the group. Each marine isolate was streaked down the middle of LPMA plates, incubated overnight at 30 o C to allow growth and production of antimicrobial compounds, and then the other 22 isolates were streaked from the edge of the plate to the central streak. The plate was incubated overnight at 37 o C and antimicrobial activity was determined based on the size of the clear zone between test and indicator bacterial streaks.</ns0:p></ns0:div>
<ns0:div><ns0:head>Well-diffusion assay</ns0:head><ns0:p>Well-diffusion assays were used to determine if antimicrobial activity was secreted into liquid culture supernatant. Muller Hilton agar plates were swabbed with a suspension of indicator bacteria (OD 600 ~0.08-0.1). Wells (6mm diameters) were punched from the agar using a sterilized cork cutter and then 50 µL of cell free supernatant (CFS) from test cultures grown for 24 hours was added to a well. The CFS in the wells was air dried and plates were incubated at the optimal conditions for growth of the indicator strain (Table <ns0:ref type='table' target='#tab_0'>1</ns0:ref>). The inhibitory effects of antimicrobial within the CFS were observed by appearance of a zone of clearing of the indicator bacteria around the well.</ns0:p><ns0:p>16S rRNA gene amplification, sequencing, and phylogenetic analysis Bacteria were sub-cultured into 5 mL of LB from a single colony. Total genomic DNA (gDNA) was extracted from the overnight culture using a guanidine thiocyanate method, as previously describes <ns0:ref type='bibr' target='#b24'>(Pitcher, Saunders & Owen, 1989)</ns0:ref>, and then used as template to amplify the 16S rRNA gene sequences. PCR was conducted using primers with the sequences (5'-3') GGCGTGCCTAATACATGCAA and TACAAGGCCCGGGAACGT. The primers were designed, based on the alignment of the 16S rDNA sequences of Bacillus isolates. PCR condition comprised initial denaturation at 98°C for 30 seconds, 30 cycles of 98°C for 5 seconds, 56°C for 10 seconds and 72°C for 20 seconds, extension at 72°C for 2 minutes; and 4°C for 10 minutes. PCR products were checked by electrophoresis in 1% agarose gel, subsequently purified by QIAquick PCR Purification Kit (Qiagen), and then Sanger-sequenced (Micromon, Monash University, Australia). The 16S rRNA gene sequences are deposited under NCBI GenBank accession numbers MT758446-MT758468. The raw reads were trimmed of unclear nucleotides at both 5' and 3' terminal ends, and subsequently blastn was used to search for homologies in the Bacterial 16S rDNA Database (https://blast.ncbi.nlm.nih.gov/Blast.cgi). The highly homologous 16S rDNA sequences were download for phylogenetic tree construction. All the sequences were aligned using ClustalW <ns0:ref type='bibr' target='#b33'>(Thompson, Higgins & Gibson, 1994)</ns0:ref>, and the phylogenetic tree was subsequently constructed in MEGA7 using the neighbor joining method with bootstrap tests performed 1000 times and pairwise detection <ns0:ref type='bibr' target='#b11'>(Kumar, Stecher & Tamura, 2016)</ns0:ref>.</ns0:p></ns0:div>
<ns0:div><ns0:head>Sensitivity of antimicrobial activities to enzyme and heat treatments</ns0:head><ns0:p>Enzymatic treatments of CFSs were conducted for 3 hours at 37°C with pronase-E from Streptomyces griseus; proteinase K; trypsin; and lipase, at final concentrations of 2 mg/mL. All enzymes were purchased from Sigma Aldrich. Heat stability of antimicrobials was determined by incubation of CFSs at 60 o C, 80 o C, and 100 o C for 30 minutes, 60 minutes, and 3 hours. The antimicrobial activities of the treated CFS preparation were evaluated by well-diffusion assay, against Clostridium perfringens.</ns0:p></ns0:div>
<ns0:div><ns0:head>Growth properties, antibiotic susceptibility testing, and enzyme production of isolates</ns0:head><ns0:p>The ability of the short-listed isolates to grow on different media was evaluated by spotting 5µL of bacterial cultures onto several different media, including low nutrition media such as marine agar (BD Difco 2216) and LPMA, and rich nutritious media such as LB agar and Muller Hilton agar, and incubated at 30 o C overnight. Aliquots of cultures were spotted on LB agar plates, and incubated under microaerophilic, aerobic, and anaerobic conditions at 30°C, and aerobically incubated at 40°C and 50°C. To measure pH tolerance, bacterial cultures were spotted onto LB plates in which the media had been adjusted to pH 5.0, 6.0, 7.0, 8.0, and 9.0. Sodium chloride tolerance was determined on LB plates supplemented with 0%, 1%, 2%, 4%, 6%, 8%, 10%, 12% and 15% (w⁄v) NaCl; while bile salt tolerance was carried out on the LB plates supplemented bile salt at final concentration of 0.1M, 0.2M, 0.3M, 0.4M, 0.5M, 0.6M, 0.7M respectively (Bile Salts Mixture No. 3; Neogen Corporation). For antibiotic susceptibility testing, LB agar plates were prepared supplemented with tetracycline, ampicillin, nalidixic acid, and kanamycin at final concentration of 50 µg/mL. The production of proteases and cellulases, amylase was evaluated respectively by spotting of 5 µL of bacterial cultures onto skim milk agar; carbon deficient media (CDM) supplemented with 1% carboxymethylcellulose (CMC), and CDM supplemented with 1% soluble starch. CDM contained 0.1 g/L yeast extract; 0.5 g/L peptone; 16.0 g/L agar. Plates were incubated at 30 o C overnight, followed by flushing the CMC plates and starch agar plates with Gram's iodine solution (Sigma Aldrich) for 1 minute. Positive reactions, indicating enzymatic activity, were noted via halo zones around the bacteria spots.</ns0:p></ns0:div>
<ns0:div><ns0:head>Statistical Analysis</ns0:head><ns0:p>A two-tailed t-test was used to assess whether there were any statistically significant differences in the rate of isolation of antimicrobial producing isolates from the different marine sources. The statistical analysis was computed in Microsoft Excel with a 5% level of probability used to indicate significance.</ns0:p></ns0:div>
<ns0:div><ns0:head>Results</ns0:head><ns0:p>Antimicrobial activity of thermally resistant, marine spore-forming bacteria A total of 389 heat resistant, spore-forming, bacterial isolates were cultured from 50 marine samples (Table <ns0:ref type='table' target='#tab_1'>2</ns0:ref>). They demonstrated a range of colony morphologies (Figure <ns0:ref type='figure' target='#fig_0'>1</ns0:ref>). Primary screening against six indicator strains, using the cross-streak assay, identified 64 isolates (16.4%) with antimicrobial activity (Figure <ns0:ref type='figure' target='#fig_1'>2A, 2C</ns0:ref>). The proportion of isolates that exhibited activity against Gram-positive indicator strains, Bacillus cereus (93.7%), Staphylococcus aureus (84.3%), Streptococcus faecalis (87.5%), was higher than the proportion that exhibited activity against the Gram-negative indicators, Escherichia coli (50%) and Pseudomonas aeruginosa (4.6%), and the yeast, Candida albicans (21.4%). A two-tailed t-test showed that the there was no statistically significant difference in the rate of isolation of antimicrobial producing bacteria from the different types of marine samples. The p values for the pairwise comparisons amongst all the sample types ranged from 0.37 to 0.77.</ns0:p><ns0:p>The spectra of antimicrobial activities, exhibited by a select group of the 23 most potent isolates from the primary screen, were determined against an expanded panel of 14 indicator strains, including important multidrug resistant pathogens. The analyses were performed using two assays: a cross-streak assay and a well-diffusion assay. There was considerable variation in the strength and spectra of antimicrobial activity across the 23 short-listed isolates. The pathogenic indicator strains most commonly affected by the antimicrobial compounds expressed by the screened isolates were the Gram-positive species (Table <ns0:ref type='table' target='#tab_3'>3; Figure 3</ns0:ref>). Of the bacterial indicators, Clostridium perfringens, B. cereus and S. aureus were inhibited by 83% (19/23) of the test isolates, whereas the proportion with activity against the Gram-negative indicators; Campylobacter jejuni, 70% (16/23), Campylobacter coli, 61% (14/23), was lower. Two antibiotic resistant Gram-positive pathogens; MRSA and VRE were inhibited by respectively 83% (19/23) and 57% (13/23) of isolates, while an antibiotic resistant Gram-negative pathogen; multidrug resistant Klebsiella pneumonia (MRKP), was inhibited by only one isolate, P. polymyxa #23. The Gram-negative bacteria P. aeruginosa was also inhibited by P. polymyxa #23. The growth of foodborne pathogens; Listeria monocytogenes, E. coli and Salmonella Enteriditis were depressed by, respectively, 78%, 44%, and 57% of the isolates. None of the selected isolates had inhibitory activity against C. albicans. The cross-streak assay was found to be more sensitive than the well-diffusion assay. In most cases antimicrobial activity was detected in both assays but occasionally the activity was less or absent in the well-diffusion assay. This difference was particularly apparent when S. aureus, MRSA and most of Gram-negative pathogens, C. jejuni, C. coli, E. coli and S. Enteriditis, were used as the indicator strains.</ns0:p></ns0:div>
<ns0:div><ns0:head>Taxonomic analysis of antimicrobial isolates</ns0:head><ns0:p>Of the 23 of the isolates selected as expressing the most antimicrobial activity, there were 22 Bacillus species isolates and 1 Paenibacillus isolate. Their 16S rRNA gene sequences shared 99% -100% identity with, mostly, terrestrially derived species (Table <ns0:ref type='table' target='#tab_5'>4</ns0:ref>). The phylogenetic tree demonstrated the relationship amongst the marine isolates and to terrestrial isolates (Figure <ns0:ref type='figure' target='#fig_3'>4</ns0:ref>). The most commonly identified species amongst the short-listed isolates were B. subtilis (10 isolates) and other members of B. subtilis group such as B. amyloliquefaciens (5 isolates), B. licheniformis (1 isolate), and B. safensis (1 isolate). In addition, members of other Bacillus groups were also identified including B. pacificus (2 isolates), belonging to B. cereus group; B. halotolerans (3 isolates), and P. polymyxa (1 isolate)</ns0:p></ns0:div>
<ns0:div><ns0:head>Many of the antimicrobials are proteinaceous compounds</ns0:head><ns0:p>It has previously been shown that protein and non-protein antimicrobial compounds are produced by some terrestrial Bacillus isolates. The proteinaceous nature of some of the antimicrobial activities identified in the marine Bacillus isolate collection was demonstrated by their susceptibility to protease action. For these assays C. perfringens was selected as the indicator strain because amongst the indictor strains tested (Table <ns0:ref type='table' target='#tab_3'>3</ns0:ref>) it displayed sensitivity to the highest percentage of the Bacillus produced antimicrobial compounds. The protease sensitivity of the anti-C. perfringens activity of the isolates was determined by digesting cell culture supernatants used in the well-diffusion assay with proteases. In 16 of the 19 isolates tested the antimicrobial activities against C. perfringens were reduced or lost after treatment with at least one proteolytic enzyme (Table <ns0:ref type='table'>5</ns0:ref>). The pronase-E enzyme completely eliminated anti-C. perfringens activity from 10 isolates, while proteinase-K removed the activity from 4 isolates. Some antimicrobials (produced by isolates #06, #11, and #21) were affected by both proteinases and a lipase. Other isolates showed reduced, but not eliminated, antimicrobial activity following protease treatment (#01, #08, #20, #21). The antimicrobial activities produced by isolates #5, #18, #19 were not affected by any of the enzymes used. Heat treatment at 60°C for 30 minutes abolished the activity of 12/19 antimicrobials while 7/19 retained activity. Of the isolates in which antimicrobial activity was not affected by proteases, isolates #05 and #19 were also resistant to heat treatment whereas anti-C. perfringens activity was abolished when #18 supernatant was heated.</ns0:p></ns0:div>
<ns0:div><ns0:head>Antimicrobial activity between the short-listed isolates</ns0:head><ns0:p>It is hypothesised that bacteria produce antimicrobial compounds to compete within an ecological niche. We therefore investigated whether this group of isolates exhibited any antimicrobial activities against each other. Growth inhibition was detected, using the cross-streak assay, between strains of the same species for isolates B. amyloliquefaciens #06, #08, and B. subtilis #05. Crossspecies antimicrobial activities were detected for many of the isolates, including, B. amyloliquefaciens #06, #08, #11, #13, B. licheniformis #03; B. safensis #10; B. halotolerans #01, and P polymyxa #23. Particularly, P. polymyxa #23 depressed growth of all the marine Bacillus assayed (Figure <ns0:ref type='figure' target='#fig_1'>2B</ns0:ref>, Table <ns0:ref type='table' target='#tab_6'>6</ns0:ref>)</ns0:p></ns0:div>
<ns0:div><ns0:head>Growth characteristics of antimicrobial-producing isolates</ns0:head><ns0:p>The basic growth parameters of the 23 isolates were qualitatively evaluated (Table <ns0:ref type='table' target='#tab_9'>7</ns0:ref>). All isolates could grow on various kinds of media. Growth was better in rich nutritious medias such as Luria-Bertani (LB), brain heart infusion (BHI), Müller Hinton agar (MH), and blood agar (BA), than growth in less nutritious media such as lab-prepared marine agar (LPMA) and marine broth (MB). All 23 isolates grew vigorously under both aerobic and microaerophilic conditions but not in anaerobic condition, under incubations temperatures of 30°C, 37 o C, 40°C, 50°C, and under pHs ranging from 6.0 to 9.0. Some isolates had reduced growth at pH5.0. All isolates could grow in media supplemented with 5% NaCl, and all but isolate #23 grew in 7% NaCl. No isolate could tolerate 12% NaCl. The marine environment, from which the isolates were derived, typically has a salt content of 3.5%. No isolates grew in LB media supplemented with bile salts, even at lowest concentration tested (0.1M). All isolates were sensitive to 3 antibiotics; nalidixic acid (50 µg/mL), kanamycin (50 µg/mL) and tetracycline (50 µg/mL), but 14 of the 23 isolates were resistant to ampicillin (50 µg/mL). All the isolates had detectable levels of cellulolytic, proteolytic or amylolytic activity (Figure <ns0:ref type='figure' target='#fig_1'>2D-F</ns0:ref>).</ns0:p></ns0:div>
<ns0:div><ns0:head>Discussion</ns0:head><ns0:p>There is growing interest in the marine environment as a potential source of bacteria that produce novel antimicrobial compounds, particularly bacteriocins. These antimicrobial peptides have been gaining interest as potential drug candidates for clinical treatment of antibiotic resistant pathogens <ns0:ref type='bibr' target='#b1'>(Cotter, Ross & Hill, 2013)</ns0:ref>. In this study, spore-forming bacteria isolated from the coastal marine environment of Nha Trang (Vietnam Sea) were screened to identify Bacillus isolates that produce antimicrobial compounds. Members of the Bacillus genus were targeted because they are wellrecognized as producers of structurally diverse bacteriocins. Of the various types of marine samples collected, the antimicrobial producing isolates were most frequently recovered from sponges, followed by sediments, seaweeds, and sea-water samples. Marine sponges have multiporous structures that may trap and maintain high bacterial densities, leading to higher recovery rates of antimicrobial producers. The recovery rate in marine sponges, at 7.6%, was at the lower end of the range reported in previous studies (5.5% to 50.0%) <ns0:ref type='bibr' target='#b12'>(Laport & Muricy, 2008)</ns0:ref>. A lower recovery rate of antimicrobial producing isolates was also seen from seaweeds, at 3.6%, whereas previous studies had identified them from seaweeds at 11.0%-16.0% <ns0:ref type='bibr' target='#b13'>(Lemos, Toranzo & Barja, 1985;</ns0:ref><ns0:ref type='bibr' target='#b23'>Penesyan et al., 2009)</ns0:ref>. The differences between this study and previous studies could be due to geographical differences, marine conditions, or heat treatment to select the spore-formers that may eliminate the metabolically active vegetative cells. The methods used to detect antimicrobial activity had different levels of sensitivity. The welldiffusion assay was less sensitive than the cross-streak assay. This effect was most obvious with the failure to detect activity against Gram-negative bacteria, S. aureus and MRSA, when the latter assay was used. This may occur because of the production of multiple antimicrobial compounds, with variable relative expression levels in liquid and solid media-based cultivation. Many studies have reported the influence of various factors on bacteriocin production in liquid culture such as, type of culture media; pH, temperature, growth phase, and quorum sensing regulation <ns0:ref type='bibr' target='#b8'>(Gutowski-Eckel et al., 1994;</ns0:ref><ns0:ref type='bibr' target='#b28'>Shanker & Federle, 2017;</ns0:ref><ns0:ref type='bibr' target='#b37'>Yang et al., 2018)</ns0:ref>. The most commonly identified species amongst the 23 shortlisted marine isolates that exhibited the strongest antimicrobial activities, were members of the Bacillus subtilis group. These species have been widely reported in both marine environments and terrestrial environments and are able to tolerate the broad environmental conditions (nutrients, pHs, chemo-physical conditions) that are typically found in marine environments. For example, a study by Ivanova et al., reported that 55.0% (11/ 20) of endospore-forming bacteria isolated from different areas of the Pacific Ocean were B. subtilis species <ns0:ref type='bibr' target='#b9'>(Ivanova et al., 1999)</ns0:ref>. B. subtilis, B. amyloliquefaciens, B. pumilus and P. polymyxa were all identified amongst aerobic spore-forming isolates from marine sources from the Gulf of Mexico, or isolated from seaweed samples collected from the Irish Sea <ns0:ref type='bibr' target='#b29'>(Siefert et al., 2000;</ns0:ref><ns0:ref type='bibr' target='#b15'>Luz Prieto et al., 2012)</ns0:ref>. Other typical marine species, for example, B. aquimaris, B. algicola and B. hwajinpoensis may require additional special nutrients or salts to recover or could have been eliminated during screening and shortlisting based on the strength of antimicrobial activity.</ns0:p><ns0:p>The marine Bacillus/Paenibacillus isolates that we have characterized had broad antimicrobial activity against a range of human, veterinary and food borne pathogens, including three antibiotic resistant pathogens (MRKP, MRSA and VRE). B. amyloliquefaciens #06, #08, #11, #13, B. halotolerans #01, #19, B. licheniformis #03, B. safensis #10, P. polymyxa #23 had broad antimicrobial activity against both Gram-positive and Gram-negative pathogen indicator strains and against other marine Bacillus of different species. This indicated production of either multiple antimicrobial compounds by a single Bacillus, or a broad-spectrum antimicrobial compound. Members of the Bacillus genus are known to produce various types of antimicrobial compounds including polyketides, lipopeptides, bacteriocins, bacilysin, and volatile compounds <ns0:ref type='bibr' target='#b21'>(Mondol, Shin & Islam, 2013)</ns0:ref>.The synergistic effects of these antimicrobials could result in a broad spectrum of antimicrobial activity, such as noted for a number of the isolates in this study. Also, production of broad-spectrum bacteriocin was recently reported for a marine Bacillus; sonorensin, identified from a marine B. sonorensis isolate, exhibited broad-spectrum antibacterial activity towards both Gram-positive and Gram-negative bacteria <ns0:ref type='bibr' target='#b0'>(Chopra et al., 2014)</ns0:ref>. Amongst Grampositive bacteria, such as Bacillus spp., the expression of bacteriocins and other antimicrobial compounds with activity against other Gram-positive bacteria is widespread and extensively studied. However, the production of compounds with activity against Gram-negative bacteria is less common, therefore, the isolates that have antimicrobial activity against Gram-negative bacteria are of particular interest. Rarely observed antimicrobial activities such as that observed against Campylobacter, P. aeruginosa, and even multidrug-resistant K. pneumonia, represent potentially novel compounds that, in future work, should be purified and structurally characterized. The P. polymyxa #23 isolate had the strongest and broadest activity and it is likely that the activity against Gram-negative bacteria results from the expression of polymyxin, which has long been known to have such activity <ns0:ref type='bibr' target='#b31'>(Stansly & Schlosser, 1947;</ns0:ref><ns0:ref type='bibr' target='#b26'>Poirel, Jayol & Nordmann, 2017)</ns0:ref>. The other isolates with activity against some of the Gram-negative bacteria tested, for example B. amyloliquefaciens #11, which had significant activity against both Campylobacter species but lesser activity against E. coli and Salmonella, appears to indicate a spectrum of activity that has not previously been reported and so the compound responsible may be novel and hence warrants further investigation. The finding that many of the compounds that had activity against C. perfringens were inactivated by proteases, indicated that the antimicrobial compounds produced probably included bacteriocins. Interestingly, the production of large quantities of these antimicrobial compounds for drug development in the future could likely be achieved as it was demonstrated that most of the isolates were well adapted to a broad range of growth conditions (variations in nutrients, pH, salt concentration, and temperature). These characteristics of the isolates represent advantages that could facilitate manufacturing processes, product storage, and the potential harnessing of these isolates for in vivo use in animal or food applications. These marine derived Bacillus/Paenibacillus isolates were also shown to have proteolytic, cellulolytic and amylolytic activity and hence may represent a promising source of important industrial enzymes such as proteases, cellulases, and amylases. Marine derived enzymes have been noted to have significant advantages in manufacturing because they commonly have high adaptability to high-salt concentration, and fluctuating temperature, pH, organic solvents, and ions <ns0:ref type='bibr' target='#b4'>(Debashish et al., 2005)</ns0:ref> In conclusion, the bacteria in this collection of marine Bacillus isolates express a range of antimicrobial activities, some of which may represent novel compounds that warrant further study.</ns0:p></ns0:div>
<ns0:div><ns0:head>Conclusions</ns0:head><ns0:p>It was hypothesized that the marine environment, particularly understudied regions abundant in varied marine habitats, such as the Vietnam Sea, would provide a rich source of bacteria that produce antimicrobial compounds. A survey of heat-resistant spore-forming bacteria found that 16.4% of isolates produced detectable levels of antimicrobial activity. Bacterial isolates were identified that had broad spectra of activity against both Gram-positive and Gram-negative pathogenic bacteria. Further analysis of a select group of isolates with the broadest activity profile showed that most of the antimicrobial compounds were sensitive to proteases, indicating that they were proteins rather than secondary metabolites. The study demonstrated that marine bacteria derived from the Vietnam Sea represent an interesting resource, producing antimicrobial compounds with activity against a range of clinically relevant bacterial pathogens, including important antibiotic resistant pathogens. Further biochemical characterization now needs to be undertaken to characterize the antimicrobial compounds, especially to define those that are novel. The neighbor-joining phylogenetic tree was constructed using the maximum composite likelihood method, bootstrap method of 1000 replication and pairwise deletion by MEGA 7. In this tree, 23 marine isolates were displayed by number value, while all reference strains bacteria were closely related bacteria identified and downloaded from NCBI. Manuscript to be reviewed Antimicrobial producing bacteria identified from marine samples</ns0:p><ns0:note type='other'>Figure Legends</ns0:note><ns0:p>PeerJ reviewing PDF | (2020:07:51002:1:0:NEW 31 Aug 2020)</ns0:p><ns0:p>Manuscript to be reviewed Manuscript to be reviewed Antimicrobial activity of short-listed isolates against 14 indicators strains.</ns0:p><ns0:p>The antimicrobial activities were evaluated by well diffusion assay (WD) and cross streak assay (CS).</ns0:p><ns0:p>+: zone of inhibition observed with clear halo of growth inhibition in at least one time point; ++ and +++: increased activity as assessed visually by the increased diameter of the inhibition zone; -: no inhibition. The percentage values were calculated based on the ratio between number of isolates with antimicrobial activity against the indicator strain and the total 23 isolates tested.</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:07:51002:1:0:NEW 31 Aug 2020)</ns0:p><ns0:p>Manuscript to be reviewed 1 Table <ns0:ref type='table' target='#tab_3'>3</ns0:ref>. Antimicrobial activity of short-listed isolates against 14 indicators strains. The antimicrobial activities were evaluated by 2 well diffusion assay (WD) and cross streak assay (CS). Manuscript to be reviewed Underlined enzyme indicated a loss of activity after enzymatic treatment. The values of temperature resistance were recorded after 3 hours of incubation. The numbers in the table represented the mean value (± standard deviation) of killing diameters from the measurement of three replicates.</ns0:p><ns0:formula xml:id='formula_0'>Lactobacillus plantarum ++ + + + + + - - - - ++ ++ - - ++ ++ - - + + ++ ++ - - Candida albicans - - - - - - - - - - - - - - - - - - - - - - - - Escherichia coli + - + - - - - - + - - - - - + + - - - - + + + - Salmonella Enteritidis + - - - - - - - - ++ - + - ++ + - - - - ++ + + - Campylobacter jejuni ++ + ++ + - - - - ++ - ++ ++ + - ++ ++ - - - - +++ ++ - - Campylobacter coli + +++ ++ ++ - - ++ +++ + - + +++ + - + +++ - - - - ++ +++ ++ ++ Pseudomonas aeruginosa - - - - - - - - + - - - - - - - - - - - - - - - MRSA +++ ++ ++ - + - - - + + +++ - - - +++ - - - ++ ++ ++ - ++ - VRE ++ ++ ++ ++ - - + + - - + - - - +++ +++ - - ++ ++ ++ ++ + ++ MRKP - - - - - - - - - - - - - - - - - - - - - -<ns0:label>3</ns0:label></ns0:formula><ns0:p>These 19 isolates showed antagonistic activities against Clostridium perfringens.</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:07:51002:1:0:NEW 31 Aug 2020)</ns0:p><ns0:p>Manuscript to be reviewed Antimicrobial activities amongst the short-listed isolates</ns0:p><ns0:p>The analysis was conducted using the cross-streak method.</ns0:p><ns0:p>+/++ zone of inhibition observed with clear halo of growth inhibition; -no inhibition. The species designations were assigned based on the 16S rRNA gene analysis in Table <ns0:ref type='table' target='#tab_3'>3</ns0:ref>.</ns0:p><ns0:p>Table <ns0:ref type='table' target='#tab_6'>6</ns0:ref>. Antimicrobial activities amongst the short-listed isolates. The analysis was conducted using the cross-streak method. Manuscript to be reviewed Manuscript to be reviewed</ns0:p><ns0:formula xml:id='formula_1'>- - - - - - - - - - + - - - - + - - - - - - #04 B. subtilis - - - - - - - - - - - + - - - - - - - - - - - #05 B. subtilis + + - + - + + + + + + + + + + + + + + - - + + #12 B. subtilis - - - - - - - - - - - - + - - - - - - - - - #14 B. subtilis - - - - - - - - - - - - - - - - - - - - - - - #16 B. subtilis - - - - - - - - - - + + - + - - - - ++ + + ++ + #17 B. subtilis - - - - - - - - - - + + - ++ - - - - ++ - ++ ++ + #18 B. subtilis - - - - - - - - - - + + - ++ - - + - + - ++ ++ ++ #20 B. subtilis - - - - - - - - - - - - - ++ - - - - - - + + - #22 B. subtilis - - - - - - - - - - - - - ++ - - + - - - ++</ns0:formula><ns0:formula xml:id='formula_2'>- + + + + #11 B. amyloliquefaciens - - - - - - - - - - - - - - - - - - + - - - - #13 B. amyloliquefaciens - - - ++ - + + + ++ ++ - - - - - ++ ++ - + - + - ++ #15 B. amyloliquefaciens - - - - - - - - - - - - - - - - - - + - - - - #07 B. pacificus - - - - - - - - - - - - - - - - - - + - - - -<ns0:label>#</ns0:label></ns0:formula><ns0:note type='other'>Figure 1</ns0:note><ns0:p>Morphological diversity of the marine isolates.</ns0:p><ns0:p>Colony morphologies of some pure marine spore-forming bacteria. The symbol # indicated the isolate number amongst the 23 short-listed isolates. 'C' is an example of a primary culture plate, in this case from a seaweed sample, from which isolates were subsequently colony purified. Photo credit: Chau Minh Khanh. </ns0:p><ns0:note type='other'>Figure 4</ns0:note><ns0:p>Phylogenetic tree of the short-listed 23 antimicrobial producing isolates.</ns0:p><ns0:p>The neighbor-joining phylogenetic tree was constructed using the maximum composite likelihood method, bootstrap method of 1000 replication and pairwise deletion by MEGA 7. In this tree, 23 marine isolates were displayed by number value, while all reference strains bacteria were closely related bacteria identified and downloaded from NCBI.</ns0:p></ns0:div><ns0:figure xml:id='fig_0'><ns0:head>Figure 1 .</ns0:head><ns0:label>1</ns0:label><ns0:figDesc>Figure1. Morphological diversity of the marine isolates. Colony morphologies of some pure marine spore-forming bacteria. The symbol # indicated the isolate number amongst the 23 shortlisted isolates. 'C' is an example of a primary culture plate, in this case from a seaweed sample, from which isolates were subsequently colony purified. Photo credit: Chau Minh Khanh.</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_1'><ns0:head>Figure 2 .</ns0:head><ns0:label>2</ns0:label><ns0:figDesc>Figure 2. Bioactivity assays. Antimicrobials activity screened by (A, B) cross-streak assay and (C) well-diffusion assay. (A) The cross-streak assay to identify antimicrobial producing bacteria against 6 indicators including; (1) S. faecalis; (2) B. cereus; (3) P. aeruginosa; (4) E. coli; (5) S. aureus; (6) C. albicans. (B) The illustration of cross-activity exhibited by P. polymyxa #23 (as a representative) against growth of other marine species. Notably, P. polymyxa killed all other marine Bacillus excepting itself #23 (as negative control). (C) The growth of MRSA as indictor was depressed by two isolates' culture supernatants. (D) Proteolytic activity screened on skim milk agar. (E) Cellulose degradation activity screened on CMC agar. (E) Amylase production screened on starch agar. Photo credit: Chau Minh Khanh.</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_2'><ns0:head>Figure 3 .</ns0:head><ns0:label>3</ns0:label><ns0:figDesc>Figure 3. Antimicrobial activities amongst the short-listed 23 isolates against 14 pathogenic indicator strains. Chart was constructed based on the value obtained from Table5. WD: Welldiffusion assay, CS: Cross-streak assay.</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_3'><ns0:head>Figure 4 .</ns0:head><ns0:label>4</ns0:label><ns0:figDesc>Figure 4. Phylogenetic tree of the short-listed 23 antimicrobial producing isolates. The neighbor-joining phylogenetic tree was constructed using the maximum composite likelihood method, bootstrap method of 1000 replication and pairwise deletion by MEGA 7. In this tree, 23 marine isolates were displayed by number value, while all reference strains bacteria were closely related bacteria identified and downloaded from NCBI.</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_4'><ns0:head>4 5</ns0:head><ns0:label>4</ns0:label><ns0:figDesc>PeerJ reviewing PDF | (2020:07:51002:1:0:NEW 31 Aug 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_5'><ns0:head /><ns0:label /><ns0:figDesc>Isolate #02 #04 #05 #12 #14 #16 #17 #18 #20 #22 #06 #08 #11 #13 #15 #07 #09 #03 #10 #01 #19 #21 #23 #02 B. subtilis -</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_6'><ns0:head>2 3 +5</ns0:head><ns0:label>3</ns0:label><ns0:figDesc>ability to grow under conditions based on colon formation; ++ and +++ increased size of colonies on plates; and -absence of growth. For enzyme 4 production the symbols indicate the relative sizes of the zones of activity.PeerJ reviewing PDF | (2020:07:51002:1:0:NEW 31 Aug 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_7'><ns0:head /><ns0:label /><ns0:figDesc>Figure 2</ns0:figDesc></ns0:figure>
<ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='35,42.52,70.87,525.00,536.25' type='bitmap' /></ns0:figure>
<ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='37,42.52,70.87,525.00,356.25' type='bitmap' /></ns0:figure>
<ns0:figure type='table' xml:id='tab_0'><ns0:head>Table 1 . List of indicators strains used for antimicrobial screening experiments</ns0:head><ns0:label>1</ns0:label><ns0:figDesc /><ns0:table><ns0:row><ns0:cell>No</ns0:cell><ns0:cell>Strain</ns0:cell><ns0:cell>Origin/ Strain storage</ns0:cell><ns0:cell>Media/ growth condition</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>Gram-positive bacteria</ns0:cell><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>1</ns0:cell><ns0:cell>Streptococcus faecalis</ns0:cell><ns0:cell>ATCC 29212</ns0:cell><ns0:cell>MH/ 37°C/aerobic</ns0:cell></ns0:row><ns0:row><ns0:cell>2</ns0:cell><ns0:cell>Lactobacillus plantarum</ns0:cell><ns0:cell>RMIT university</ns0:cell><ns0:cell>MRS/37°C/aerobic</ns0:cell></ns0:row><ns0:row><ns0:cell>3</ns0:cell><ns0:cell>Bacillus cereus</ns0:cell><ns0:cell>ATCC 10876</ns0:cell><ns0:cell>MH/ 30°C/aerobic</ns0:cell></ns0:row><ns0:row><ns0:cell>4</ns0:cell><ns0:cell>Staphylococcus aureus</ns0:cell><ns0:cell>ATCC25923</ns0:cell><ns0:cell>MH/37°C/aerobic</ns0:cell></ns0:row><ns0:row><ns0:cell>5</ns0:cell><ns0:cell>Listeria monocytogenes</ns0:cell><ns0:cell>Human pathogen, RMIT</ns0:cell><ns0:cell>BA /37°C/aerobic</ns0:cell></ns0:row><ns0:row><ns0:cell>6</ns0:cell><ns0:cell>Clostridium perfringens</ns0:cell><ns0:cell>Chicken pathogen, RMIT</ns0:cell><ns0:cell>MH /37°C/anaerobic</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>Gram-negative bacteria</ns0:cell><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>7</ns0:cell><ns0:cell>Salmonella Enteritidis</ns0:cell><ns0:cell>ATCC 13076</ns0:cell><ns0:cell>MH/ 37°C/aerobic</ns0:cell></ns0:row><ns0:row><ns0:cell>8</ns0:cell><ns0:cell>Escherichia coli</ns0:cell><ns0:cell>ATCC 25922</ns0:cell><ns0:cell>MH/ 37°C/aerobic</ns0:cell></ns0:row><ns0:row><ns0:cell>9</ns0:cell><ns0:cell>Pseudomonas aeruginosa</ns0:cell><ns0:cell>ATCC 15442</ns0:cell><ns0:cell>MH/ 37°C/aerobic</ns0:cell></ns0:row><ns0:row><ns0:cell>10</ns0:cell><ns0:cell>Campylobacter jejuni</ns0:cell><ns0:cell>Chicken pathogen, RMIT</ns0:cell><ns0:cell>BA/37°C/microaerophilic</ns0:cell></ns0:row><ns0:row><ns0:cell>11</ns0:cell><ns0:cell>Campylobacter coli</ns0:cell><ns0:cell>Chicken pathogen, RMIT</ns0:cell><ns0:cell>BA/37°C/ microaerophilic</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>Yeast</ns0:cell><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>12</ns0:cell><ns0:cell>Candida albicans</ns0:cell><ns0:cell>ATCC 10231</ns0:cell><ns0:cell>MH/30°C/aerobic</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>Antibiotic resistant pathogens</ns0:cell><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>13</ns0:cell><ns0:cell>Methicillin resistant</ns0:cell><ns0:cell>Human pathogen, RMIT</ns0:cell><ns0:cell>MH/ 37°C/aerobic</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>Staphylococcus aureus (MRSA)</ns0:cell><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>14</ns0:cell><ns0:cell>Vancomycin resistant</ns0:cell><ns0:cell>Human pathogen, RMIT</ns0:cell><ns0:cell>MRS/ 37°C/aerobic</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>Enterococcus faecalis (VRE)</ns0:cell><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>15</ns0:cell><ns0:cell>Multidrug resistant Klebsiella</ns0:cell><ns0:cell>Human pathogen, RMIT</ns0:cell><ns0:cell>MH/37°C/aerobic</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>pneumonia (MRKP)</ns0:cell><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell cols='4'>Standard media were MRS, De Man, Rogosa and Sharpe agar; MH, Muller Hilton agar; BA, Muller Hilton</ns0:cell></ns0:row><ns0:row><ns0:cell cols='4'>agar supplemented with 5% sheep blood. Microaerophilic and anaerobic condition were obtained with Gas-</ns0:cell></ns0:row></ns0:table><ns0:note>PakPeerJ reviewing PDF | (2020:07:51002:1:0:NEW 31 Aug 2020)</ns0:note></ns0:figure>
<ns0:figure type='table' xml:id='tab_1'><ns0:head>Table 2 (on next page)</ns0:head><ns0:label>2</ns0:label><ns0:figDesc /><ns0:table /></ns0:figure>
<ns0:figure type='table' xml:id='tab_2'><ns0:head>Table 2 . Antimicrobial producing bacteria identified from marine samples</ns0:head><ns0:label>2</ns0:label><ns0:figDesc /><ns0:table><ns0:row><ns0:cell cols='2'>1 Number of marine</ns0:cell><ns0:cell>Number of isolated</ns0:cell><ns0:cell>Number of</ns0:cell><ns0:cell>Percentage of isolates</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>samples collected</ns0:cell><ns0:cell>spore-forming bacteria</ns0:cell><ns0:cell>antimicrobial isolates</ns0:cell><ns0:cell>with antimicrobial</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell>activity</ns0:cell></ns0:row><ns0:row><ns0:cell>Sponges</ns0:cell><ns0:cell>16</ns0:cell><ns0:cell>183</ns0:cell><ns0:cell>28</ns0:cell><ns0:cell>14.8%</ns0:cell></ns0:row><ns0:row><ns0:cell>Seaweeds</ns0:cell><ns0:cell>13</ns0:cell><ns0:cell>92</ns0:cell><ns0:cell>15</ns0:cell><ns0:cell>16.3%</ns0:cell></ns0:row><ns0:row><ns0:cell>Sediments</ns0:cell><ns0:cell>13</ns0:cell><ns0:cell>81</ns0:cell><ns0:cell>16</ns0:cell><ns0:cell>19.8%</ns0:cell></ns0:row><ns0:row><ns0:cell>Sea water</ns0:cell><ns0:cell>8</ns0:cell><ns0:cell>33</ns0:cell><ns0:cell>6</ns0:cell><ns0:cell>18.2%</ns0:cell></ns0:row><ns0:row><ns0:cell>Total</ns0:cell><ns0:cell>50</ns0:cell><ns0:cell>389</ns0:cell><ns0:cell>64</ns0:cell><ns0:cell>16.5%</ns0:cell></ns0:row></ns0:table><ns0:note>2PeerJ reviewing PDF | (2020:07:51002:1:0:NEW 31 Aug 2020)</ns0:note></ns0:figure>
<ns0:figure type='table' xml:id='tab_3'><ns0:head>Table 3 (on next page)</ns0:head><ns0:label>3</ns0:label><ns0:figDesc /><ns0:table /></ns0:figure>
<ns0:figure type='table' xml:id='tab_5'><ns0:head>Table 4 . Enzymatic sensitivity and heat stability profile of antimicrobial activities.</ns0:head><ns0:label>4</ns0:label><ns0:figDesc>+: zone of inhibition observed with clear halo of growth inhibition in at least one time point; ++ and +++: increased activity as assessed 8 visually by the increased diameter of the inhibition zone; -: no inhibition. The percentage values were calculated based on the ratio 9 between number of isolates with antimicrobial activity against the indicator strain and the total 23 isolates tested. Value indicated the diameter of halo zone of antimicrobial activity in millimetres including well diameter of 6 mm. '-'indicates no zone of clearing.</ns0:figDesc><ns0:table><ns0:row><ns0:cell>6</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell>#13</ns0:cell><ns0:cell /><ns0:cell>#14</ns0:cell><ns0:cell /><ns0:cell>#15</ns0:cell><ns0:cell /><ns0:cell>#16</ns0:cell><ns0:cell /><ns0:cell>#17</ns0:cell><ns0:cell /><ns0:cell>#18</ns0:cell><ns0:cell /><ns0:cell>#19</ns0:cell><ns0:cell /><ns0:cell>#20</ns0:cell><ns0:cell /><ns0:cell>#21</ns0:cell><ns0:cell /><ns0:cell>#22</ns0:cell><ns0:cell>#23</ns0:cell><ns0:cell /><ns0:cell cols='2'>Percent (%)</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>CS</ns0:cell><ns0:cell>WD</ns0:cell><ns0:cell>CS</ns0:cell><ns0:cell cols='2'>WD CS</ns0:cell><ns0:cell cols='2'>WD CS</ns0:cell><ns0:cell>WD</ns0:cell><ns0:cell cols='2'>CS WD</ns0:cell><ns0:cell cols='2'>CS WD</ns0:cell><ns0:cell>CS</ns0:cell><ns0:cell>WD</ns0:cell><ns0:cell>CS</ns0:cell><ns0:cell>WD</ns0:cell><ns0:cell>CS</ns0:cell><ns0:cell>WD</ns0:cell><ns0:cell>CS</ns0:cell><ns0:cell cols='2'>WD CS</ns0:cell><ns0:cell cols='2'>WD CS</ns0:cell><ns0:cell>WD</ns0:cell></ns0:row><ns0:row><ns0:cell>Clostridium perfringens</ns0:cell><ns0:cell cols='2'>+++ +++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>+</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>+</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>+</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>+</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>+</ns0:cell><ns0:cell cols='2'>+++ +++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>+</ns0:cell><ns0:cell>86.7</ns0:cell><ns0:cell>82.6</ns0:cell></ns0:row><ns0:row><ns0:cell>Staphylococcus aureus</ns0:cell><ns0:cell cols='2'>+++ -</ns0:cell><ns0:cell cols='2'>+++ -</ns0:cell><ns0:cell cols='2'>+++ -</ns0:cell><ns0:cell>+++</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>+++</ns0:cell><ns0:cell cols='2'>+++ -</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell cols='2'>+++ -</ns0:cell><ns0:cell>+++</ns0:cell><ns0:cell>+</ns0:cell><ns0:cell>82.6</ns0:cell><ns0:cell>21.7</ns0:cell></ns0:row><ns0:row><ns0:cell>Bacillus cereus</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>+++</ns0:cell><ns0:cell cols='2'>+++ ++</ns0:cell><ns0:cell cols='2'>+++ ++</ns0:cell><ns0:cell>+++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>82.6</ns0:cell><ns0:cell>82.6</ns0:cell></ns0:row><ns0:row><ns0:cell>Listeria monocytogenes</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell cols='4'>+++ +++ +++ ++</ns0:cell><ns0:cell>+++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>+++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell cols='2'>+++ +++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>78.2</ns0:cell><ns0:cell>60.8</ns0:cell></ns0:row><ns0:row><ns0:cell>Lactobacillus plantarum</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>+++</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>+</ns0:cell><ns0:cell>+</ns0:cell><ns0:cell>47.8</ns0:cell><ns0:cell>47.8</ns0:cell></ns0:row><ns0:row><ns0:cell>Candida albicans</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>0</ns0:cell><ns0:cell>0</ns0:cell></ns0:row><ns0:row><ns0:cell>Escherichia coli</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>+</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>43.5</ns0:cell><ns0:cell>17.3</ns0:cell></ns0:row><ns0:row><ns0:cell>Salmonella Enteritidis</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>+</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>+</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>+</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>+</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell cols='2'>+++ +</ns0:cell><ns0:cell>+</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>+++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>56.5</ns0:cell><ns0:cell>17.3</ns0:cell></ns0:row><ns0:row><ns0:cell>Campylobacter jejuni</ns0:cell><ns0:cell>+</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>+</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>+</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>+</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>69.5</ns0:cell><ns0:cell>26</ns0:cell></ns0:row><ns0:row><ns0:cell>Campylobacter coli</ns0:cell><ns0:cell>+</ns0:cell><ns0:cell cols='2'>++++ -</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>+</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>+</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++++</ns0:cell><ns0:cell>+</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>+++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>60.8</ns0:cell><ns0:cell>43.5</ns0:cell></ns0:row><ns0:row><ns0:cell cols='2'>Pseudomonas aeruginosa -</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>4.3</ns0:cell><ns0:cell>4.3</ns0:cell></ns0:row><ns0:row><ns0:cell>MRSA</ns0:cell><ns0:cell cols='2'>+++ -</ns0:cell><ns0:cell cols='2'>+++ ++</ns0:cell><ns0:cell cols='2'>+++ ++</ns0:cell><ns0:cell>+++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>+++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>+</ns0:cell><ns0:cell cols='2'>+++ ++</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell cols='2'>+++ ++</ns0:cell><ns0:cell>+++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>82.6</ns0:cell><ns0:cell>21.7</ns0:cell></ns0:row><ns0:row><ns0:cell>VRE</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>+++</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>+</ns0:cell><ns0:cell>+</ns0:cell><ns0:cell>+</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>+</ns0:cell><ns0:cell>+</ns0:cell><ns0:cell>56.5</ns0:cell><ns0:cell>52.2</ns0:cell></ns0:row><ns0:row><ns0:cell>MRKP</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>4.3</ns0:cell><ns0:cell>4.3</ns0:cell></ns0:row><ns0:row><ns0:cell>7 10</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row></ns0:table><ns0:note>*:</ns0:note></ns0:figure>
<ns0:figure type='table' xml:id='tab_6'><ns0:head>Table 6 (on next page)</ns0:head><ns0:label>6</ns0:label><ns0:figDesc /><ns0:table /></ns0:figure>
<ns0:figure type='table' xml:id='tab_9'><ns0:head>Table 7 (on next page)</ns0:head><ns0:label>7</ns0:label><ns0:figDesc /><ns0:table /></ns0:figure>
<ns0:figure type='table' xml:id='tab_10'><ns0:head>Table 7 . Characterization of short-listed bacterial isolates</ns0:head><ns0:label>7</ns0:label><ns0:figDesc>Characterization of short-listed bacterial isolates + ability to grow under conditions based on colon formation; ++ and +++ increased size of colonies on plates; and -absence of growth. For enzyme production the symbols indicate the relative sizes of the zones of activity. # #01 #02 #03 #04 #05 #06 #07 #08 #09 #10 #11 #12 #13 #14 #15 #16 #17 #18 #19 #20 #21 #22 #23</ns0:figDesc><ns0:table><ns0:row><ns0:cell cols='2'>Strains Salinity tolerance (NaCl)</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>3%-5%</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell></ns0:row><ns0:row><ns0:cell>6%-7%</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>-</ns0:cell></ns0:row><ns0:row><ns0:cell>8%</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>+</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>+</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>+</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>+</ns0:cell><ns0:cell>-</ns0:cell></ns0:row><ns0:row><ns0:cell>9%</ns0:cell><ns0:cell>+</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell></ns0:row><ns0:row><ns0:cell>10%</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>+</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>+</ns0:cell><ns0:cell>+</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>+</ns0:cell><ns0:cell>+</ns0:cell><ns0:cell>+</ns0:cell><ns0:cell>+</ns0:cell><ns0:cell>+</ns0:cell><ns0:cell>+</ns0:cell><ns0:cell>+</ns0:cell><ns0:cell>+</ns0:cell><ns0:cell>+</ns0:cell><ns0:cell>+</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>+</ns0:cell><ns0:cell>+</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell></ns0:row><ns0:row><ns0:cell>12%</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell></ns0:row><ns0:row><ns0:cell cols='2'>Bile salt tolerance (mole/L)</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>0.1M -0.7M</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell></ns0:row><ns0:row><ns0:cell cols='2'>Antibiotic susceptibility</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>Kanamycin</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell></ns0:row><ns0:row><ns0:cell>Ampicillin</ns0:cell><ns0:cell>+</ns0:cell><ns0:cell>+</ns0:cell><ns0:cell>+</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>+</ns0:cell><ns0:cell>+</ns0:cell><ns0:cell>+</ns0:cell><ns0:cell>+</ns0:cell><ns0:cell>+</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>+</ns0:cell><ns0:cell>+</ns0:cell><ns0:cell>+</ns0:cell><ns0:cell>+</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>+</ns0:cell><ns0:cell>+</ns0:cell><ns0:cell>-</ns0:cell></ns0:row><ns0:row><ns0:cell>Tetracycline</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell></ns0:row><ns0:row><ns0:cell>Nalidixic acid</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell></ns0:row><ns0:row><ns0:cell>pH tolerance</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>5</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>+</ns0:cell><ns0:cell>+</ns0:cell><ns0:cell>+</ns0:cell><ns0:cell>+</ns0:cell><ns0:cell>+</ns0:cell><ns0:cell>+</ns0:cell><ns0:cell>+</ns0:cell><ns0:cell>+</ns0:cell><ns0:cell>+</ns0:cell><ns0:cell>+</ns0:cell><ns0:cell>+</ns0:cell><ns0:cell>+</ns0:cell><ns0:cell>+</ns0:cell><ns0:cell>+</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>+</ns0:cell></ns0:row><ns0:row><ns0:cell>6, 7, 8, 9</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell></ns0:row><ns0:row><ns0:cell>Thermal tolerance</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>30°C; 40 o C; 50 o C</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell></ns0:row><ns0:row><ns0:cell cols='3'>Oxygen requirement for growth</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>Aerobic</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell></ns0:row><ns0:row><ns0:cell>Microaerophillic</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell></ns0:row><ns0:row><ns0:cell>Anaerobic</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell></ns0:row><ns0:row><ns0:cell>Enzyme production</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>Protease</ns0:cell><ns0:cell>+++</ns0:cell><ns0:cell>+</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>+</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>+++</ns0:cell><ns0:cell>+</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>+</ns0:cell><ns0:cell>+</ns0:cell><ns0:cell>+</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>+</ns0:cell><ns0:cell>+</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>+++</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>+</ns0:cell><ns0:cell>++</ns0:cell></ns0:row><ns0:row><ns0:cell>Cellulase</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>+</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>+</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>++</ns0:cell><ns0:cell>++</ns0:cell></ns0:row><ns0:row><ns0:cell>Amylase</ns0:cell><ns0:cell>+</ns0:cell><ns0:cell>+</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>+</ns0:cell><ns0:cell>+</ns0:cell><ns0:cell>+</ns0:cell><ns0:cell>+</ns0:cell><ns0:cell>+</ns0:cell><ns0:cell>+</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>+</ns0:cell><ns0:cell>+</ns0:cell><ns0:cell>+</ns0:cell><ns0:cell>+</ns0:cell><ns0:cell>+</ns0:cell><ns0:cell>+</ns0:cell><ns0:cell>+</ns0:cell><ns0:cell>+</ns0:cell><ns0:cell>+</ns0:cell><ns0:cell>+</ns0:cell><ns0:cell>+</ns0:cell><ns0:cell>+</ns0:cell><ns0:cell>+</ns0:cell></ns0:row></ns0:table><ns0:note>PeerJ reviewing PDF | (2020:07:51002:1:0:NEW 31 Aug 2020)Manuscript to be reviewed 1</ns0:note></ns0:figure>
<ns0:note place='foot'>PeerJ reviewing PDF | (2020:07:51002:1:0:NEW 31 Aug 2020)</ns0:note>
</ns0:body>
" | "Dear Editor,
Thank you for the comments from the two reviewers. We have made changes to the manuscript to address the concerns that they raised.
Our detailed responses to all remarks are listed below.
Regards,
Robert Moore
Please take into consideration the reviewer’s comments and provide back a point-by-point rebuttal letter addressing those concerns. In particular, please show the relevance and timeliness of your research since it is questioned its novelty by one of the reviewers. More in-depth experimental evidence of novel findings is also critical for further consideration.
[# PeerJ Staff Note: While the reviewer(s) mentioned novelty in their reviews, this is not a criterion for rejection. However, a submission must make a contribution to the field as described in the editorial criteria (https://peerj.com/about/editorial-criteria/). #]
REPLY: We have indicated that the study is novel and does make a significant contribution. The geographical source of the strains, the Vietnam Sea, has not been previously assessed. An extensive survey was undertaken, encompassing 389 isolates and screening for activity against a much wider range of indicator strains than normally used for such screening. Isolates with unusually broad spectra of activity were identified.
Response to reviewers
Reviewer 1
Experimental design
Include more detail in enzymatic assays protocol performed in this manuscript.
Include more detail in the experimental design, highlight statistical factors and variables of response in the statistical analyses applied to the findings of this research.
REPLY: Statistical analysis has been added for the rate of isolation of antimicrobial producing isolates from the different marine sources. Standard deviations have been added to Table 4.
Validity of the findings
Include a possible mode of action of spore-forming strains against pathogenic bacteria tested.
Try to compare the obtained findings with similar assays were spore-forming bacteria were applied to inhibit pathogenic bacteria growth.
REPLY: Because of the sensitivity of some of the antimicrobial activities to protease treatment we speculated that some of the activities are likely to be caused by bacteriocins. We have also indicated that some of the activity form the Paenibacillus isolates is likely to be caused by polymyxin. We have also emphasised the relatively unusual finding of activity against important Gram-negative pathogens such as Campylobacter but have no basis on which to speculate about the possible mode of action. Further speculation on modes of action are best left until subsequent studies have been completed to purify some of the compounds.
Comments for the Author
Dear Author, I reviewed the manuscript (51002v1) entitled Broad spectrum antimicrobial activities from spore-forming bacteria isolated from the Vietnam Sea. This manuscript presents relevant information about the use of spore-forming bacteria to inhibit pathogenic bacteria growth. However, some sections of the presented data can be improved. For this reason, I considered that this manuscript needs minor changes for being considered for its publication in this journal.
REPLY: Thank you for your advice.
Additional comments.
Highlight the advantages of using nowadays these spore-forming strains against pathogenic bacteria.
REPLY: A brief paragraph has been added to the Introduction to highlight the advantages. The advantages had also already been mentioned towards the end of the Discussion.
Check paragraphs extension in this manuscript.
REPLY: Sorry, I don’t know what this refers to.
Include more detail in enzymatic assays protocol performed in this manuscript.
REPLY: Full details of the conditions are included.
Include more detail in the experimental design, highlight statistical factors and variables of response in the analyses applied to the findings of this research.
(Repeating the reply above.) REPLY: Statistical analysis has been added for the rate of isolation of antimicrobial producing isolates from the different marine sources. Standard deviations have been added to Table 4.
Include a possible mode of action of spore-forming strains against pathogenic bacteria tested.
(Repeating the reply above.) REPLY: Because of the sensitivity of some of the antimicrobial activities to protease treatment we speculated that some of the activities are likely to be caused by bacteriocins. We have also indicated that some of the activity form the Paenibacillus isolates is likely to be caused by polymyxin. We have also emphasised the relatively unusual finding of activity against important Gram-negative pathogens such as Campylobacter but have no basis on which to speculate about the possible mode of action. Further speculation on modes of action are best left until subsequent studies have been completed to purify some of the compounds.
Try to compare the obtained findings with similar assays were spore-forming bacteria were applied to inhibit pathogenic bacteria growth.
REPLY: We have done some of this sort of comparison and commented on the unusual spectra of activity that we found for some of our isolates. This emphasises the value of the work we report and the possibility of novel compounds being present.
Include future trends to keep working with the obtained data.
REPLY: This is the included as the last sentence of the Conclusions section.
Try to conclude with a general statement of the most relevant part of this study.
REPLY: This is what is written in the whole Conclusion section.
Reviewer 2
Basic reporting
The article needs to improve, it presents an insufficient introduction and a background is missing to demonstrate how the work fits into the broader field of knowledge. Current references are missing.
REPLY: The introduction concisely states the large picture concerns about antibiotic resistance and the need for alternatives. It then outlines the rational for looking in novel locations and justifies the approach taken in this study. The references used cite the most relevant articles in the scientific literature, without turning the work into a review.
Contaminated culture in figure 1; Low quality of figure 3; Figure 4 is a catastrophe, the phylogenetic tree must be rebuilt following all the required standards.
REPLY: The reviewer has misunderstood what is shown in Figure 1. The central panel (C) is not ‘contaminated’, it is, as the legend details, an example of a primary isolation plate and therefore has a mix of different sorts of colonies. The surrounding panels show the different morphologies of purified colonies.
The quality of the graph shown in Figure 3 has been improved.
The basis of labelling Figure 4 ‘a catastrophe’ is not detailed so it is difficult to respond to. We have modified the figure to include an outgroup, have corrected the italicization of two of the names of the type strains, and have expanded the isolate labels.
There are many tables, some unnecessary and incomplete essential information.
REPLY: The tables contain essential details of the characterisation of the isolates and their antimicrobial activity.
Experimental design
The results are very preliminaries. The information is not new in relation to marine bacteria, mainly for the genus Bacillus. Additional experiments are needed for the study, such as isolation and identification of bioactive compounds, tests for minimum inhibitory concentration and minimum bactericidal concentration, and cytotoxicity tests.
REPLY: The information we provide is, as far as we are aware, novel. This is the first survey of spore-forming bacteria derived from the Vietnam Sea and represents the most comprehensive list of bacteria against which initial screening for antimicrobials has been applied. The spectra of activity that we have identified indicates that some of the isolates are likely to produce novel antimicrobials. We agree that additional experiments are, of course, needed to isolate and fully identify the compounds. It is only at that stage, with purified compounds, that it would make sense to measure MICs. That will be the subject of our ongoing research into these isolates. The reviewer has already commented on the fact that there are already many tables – this represents a solid piece of research and to include further work would make the manuscript to large.
The reference list has many errors, this demonstrates a lack of care in preparing the document.
REPLY: This is unhelpful without providing some details of what problems were identified. We could only identify a single issue – one of the references had words starting with capital letters in the article title – that has now been corrected.
Validity of the findings
This study does not bring news in the area, it is a weak repetition of other publications.
In view of its lack of new information, I suggest that the manuscript be rejected.
REPLY: As we have indicated above, this work is novel. here aren’t previous reports of screening such bacteria derived from this geographic location and no other reports have carried out initial screening for activity against the range of bacteria that we have used. The spectra of activity identified from some isolates indicates that they may produce novel antimicrobial compounds that justify further investigation.
Comments for the Author
The study would be interesting if more results were presented. The work must be continued in order to isolate the bioactive compound and show that it has some potential to be used in the treatment of any infection.
REPLY: Many results, the product of a large amount of work, are presented. We start by postulating that the marine environment of the Vietnam Sea is a worthwhile location to investigate. Many environmental samples are recovered and then hundreds of spore-forming bacterial isolates were recovered and screened against a variety of bacteria, including clinically relevant highly antibiotic resistant pathogens. A range of different spectra of antimicrobial activity were identified, demonstrating the value of our whole approach and the importance of continuing with further work to characterise the antimicrobial compounds – work to be reported in subsequent publications.
" | Here is a paper. Please give your review comments after reading it. |
9,755 | "<ns0:abstract xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:p>Background: Experimental evolution of microbes can be used to empirically address a wide range of questions about evolution and is increasingly employed to study complex phenomena ranging from genetic evolution to evolutionary rescue. Regardless of experimental aims, fitness assays are a central component of this type of research, and low-throughput often limits the scope and complexity of experimental evolution studies. We created an experimental evolution system in Saccharomyces cerevisiae that utilizes genetic barcoding to overcome this challenge.</ns0:p></ns0:div>
<ns0:div><ns0:head>Results:</ns0:head><ns0:p>We first confirm that barcode insertions do not alter fitness and that barcode sequencing can be used to efficiently detect fitness differences via pooled competition-based fitness assays. Next, we examine the effects of ploidy, chemical stress, and population bottleneck size on the evolutionary dynamics and fitness gains (adaptation) in a total of 76 experimentally evolving, asexual populations by conducting 1,216 fitness assays and analyzing 532 longitudinal-evolutionary samples collected from the evolving populations. In our analysis of these data we describe the strengths of this experimental evolution system and explore sources of error in our measurements of fitness and evolutionary dynamics.</ns0:p></ns0:div>
<ns0:div><ns0:head>Conclusions:</ns0:head><ns0:p>Our experimental treatments generated distinct fitness effects and evolutionary dynamics, respectively quantified via multiplexed fitness assays and barcode lineage tracking. These findings demonstrate the utility of this new resource for designing and improving high-throughput studies of experimental evolution. The approach described here provides a framework for future studies employing experimental designs that require high-throughput multiplexed fitness measurements.</ns0:p></ns0:div>
</ns0:abstract>
<ns0:body xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:head>Introduction</ns0:head><ns0:p>Experimental evolution in microorganisms such as yeast, bacteria, and viruses has been used to answer evolutionary questions that are experimentally intractable in organisms with longer generation times (de Varigny 1892; Garland and Rose 2009; Kassen 2014; <ns0:ref type='bibr'>Van den Bergh et al. 2018)</ns0:ref>. A central benefit of most microbial experimental evolution systems is the ability to replicate and repeat evolution as well as the ability to store and compete evolved, intermediate, and ancestral strains <ns0:ref type='bibr'>(Van den Bergh et al. 2018)</ns0:ref>. Indeed, if there is a single unifying theme to what we have learned from experimental evolution it is that adaptation is universal and often repeatable due to parallel changes down to the molecular level <ns0:ref type='bibr'>(Burke, Liti, and Long 2014;</ns0:ref><ns0:ref type='bibr'>Kohn and Anderson 2014;</ns0:ref><ns0:ref type='bibr' target='#b1'>Bailey et al. 2017;</ns0:ref><ns0:ref type='bibr'>Graves et al. 2017;</ns0:ref><ns0:ref type='bibr' target='#b2'>Bailey, Guo, and Bataillon 2018)</ns0:ref>. The power of this approach has led to numerous investigations of the effects of population size <ns0:ref type='bibr'>(Schoustra et al. 2009;</ns0:ref><ns0:ref type='bibr'>A. C. Gerstein et al. 2011;</ns0:ref><ns0:ref type='bibr' target='#b1'>Bailey et al. 2017</ns0:ref>) and structure <ns0:ref type='bibr'>(Bell and Gonzalez 2011;</ns0:ref><ns0:ref type='bibr'>Kryazhimskiy, Rice, and Desai 2012;</ns0:ref><ns0:ref type='bibr'>Low-Décarie et al. 2015)</ns0:ref>, mutation rate the effects of thousands of single gene deletions on fitness using this approach. Barcodes can also be used to track lineages during experimental evolution and measure fitness improvements via competition assays. Crucially, utilization of barcodes vastly increases the throughput of experimental evolution fitness assays. For example, Blundell, Levy, and colleagues utilized a pool of half a million barcoded yeast strains to detect and track the fate of lineages and to simultaneously measure fitness of thousands of strains relative to one another over a short (<200 generation) time period <ns0:ref type='bibr'>(Blundell and Levy 2014;</ns0:ref><ns0:ref type='bibr'>Levy et al. 2015)</ns0:ref>. In subsequent work, the fitness of isolates from the evolved population of barcoded strains was measured using shortterm competition assays with a common barcoded reference strain <ns0:ref type='bibr'>(Venkataram et al. 2016;</ns0:ref><ns0:ref type='bibr'>Y. Li et al. 2018</ns0:ref>). While the above system leverages the throughput of pooled competition assays, it is limited to a small number of evolving populations over short time-scales. However, a similar system could accommodate a large number of evolving populations while retaining the throughput of pooled competition assays if each population were started with a unique barcoded strain(s). Such a system is quite desirable as the number of evolving populations determines the number of treatments, e.g. environments, and the number of replicates.</ns0:p><ns0:p>In this study we describe an experimental evolution system in Saccharomyces cerevisiae (hereafter, yeast) that utilizes genetic barcodes to enable high-throughput assessment of adaptation. The system is composed of a library of isogenic, diploid strains that only differ by a unique 20 bp barcode inserted upstream of the HO locus. The main advantage of this system is that it enables pooled fitness assays of replicate or differently treated lines marked by unique barcodes. Thus, populations can each be initiated with a single barcode and fitness of the resulting evolved lines can be measured in a single pooled fitness assay. However, the system is PeerJ reviewing PDF | (2020:06:49864:1:1:NEW 14 Sep 2020)</ns0:p><ns0:p>Manuscript to be reviewed also quite flexible, if populations are initiated with multiple barcodes they can be used to track the adaptive dynamics of different lineages that occur during the experimental evolution period.</ns0:p><ns0:p>While our system is limited in the number of lineages it can track compared to other studies <ns0:ref type='bibr'>(Kao and Sherlock 2008;</ns0:ref><ns0:ref type='bibr'>Blundell and Levy 2014;</ns0:ref><ns0:ref type='bibr'>Levy et al. 2015;</ns0:ref><ns0:ref type='bibr'>Selmecki et al. 2015)</ns0:ref>, crosscontamination between populations with different barcodes can be detected and monitored and multiple treatments can be directly compared even if represented by low-relative-fitness lineages.</ns0:p><ns0:p>To demonstrate the capabilities of our system, we began by conducting a series of proof-ofconcept fitness assays and subsequently applied what we learned in a short, 25-day (~250generation) experimental evolution. To take advantage of the flexibility of the system, the experimental design consisted of 76 populations spanning six different treatments, each initiated with two barcodes per population. We confirm that barcode insertions (1) do not alter the fitness of our source strains, and (2) provide a high-throughput means of measuring fitness differences and evolutionary dynamics for individual lineages from pooled samples obtained at different stages of experimental evolution. We highlight the advantages of multiplexing samples with indexing as well as the importance of limiting molecular contamination between initial sampling and library construction. This system represents a valuable resource for designing and improving high-throughput studies in experimental evolution.</ns0:p></ns0:div>
<ns0:div><ns0:head>Materials & Methods</ns0:head></ns0:div>
<ns0:div><ns0:head>Strains, media and culture methods</ns0:head><ns0:p>Barcoded yeast strains were constructed using two isogenic haploid derivatives of a strain collected from an Oak tree in Pennsylvania (YPS163) <ns0:ref type='bibr'>(Sniegowski, Dombrowski, and Fingerman 2002)</ns0:ref>: YJF153 (MATa, HO::dsdAMX4) and YJF154 <ns0:ref type='bibr'>(MATalpha,</ns0:ref><ns0:ref type='bibr'>HO::dsdAMX4)</ns0:ref> PeerJ reviewing PDF | (2020:06:49864:1:1:NEW 14 Sep 2020)</ns0:p><ns0:p>Manuscript to be reviewed (X. C. Li and Fay 2017). To initiate the barcoding process, barcoded kanMX deletion cassettes were amplified from the MoBY plasmid collection <ns0:ref type='bibr'>(Ho et al. 2009</ns0:ref>) with primers containing homology to the promoter region (-1,129 to -1,959) of HO. This region was selected because without HO it is unlikely to be functional. A set of 184 of these barcoded cassettes were selected based on confirmation of correct barcodes by sequencing (Supplemental Table <ns0:ref type='table'>S1</ns0:ref> contains entries for barcodes used in this project), then transformed into YJF153 and confirmed by PCR.</ns0:p><ns0:p>Barcoded diploid strains were made by mating these barcoded haploids (YJF153) to YJF154, diploids were confirmed by mating-type PCR <ns0:ref type='bibr'>(Huxley, Green, and Dunbam 1990)</ns0:ref>. The final barcoded yeast strain library consisted of 184 haploid and 184 diploid strains isogenic except for their barcodes; 92 strains from this set were utilized in the proof-of-concept fitness assays and 77 strains from the set were utilized in the 250-generation experiment. The 'ancestral reference strain,' discussed below, was arbitrarily selected from amongst the 184 candidate barcodes in the library; the same barcode sequence (1H10) is used for haploids (strain h1H10) and diploids (strain d1H10). Strain construction is a straightforward process that follows accepted protocols; additional strains can be constructed from the MOBY plasmid collection <ns0:ref type='bibr'>(Ho et al. 2009)</ns0:ref> as needed. Strains were stored at -80°C as 15% glycerol stocks.</ns0:p><ns0:p>All evolution and fitness assays were conducted in complete medium (CM; 20 g/l dextrose, 1.7 g/l yeast nitrogen base without amino acid and ammonium sulfate, 5.0 g/l ammonium sulfate, 1.3 g/l dropout mix complete without yeast nitrogen base) with or without additional stresses in 96deep well plates (2.2-ml poly-propylene plates, square well, v-conical bottom; Abgene AB-0932) covered with rayon acrylate breathable membranes <ns0:ref type='bibr'>(Thermo Scientific, 1256705)</ns0:ref>. Growth plates were incubated at 30°C for 24 hours inside an incubator (VWR, Forced Air Incubator, basic, PeerJ reviewing PDF | (2020:06:49864:1:1:NEW 14 Sep 2020)</ns0:p><ns0:p>Manuscript to be reviewed 120v, 7 cu. ft.) with agitation using a horizontal electromagnetic microplate shaker (Union Scientific LLC, 9779-TC). Saturated (stationary phase) 24-hour culture was diluted (1:1000) into fresh medium at the same time each day to initialize the next round of growth for all evolution and fitness assays.</ns0:p><ns0:p>Starting material for all evolution and fitness assays originated from -80°C freezer stocks of the constructed barcoded yeast strains. Yeast were revived from -80°C freezer stocks via a single round of growth (24 hours, 10 generations) under standard culture conditions (see previous paragraph) for these assays. Yeast samples collected during our experiments were stored both as (1) 15% glycerol stocks at -80°C to maintain viable freezer stocks of yeast populations [i.e., a 'frozen time vault', <ns0:ref type='bibr'>Van den Bergh et al. 2018</ns0:ref>] and ( <ns0:ref type='formula'>2</ns0:ref>) pelleted samples at -20°C for DNA extraction.</ns0:p></ns0:div>
<ns0:div><ns0:head>Experimental design</ns0:head><ns0:p>Proof-of-concept fitness assays: The design of this new system for experimental evolution began with a proof-of-concept analysis that allowed us to optimize our methods and to confirm that the fitness of multiple strains in pooled samples could be simultaneously and accurately measured with sequencing-based competition assay methods. This initial step involved measuring the fitness of 91 barcoded yeast strains relative to an ancestral reference strain simultaneously using a sequencing-based fitness assay (Figure <ns0:ref type='figure'>1</ns0:ref> Manuscript to be reviewed freezer stocks, mixed in equal proportions (i.e., such that each strain comprised 1/92 of the pooled population). The pooled strains were then diluted (1:1000) into fresh medium and grown for two-days, approximately 20-generations, in ten separate wells for the proof-of-concept fitness assays. Samples were obtained from the undiluted initial mixtures (Time 0-hours) and from the final overnight population cultures (Time 48-hours). Fitness was measured by the change in barcode abundance relative to a 'reference' strain (d1H10) from 10 replicates for a total of 910 fitness assays (91 focal barcodes x 10 replicates). See Fitness calculations, below, for a full description of the competition-based fitness assay methodology and for calculations of fitness from barcode abundance data. DNA was isolated separately for each fitness assay sample using a ZR Fungal/Bacterial DNA Kit (Zymo Research D6005) in individual 2.0 mL screw-cap tubes following the manufacturer's instructions. Physical cell disruption by bead-beating was carried out in a mixer mill (Retsch, MM 300) at 30 Hz (1800 min -1 ) for ten minutes (1-minute on, 1minute off, times ten cycles). Following extraction, DNA was amplified with forward/reverse primers containing a 9-12 bp index for multiplex sequencing. PCR products were quantified, pooled and purified to form a single multiplexed library for sequencing. Additional control samples were also included in the library to track barcode cross-contamination. See Library construction and sequencing, below, for a detailed description of the library preparation protocol used for these and all other samples. 250-Generation evolution experiment: After establishing the feasibility of the analysis strategy and optimizing culture, assay, and processing methods, 152 barcoded yeast strains were evolved for 25 days (i.e., ca. 250 generations at 9.97 generations per day -calculated from number of doublings based on optical density data) under different scenarios of selection in a second set of PeerJ reviewing PDF | (2020:06:49864:1:1:NEW 14 Sep 2020)</ns0:p><ns0:p>Manuscript to be reviewed experiments. These 152 strains were paired into groups of two to form 76 populations (wells) for evolution in order to facilitate examination of the evolutionary dynamics within each (see below). Specifically, yeast were evolved by serial dilution in one of six different treatments (Figure <ns0:ref type='figure'>1</ns0:ref>, B1.; See Supplemental Table <ns0:ref type='table'>S2</ns0:ref> for treatment descriptions). Evolutionary treatments involved growth in either complete media (CM), CM with ethanol (8% by volume) or CM with NaCl (0.342 M). Serial transfers were achieved either through standard dilution (1:1000), reduced dilution (1:250), or increased dilution (1:4000). Note that the 1:250 dilution treatment and the 1:4000 dilution treatment are expected to have undergone 200 and 300 generations of evolution, respectively, rather than the 250 generations of evolution expected in the treatments with a standard 1:1000 dilution. This difference in number of generations is caused by variation in the number of possible doubling events when different numbers of cells are added to media with the same (limiting) amount of glucose. Additionally, haploid and diploid yeast were evolved under standard culture conditions (see Strains, media and culture methods) to assess the effects of ploidy. To initialize this evolution experiment, yeast were revived from -80°C freezer stocks. Yeast strain pairs slated to evolve in competition were then mixed in equal proportions and diluted in fresh medium, according to the experimental design, and allowed to evolve for 25 days.</ns0:p><ns0:p>Yeast samples were obtained from the initial undiluted mixtures and on day 25 to use as the starting material for the Generation-0 and the Generation-250 fitness assays. These end-point assays of relative fitness were developed to characterize fitness outcomes in the 250-generation experiment (See End-point assay of relative fitness change). Additional samples were obtained at PeerJ reviewing PDF | (2020:06:49864:1:1:NEW 14 Sep 2020)</ns0:p><ns0:p>Manuscript to be reviewed regular intervals to characterize evolutionary dynamics (See Evolutionary dynamics). These two sets of assays are described in detail in the two following sections.</ns0:p><ns0:p>Evolutionary Dynamics: Relative proportions of strain pairs evolved in competition (within the same well) were quantified at generations 0, <ns0:ref type='bibr'>100,</ns0:ref><ns0:ref type='bibr'>150,</ns0:ref><ns0:ref type='bibr'>200,</ns0:ref><ns0:ref type='bibr'>220,</ns0:ref><ns0:ref type='bibr'>240,</ns0:ref><ns0:ref type='bibr'>and 250 (Days 0,</ns0:ref><ns0:ref type='bibr'>10,</ns0:ref><ns0:ref type='bibr'>15,</ns0:ref><ns0:ref type='bibr'>20,</ns0:ref><ns0:ref type='bibr'>22,</ns0:ref><ns0:ref type='bibr'>24 and 25)</ns0:ref>. Yeast samples from each timepoint were pooled for DNA extraction such that there was no barcode overlap within pools. DNA was subsequently extracted (see the Proof-ofconcept fitness assays section for details). Libraries were constructed for sequencing as described in the Library construction and sequencing section. From these data, the relative proportions of strains within each pair were measured at each time-point. Using this information, the time-point (t-max) and magnitude (m-max) of the maximum change in relative abundance in comparison to the starting conditions, the time-point (t-max-rate) and magnitude (m-max-rate) of the maximum rate of change between adjacent time-points, the time-point (t-max-diff) and maximum difference (m-max-diff) in barcode proportions, and the total cumulative change in barcode relative abundance were calculated for each two-barcode population. Barcodes approaching fixation (hereafter referred to as 'fixed') were also noted and were defined as cases in which a single barcode in the pair obtained (and maintained) a proportion of 0.95 or greater by (through) generation 250 of the evolution experiment.</ns0:p><ns0:p>End-point assays of relative fitness: This second type of assay was designed to assess the fitness of each focal strain relative to a static reference using a barcoded competition-based assay. Using a static reference enabled us to quantify the change in fitness for each strain (relative to the reference) between the start and end of an experimental evolution. In this case, the assays ). Briefly, Generation-0 and Generation-250 yeast strains were revived and samples from each time point were pooled, separately, such that there was no overlap in barcoded yeast strain identity within each pool. Pools contained 8-22 unique barcodes.</ns0:p><ns0:p>Pools were independently combined in equal proportion (50% pooled-yeast : 50% ancestral reference) with an ancestral reference strain (re: an 'unevolved' barcoded yeast strain), then diluted into fresh medium to initialize the fitness assays. Four replicate fitness assays were conducted for each pool of generation-0 and generation-250 yeast. Fitness assays employed a standard 1:1000 transfer dilution across all samples (see Strains, media and culture methods).</ns0:p><ns0:p>Samples evolved in CM plus additional stresses were assayed in the same media type in which they were evolved. Diploids were competed against a diploid reference strain (strain ID: d1H10), while haploids were competed against the haploid version of this same reference (strain ID: h1H10). In these assays, yeast samples for DNA extraction were obtained at hour-0 from the undiluted initial mixtures (fitness assay starting material), and, 20 generations later, at hour-48 from the final overnight population cultures (fitness assay end) to calculate fitness. See Fitness Calculations, below, for a full description of fitness assay methodology and for calculations of fitness from fitness assay barcode abundance data. Libraries were constructed for sequencing as described in the Library construction and sequencing section, below. <ns0:ref type='table'>2020:06:49864:1:1:NEW 14 Sep 2020)</ns0:ref> Manuscript to be reviewed Ion Torrent adaptors with indexes assigned to distinguish samples from one another (Supplemental Table <ns0:ref type='table'>S3</ns0:ref>).</ns0:p></ns0:div>
<ns0:div><ns0:head>Library construction and sequencing:</ns0:head><ns0:p>PCR products for library construction were generated using 25 cycles and were subsequently quantified with a Qubit 3.0 Fluorometer, high sensitivity assay kit. These products were then combined at equimolar concentrations and purified using a Zymo DNA Clean & Concentrator kit (Zymo Research D4014) to create a single library for sequencing. The DNA extraction and PCR steps were repeated for samples that did not attain sufficient DNA concentrations (no samples were exposed to > 25 rounds of PCR). We note that PCR jackpotting can be an issue in DNA library construction and that it can add noise to measurements of barcode frequency. To minimize the influence of this phenomenon on our results, we used a large amount of template DNA (Cha and Thilly 1993), and conducted all fitness assays in four replicates. Additionally, we avoided misclassification of barcodes by using high fidelity Taq Plus DNA Polymerase (Lambda Biotech, Taq Plus Master mix Red), and MOBY barcode sequences that differ from one another by at least six substitutions.</ns0:p><ns0:p>DNA libraries were sequenced using an Ion Torrent sequencer (Ion Proton System, Ion Torrent) at the Genomics Core Facility at Saint Louis University with a customized parameter to assess polyclonality after 31bp (the start position of the forward Ion Torrent adapter index sequence). A single sequencing run was used for each pooled library (library 1 -proof-of-concept fitness assays, library 2 -evolution experiment: evolutionary dynamics samples, Generation-0 fitness assays, and Generation-250 fitness assays). An additional library was constructed and sequenced for a set of samples from library 2 with elevated barcode cross-contamination rates. <ns0:ref type='table'>PDF | (2020:06:49864:1:1:NEW 14 Sep 2020)</ns0:ref> Manuscript to be reviewed</ns0:p></ns0:div>
<ns0:div><ns0:head>PeerJ reviewing</ns0:head></ns0:div>
<ns0:div><ns0:head>Sequence data processing & calculations</ns0:head><ns0:p>Sequence datasets: Sequence data in FASTQ format were parsed and demultiplexed using custom scripts in R (see the Availability of data and materials statement, below). A total of 142,243,245 raw reads that matched the forward Ion Torrent adapter indices included in our experiment (omitting reads that matched no forward adapter, polyclonal reads, low quality reads, and adapter dimer reads) were recovered across the sequenced libraries. 104,365,740 reads (73.4%) were retained for analysis that perfectly matched a forward sequencing adapter index (9-12 bp), reverse sequencing adapter index (9-12 bp) pair, and a MOBY genetic barcode (20 bp) included in the full experimental design.</ns0:p></ns0:div>
<ns0:div><ns0:head>Barcode cross-contamination rate:</ns0:head><ns0:p>The barcoded yeast experimental evolution system has an innate ability to detect and track barcode cross-contamination that could arise during evolution, over the course of short-term fitness assays, or during DNA library preparation. Barcode crosscontamination rate was defined as the mean number of counts mapping to any given barcoded yeast strain included in the full experimental design (sequenced library), but not expected to be present in the given sample (pair of forward and reverse IonTorrent adapter Indices). Barcode cross-contamination rates were calculated separately for each unique forward-reverse index pair included in each sequencing library and represent the amount of noise an average contaminating barcode strain contributes to each sample. The barcode cross-contamination rate of sample (primer pair) j, B j , was measured as, Equation <ns0:ref type='formula'>1</ns0:ref>.</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:06:49864:1:1:NEW 14 Sep 2020)</ns0:p><ns0:p>Manuscript to be reviewed</ns0:p><ns0:formula xml:id='formula_0'>𝐵 𝑗 = 1 𝑚 𝑚 ∑ 𝑖 = 1 𝐶 𝑖 𝑇 𝑗</ns0:formula><ns0:p>where m is the number of barcoded strains that could potentially contribute to barcode crosscontamination in experiment j (i.e., the number of unique strain IDs included in the full library, but not expected in sample j), C i is the number of barcoded yeast strain counts recovered for contaminating barcode i, and T j is the total number of counts recovered in sample j across all barcoded yeast strains included in the full experimental design.</ns0:p><ns0:p>Contamination rate summaries are reported separately for the proof-of-concept fitness assays and the evolution experiment (the latter containing the evolutionary dynamics, generation-0 fitness assay and generation-250 fitness assays samples). For the subset of evolution experiment samples that were sequenced in two separate libraries, only the replicate with a lower contamination rate was retained for contamination rate summary reporting and all downstream analyses. In all statistical analyses reported below, contamination rate is initially included as a potential predictor and subsequently removed from the model if its effect was found to be nonsignificant.</ns0:p><ns0:p>Fitness calculations: Fitness was evaluated via sequencing-based assays that involved competing barcoded yeast strains against a common ancestral reference strain for 48 hours (two overnight cultures; 20 generations). Reported fitness values in all fitness assays are relative to the same ancestral reference strain (strain ID: d1H10 for diploids; strain ID: h1H10 for haploids). where w i g0 and w i g250 are the strain's fitness relative to the ancestral reference at generation 0 and 250 as measured from Equation <ns0:ref type='formula'>2</ns0:ref>. Thus, we assumed no frequency dependent selection.</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:06:49864:1:1:NEW 14 Sep 2020)</ns0:p><ns0:p>Manuscript to be reviewed Evolutionary dynamics: For all analyses of evolutionary dynamics, only the first barcoded yeast strain from each pair was included to ensure that data were statistically independent. A set of weighted linear mixed effects models, with treatment as the predictor variable and either tmax, m-max, t-max-rate, m-max-rate, t-max-diff, m-max-diff, or total cumulative change in barcoded relative abundance across all time-points as the dependent variable, were employed to assess treatment differences in evolutionary dynamics. Full models included initial barcode abundance as a predictor term because initial barcode abundance could impact subsequent dynamics (e.g., (t-max ~ Treatment + Initial barcode abundance; family = gaussian)). Initial barcode abundance was subsequently removed from models when deemed nonsignificant, resulting in removal from all but one model (t-max-diff). Barcode fixation rate was not assessed statistically due to the small number of fixation events observed (n=7/76 populations).</ns0:p></ns0:div>
<ns0:div><ns0:head>Availability of data and materials</ns0:head><ns0:p>PeerJ reviewing PDF | (2020:06:49864:1:1:NEW 14 Sep 2020)</ns0:p><ns0:p>Manuscript to be reviewed </ns0:p></ns0:div>
<ns0:div><ns0:head>Results</ns0:head></ns0:div>
<ns0:div><ns0:head>Proof-of-concept fitness assays</ns0:head><ns0:p>We constructed 92 diploid yeast strains, all genetically identical except for a unique 20 bp barcode inserted upstream of the deleted HO gene. Proof-of-Concept fitness assays were then used to measure any fitness differences among these strains, to estimate our power to detect small fitness differences, and to assess our ability to measure fitness using multiplexed barcode sequencing. We assayed fitness simultaneously for our pool of 91 barcodes by measuring changes in barcode abundance relative to a 'reference' strain (barcode ID: d1H10) over a twoday period of approximately 20 generations.</ns0:p><ns0:p>A few of the barcoded strains showed significant differences in fitness (3/91 at a 5% false discovery rate, FDR; 1/91 at a 1% FDR) (Figure <ns0:ref type='figure' target='#fig_12'>2</ns0:ref>, Supplemental Table <ns0:ref type='table'>S4</ns0:ref>), possibly due to</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:06:49864:1:1:NEW 14 Sep 2020)</ns0:p><ns0:p>Manuscript to be reviewed mutations arising during the barcoding process. The root mean squared error (rMSE) among replicated measures of fitness was 1.76e-2, indicating good power (80%) to detect fitness differences of 1.2% between strains with four replicates at a nominal significance level of P = 0.05 (Figure <ns0:ref type='figure' target='#fig_13'>3a</ns0:ref> -green lines, Supplemental Table <ns0:ref type='table'>S5</ns0:ref>). We subsequently used these data to estimate power to detect treatment effects with end-point fitness assays in our proposed experimental design for the 250-generation experiment. With a modest number of barcoded yeast strains per treatment (n=22), we estimate high power (99.8%) to detect 1% fitness differences between treatments at a nominal significance level of P = 0.05 with four replicate end-point fitness assays (Figure <ns0:ref type='figure' target='#fig_13'>3b</ns0:ref> -green lines, Supplemental Table <ns0:ref type='table'>S6</ns0:ref>). With these promising results, we proceeded to conduct a more comprehensive test of the utility of barcoded strains in a practical experimental evolution context (i.e., the utility to evaluate the effects of treatment on fitness outcomes and evolutionary dynamics).</ns0:p></ns0:div>
<ns0:div><ns0:head>Experimental evolution</ns0:head><ns0:p>To empirically evaluate the strengths of this barcoded-strain system for studies of experimental evolution, we conducted a 25-day evolution experiment, which is approximately equal to 250 generations of evolution. Our design included 76 populations, each initiated with two barcodes per well. We used two barcodes per well to monitor evolutionary dynamics <ns0:ref type='bibr'>(Kao and Sherlock 2008;</ns0:ref><ns0:ref type='bibr'>Selmecki et al. 2015)</ns0:ref> while minimizing the chance of barcode loss (Blundell and Levy 2014). Samples spanned six treatments, which varied in yeast strain ploidy, growth medium, and daily transfer dilution (Supplemental Table <ns0:ref type='table'>S2</ns0:ref>). To measure any changes in fitness we competed the evolved generation-250 strains or their generation-0 ancestors against a common reference strain over a two-day period of approximately 20 generations. We also measured the change in Manuscript to be reviewed barcode frequency of barcoded strains within each evolving population (barcode pair) from longitudinal-evolutionary dynamics samples collected at 7 timepoints across the evolution. From these samples we quantified the magnitude and timing of changes in barcode frequency, which should be influenced by changes in the fitness of the evolving barcoded strains present within each population.</ns0:p></ns0:div>
<ns0:div><ns0:head>Multiplex barcode sequencing:</ns0:head><ns0:p>We utilized a two-step multiplexed design to obtain high throughput estimates of fitness based on barcode sequencing in our fitness assays. In the first step we leveraged the strain-identifying barcodes by pooling multiple strains together and simultaneously competing them against an ancestral reference strain to estimate relative fitness.</ns0:p><ns0:p>In our second multiplexing step we PCR-amplified each fitness assay sample with unique forward and reverse indexed sequencing adaptors. This latter step enabled us to assign sequencing reads to the appropriate fitness assay after sequencing many samples in concert as a single library. We constructed one library for each experiment (Library 1 -Proof-of-Concept fitness assays; Library 2 -250-generation evolution experiment evolutionary dynamics and fitness assays samples).</ns0:p><ns0:p>From 84,776,627 demultiplexed reads, we obtained a median of 2,961 reads per barcode in each sample. However, we also detected barcode cross-contamination, defined as barcodes with sample indexes that shouldn't exist in our sequenced libraries. The average rate of barcode crosscontamination per sample was low, 0.04%, and consistently present in nearly all of our samples (Supplemental Figure <ns0:ref type='figure'>S1 A</ns0:ref> ). However, the low but uniform rate of cross contamination we observe is more consistent with library preparation and sequencing than yeast contamination. This is further evidenced by the fact that no growth was observed in culture blanks during any fitness assays nor during the 25-day experimental evolution experiment.</ns0:p><ns0:p>To test whether the barcode cross-contamination rate depends on liquid handling during library preparation, we reprocessed a subset of samples with high cross contamination rates using identical starting material (multiple yeast sample aliquots were created at the time of sample collection). Cross contamination rates decreased significantly in these reprocessed samples (t = 22.3, df = 65, p < 10e-6), but a low level of background contamination remained (Original Samples: mean = 0.75%; reprocessed samples: mean = 0.05%) (Supplemental Figure <ns0:ref type='figure' target='#fig_12'>S2</ns0:ref>).</ns0:p><ns0:p>In our analyses, the presence of low abundance barcode cross-contamination was removed from all samples in which a barcode is not expected to occur. However, this doesn't eliminate contamination in samples where a barcode is expected to occur. In such cases, error in estimates of barcode frequency is highest when a barcode's frequency is low and approaches the crosscontamination rate. The effects of contamination rate and other system-specific covariates on experimental outcomes are presented in detail in the next section.</ns0:p></ns0:div>
<ns0:div><ns0:head>End-point fitness assays:</ns0:head><ns0:p>We assessed fitness in the 152 evolved strains (evolution experiment generation-250 strains) as well as their ancestors (evolution experiment generation-0 strains). For each strain and timepoint, fitness was measured in comparison to a common ancestral reference</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:06:49864:1:1:NEW 14 Sep 2020)</ns0:p><ns0:p>Manuscript to be reviewed strain via a two-day competition-based fitness assay in the same media type under which the strain had been evolved. The root mean squared error (rMSE) among replicated measures of fitness was 4.31e-2 for these assays. This is slightly higher than our proof-of-concept fitness assays (rMSE = 1.76e-2) but is still expected to provide reasonable power to detect fitness changes for individual barcodes (80% power to detect fitness changes of 2.99%) and fitness differences among treatments (80% power to detect treatment differences of 0.81%) (Figure <ns0:ref type='figure' target='#fig_13'>3</ns0:ref> orange lines, Supplemental Table <ns0:ref type='table'>S7</ns0:ref>, Supplemental Table <ns0:ref type='table'>S8</ns0:ref>). Among a number of covariates tested we found that only the cross-contamination rate (P = 0.01) and magnitude fitness change (P = 6.03e-4) were associated with error among replicates (Supplemental Table <ns0:ref type='table'>S9</ns0:ref>).</ns0:p><ns0:p>Twenty three percent (35/152) of the strains showed significant increases in fitness between generation 0 and generation 250 at a 1% false discovery rate, FDR (Figure <ns0:ref type='figure' target='#fig_14'>4</ns0:ref>, Supplemental Table <ns0:ref type='table'>S10</ns0:ref>). Within this subset, the average fitness increase was 6.80% with a range of 2.75% to 23.5%. Relatively few strains with significant increases in fitness were found in the CM treatment (3/42; 7.14%) and ethanol stress treatment (2/22; 9.09%). A higher proportion of strains exhibited significant increases in fitness in the lower 1:250 dilution treatment (9/22; 40.9%), the salt stress treatment (11/22; 50.0%), and the haploid treatment (10/22; 45.5%). No strains in the 1:4000 dilution treatment (0/22) exhibited increases in fitness. No strains (0/152) exhibited a significant decrease in fitness.</ns0:p><ns0:p>Next, we evaluated the effect of evolutionary treatment on fitness change. We found significant variation in fitness among treatments (P < 10e-6, Figure <ns0:ref type='figure' target='#fig_15'>5</ns0:ref>, Supplemental Table <ns0:ref type='table'>S11</ns0:ref>). Relative to our standard culture conditions, diploids in CM ('no stress') with 1:1000 transfer dilution, we</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:06:49864:1:1:NEW 14 Sep 2020)</ns0:p><ns0:p>Manuscript to be reviewed found greater (more positive) fitness change in diploid strains evolved in salt stress (P < 10e-6), in haploid strains evolved in standard culture conditions (P = 0.01), and in diploid strains evolved with a less extreme (1:250) daily transfer dilution (P = 9.49e-4). Relative to the CM ('no stress') diploid treatment (our standard culture conditions), we found less (less positive) fitness change in diploid strains evolved in ethanol stress (p = 3.99e-3). We found no significant effect of a more extreme (1:4000) daily transfer dilution on fitness change (p=0.39). In addition to the treatment effects, there was also a negative effect of barcode cross contamination rate on fitness change (P < 10e-6).</ns0:p></ns0:div>
<ns0:div><ns0:head>Evolutionary dynamics:</ns0:head><ns0:p>In experimental evolution, adaptation can influence the relative abundance of barcodes evolving in competition <ns0:ref type='bibr'>(Kao and Sherlock 2008;</ns0:ref><ns0:ref type='bibr'>Selmecki et al. 2015)</ns0:ref>.</ns0:p><ns0:p>In addition to characterizing fitness change for the evolved barcodes, we tracked the relative proportions of the barcoded-yeast pairs at seven timepoints during our evolution experiment to examine the evolutionary dynamics generated by adaptation and other processes (e.g., drift). We Out of the seven measures of adaptive dynamics that were amenable to statistical testing (i.e., all adaptive dynamics measures other than fixation rate, which had sparse imbalanced binary data),</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:06:49864:1:1:NEW 14 Sep 2020)</ns0:p><ns0:p>Manuscript to be reviewed six significantly differed by treatment (Table <ns0:ref type='table'>1</ns0:ref>; see Supplemental Figures S3:S9 and Supplemental Tables S12:S18 for individual adaptive dynamics plots and results tables, respectively). Relative to the CM-diploid treatment (our standard culture conditions) both the salt and haploid treatments were associated with a larger maximum change in relative abundance in comparison to the start (p = 5.24e-5, p = 5.85e-3, respectively) and a more extreme maximum difference in barcode proportion (p = 2.77e-4, p = 6.01e-6, respectively). The salt treatment was also associated with an earlier time-point of the maximum change in abundance relative to the start (p = 0.01), and the haploid treatment was associated with an earlier maximum rate of change (p = 1.35e-5). The total cumulative change in barcode abundance was greater for the 1:4000 dilution treatment than the 1:1000 dilution treatment (p = 1.77e-3). Total cumulative change in barcode abundance was significantly less for haploids than diploids (p = 4.9e-2), and significantly less in the ethanol stress treatment than in the CM with no stress treatment (p = 2.98e-2). Finally, we observed barcodes that approached fixation in eleven percent (9/76) of our populations. Most fixation events were observed in the salt (3/11; 27.3%) and haploid (5/11; 45.5%) treatments, one instance was found in the CM-diploid treatment and no instances of fixation were observed in the ethanol, 1:4000 dilution, nor 1:250 dilution treatments.</ns0:p></ns0:div>
<ns0:div><ns0:head>Discussion</ns0:head><ns0:p>Experimental evolution has proven to be a valuable approach for studying a range of evolutionary questions. In this study we implemented a genetic barcoding system in S. cerevisiae to increase the efficiency and throughput of measuring fitness. Through barcoding we were able to employ a complex experimental design, increase throughput of fitness measurement by multiplexing, and provide a relatively simple means to track the evolutionary dynamics of Manuscript to be reviewed barcodes evolving in competition during the experimental evolution period. Accordingly, we have demonstrated the potential for detecting fitness differences in six experimental evolution treatments and have shown that, while the typically low levels of barcode cross-contamination we observe cannot be completely eliminated, their effects on inference can be minimized through simple statistical procedures. Below we discuss the merits and drawbacks of the barcoding system, and its capabilities for efficient and high throughput quantification of fitness changes that occur during experimental evolution.</ns0:p><ns0:p>One important consideration for barcoding systems like the one we examine here and the system described by <ns0:ref type='bibr'>Levy, Blundell, and colleagues (Levy et al. 2015)</ns0:ref>, is that barcoded strains are not necessarily identical to one another at the beginning of an experiment (i.e., have equal fitness) even when all barcoded variants are produced from a single ancestral clone. Although we found no significant fitness differences among the majority of barcoded strains, we note that we did indeed observe a few strains with significant deviations from the population mean fitness. Given the location of our barcode insertions (i.e., a non-functional region of the genome) it is unlikely that the barcodes themselves generated these fitness differences. A perhaps more likely explanation is that these differences arose from mutations that occurred during transformation (Giaever and Nislow 2014) or shortly thereafter. Regardless, this limitation could be mitigated by (1) removing strains with unequal generation-0 fitness entirely from their analyses, (2) quantifying initial fitness differentials and including them as a covariate in downstream analyses, or (3) looking explicitly at change in fitness for each individual barcode as we do here. We note that initial fitness differentials may be of interest themselves, given that they can potentially impact evolutionary outcomes <ns0:ref type='bibr' target='#b3'>(Barrick et al. 2010;</ns0:ref><ns0:ref type='bibr'>Kryazhimskiy et al. 2014;</ns0:ref><ns0:ref type='bibr'>Jerison et al. 2017)</ns0:ref> PeerJ reviewing PDF | (2020:06:49864:1:1:NEW 14 Sep 2020)</ns0:p><ns0:p>Manuscript to be reviewed and the types of mutations that successfully spread through an experimental population (MacLean, Perron, and Gardner 2010).</ns0:p><ns0:p>In addition to throughput, the power to detect small changes in fitness in evolved strains is a critical parameter of any experimental evolution system. We estimated 80% power to detect a fitness difference of 1.2% in the proof-of-concept assays and 80% power to detect a fitness change of 3.0% in the 250-generation evolution experiment for any individual barcode. While a formal power analysis is not often reported, a review of the literature suggests that prior studies have detected fitness differences between 0.5% and 5% <ns0:ref type='bibr'>(Gresham et al. 2008;</ns0:ref><ns0:ref type='bibr'>Lang, Botstein, and Desai 2011;</ns0:ref><ns0:ref type='bibr'>McDonald, Rice, and Desai 2016;</ns0:ref><ns0:ref type='bibr'>Venkataram et al. 2016;</ns0:ref><ns0:ref type='bibr'>Fisher et al. 2018;</ns0:ref><ns0:ref type='bibr'>Marad, Buskirk, and Lang 2018)</ns0:ref>. For example, one previous barcoded fitness assay found between replicate deviations of 1-2% and between batch deviations of up to 5% <ns0:ref type='bibr'>(Venkataram et al. 2016)</ns0:ref>. Some studies report confidence intervals significantly lower than 0.5%, but these studies tend to employ a very large number of replicates <ns0:ref type='bibr'>(MacLean, Perron, and Gardner 2010;</ns0:ref><ns0:ref type='bibr'>Gallet et al. 2012;</ns0:ref><ns0:ref type='bibr'>Kryazhimskiy, Rice, and Desai 2012)</ns0:ref>. We also note that our power to detect treatment effects was much higher than our ability to detect changes in fitness for individual barcoded strains. Specifically, we estimated 99.8% power to detect fitness differences of 1% between treatments using the error in our proof-of-concept fitness assays, and 87.7% power to detect fitness differences of 1% between treatments using the actual 250-generation experiment data, indicating that the power to detect treatment effects in this system is similar to traditional systems that utilize competition-based assays to quantify fitness <ns0:ref type='bibr'>(Lang, Botstein, and Desai 2011;</ns0:ref><ns0:ref type='bibr'>McDonald, Rice, and Desai 2016;</ns0:ref><ns0:ref type='bibr'>Fisher et al. 2018;</ns0:ref><ns0:ref type='bibr'>Marad, Buskirk, and Lang 2018)</ns0:ref>. We note that power to detect small fitness effects can be increased by measuring the log-linear slope <ns0:ref type='bibr'>Salit, and Levy 2018)</ns0:ref>. While these strategies were not employed here, they could be incorporated into our system.</ns0:p><ns0:p>While the error measured in our fitness assays is consistent with most prior studies, pooled fitness assays do have additional sources of error. Prior studies using pooled fitness assays estimated that pools containing 1000 cells per strain are expected to have an error rate of 4.4% <ns0:ref type='bibr'>(Pierce et al. 2007)</ns0:ref> after two rounds of dilution due to sampling error. The expected number of cells per barcoded strain at inoculation in our system is 9,000 (3.0x10^8 cells/ml at saturation in 0.6ml of media, diluted 1:1000 and divided by up to 20 strains per assay). However, the number of cells could be much lower due to sampling or to low frequency of the barcode of interest in a well at the end of the 250 generations of evolution. Lineages with large changes in fitness are expected to deviate the most from equal numbers of cells in a well. Consistent with this possibility, the RMSE of fitness in generation 250 (4.34) is higher than the RMSE of fitness in generation 0 (2.92), where barcode frequencies are much more uniform. This provides one explanation for our finding that error among replicates is associated with the magnitude of the fitness change. Thus, while there are some limitations to pooled fitness assays, these limitations can be compensated by examination a modest number of strains per treatment -as done in this study-or by increasing the number of replicate assays for each individual strain.</ns0:p><ns0:p>Another important consideration when designing an experimental evolution system is that it must be robust to contamination. While culture contamination is rare in experimental evolution Manuscript to be reviewed Manuscript to be reviewed Manuscript to be reviewed abundance over long time scales may not be directly related to final fitness as these measures are calculated over 250 generations of evolution, during which relative fitness relationships between evolving barcodes could change significantly. For example, the 1:4000 transfer dilution bottleneck treatment had elevated rates of change in barcode abundance and a high amount of total change in barcode abundance without a concomitant increase in fitness, possibly as a result of the potentially strong effects of drift when severe bottlenecks are frequent. We suggest that future studies employ a denser and more even longitudinal-evolutionary dynamics sampling scheme, with replication, to maximize the value and precision of this type of lineage tracking or evolutionary dynamics data.</ns0:p></ns0:div>
<ns0:div><ns0:head>Conclusions</ns0:head><ns0:p>In summary, we conclude that the barcoded yeast system that we describe here offers a flexible, yet high-throughput means of fitness measurements and provides a relatively simple means for lineage tracking. These two characteristics support our ability to design more complex and potentially more informative evolutionary experiments. We note that although barcode crosscontamination imposes some limitations on the implementation of this system, it is possible to track the origin and rates of such contamination and, therefore, to statistically consider its effects on experimental outcomes. While prior barcoding systems leverage massive lineage tracking over short time-periods <ns0:ref type='bibr'>(Blundell and Levy 2014;</ns0:ref><ns0:ref type='bibr'>Levy et al. 2015;</ns0:ref><ns0:ref type='bibr'>Venkataram et al. 2016;</ns0:ref><ns0:ref type='bibr'>Y. Li et al. 2018)</ns0:ref>, our system uses far fewer barcodes but is more flexible at handling different experimental designs, such as the multiple treatments that we employed in this study. Thus, we conclude that this system represents an informative step towards the design and implementation of experimental evolution systems capable of both long-term evolution studies and high throughput fitness measurements.</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:06:49864:1:1:NEW 14 Sep 2020)</ns0:p><ns0:p>Manuscript to be reviewed </ns0:p></ns0:div><ns0:figure xml:id='fig_0'><ns0:head /><ns0:label /><ns0:figDesc>, A.). Ten replicate pooled fitness assays were conducted under standard culture conditions (see Strains, media and culture methods) with the null expectation that all strains should have identical fitness values if the barcode insertion process does not affect fitness. Briefly, 92 strains were revived from -80°C PeerJ reviewing PDF | (2020:06:49864:1:1:NEW 14 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_1'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:06:49864:1:1:NEW 14 Sep 2020) Manuscript to be reviewed involved quantifying barcode fitness using pooled samples from generation-0 (Figure 1, B2.) and separately pooled samples from generation-250 (Figure 1, B3.) from the 250-generation evolution experiment (Figure 1, B1.</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_2'><ns0:head /><ns0:label /><ns0:figDesc>Barcode sequencing libraries for the proof-of-concept fitness assays and all samples from both components of the 250-generation evolution experiment were constructed by amplification of MoBY barcodes (Ho et al. 2009) with primers containing PeerJ reviewing PDF | (</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_3'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:06:49864:1:1:NEW 14 Sep 2020)Manuscript to be reviewedThe relative Malthusian fitness of barcoded yeast strain i at generation gn, m i gn , was measured as, Equation2.𝑚 𝑖 𝑔𝑛 = ( 𝑙𝑛 𝐶𝑖 48 -ℎ𝑜𝑢𝑟𝑠 𝑅 48 -ℎ𝑜𝑢𝑟𝑠 -𝑙𝑛 𝐶𝑖 0 -ℎ𝑜𝑢𝑟𝑠 𝑅 0 -ℎ𝑜𝑢𝑟𝑠 ) 20 where Ci and R refer to barcode counts for the focal barcode and reference barcode at fitness assay time 0-hours (initial mixtures; fitness assay initial measurement) and time 48-hours (final overnight cultures; fitness assay final measurement), and 20 is the number of generations over 48 hours (two overnight cultures at 9.97 generations each -calculated from number of doublings based on optical density data) (Hartl and Clark 1997; Chevin 2011). We use the standard equation, m=ln(w) to convert Malthusian fitness values to Wrightian fitness values (Orr 2009; Wu et al. 2013; Passagem-Santos and Perfeito 2018). Herafter, fitness, denoted by a w will refer to Wrightian fitness. The change in fitness of strain i between Day 0 and Day 25 in for our 250generation evolution experiment, , we therefore computed as, ∆𝑤 𝑖 Equation 3. ∆𝑤 𝑖 = 𝑤 𝑖 𝑔250 -𝑤 𝑖 𝑔0</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_4'><ns0:head /><ns0:label /><ns0:figDesc>tools: All Analyses and statistical test were conducted in R version 3.5.2 (R Core Team 2012). Data processing uses primarily base-R functionality; the plyr package(Hadley Wickham 2011) is used in some cases for data frame manipulation. All Statistical models with linear mixed effects were generated using the lme4<ns0:ref type='bibr' target='#b5'>(Bates et al. 2015)</ns0:ref>, and lmerTest(Kuznetsova, Brockhoff, and Christensen 2017) packages. Weighted t-tests were assessed with the weights package (Pasek 2018). Power analyses were conducted using the pwr package (Champely 2018). Finally, figures and tables were generated with the ggplot2(Wikham 2009) and sjPlot (Ludecke 2019) packages, respectively; multi-panel figures were built using the gridExtra package<ns0:ref type='bibr' target='#b0'>(Auguie 2017)</ns0:ref>. Raw p-values are reported unless otherwise noted; some tables included in the supplement report raw and corrected p-values (i.e., false discovery rate) for cases where only corrected values are presented in the main text.Reads: Analysis of multiplex barcode sequencing data requires careful consideration of sample size. Because count data are ultimately handled as relative frequencies (proportions), it was necessary to consider underlying sample size or 'confidence' in each piece of data within the full dataset for all calculations and analyses. That is, entries with more reads were explicitly assumed to contribute more to summary calculations and statistical models. Variation in sample size was thus controlled for by weighting all calculations by the read sample size and by including such weights in downstream statistical models. This sample size metric considers both the total counts recovered for a multiplexed sample (unique forward and reverse index sequence adapter pair) and the number of counts recovered for the focal barcoded yeast strain(s) in that sample. In the analyses presented here, read calculations utilize harmonic means rather thanMultiplex barcode sequencing and cross-contamination: Barcode cross-contamination rate was assessed (Equation 1) and summarized, separately for the proof-of-concept fitness assays and the 250-generation experiment. The effects of barcode cross-contamination rate on the change in fitness and the error among replicate fitness measures for each barcode were assessed by including barcode cross-contamination as a predictor in two sets of linear mixed-effects models with either change in fitness over 250 generations of experimental evolution or standard error in fitness change over 250 generations (across four replicate measures for each barcode) as the response variable. Additionally, a weighted 1-tailed t-test was utilized to assess whether barcode cross-contamination rates for samples collected in the 250-generation experiment decreased after re-extracting DNA and resequencing. See the following sections for linear model and linear mixed-effects model specifications. Fitness change in 250 generations of experimental evolution: Individual barcoded yeast strains that exhibited a change in fitness over the 250-generation fitness experiment were identified using a weighted linear model with change in fitness as the response variable and a strain identifier (treatment plus yeast strain Barcode ID) as the predictor variable (fitness change ~ strain ID; family = gaussian). Next, the effects of evolutionary treatment (medium type, ploidy, and transfer dilution) on change in fitness over 250 generations of experimental evolution was assessed using a weighted linear mixed-effects model with cross-contamination rates as a covariate. A random effect of strain identifier (treatment plus yeast strain barcode ID) was placed on the model intercept (Fitness change ~ treatment + cross-contamination + (1|Strain identifier); family = gaussian). A PeerJ reviewing PDF | (2020:06:49864:1:1:NEW 14 Sep 2020) Manuscript to be reviewed second weighted linear model was fit to assess potential sources of error in the measurement of fitness in this system. Here, a model was constructed using a comprehensive set of factors that could have impacted the standard error in fitness measurement among replicates. These included: treatment, cross-contamination rate, magnitude fitness change, median count (median number of counts returned for a focal barcode across all sampling points that contribute to the fitness change calculation), and median proportion (median proportion of the focal barcode in its mixed 2-barcode well across all sampling points); (SE-Fitness change ~ treatment + crosscontamination + fitness change + median count + median proportion; family = gaussian).</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_5'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:06:49864:1:1:NEW 14 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_6'><ns0:head /><ns0:label /><ns0:figDesc>, B). Contamination could occur during culturing of yeast strains, liquid handling during preparation of libraries (Lenski et al. 1991; Van den Bergh et al. 2018), or PeerJ reviewing PDF | (2020:06:49864:1:1:NEW 14 Sep 2020)Manuscript to be reviewed potentially via index switching during sequencing procedures(Illumina 2017; Sinha et al. 2017; Costello et al. 2018</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_7'><ns0:head /><ns0:label /><ns0:figDesc>collected 8 measurements from these evolutionary dynamics: (1) barcode fixation, (2) the timepoint and (3) magnitude of the maximum change in relative abundance in comparison to the starting conditions, (4) the time-point and (5) magnitude of the maximum rate of change between adjacent time-points, (6) the time-point and (7) magnitude of the maximum difference in BC proportions, and (8) the total cumulative change in barcoded relative abundance summed across all time-points.</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_8'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:06:49864:1:1:NEW 14 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_9'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:06:49864:1:1:NEW 14 Sep 2020) Manuscript to be reviewed of barcode frequency change across several time-points and by accounting for PCR duplicates when estimating barcode frequencies (Blundell and Levy 2014; Venkataram et al. 2016; F. Li,</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_10'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:06:49864:1:1:NEW 14 Sep 2020)Manuscript to be reviewed(Lenski et al. 1991; T. F. Cooper and Lenski 2010), barcode cross-contamination is possible. Our results are consistent with prior work on this topic. Specifically, while we found no evidence of culture contamination in any growth blanks, we detected a uniformly low level of diffuse barcode cross-contamination in multiplexed fitness assays and evolutionary dynamics samples.Because there was no conspicuous pattern of cross-contamination, we suggest that the observed contamination was likely introduced during sample preparation and/or DNA sequencing. DNA extraction is a likely source of cross-contamination in samples processed in strip-tubes or 96well plate formats that prioritize throughput. However, we minimized the chance for contamination in the DNA isolation step in our experiments by isolating DNA with individual reaction tubes for each sample. Furthermore, contamination in the DNA extraction step would not be consistent with the low-level diffuse contamination observed, a pattern that tends to be more indicative of contamination occurring downstream of DNA extraction. Several other sources of contamination are possible, including primer contamination during the index addition for PCR(Lo, Mehal, and Fleming 1988) and index switching during library construction or sequencing(Illumina 2017; Sinha et al. 2017; Costello et al. 2018). The latter possibility seems nevertheless less likely because all PCR steps were performed separately prior to pooling of libraries and because no previous studies report index switching for IonTorrent sequence data to our knowledge.The strength and efficiency of barcoded experimental evolution is further evidenced by the biological results of this study. Here, barcoding enabled us to evolve a moderate number of strains (152) for 250 generations across six treatments, and to conduct a total of 532 evolutionary PeerJ reviewing PDF | (2020:06:49864:1:1:NEW 14 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_11'><ns0:head /><ns0:label /><ns0:figDesc>In addition to high-throughput fitness assays, barcoding enabled us to track the relative proportions of barcoded lineage pairs evolving in competition during the experimental evolution period(Hegreness et al. 2006; Blundell and Levy 2014; Levy et al. 2015; V. S. Cooper 2018). As expected, we found general agreement between the evolutionary dynamics results and endpoint fitness assay results in our six experimental treatments. Our haploid and NaCl stress treatments both displayed dynamics consistent with more extreme increases in fitness, including greater change in barcode abundance relative to the starting conditions and a greater maximum difference in barcode abundance than diploids in CM. Interestingly, haploids also showed signs of earlier adaptation than diploid strains in similar conditions, as evidenced by an earliergeneration of maximum rate of change in barcode abundance (also noted in (Blundell and Levy 2014; Levy et al. 2015; Selmecki et al. 2015)) and a lower total change in barcode abundance over 250 generations. Although this latter result may seem paradoxical, it is consistent with the observation that haploids adapt earlier than diploid strains (A. C. Gerstein et al. 2011), and is expected if a greater proportion of the total change in barcode abundance of haploids in our experiments happened in the first 100 generations (before we collected any evolutionary dynamics data).Despite significant differences in barcode dynamics, there are some limitations to interpreting these results. Because we mostly assessed abundance every 50 generations, it is possible that we missed some of the adaptive dynamics; dense temporal sampling is ideal for a full picture of evolutionary dynamics in populations subjected to experimental evolution(Hegreness et al. 2006). Furthermore, due to projected read depth constraints, barcode frequencies over time were not measured in replicate for the 250-generation evolution project. Finally, changes in barcode PeerJ reviewing PDF | (2020:06:49864:1:1:NEW 14 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_12'><ns0:head>Figure 2</ns0:head><ns0:label>2</ns0:label><ns0:figDesc>Figure 2</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_13'><ns0:head>Figure 3</ns0:head><ns0:label>3</ns0:label><ns0:figDesc>Figure 3</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_14'><ns0:head>Figure 4</ns0:head><ns0:label>4</ns0:label><ns0:figDesc>Figure 4</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_15'><ns0:head>Figure 5</ns0:head><ns0:label>5</ns0:label><ns0:figDesc>Figure 5</ns0:figDesc></ns0:figure>
<ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='49,42.52,459.37,525.00,224.25' type='bitmap' /></ns0:figure>
<ns0:figure type='table' xml:id='tab_0'><ns0:head /><ns0:label /><ns0:figDesc>The dataset supporting the conclusions of this article is available in the NCBI Sequence ReadArchive (SRA) repository, BioProject Number PRJNA555990, https://www.ncbi.nlm.nih.gob.bioproject/PRJNA555990 (Fasanello et al. 2019). Data formatted for analysis, intermediate data frames, as well as the custom R scripts utilized for all data processing, statistical analysis, and figure generation are available from GitHub (github.com/VinceFasanello/MM_Code_Supplement); a static version of the repository is available from Zenodo (Fasanello 2020). A readme file is available in the repository with the instructions necessary to reproduce the analyses and to confirm the results presented in this article. Supplementary figures, tables, and files referenced throughout the main text are available as 'Supplemental Files'. Strains are available from the Justin C. Fay Lab at The University of</ns0:figDesc><ns0:table /><ns0:note>Rochester; contact James Miller (e: j.h.miller@rochester.edu).</ns0:note></ns0:figure>
<ns0:figure type='table' xml:id='tab_1'><ns0:head /><ns0:label /><ns0:figDesc>dynamics measurements and 1,216 fitness assays related to these manipulations in a relatively short amount of time and with limited resources.Reassuringly, our biological results are largely consistent with prior work. We find greater fitness increases, i.e., a greater rate of adaptation, in haploids than diploids. This finding agrees with another study that found faster rates of adaptation and larger effective population sizes in haploids relative to diploids(A. C. Gerstein et al. 2011). In a related study, Selmecki et al.,(Selmecki et al. 2015) found faster adaptation in tetraploids than diploids, but no difference in rate of adaptation between haploids and diploids, potentially suggesting a trend opposite to ours of increasing adaptive rate with increasing ploidy. Further exploration of the evolutionary differences between haploid and diploid yeast reveals more support for the observation that diploids tend to evolve slower than haploids(Fisher et al. 2018; Marad, Buskirk, and Lang Sahin, Alkim, and Sezgin 2018). In contrast, we were surprised to detect less fitness increase in CM plus EtOH stress than in CM alone. There are several non-mutually exclusive explanations for this result. It is possible that ethanol did not present a significant stress (selective pressure) to the cells once they had attained physiological adaptation to the medium, i.e., acclimation(Huang et al. 2018). It is also possible that adaptations to CM and adaptations to ethanol exhibitantagonistic pleiotropy, similar to what has been found in experiments contrasting rich and poor media (Minty et al. 2011) or exploring adaptation to other chemical stressors (Reyes, Abdelaal, and Kao 2013). Pleiotropy could also shed light on the marked adaptation observed in the NaCl treatment given that adaptation to CM and NaCl stress may exhibit complementarity via synergistic or positive pleiotropy (Ostman, Hintze, and Adami 2012; Riddhiman Dhar et al.</ns0:figDesc><ns0:table><ns0:row><ns0:cell>2018), and indicates that haploids and diploids have different targets and types of mutations as</ns0:cell></ns0:row><ns0:row><ns0:cell>well as access to differing spectrums of beneficial mutations (Fisher et al. 2018; Marad, Buskirk,</ns0:cell></ns0:row><ns0:row><ns0:cell>and Lang 2018). However, differences between haploids and diploids must be treated with</ns0:cell></ns0:row><ns0:row><ns0:cell>caution because there is mounting evidence that haploid yeast evolves towards diploidy under 1999; De Visser and Rozen 2006; Kryazhimskiy, Rice, and Desai 2012), they nevertheless</ns0:cell></ns0:row><ns0:row><ns0:cell>experimental evolution conditions (Aleeza C. Gerstein and Otto 2011; R. Dhar et al. 2011; support earlier findings that less extreme bottlenecks favor the maintenance of adaptive mutants</ns0:cell></ns0:row><ns0:row><ns0:cell>Selmecki et al. 2015), potentially due to an initial spike in fitness associated with autodiploidy (Wahl, Gerrish, and Saika-Voivod 2002) and that large populations are less adaptively</ns0:cell></ns0:row><ns0:row><ns0:cell>(Fisher et al. 2018). We also note that neither haploids or diploids were put through a sexual constrained than small ones in simple environments (De Visser and Rozen 2006). They are also</ns0:cell></ns0:row><ns0:row><ns0:cell>cycle; allowing for sexual reproduction speeds the rate of adaptation (McDonald, Rice, and consistent with previous evidence for greater mean fitness increase in wide (relatively</ns0:cell></ns0:row><ns0:row><ns0:cell>Desai 2016). We also find greater fitness increase in complete medium (CM) plus NaCl stress unrestricted) versus narrow (relatively restricted) bottleneck populations (Schoustra et al. 2009).</ns0:cell></ns0:row><ns0:row><ns0:cell>than in CM alone, which was not surprising given what is known about adaptation to NaCl stress</ns0:cell></ns0:row></ns0:table><ns0:note>in S. cerevisiae(Blomberg 1995; R. Dhar et al. 2011; Park, Yang, and Kim 2015; Tekarslan-PeerJ reviewing PDF | (2020:06:49864:1:1:NEW 14 Sep 2020) Manuscript to be reviewed 2013; McGee et al. 2016; K. A. Hughes and Leips 2017). We find no difference in fitness change between our standard (1:1000) dilution treatment and a treatment with a more extreme (1:4000) daily transfer dilution. We do, however, find a greater increase in fitness when a less extreme (1:250) daily transfer dilution was used instead of the standard (1:1000) dilution. While these transfer dilution findings are not necessarily expected (Gerrish and Lenski 1998; De Visser et al. PeerJ reviewing PDF | (2020:06:49864:1:1:NEW 14 Sep 2020)</ns0:note></ns0:figure>
<ns0:note place='foot' n='2'>PeerJ reviewing PDF | (2020:06:49864:1:1:NEW 14 Sep 2020)Manuscript to be reviewed</ns0:note>
</ns0:body>
" | "Editor comments (Jürg Bähler)
All three reviewers raise a series of important issues about the narrative, missing details on the methods, experimental design, and framing of the research question, along with several minor corrections. These shortcomings should be carefully addressed to improve the paper.
Reviewer 1 (Anonymous)
Basic reporting:
no comment
Experimental design:
no comment
Validity of the findings:
no comment
Comments for the Author:
GENERAL COMMENTS
The main challenge I found with understanding this manuscript was to follow the narrative. There appear to have been multiple apporaches taken, at various times, for various reasons. For example, in the Experimental evolution section, lines 459-472, there was a 25-day evolution experiment (two barcoded per well), and also mention of '532 longitudinal-evolutionary dynamics samples'. I was confused. The dicsssino is much clearer, and does not need to change.
We have revised the methods to improve clarity on these points and provide context necessary to understand these specifics more clearly.
The Strains, media and culture methods section P9, line 109- do not explain well what the 92 diploid barcoded strains are. For example, are they geetically different from noe another? Have they been sequenced? How heterozygous was the parent YPS163? How different are they genetically from the ‘reference’ strain (barcode ID: d1H10) This is also the case in the results: P23, line 437. STill at this point I don't know what the strains are. It was only in the discussion that I was sure that the barcoded strains started as clocal replicates. hHis need to be explciit earlier, as it is critical to any readers understanding the experiment.
The strains are isogenic except for the barcodes, but were not sequenced. The barcoded strains are isogenic because they were all inserted into the same haploid strain, then mated to an isogenic strain to create diploids. The two isogenic haploids used in this study were derived by knocking out HO in a monosporic clone and so the starting diploid was homozygous (Li and Fay 2017). We have clarified this section of the methods and also noted that they are isogenic in the results:
'Barcoded yeast strains were constructed using two isogenic haploid derivatives of a strain collected from an Oak tree in Pennsylvania (YPS163) (Sniegowski, Dombrowski, and Fingerman 2002): YJF153 (MATa, HO::dsdAMX4) and YJF154 (MATalpha, HO::dsdAMX4)(Li and Fay 2017). ...The final barcoded yeast strain library consisted of 184 haploid and 184 diploid strains isogenic except for their barcodes'
The 143 Experimental design section (P10, lines 143-) could be clearer. Even with Figure 1 (which is good), I was not always clear about how many strains per well. Perhaps show this to a colleage (that was not inolved) and check that they can understand samples, pools etc.
There were 92 strains per well in the Proof-of-concept experiment and two barcodes per well in the experimental evolution. The pooled fitness assays for strains used in experimental evolution had between 8 and 22 barcodes per pool. We have revised the methods section to improve clarity. We also made edits throughout the Abstract, Introduction, and Materials & Methods to improve clarity regarding experimental design. Furthermore, targeted edits to the figure 1 description provide a clear roadmap to the strains per well.
Data: there are no issues with the availabily of the raw data, or the code. All these are aspects are at or above the current standard. Data analysis is also excellent. Sequence analysis and laboratory cuture also appear rigorous.
P24, line 453: 1% fitness differences does not seem very high, given that new alleles typically have effect sizes much smaller. How does this compare to other methods?
We discuss this finding in relation to prior studies in the Discussion:
'We estimated 80% power to detect a fitness difference of 1.2% in the proof-of-concept assays and 80% power to detect a fitness change of 3.0% in the 250-generation evolution experiment for any individual barcode. While a formal power analysis is not often reported, a review of the literature suggests that prior studies have detected fitness differences between 0.5% and 5% (Gresham et al. 2008; Lang, Botstein, and Desai 2011; McDonald, Rice, and Desai 2016; Venkataram et al. 2016; Fisher et al. 2018; Marad, Buskirk, and Lang 2018)...' Additional information has been added to this paragraph to complete this comparison.
PS: Please provide figures in line with the text in future. Not at the end. Placing figures at the end is counter-productive as it makes the reviewers job harder. The journal should insist on this, as well (I know that most don't).
Figures & Tables will be placed according to final submission guidelines for PeerJ.
P24, line 453 '(n=22), we estimate high power (99.8%) to detect 1% fitness differences between treatments at a' You estimate power for 22 strains, at this point in the ms. Why did you tehn conduct the 250 day experiments with two barcodes per well? This should be explained in the ms.
The power calculations for treatments (n=22) were made to match our experimental design. This calculation is the same regardless of the number of strains per well. We chose 2 barcodes per well since prior studies have shown high rates of loss (extinction) when there are many barcodes per population. We have now stated our choice explicity: 'We used two barcodes per well to monitor evolutionary dynamics (Selmecki et al. 2015, Kao and Sherlock 2008) while minimizing the chance of barcode loss (Levy et al. 2015).'
SMALL CORRECTIONS
All page numbers ref to the page in the pdf.
P4, 'fitness gains (adaptation) in a total of 76 experimentally evolving populations by conducting 1,216 fitness'. It would be useful to define what kind of populations here. Mitotic only or with meiosis? Starting from clonal, or a genetically heterogenous mix?
We now state these are asexual populations: 'in a total of 76 experimentally evolving asexual populations'. We also now state the genetic origin (isogenic strains) in the methods and results.
P7, line 69 '10,26,27]. However, the number of fitness assays, and thereby the resolution of those assays,' reference style differs
Correction made.
P8, line 90: 'Saccharomyces cerevisiae (hereafter, yeast) strains,' At an early stage, please tell the reader what the strains are. A gene knockouts? A population of wild strains? etc. Otherwise all the futher reading I am wonder what the experiments are doing to show. (at the this point in the ms I am already convinced that bar codes are a good idea).
Edits have been made to the relevant paragraph: 'The system is composed of a library of isogenic, diploid strains that only differ by a unique 20 bp barcode inserted upstream of the HO locus.'
The Strains, media and culture methods section P9, line 109- still does not explain well what the strains in use are. This is critical to understandin the experiment.
We have now clarified the provenance of the strains as described in the response above.
P9: line 134 'Saturated 24-hour culture was diluted (1:1000) into fresh medium at 134 the same time each day to initialize the next round of growth for all evolution and fitness assays.' What was the OD at this point? Did they reach stationary phase?
Yes, saturated is synonymous with stationary phase. OD for these strains is typically ~20 at saturation but depends on the environment and the instrument used. Because of the later dependency we have just noted that the culture is in stationary phase rather than the specific OD.
P13, line 201 '20, 22, 24 and 25) for a total of 532 evolutionary dynamics samples across 76 evolving two-' Adding numbers such as '532 evolutionary dynamics samples across 76' does not help to clarify the experiment.
This text has been removed.
P11, line 170 Why 152 barcoded yeast strains used here?. At each point, please explain why certain numbers of strauis were chosen.
This is simply a constraint of the 96-well format that we were using and a choice regarding the size of the experiment: 152 barcodes fit into 76 wells of a 96 well plate. The remaining wells were seeded with a different number of barcodes. However, we decided not to include these results as they were not informative and complicated an already complex experimental design. While we want to provide complete information regarding the sample size and design of the experiments included in the study, we feel adding in explanations of wells and experimental design for data not included would make the methods more difficult to understand.
There are many sentences descrbing how many fitness assays can be conducted, for example:
'With four replicate fitness assays for each barcode, this amounted to 1,216 fitness assays (2 barcodes * 2 time-points 468 * 76 populations * 4 replicates).' P 24, line 267. And again P24, line 467, 'fitness assays for each barcode, this amounted to 1,216 fitness assays (2 barcodes * 2 time-points'.
I wonder how many of these statements help the reader to understand the experiment? We all apreciate, from an eary stage in them manuscipt, that the system allows for many repeats. That bar-seq is a cheaper option that full genome sequencing is well-known.
Statements of this nature have been removed from the results section and in several other locations where they were extraneous.
Reviewer 2 (Anonymous)
Basic reporting :
I felt that the article came a little short in terms of the background/context provided, specifically with reference to how this methods fits in with other similar methods. The authors should more clearly describe the similarities and differences. Please see more comments below.
Experimental design :
Experimental design is sound.
Validity of the findings :
The findings are supported by the data.
Comments for the Author :
The manuscript “High-throughput analysis of adaptation using barcoded strains of Saccharomyces cerevisiae” presents a system to both measure fitness and conduct evolution experiments in budding yeast. The authors perform detailed analysis to assess the power of their system for detecting small differences in fitness. They also describe new experimental results confirming observations seen in previous studies, for example, that the rate of adaptation is faster in haploids than diploids. All in all, this is a nice study. My two major comments turned out to be very related, and so I really only have one main piece of advice that I hope improves the paper.
Major comments:
1. Another similar barcode system (the Levy/Blundell system cited by this study) achieves many of the same goals and the authors should take the time to describe all of the similarities and differences with that system. Here are some examples:
Thanks for pointing this out. We have revised parts of the introduction to make this clear and we address the specific examples below. In the introduction: 'The main advantage of this system is that it enables pooled fitness assays of replicate or differently treated lines marked by unique barcodes. Thus, populations can each be initiated with a single barcode and fitness of the resulting evolved lines can be measured in a single pooled fitness assay. However, the system is also quite flexible, if populations are initiated with multiple barcodes they can be used to track the adaptive dynamics of different lineages that occur during the experimental evolution period'
a. These systems have barcodes inserted into different regions of the genome. Is there any benefit to inserting into HO? Does the observation that barcode systems using a different barcode insertion locus are successful create possibilities for future multiplexing?
The insertion site doesn't matter as long as it doesn't affect fitness. We chose to insert at the HO promoter because HO is deleted. Combining these barcodes with those at another locus wouldn't provide a big advantage since barcode reads from both loci would have to be paired to maximize multiplexing.
b. The previous system measures fitness by finding the log-linear slope of barcode frequency change across multiple timepoints (see cited Venkataram et al reference from 2016). The system in this paper uses only the start and end point, which has been shown to yield reduced power in the following reference: https://www.cell.com/cell-systems/pdfExtended/S2405-4712(18)30390-9. Is there a future plan to gain power by considering more timepoints in the fitness measurements?
We expect that taking multiple time-points would improve estimates of fitness. We now make a note regarding this distinction and the advantage of using multiple time-points the discussion: 'We note that power to detect small fitness effects can be increased by measuring the log-linear slope of barcode frequency change across several time-points and by accounting for PCR duplicates when estimating barcode frequencies (Venkataram et al. 2016; F. Li, Salit, and Levy 2018). While these strategies were not employed here, they could be incorporated into our system'
c. The previous system does not pair the barcodes but combines hundreds of thousands. Is that something that you can do with your system? Is the max throughput of your system presumably similar to that seen in previous studies? Why did you choose to pair barcodes in the evolutions rather than combine hundreds of barcodes?
The system is flexible and can combine multiple barcodes into a single well. However, it is limited by the library size. Because each barcoded strain is generated individually, libraries of >1000 strains are unrealistic using our approach and the Levy/Blundell would be better. We chose to use only pairs so that we could simultaneously measure fitness from strains evolved in different wells. Thus, one barcode per well maximizes the number of treatments/replicates during experimental evolution. Note that processing 100 wells is relatively easy in our system since they just need to be pooled for the competition assay. We have revised the introduction to clarify the difference between many barcodes in one evolving pool versus one/two barcodes in many evolving pools:'While the above (Blundell/Levy) system leverages the throughput of pooled competition assays, it is limited to a small number of evolving populations over short time-scales. However, a similar system could accommodate a large number of evolving populations while retaining the throughput of pooled competition assays if each population started with a unique set of barcoded strains. Such a system is quite desirable as the number of evolving populations determines the number of treatments, e.g. environments, and the number of replicates....The main advantage of this system is that it enables pooled fitness assays of replicate or differently treated lines marked by unique barcodes. Thus, populations can each be initiated with a single barcode and fitness of the resulting evolved lines can be measured in a single pooled fitness assay. However, the system is also quite flexible, if populations are initiated with multiple barcodes they can be used to track the adaptive dynamics of different lineages that occur during the experimental evolution period'
d. Would having performed a two-step PCR, as in the previous system, have decreased noise due to PCR jackpotting? Is this going to become a requirement if the current system is used for evolutions tracking pools of thousands of barcodes?
PCR jackpotting is thought to be a much greater issue when the number of templates per barcode is small. Our PCR reactions would roughly correspond to 1.5e6 templates and so even with 100 barcodes the number of templates per barcode would be quite high. Blundell estimated 2000 templates per barcode/lineage, mostly a consequence of examining 500k barcodes. We now note that removing PCR duplicates could increase accuracy in the discussion.
2. The authors argue that their system increases the throughput of fitness measurements and experimental evolutions. But a previous system (see cited Venkataram et al study 2016 and Levy & Blundell et al 2015) is actually much higher-throughput. The authors should be clear about this by making the following changes:
a. In the beginning of the conclusion, instead of saying that this system improves throughput, perhaps add a qualifier, something like, “improves throughput relative to other evolution experiments using paired strains with fluorescent markers”.
We have edited the conclusions to be more precise: 'In summary, we conclude that the barcoded yeast system that we describe here offers a flexible, yet high-throughput means of fitness measurements and provides a relatively simple means for lineage tracking.'
b. Are there specific questions that can be addressed with this system that are less amenable for study with the Levy/Blundell system? Can you please describe these? Is your system basically another version of that system, or are there some nuances here that are different? Either way, this manuscript represents a valuable contribution. However, I wish this were clearer.
This comment is now addressed in the final two paragraphs of the introduction and copied into the response to the first major concern. We hope it is now clear that the main advantage is that our system enables many different treatments/replicates (stresses, dilutions, ploidy, etc) while retaining the ability of pooled fitness assays.
c. I was very confused while reading the methods because I am familiar with the Levy/Blundell system and I made some assumptions about the system in this manuscript that were untrue. This was partly because the introduction reads as though the system proposed here improves throughput over Levy/Blundell. My confusion was about the hundreds of barcodes used for the experimental evolution studies. I assumed each of these barcoded strains was in fact a library, containing hundreds of thousands of unique barcodes. But this is not the case, and this system, as depicted, does not seem to improve throughput relative to Levy Blundell. I think the introduction needs to clarify what this systems does and does not do and how it is similar/dissimilar to Levy/Blundell’s.
We understand that other readers may have a similar background and so have endeavored to make this clear throughout the manuscript.
d. Pooling samples with different barcodes before DNA extraction is a clever idea! This is the kind of thing that might appear earlier in the paper, in some kind of comparison to previous systems. Alternately this could appear in the discussion. Performing the evolutions in 96-well plates is also different from Levy/Blundell and gives the opportunity to study many environments. This might help avoid batch effects, which were very high in Venkatarm et al.
Thanks, we hope the revision clarifies this advantage from the beginning.
Minor comments:
1. The authors mention Blundell and colleagues in the introduction, but I believe Levy is the co-first author on at least one of the cited papers, maybe both. The text should probably read Blundell, Levy and colleagues.
Edit Completed.
2. I was mildly confused when reading the methods section referring to the proof of concept experiments because it seemed to me that all of the strains should have the same fitness. Indeed, this was the null expectation. The authors should say so in this section of the methods, perhaps more than once.
This null expectation is now explicitly stated in the appropriate methods section.
3. I love the power analysis! It is exactly the kind of thing I’ve been missing from previous studies.
4. I’m confused about the orange lines in figure 3. It is a little strange that this figure appears before the experimental evolutions are discussed in the results section. Perhaps give a heads up about why the barcodes have different fitness (I assume because they receive beneficial mutations) and why power is weaker than in fitness measurements?
Additional text has been added to all references to figure 3 in order to clarify this point. Parenthetical also added to figure 3 description.
5. Label the control group in figure 5.
The control group is now appropriately labeled.
Reviewer 3 (Anonymous)
Basic reporting:
Lines 62 – In this paragraph, it might be helpful to address what is actually being measured during these fitness assays (what is fitness in a microbial context and in the traditional fitness assay context?)
We now state fitness is measured by growth rate: 'Measuring adaptation by changes in fitness (i.e., growth rate)'
Line 69 – correct references to proper format
Correction Made.
Line 225-228 – Am I correct that for each evolved population, you pooled the time point 0 and competed it against an unevolved reference strain, and then pooled timepoint 250 and competed the pool against the unevolved reference strain? The wording here is a little confusing.
This is correct. Deletions made in this section improve clarity on this point.
Figure 1 – could panels be included that illustrates the concept of the barcoding and pooled fitness assays? I think these aspects of the experimental design are more important to illustrate visually than a sampling regime.
This figure is intentionally kept to a timeline as the barcoding and pooled fitness design is complicated and would not be a simple diagram – at least for the experimental evolution fitness assays.
There are many papers from the Lang and Desai labs that might be helpful to address sensitivity of fitness measurements, ploidy. For example:
The suggested references are now included in the discussion of sensitivity. The latter three references have also been used to further describe our ploidy results in context in the discussion section.
Lang GI, Botstein D, and Desai MM. 2011. Genetic variation and the fate of beneficial mutations in asexual populations. Genetics. Jul;188(3):647-61.
Marad DM, Buskirk SW, Lang GI. 2018. Altered access to beneficial mutations slows adaptation and biases fixed mutations in diploids. Nature Ecology & Evolution. May;2(5):882-889.
Fisher KJ, Buskirk SW, Vignogna RC, Marad DM, Lang GI. 2018. Adaptive genome duplication affects patterns of molecular evolution in Saccharomyces cerevisiae. PLoS Genetics. May 25;14(5):e1007396.
McDonald, M., Rice, D. & Desai, M. Sex speeds adaptation by altering the dynamics of molecular evolution. Nature 531, 233–236 (2016).
A particular comment on the correspondence of barcoding and fluorescence based fitness measurements from Levy et al. 2015 would be helpful
The authors of the study did due diligence in their comparisons and offer quite a few insights, but stated 'Other discrepancies between barcode-sequencing fitness and the fluorescent fitness however are more of a mystery'. We don't feel our results provide any further insights into the numerous issues related to why these two technologies may yield different results.
Experimental design :
What was the actual representation of the different barcoded strains in the timepoint 0 of the pooled fitness assays (e.g., how easy is it to ensure equal representation across all barcoded strains)? How does the initial abundance influence relative abundance over the course of the assay?
At time-point 0 the actual barcode frequency was similar to what one might expect given slight differences in density and liquid handling (see table below). While we did examine whether initial abundance was related to relative abundance, i.e. fitness, we don't believe any reliable conclusions can be made. Mathematically, even if two variables X and Y are entirely independent, COV(X, Y-X) is not zero if there is error in estimating X and Y. For example if X is slighly underestimated then by definition Y-X is slightly over-estimated. This is essentially the same statistical issue that has gone mostly unnoticed in cis/trans expression studies (PMID: 30509787). This issue is relevant since initial abundance at generation 0 is used to calculate fitness at generation 0, and similarly for generation 250. These statistical inter-dependencies are also conflated with biological relationships that we already expect. For example, a barcode that has outcompeted its pair and so starts at slightly higher frequency in the overall pool is more likely to have higher fitness (or a larger increase in relative abundance) than a barcode that had not outcompeted its pair. Presumably, the concern is that there may be frequency dependent selection such that initial frequency and fitness are dependent, or that the accuracy of our fitness assays depends on initial frequency. As stated in the manuscript, we assume no frequency-dependent selection in our calculation of fitness, and we used estimates of fitness weighted by underlying counts to control for the uncertainty of barcode frequency related to the underlying barcode counts in the pool.
Pool1
Pool2
Pool3
Pool4
Pool5
Pool6
Pool7
Pool8
Pool9
Pool10
Pool11
Pool12
Mean
0.055
0.060
0.071
0.058
0.074
0.064
0.056
0.061
0.073
0.071
0.029
0.025
SD
0.026
0.024
0.049
0.026
0.029
0.029
0.021
0.030
0.042
0.031
0.012
0.013
Barcode initial proportion mean and standard deviation for generation-0 fitness assay pools.
Can the authors address their rationale for including an unevolved reference strain in the competition vs. just measuring relative fitness of the pool by not including a reference strain and just looking at barcode abundance itself?
If all evolved strains improved we'd have no way measuring this in comparison to the unevolved strains without a common reference. This is a consequence of the evolved strains and their ancestors having the same barcodes and so their fitness must be measured in separate pools with a common reference. We have edited the methods to explain this rationale: 'Using a static reference enabled us to quantify the change in fitness for each strain (relative to the reference) between the start and end of an experimental evolution.'
Line 315 – how do you know there were 20 generations?
This is stated once in a parenthetical on former line 171. This information is now repeated in the description for Equation 2., as suggested. '20 is the number of generations over 48 hours (two overnight cultures at 9.97 generations each – calculated from number of doublings based on optical density data)'
Can the authors address their rationale for the paired barcoding in the experimental evolutions? What kind of considerations should be made for how many barcoded strains go in each population to track evolutionary dynamics?
The main advantage of our system is enabling pooled fitness assays of multiple evolved populations. We used 2 rather than multiple barcodes because more barcodes per well limits the number of evolutionary treatments. Despite our effort to include an analysis of evolutionary dynamics, we believe not much was gained, mostly because more time-points were needed: 'We suggest that future studies employ a denser and more even longitudinal-evolutionary dynamics sampling scheme, with replication, to maximize the value and precision of this type of lineage tracking or evolutionary dynamics data.'
Validity of the findings :
There is some interesting data presented here about how bottlenecking and different environments influence evolutionary dynamics, but this section is very brief and all the data are buried in a bunch of tables and figures in the supplemental. I would suggest a Figure that includes some of this data in the main text. I would also suggest clarifying the biological relevance of the different measurements that were obtained.
Thank you for appreciating that there are some potential insights and lots of potential for future studies to examine the best way to handle evolutionary dynamics. Our emphasis is intentionally steered away from evolutionary dynamics results as i) they are not the main focus of the study but rather an additional feature that is shared with other well know systems, and ii) the data backing the evolutionary dynamics results is quite sparse compared to the data for the fitness assay results (the main focus of the manuscript). We view these results as confirmatory but agree that future studies will better reveal the potential of evolutionary dynamics data that extends the lineage tracking results from Blundell, Levy and colleagues. We opted away from including a figure illustrating these results in the main text for this reason as well. We feel that the table provided is sufficient to highlight these results without overemphasizing their importance to the study.
Comments for the Author :
Barcoding has been used to track fitness and evolutionary dynamics for many years now, and there are a variety of different systems in use. There is a tradeoff with the proposed system (or any barcoding system), the upfront labor to generate barcoded strains (e.g., transforming and verifying presence of barcodes, validation of neutral fitness) vs. the downstream payoff of pooled fitness assays. This compares to other commonly used systems, like competition against a fluorescently tagged ancestor strain, which requires no upfront labor, but fitness assays are more laborious. Ultimately, it comes down to what questions one is interested in asking. For example, if you were interested in different strain backgrounds, barcoding approaches are challenging. However, the ability to request the strains and use the scripts provided here to analyze data is a big benefit. I can particularly imagine the use of this system in undergraduate courses being very impactful.
Thank you, we hope the revised manuscript conveys this point well.
Addressing the sensitivity of fitness assays for experimental evolution is an important problem in the field. The authors approach these issues very thoughtfully and methodically. It appears that the sensitivity of barcoded fitness assays presented here is similar to the sensitivity of fluorescent based approaches.
This is correct. We have made edits to improve clarity in comparison between these systems.
In my opinion, the greatest utility of this approach comes from being able to pool strains for fitness assays. This could be particularly helpful for researchers interested in tracking fitness over the course of evolution experiments (e.g., questions about evolvability), which is very difficult to assay with current protocols.
I think my struggle comes down to the framing of the manuscript. The framing is very technical and methods based, but I think it could benefit from highlighting the results and how they are innovative in a more biological context.
The framing is intended to be technical and methodological while still giving appropriate attention to biological results that we see as confirmatory of expectations and previous findings. A richer discussion of our ploidy findings has been added to the discussion to better contextualize these results.
" | Here is a paper. Please give your review comments after reading it. |
9,756 | "<ns0:abstract xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:p>Background: Advances in sequencing, assembly, and assortment of contigs into species-specific bins has enabled the reconstruction of genomes from metagenomic data (MAGs). Though a powerful technique, it is difficult to determine whether assembly and binning techniques are accurate when applied to environmental metagenomes due to a lack of complete reference genome sequences against which to check the resulting MAGs.</ns0:p></ns0:div>
<ns0:div><ns0:head>Methods:</ns0:head><ns0:p>We compared MAGs derived from an enrichment culture containing ~20 organisms to complete genome sequences of 10 organisms isolated from the enrichment culture. Factors commonly considered in binning software -nucleotide composition and sequence repetitiveness -were calculated for both the correctly binned and not-binned regions. This direct comparison revealed biases in sequence characteristics and gene content in the not-binned regions. Additionally, the composition of three public data sets representing MAGs reconstructed from the Tara Oceans metagenomic data was compared to a set of representative genomes available through NCBI RefSeq to verify that the biases identified were observable in more complex data sets and using three contemporary binning software packages.</ns0:p><ns0:p>Results: Repeat sequences were frequently not binned in the genome reconstruction processes, as were sequence regions with variant nucleotide composition. Genes encoded on the not-binned regions were strongly biased towards ribosomal RNAs, transfer RNAs, mobile element functions and genes of unknown function. Our results support genome reconstruction as a robust process and suggest that reconstructions determined to be >90% complete are likely to effectively represent organismal function, however, population-level genotypic heterogeneity in natural populations, such as uneven distribution of plasmids, can lead to incorrect inferences.</ns0:p></ns0:div>
</ns0:abstract>
<ns0:body xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:head>INTRODUCTION</ns0:head><ns0:p>High-throughput sequencing has revolutionized microbiology by circumventing 'the great plate count anomaly' (1) and allowing direct investigation of natural communities in a cultureindependent manner <ns0:ref type='bibr' target='#b1'>(2)</ns0:ref><ns0:ref type='bibr' target='#b2'>(3)</ns0:ref><ns0:ref type='bibr' target='#b3'>(4)</ns0:ref><ns0:ref type='bibr' target='#b4'>(5)</ns0:ref><ns0:ref type='bibr' target='#b5'>(6)</ns0:ref><ns0:ref type='bibr' target='#b6'>(7)</ns0:ref><ns0:ref type='bibr' target='#b7'>(8)</ns0:ref>. One goal of metagenomics has always been to obtain organismspecific, complete, genomic information from the complex mixture of sequence data generated from environmental samples. Having a complete genome sequence provides a platform for understanding the range of metabolic roles an organism can play within a community and the interactions it has with other organisms <ns0:ref type='bibr' target='#b8'>(9)</ns0:ref><ns0:ref type='bibr' target='#b9'>(10)</ns0:ref><ns0:ref type='bibr' target='#b11'>(11)</ns0:ref>, and it can provide specific context for interpretation of transcriptomics and proteomics <ns0:ref type='bibr' target='#b12'>(12,</ns0:ref><ns0:ref type='bibr' target='#b13'>13)</ns0:ref>. Metagenome-assembled genomes (MAGs) are produced by segregating assembled contigs/scaffolds into organism-specific 'bins'. This process of genome reconstruction has benefitted from continuing advances in sequencing technologies, sequence assembly algorithms, and segregation methods <ns0:ref type='bibr' target='#b14'>(14)</ns0:ref>. Early success assembling genomes from a simple community <ns0:ref type='bibr' target='#b15'>(15)</ns0:ref> has led to more recent studies reconstructing many organisms from complex environments <ns0:ref type='bibr' target='#b16'>(16)</ns0:ref><ns0:ref type='bibr' target='#b17'>(17)</ns0:ref><ns0:ref type='bibr' target='#b18'>(18)</ns0:ref><ns0:ref type='bibr' target='#b19'>(19)</ns0:ref><ns0:ref type='bibr' target='#b20'>(20)</ns0:ref><ns0:ref type='bibr' target='#b21'>(21)</ns0:ref><ns0:ref type='bibr' target='#b22'>(22)</ns0:ref><ns0:ref type='bibr' target='#b23'>(23)</ns0:ref><ns0:ref type='bibr' target='#b24'>(24)</ns0:ref><ns0:ref type='bibr' target='#b25'>(25)</ns0:ref><ns0:ref type='bibr' target='#b26'>(26)</ns0:ref><ns0:ref type='bibr' target='#b27'>(27)</ns0:ref><ns0:ref type='bibr' target='#b28'>(28)</ns0:ref><ns0:ref type='bibr' target='#b29'>(29)</ns0:ref><ns0:ref type='bibr' target='#b30'>(30)</ns0:ref>. The accuracy of these techniques in the context of a complex environmental community is difficult to gauge, however, because most available complete microbial genome sequences that could serve as references are from cultured isolates, and these isolates are rarely present in environmental metagenomes. Techniques that have been developed to evaluate the accuracy of the binning process rely on conserved genes and consistency of nucleotide composition <ns0:ref type='bibr' target='#b31'>(31)</ns0:ref><ns0:ref type='bibr' target='#b32'>(32)</ns0:ref><ns0:ref type='bibr' target='#b33'>(33)</ns0:ref><ns0:ref type='bibr' target='#b34'>(34)</ns0:ref><ns0:ref type='bibr' target='#b35'>(35)</ns0:ref>. These techniques, however, cannot make accurate determinations of how much sequence is missing or the functional potential of missing content. Genome reconstruction techniques have been tested using synthetic communities of cultured organisms <ns0:ref type='bibr' target='#b36'>(36)</ns0:ref> and simulated metagenomic datasets. Over time, increasingly sophisticated methods have been developed to simulate metagenomic read data sets, from the Manuscript to be reviewed earlier Grinder <ns0:ref type='bibr' target='#b37'>(37)</ns0:ref>, MetaSim <ns0:ref type='bibr' target='#b38'>(38)</ns0:ref>, GemSIM <ns0:ref type='bibr' target='#b39'>(39)</ns0:ref>, BEAR <ns0:ref type='bibr' target='#b40'>(40)</ns0:ref>, and NeSSM <ns0:ref type='bibr' target='#b41'>(41)</ns0:ref>, to the more recent CAMISIM <ns0:ref type='bibr' target='#b42'>(42)</ns0:ref>, which was developed as part of the community effort to address standards in metagenome analysis software development <ns0:ref type='bibr' target='#b43'>(43)</ns0:ref>. Generally these simulators concern themselves with modeling community structure and sequencing attributes, such as read length and error rates, but are limited to presenting data generated from a reference genomic database, thus cannot model the genetic diversity found in most environments, although CAMISIM addresses this issue by implementing the genome evolution simulator sgEvolver <ns0:ref type='bibr' target='#b44'>(44)</ns0:ref>.</ns0:p><ns0:p>Because genetic variability within natural populations is, as yet, ill-defined <ns0:ref type='bibr' target='#b45'>(45)</ns0:ref>, it is unlikely that such test data can accurately replicate the type and amount of variability found in natural communities, and the complications this variability causes.</ns0:p><ns0:p>Unicyanobacterial consortia (UCC) were developed as model systems to investigate the mechanisms of metabolic interaction between cyanobacteria and heterotrophs. These systems provide an opportunity to compare MAGs against a matching reference genome set and learn about potential gaps and pitfalls of current reconstruction processes. Two consortia, each containing a single unique cyanobacterial species and sharing an additional 18 heterotrophic species, were derived from a natural mat community <ns0:ref type='bibr' target='#b46'>(46)</ns0:ref>. The communities have been sequenced, and genome reconstruction has been performed <ns0:ref type='bibr' target='#b47'>(47)</ns0:ref>, yielding near-complete genome sequences revealing the presence and maintenance of microdiversity, such as might be found within an intact environmental sample. Thus, this system more accurately reflects in situ community diversity compared to synthetic communities constructed from isolated organisms. In parallel, isolates of 10 of the member species have also been sequenced <ns0:ref type='bibr' target='#b47'>(47,</ns0:ref><ns0:ref type='bibr' target='#b48'>48)</ns0:ref>. This paired genomic and metagenomic data set allows direct comparison of MAGs from diverse organisms against 'ground truth' genomic data. Previously, we have shown that common aspects of the Manuscript to be reviewed genome reconstruction process (assembly from a complex sequence space and segregation of contigs based on read depth profiles and sequence composition) to be both specific and sensitive <ns0:ref type='bibr' target='#b47'>(47)</ns0:ref>.</ns0:p><ns0:p>We have investigated the nature of genomic regions that under current standard genome reconstruction techniques are not recovered (herein referred to as not-binned regions, or NRs) to evaluate how these regions differ from recovered regions (correctly binned regions, or CRs),</ns0:p><ns0:p>and to what extent the missing genomic information might impact conclusions drawn from analysis of MAGs. Two common elements of current sequence segregation protocols are analysis of sequence composition and comparison of coverage profiles between samples, so we compared the nucleotide content of NRs vs CRs, examining both %G+C and tetranucleotide content, and the redundancy of sequence information both within the individual genome (i.e., repetitiveness within the genome) and across the entire metagenomic data set (i.e., sequence shared between populations). To determine the impact on downstream functional analyses, the gene content was examined for biases in the cellular roles of genes found within NRs and CRs.</ns0:p><ns0:p>To verify that the biases observed extended to more complex metagenomic datasets and across binning algorithms, the Tara Oceans metagenome, which has been binned by different groups using MetaBAT <ns0:ref type='bibr' target='#b22'>(22,</ns0:ref><ns0:ref type='bibr' target='#b49'>49)</ns0:ref>, Anvi'o <ns0:ref type='bibr' target='#b31'>(31,</ns0:ref><ns0:ref type='bibr' target='#b50'>50)</ns0:ref>, and BinSanity <ns0:ref type='bibr' target='#b21'>(21,</ns0:ref><ns0:ref type='bibr' target='#b51'>51)</ns0:ref>, was subjected to similar sequence and repeat compositional analysis.</ns0:p></ns0:div>
<ns0:div><ns0:head>MATERIALS & METHODS</ns0:head></ns0:div>
<ns0:div><ns0:head>Data and Code Availability.</ns0:head><ns0:p>The UCC MAG and genome data analyzed are available in the GenBank repository as listed in Manuscript to be reviewed SRA (accessions SRX1063989 and SRX1065184). MAGs reconstructed from the Tara Oceans metagenomic data <ns0:ref type='bibr' target='#b21'>(21,</ns0:ref><ns0:ref type='bibr' target='#b22'>22)</ns0:ref> are available in the GenBank repository. MAGs from Delmont et al. <ns0:ref type='bibr' target='#b50'>(50)</ns0:ref> are available through figshare (doi: 10.6084/m9.figshare.4902923). A list of MAGs and corresponding identifiers are available in Supplemental Table <ns0:ref type='table'>1</ns0:ref>. Complete bacterial and archaeal genomes were collected from NCBI RefSeq (52) (accessed Aug 2019) based on assignment as either 'reference genome' or 'representative genome' for the data column 'refseq_category' and 'Complete Genome' in the 'assembly_level' column. A list of genomes used in the analysis are available in Supplemental Table <ns0:ref type='table' target='#tab_3'>2</ns0:ref>. All analysis scripts are available at http://github.com/wichne/biases_in_genome_reconstruction.</ns0:p></ns0:div>
<ns0:div><ns0:head>Identification of CR and NR regions.</ns0:head><ns0:p>The UCC scaffolds comprising each MAG were searched against their cognate complete genome sequence using nucmer using the maxmatch option <ns0:ref type='bibr' target='#b53'>(53)</ns0:ref>. Regions of the genomes that aligned end-to-end to MAG scaffolds at ≥99% identity were cataloged as CR regions. All other genome regions were considered NR regions.</ns0:p></ns0:div>
<ns0:div><ns0:head>Compositional analysis.</ns0:head><ns0:p>For the UCC MAGs and genomes, %G+C calculation and tetranucleotide frequency (TNF) chisquare test were performed using custom Perl scripts (available at http://github.com/wichne/biases_in_genome_reconstruction). Compositional analysis was restricted to CR or NR regions longer than 1000 bp to ensure sufficient sequence for meaningful results. For TNF, the chi-squared statistic was calculated for each region using the TNF for the whole genome as the expected values, and the mean and standard deviation for the CR and NR pools calculated. For %G+C analysis, the mean %G+C for the CR and NR regions was calculated, and the absolute difference was calculated between each region and the genome Manuscript to be reviewed average, and average differences determined for CR and NR pools. To estimate p-values for the %G+C and TNF analyses, one thousand random coordinate sets yielding the same number and length of fragments as in each genome's CR or NR set were generated from the genome sequence and evaluated.</ns0:p><ns0:p>For comparison of the UCC data set to the Tara Oceans MAGs and RefSeq genome data sets, sequence composition variance (i.e., deviation from the mean) was calculated for the %G+C and tetranucleotide frequency using a custom Python script. The %G+C was calculated for 2kb segments (sliding window of 500bp) for each MAG or genome. A genome-wide variance value was calculated for each MAG or genome based on the segments and plotted as a box plot per source data set. TNF was calculated for 10kb segments (sliding window 5kb) for each MAG or genome. Using the calculation described in Teeling <ns0:ref type='bibr' target='#b54'>(54)</ns0:ref>, each segment had a Z-score calculated for each tetranucleotide based on the observed-vs-expected frequency of the tetranucleotide in the 10kb segment. A Pearson correlation was then calculated in a pairwise fashion for all segments. Variance of the Pearson correlation values within a MAG or genome was calculated and plotted as a box plot per source data set.</ns0:p></ns0:div>
<ns0:div><ns0:head>Repetitiveness analysis</ns0:head><ns0:p>To calculate intragenome sequence repetitiveness, we determined the fraction of each genome that was comprised of repeat sequence. Each genome sequence was searched against itself using nucmer v3.0 <ns0:ref type='bibr' target='#b53'>(53)</ns0:ref> with the maxmatch option, and the lengths of regions that aligned to another part of the genome/MAG with 97% identity were summed and divided by the length of the ≥ genome/MAG.</ns0:p><ns0:p>To determine the repetitiveness of sequences across the entire metagenomic data set, metagenome reads were searched against genome sequences using Bowtie2 <ns0:ref type='bibr' target='#b55'>(55)</ns0:ref>. Per-base coverage was calculated using the samtools (56) depth command, and average coverage values for the genomes, NRs and CRs were determined. One thousand sets of random coordinate regions of the same number and lengths as in each set were analyzed to estimate p-values.</ns0:p><ns0:p>Results are reported as average coverage depth of NRs and CRs and the average difference from the genome depth-of-coverage.</ns0:p></ns0:div>
<ns0:div><ns0:head>Gene function analysis</ns0:head><ns0:p>UCC complete genome sequences were annotated by the IMG pipeline <ns0:ref type='bibr' target='#b57'>(57)</ns0:ref>, which included COG assignment based on the December 2014 release of the 2003-2014 COGs <ns0:ref type='bibr' target='#b58'>(58)</ns0:ref>. COGs assigned to more than one functional category were counted for each assigned category. Genes not assigned to a COG category were classified as 'unassigned'. Ribosomal RNA (rRNA) gene features were identified by the IMG pipeline (59); transfer RNAs (tRNA) were identified with tRNAscan-SE ( <ns0:ref type='formula'>60</ns0:ref>); other non-coding RNAs (ncRNA) were identified using the Rfam database v11.0 (61) and infeRNAl v1.1 software <ns0:ref type='bibr' target='#b62'>(62)</ns0:ref>. For each gene set, the category counts were normalized to the total feature counts. Principle component analysis was performed and biplot of gene categories was generated using R package bpca v.1.2-2 (http://cran.rproject.org/web/packages/bpca/).</ns0:p></ns0:div>
<ns0:div><ns0:head>Statistical analysis.</ns0:head><ns0:p>Statistical tests were performed using modules within the Python package SciPy <ns0:ref type='bibr' target='#b63'>(63)</ns0:ref>. The normality of the calculated variance distributions for each set of genomes was determined using the Shapiro-Wilk test <ns0:ref type='bibr' target='#b64'>(64)</ns0:ref>. Genome sets with a normal distribution were compared to each other with the T-test for two independent variables <ns0:ref type='bibr' target='#b65'>(65)</ns0:ref>. Genome sets without a normal distribution were compared to each other with the Mann-Whitney U test <ns0:ref type='bibr' target='#b66'>(66)</ns0:ref>. p-values were adjusted for Manuscript to be reviewed multiple comparisons with the Benjamini-Hochberg procedure (67) correction with a false discovery rate of 25% (Supplemental Table <ns0:ref type='table' target='#tab_7'>3</ns0:ref>).</ns0:p></ns0:div>
<ns0:div><ns0:head>RESULTS AND DISCUSSION</ns0:head><ns0:p>The power of metagenomics is that it allows exploration of diverse communities from which we Manuscript to be reviewed sequenced. Comparison of the MAGs to the isolate genomes showed recovery of >90% of sequence for genomes with at least 10x coverage, with one exception, Halomonas sp. HL-93, which had 85% recovery from 11x coverage (Table <ns0:ref type='table'>1</ns0:ref>). Co-linear sequence alignments indicated there were no assembly errors in the binned contigs (47, and data not shown). Based on the isolate-MAG comparisons, NRs were identified. Porphyrobacter HL-46 had the lowest metagenome coverage (3.6x). Its MAG comprised hundreds of short contigs and was determined to be ~40% complete. Thus, the NRs for HL-46 are assumed to be primarily caused by the random sampling of the shotgun sequencing methodology and not by any inherent content biases, allowing the HL-46 analyses to serve as a control.</ns0:p><ns0:p>To determine if NRs were not binned due to lack of assembly, we mapped the contigs from the assembly to the CR and NR regions of the genomes and looked at the contig coverage of the regions. As expected, the CRs showed an average contig coverage of 1.04±0.14, and most regions had only a single contig map to them (Fig <ns0:ref type='table'>S1</ns0:ref>). Many of the cases of multiple contigs mapping to a CR were due to short (<200 bp) contigs of repeat sequence which might be an artifact of the assembler (IDBA_ud). NRs show a strong positive correlation between region length and number of contigs mapping, with an average coverage of 0.94±0. <ns0:ref type='bibr'>71 (Fig S2)</ns0:ref>. This suggests poorer assembly of the NRs and higher repeat content, but also indicates that most NR sequence is present in the contig set, and thus the binning process is the main determinant of NRs.</ns0:p></ns0:div>
<ns0:div><ns0:head>Nucleotide composition of NRs frequently differs from the genome average</ns0:head><ns0:p>Bacteria and Archaea have evolved to have a fairly consistent %G+C across their genome (69), so much so that it has been proposed as a metric of classification at higher taxonomic levels <ns0:ref type='bibr' target='#b70'>(70)</ns0:ref>.</ns0:p><ns0:p>PeerJ reviewing PDF | (2017:03:17041:1:1:NEW 4 Sep 2020)</ns0:p><ns0:p>Manuscript to be reviewed</ns0:p><ns0:p>It is not uncommon, however, to observe regions within a genome that differ significantly from the genome average <ns0:ref type='bibr' target='#b71'>(71)</ns0:ref>. This variation can be the result of selective pressure for structural properties in non-coding genes, for instance ribosomal RNAs and other functional RNAs have been shown to vary in nucleotide composition in correlation with optimal growth temperature <ns0:ref type='bibr' target='#b72'>(72)</ns0:ref>. In other cases, divergent %G+C indicates a region which has been acquired recently (in evolutionary time) from a non-related source (i.e., horizontal gene transfer) <ns0:ref type='bibr' target='#b73'>(73)</ns0:ref>. To investigate whether variant G+C confounds genome reconstruction, we compared the %G+C of NRs to that of CRs and the complete genome.</ns0:p><ns0:p>The genomes in this study had a range of %G+C values, from 42% (A. marincola HL-49)</ns0:p><ns0:p>to 68% (Erythrobacteraceae bacterium HL-111), with most skewing toward the higher values (Table <ns0:ref type='table' target='#tab_3'>2</ns0:ref>). We determined the %G+C for each CR and NR ≥200 bp in length and compared them to the %G+C for the complete genome. For genomes with more than one genomic element, each molecule was considered separately since extrachromosomal elements may have distinct nucleotide composition. For seven of the genomes, the %G+C for the NRs differed significantly (p≤0.005) from the genome average, while the CRs generally reflected the genome average ( Manuscript to be reviewed %G+C profiles (Figure <ns0:ref type='figure' target='#fig_8'>1</ns0:ref>). There was a slight bias toward higher %G+C for the NRs and lower %G+C in the CRs, which could reflect a bias in the assembly algorithm.</ns0:p><ns0:p>Tetranucleotide frequency (TNF) has been shown to be capable of distinguishing higher taxonomic classifications, up to species <ns0:ref type='bibr' target='#b54'>(54,</ns0:ref><ns0:ref type='bibr' target='#b68'>68)</ns0:ref>. This resolving power has been leveraged in binning protocols <ns0:ref type='bibr' target='#b15'>(15,</ns0:ref><ns0:ref type='bibr' target='#b74'>(74)</ns0:ref><ns0:ref type='bibr' target='#b75'>(75)</ns0:ref><ns0:ref type='bibr' target='#b76'>(76)</ns0:ref>. To investigate whether genomic regions with divergent TNF are poorly recovered in genome reconstruction, we compared the TNFs of CRs and NRs to that of the cognate complete genome using chi-squared analysis. In most cases, the chi-squared statistic was an order of magnitude higher for NRs versus CRs, and the differences were significant for all chromosomal sequences except for HL-46, HL-109, HL-93 and the small chromosome of HL-91 (Table <ns0:ref type='table' target='#tab_7'>3</ns0:ref>).</ns0:p><ns0:p>One factor that could affect nucleotide composition effects on binning is the length of the region with divergent composition versus the length of the contig. If the variant region comprises most of the length of the contig being evaluated, the difference from the genome average will be pronounced, whereas if the divergent region is only a small percentage of the contig length, the signal will be muted. An examination of CR/NR length versus compositional variance (Fig. <ns0:ref type='figure'>S3</ns0:ref>) revealed a strong, significant negative correlation between contig length and TNF chi square for CRs (R 2 =0.64, p-value<2.2x10 -16 ) and a weaker relationship for NRs (R 2 =0.14, p-value=4.9x10 -12 ). Taken together, the %G+C and TNF results show that genomic regions with divergent nucleotide composition are more likely to be missed during binning, and this effect is stronger for short contigs. The most effective way to overcome this problem is to enhance assembly such that regions with unusual content are included in significantly longer contigs, or, through clone linkage, identify strong, unique connections to binned contigs.</ns0:p><ns0:p>PeerJ reviewing PDF | (2017:03:17041:1:1:NEW 4 Sep 2020)</ns0:p><ns0:p>Manuscript to be reviewed</ns0:p></ns0:div>
<ns0:div><ns0:head>Repeated sequences segregate aberrantly</ns0:head><ns0:p>Sequence coverage profiles are frequently effective in discriminating contigs from different organisms <ns0:ref type='bibr' target='#b15'>(15)</ns0:ref>. Samples taken under different conditions or at different times capture community states which have similar organismal composition but differing relative abundances.</ns0:p><ns0:p>This difference translates to distinct coverage profiles for assembled contigs, and thus contigs with similar coverage profiles are assumed to originate from the same organism. In this data set, for example, we compared two cultures with near-identical heterotroph species composition, but different cyanobacteria acting as a conduit for energy and carbon <ns0:ref type='bibr' target='#b46'>(46,</ns0:ref><ns0:ref type='bibr' target='#b47'>47)</ns0:ref>. Other studies have compared samples taken at different times <ns0:ref type='bibr' target='#b75'>(75)</ns0:ref>. Coverage analysis is more difficult for repeated regions of a genome, which will yield higher coverage values than the genome average and thus are more likely to be either not binned or binned improperly. Differential coverage analysis can mitigate this problem by identifying correlated changes in abundance of contigs with different coverage. Unlike nucleotide composition variance, however, unusual high-coverage signal due to repeat sequence is less likely to be diluted by incorporation into a larger contig because assemblers (especially standard de Bruijn graph assemblers using short-read data) tend to terminate contigs when repeats are encountered and/or assemble repeats into separate contigs <ns0:ref type='bibr' target='#b77'>(77)</ns0:ref>.</ns0:p><ns0:p>To examine the impact of repeated sequences on genome reconstruction, we determined the repetitiveness of sequence information across CRs and NRs, determined from a self-versusself similarity search, and compared those values to the genome average. Correspondence of repeated regions and NRs was strong (Figures <ns0:ref type='figure' target='#fig_9'>2 and 3</ns0:ref>, Figure <ns0:ref type='figure'>S4</ns0:ref>). In HL-111, all NRs save one were present in at least two copies (Figure <ns0:ref type='figure' target='#fig_9'>2</ns0:ref>). For all reconstructions, save HL-46, the CRs had repeat content equal to or lower than the genome average. Manuscript to be reviewed</ns0:p><ns0:p>Another phenomenon that can affect contig coverage in metagenomic assembly is multiple organisms sharing identical regions of DNA. Some regions are highly conserved between related species, an example being the ribosomal RNA operon, which is known to confound assemblers and segregation strategies <ns0:ref type='bibr' target='#b78'>(78)</ns0:ref>. Alternatively, mobile elements such as plasmids or transposons can have a broad host range and invade and inhabit closely or even distantly related organisms <ns0:ref type='bibr' target='#b79'>(79)</ns0:ref>. Such regions, even if not repeated within a genome, will exhibit anomalous coverage and thus could be either excluded or mis-binned. We examined the metagenomic read coverage depth to determine if NRs had anomalous profiles relative to the whole genome and the CRs. For most reconstructions, the NRs' coverage differed from the genome average and that of the CRs ( <ns0:ref type='figure'>S4</ns0:ref>). A likely explanation for this is the presence in the consortia of sub-populations of these organisms that lack the plasmids.</ns0:p></ns0:div>
<ns0:div><ns0:head>Functional assessment of NR genes</ns0:head><ns0:p>To determine the extent to which regions missing from reconstructions might affect downstream metabolic or functional analyses and predictions for organisms and communities, we examined the gene content of the NRs and the functional roles of those genes. COG categorization was used as a basis for comparison because of its ability to identify, in particular, genes associated with mobile elements such as plasmids, phage and insertion sequences. In addition, we evaluated the distribution of non-coding RNA genes since some are known to be repeated within genomes Manuscript to be reviewed (multiple rRNA operons, for example), and others (tRNAs) are commonly associated with mobile elements <ns0:ref type='bibr' target='#b80'>(80)</ns0:ref>.</ns0:p><ns0:p>For all the reconstructions, the gene content of the NRs differed from that of the CRs and complete genomes. Functional analysis of gene sequences shows that this difference was largely driven by genes encoding mobile element functions (COG category X) and RNA genes (Figure <ns0:ref type='figure'>4</ns0:ref>). The mobile element genes in the NR regions were predominantly transposases with some contribution from bacteriophage and plasmid genes (HL-91; HL-93). Most of the identified rRNA genes fell within NRs, with only HL-48 and HL-53 each having one rRNA contained in a CR. In addition, the NRs, including the two entire plasmids from HL-91 which were not binned, contained a higher percentage of genes that were not assigned to a COG category.</ns0:p></ns0:div>
<ns0:div><ns0:head>Evaluation of a complex metagenomic data set and common automated binning tools</ns0:head><ns0:p>To verify that our conclusions of genome reconstruction bias in the highly curated UCC data set were extendable to more complex data sets and for alternate, widely-used binning tools, we applied similar analyses to MAGs generated from the Tara Oceans metagenomic data using distinct genome reconstruction protocols. For this comparison, 4,557 MAGs generated from the Tara Oceans microbial metagenomic data reconstructed using three complementary methods were collected and analyzed. Three different automated binning methodologies were employed to generate the MAG data set: MetaBat (v0.26.3) <ns0:ref type='bibr' target='#b22'>(22,</ns0:ref><ns0:ref type='bibr' target='#b49'>49)</ns0:ref>, BinSanity (v1.0) <ns0:ref type='bibr' target='#b21'>(21,</ns0:ref><ns0:ref type='bibr' target='#b51'>51)</ns0:ref>, and CONCOCT (with manual refinement in anvi'o) <ns0:ref type='bibr' target='#b31'>(31,</ns0:ref><ns0:ref type='bibr' target='#b50'>50)</ns0:ref>. All three automated binning algorithms utilized read coverage and TNF to identify congruent contigs, with the intended role of the algorithms to reconstruct high confidence environmental genomes while avoiding over-binning (i.e., removing elements that deviate from the mean values of the binned contigs). The MAGs <ns0:ref type='bibr' target='#b21'>(21,</ns0:ref><ns0:ref type='bibr' target='#b50'>50)</ns0:ref>. The MAGs also had lower variance with regards to TNF compared to the RefSeq genomes (p < 0.001) (Figure <ns0:ref type='figure' target='#fig_11'>5B</ns0:ref>), again, likely due to genomic elements that deviated from the average value of the binned contigs having been removed during the binning steps. These observations support our conclusions regarding genome regions having divergent nucleotide composition being underrepresented in MAGs.</ns0:p><ns0:p>The Tara and NCBI Refseq data sets were then evaluated for repeat sequence content.</ns0:p><ns0:p>Each MAG and isolate genome was compared to itself using NUCmer to identify the fraction of the genome composed of repeat regions (regions with 97% sequence identity). MAGs ≥ universally had a smaller fraction of genomic information in repeat regions compared to isolate genomes (p < 0.01; Figure <ns0:ref type='figure' target='#fig_12'>6</ns0:ref>). The lack of repeat regions in MAGs is likely the result of repeated regions having inflated or depressed read coverage values relative to the mean of the genome, depending on the number of copies of the repeat region present in the genome and how stable </ns0:p></ns0:div>
<ns0:div><ns0:head>What's missing from reconstructed genomes?</ns0:head><ns0:p>Analysis of regions that were not recovered from genome reconstruction (NRs) showed both nucleotide compositional variance and intragenome repetitiveness. The %G+C and tetranucleotide frequencies of NRs tended to differ from that of complete genomes (Tables <ns0:ref type='table' target='#tab_7'>2 and 3</ns0:ref>, Figure <ns0:ref type='figure' target='#fig_8'>1</ns0:ref>), and the sequence coverage differed. This met expectations since, in general, binning tools are designed around the assumption that sequences with similar properties belong together, thus any genome region that varies significantly from the genome average is likely be incorrectly binned if it comprises the majority of a contig under consideration. Regions with atypical nucleotide content have been observed to contain genes upon which selective pressures are acting on nucleic acid structure, such as ribosomal RNAs and tRNAs <ns0:ref type='bibr' target='#b72'>(72,</ns0:ref><ns0:ref type='bibr' target='#b81'>81,</ns0:ref><ns0:ref type='bibr' target='#b82'>82)</ns0:ref>, and exogenously introduced segments such as mobile elements <ns0:ref type='bibr' target='#b83'>(83,</ns0:ref><ns0:ref type='bibr' target='#b84'>84)</ns0:ref>. It is significant that many of the NRs displayed lower %G+C than the genome average, since it has been observed that laterally acquired regions tend to have lower %G+C than their hosts <ns0:ref type='bibr' target='#b83'>(83)</ns0:ref>, as phage and insertion sequences tend to have A+T-enriched genomes <ns0:ref type='bibr' target='#b85'>(85)</ns0:ref>. Notably, many genome regions with variant nucleotide composition were incorporated into longer contigs by the assembler, masking the variance and allowing correct binning. Conversely, the assembler collapsed repeated region sequences into single contigs, and thus they were not binned due to the inflated sequence coverage values. Often, repeated sequences displayed divergent nucleotide composition, but the reciprocal was less frequent, indicating that repetitiveness is the stronger driver of binning failure. These results demonstrate that assembly efficiency is an important determining factor for correct binning, or conversely, any factor that results in shorter assemblies will result in poorer recovery of anomalous regions. Thus, it is advisable to include replication and positive controls in metagenomic sequencing protocols, particularly for highly diverse communities such as soils and riverbed sediments, to allow evaluation of assembly efficiency and accuracy.</ns0:p><ns0:p>Repeat regions identified in this study appeared to largely consist of insertion elements based on functional analysis and their relatively short size (1-2 kb). Failure of these regions to be correctly binned is unlikely to meaningfully affect functional predictions for a reconstructed genome. Their presence in a genome is more likely to affect metabolic reconstruction analysis by reducing assembly efficiency, resulting in more, shorter contigs and increasing the chance that these shorter contigs are not binned or incorrectly binned. Technological advances increasing read length beyond 2 kb will increase contig lengths, binning accuracy, and the likelihood of yielding closed genomes from environmental samples <ns0:ref type='bibr' target='#b7'>(8,</ns0:ref><ns0:ref type='bibr' target='#b86'>86,</ns0:ref><ns0:ref type='bibr' target='#b87'>87)</ns0:ref>.</ns0:p><ns0:p>NRs were generally observed to be short, with a median length of less than 5 kb (Table <ns0:ref type='table'>1</ns0:ref>) and containing only a handful of genes. Thus, even a MAG with many gaps (indicating a large number of NRs) may be missing only a small percentage of its genome. The conserved singlecopy gene (CSCG) estimations for completeness appear for all intents and purposes to be a reasonable indication of how much information is absent <ns0:ref type='bibr' target='#b47'>(47)</ns0:ref>. One caveat to this conclusion, however, is that extrachromosomal elements, plasmids and phages (integrated or otherwise) typically do not carry CSCG markers, and thus are essentially invisible in such analyses. The longer NRs observed in our analysis appear to comprise integrated plasmids or phage, and thus any gap in a reconstruction could represent up to 50 kb (or more) of genetic material.</ns0:p><ns0:p>Importantly, these represent introduced genetic material, which, while likely conveying a beneficial trait, are unlikely to carry functions that are integral to host metabolic function.</ns0:p></ns0:div>
<ns0:div><ns0:head>CONCLUSIONS</ns0:head><ns0:p>This analysis indicates that reconstructed genomes estimated to be near-complete can be assumed to contain nearly all genes important to metabolic reconstruction. The majority of identifiable genes present on NRs appear to be either highly conserved, non-coding genes that can be assumed to be present (such as the rRNA genes and tRNA genes) or are associated with mobile genetic elements. While many of these genes may be not be directly related to cellular metabolism (transposases, toxin/antitoxin systems, phage and plasmid functions), it should be noted that entire extrachromosomal elements may be missed by the binning process due to either alternate nucleotide composition, a higher number of copies per cell than the genome, or occupancy in only a subset of the population (such as the two molecules in HL-109). These elements frequently carry genes that alter the physiology or resistance of the host organism. For example, HL-109 and HL-111 have NRs that includes genes involved in glycan biosynthesis, suggesting alterations to the cell wall, while HL-91 has picked up a multidrug efflux transporter.</ns0:p><ns0:p>As such, reconstructed genomes can be considered reliable foundations for metabolic reconstruction but should not be assumed to be comprehensive for the function of the organism. </ns0:p><ns0:note type='other'>Figure 4</ns0:note><ns0:p>Functional categorization of genes present on MDRs. </ns0:p></ns0:div><ns0:figure xml:id='fig_0'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2017:03:17041:1:1:NEW 4 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_1'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2017:03:17041:1:1:NEW 4 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_2'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2017:03:17041:1:1:NEW 4 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_3'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2017:03:17041:1:1:NEW 4 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_4'><ns0:head /><ns0:label /><ns0:figDesc>cannot culture the component populations either because the proper growth conditions are unknown or difficult to replicate in a laboratory environment, or simply because there are too many organisms present to have the resources or time to pursue the effort. Because of this, there are very few examples of sequenced organisms isolated from the same sample from which metagenomic sequencing and binning has been done to generate MAGs. As such, a 'gold standard' for evaluation of MAG content has been difficult to come by. We have taken advantage of two enrichment cultures from which MAGs and isolate genomes have been derivedto generate just such a 'gold standard' comparison framework. We have previously generated two unicyanobacterial consortial cultures (UCC) -enrichment cultures each containing a distinct cyanobacterial population and different, yet overlapping, communities of associated heterotrophs, each numbering <20 species -and performed metagenomic sequencing, assembly and binning.<ns0:ref type='bibr' target='#b47'>(47,</ns0:ref><ns0:ref type='bibr' target='#b48'>48)</ns0:ref>. Illumina 150 bp paired-end reads were generated from each community, and IDBA_ud was used to assemble the read sets separately and in co-assembly. The abundances of the organisms differed between the two communities, allowing us to bin the sequences by comparing sequence coverage values of contigs between the two UCCs in a predominantly manual process (inspired by the work of Dick, et al<ns0:ref type='bibr' target='#b68'>(68)</ns0:ref>). The resulting MAGs were manually curated to eliminate contaminating contigs and identify mis-binned contigs, correctly placing them when possible. In parallel, ten organisms were isolated from the UCCs and completelyPeerJ reviewing PDF | (2017:03:17041:1:1:NEW 4 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_5'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2017:03:17041:1:1:NEW 4 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_6'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2017:03:17041:1:1:NEW 4 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_7'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2017:03:17041:1:1:NEW 4 Sep 2020) Manuscript to be reviewed this number is across the population. Compared to the other Tara MAGs, the Tully et al. MAGs had a larger fraction of redundant genomic elements. It is unclear what aspect of the assembly and binning methodology has influenced these results. On average, the lengths of the repeat regions from the Tully et al. MAGs are longer than the repeat regions in the RefSeq genomes (mean: 1,052bp vs 868bp, respectively).</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_8'><ns0:head>Figure 1 Distributions</ns0:head><ns0:label>1</ns0:label><ns0:figDesc>Figure 1</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_9'><ns0:head>Figure 2 Analysis</ns0:head><ns0:label>2</ns0:label><ns0:figDesc>Figure 2</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_10'><ns0:head /><ns0:label /><ns0:figDesc>COG categories: A -RNA processing and modification; B -Chromatic structure and dynamics; C -Energy production and conversion; D -Cell cycle control, cell division, chromosome partitioning; E -Amino acid transport and metabolism; F -Nucleotide transport and metabolism; G -Carbohydrate transport and metabolism; H -Coenzyme transport and metabolism; I -Lipid transport and metabolism; J -Translation, ribosomal structure and biogenesis; K -Transcription; L -DNA replication, recombination and repair; M -Cell wall/membrane/envelope biogenesis; N -Cell motility; O -Post-translational modification, protein turnover, chaperones; P -Inorganic ion transport and metabolism; Q -Secondary metabolites biosynthesis, transport and catabolism; R -General function prediction; S -Function unknown; T -Signal transduction mechanisms; U -Intracellular trafficking, secretion and vesicular transport; V -Defense mechanisms; W -Extracellular structures; X -Mobilome, transposons, phages; Y -Nuclear structure; Z -Cytoskeleton.</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_11'><ns0:head>Figure 5 Tara</ns0:head><ns0:label>5</ns0:label><ns0:figDesc>Figure 5</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_12'><ns0:head>Figure 6 Tara</ns0:head><ns0:label>6</ns0:label><ns0:figDesc>Figure 6</ns0:figDesc></ns0:figure>
<ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='39,42.52,70.87,525.00,402.75' type='bitmap' /></ns0:figure>
<ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='45,42.52,70.87,525.00,264.00' type='bitmap' /></ns0:figure>
<ns0:figure type='table' xml:id='tab_0'><ns0:head>Table</ns0:head><ns0:label /><ns0:figDesc /><ns0:table /></ns0:figure>
<ns0:figure type='table' xml:id='tab_2'><ns0:head /><ns0:label /><ns0:figDesc>. The metagenomic data used to construct the UCC MAGs is available from the NCBI</ns0:figDesc><ns0:table /><ns0:note>PeerJ reviewing PDF | (2017:03:17041:1:1:NEW 4 Sep 2020)</ns0:note></ns0:figure>
<ns0:figure type='table' xml:id='tab_3'><ns0:head>Table 2</ns0:head><ns0:label>2</ns0:label><ns0:figDesc /><ns0:table><ns0:row><ns0:cell>average %G+C values (47.5%, 56.0% and 64.1% respectively). Extrachromosomal elements</ns0:cell></ns0:row><ns0:row><ns0:cell>analyzed did not display a significant difference in the %G+C of their NRs from the molecule</ns0:cell></ns0:row><ns0:row><ns0:cell>average. As expected, the values for the NRs and CRs of HL-46 showed no significant difference</ns0:cell></ns0:row></ns0:table><ns0:note>). The %G+C averages for NRs from HL-48 and HL-111 were significantly lower (45.76% and 64.26%, respectively) than the genomes' averages (58.98% and 68.12% respectively). Other genomes (HL-53, HL-55, HL-109) had some NRs with %G+C higher than the genome average and some NRs with lower values (Figure1), despite having different from the genome average (Table 2), however, HL-46's CRs and NRs did not display identical PeerJ reviewing PDF | (2017:03:17041:1:1:NEW 4 Sep 2020)</ns0:note></ns0:figure>
<ns0:figure type='table' xml:id='tab_4'><ns0:head>Table 4</ns0:head><ns0:label>4</ns0:label><ns0:figDesc /><ns0:table /><ns0:note>, Fig 2, Fig S4). Only HL-46 and one of the HL-109 molecules did not have significant differences. Most NRs displayed higher or equivalent coverage values, however, several NRs in HL-48 and the two small plasmids associated with HL-91 showed lower metagenomic coverage values (Figure</ns0:note></ns0:figure>
<ns0:figure type='table' xml:id='tab_5'><ns0:head /><ns0:label /><ns0:figDesc>Our results above predicted that the MAGs would have lower %G+C variance and TNF variance than the isolate complete genome data set. For the observed %G+C, MAGs tended to have lower variance (p < 0.001) than isolate genomes (Figure5A). The exception was the Parks et al. MAGs, which had a much larger variance, even compared to the RefSeq genome set (mean vs mean, p < 0.001). This may be the result of the additional step applied to the MAGs by Parks</ns0:figDesc><ns0:table /><ns0:note>PeerJ reviewing PDF | (2017:03:17041:1:1:NEW 4 Sep 2020) Manuscript to be reviewed had a mean estimated completeness and contamination of 76.6% and 2.2%, respectively, as determined by CheckM v.1.1.1 (32). In comparison, 1,736 'representative' and 'reference' complete genomes were collected from NCBI RefSeq. et al., whereby related MAGs with <3% mean %G+C difference were merged into a single representative MAG (22). For the Tully et al. and Delmont et al. MAGs, the lower variance observed compared to the RefSeq genomes is likely due to removal of contigs with deviant %G+C values during binning</ns0:note></ns0:figure>
<ns0:figure type='table' xml:id='tab_6'><ns0:head>Table 2 . Comparison of %G+C for genomes, CRs and NRs</ns0:head><ns0:label>2</ns0:label><ns0:figDesc /><ns0:table><ns0:row><ns0:cell /><ns0:cell /><ns0:cell>Genome</ns0:cell><ns0:cell /><ns0:cell>CRs</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell>NRs</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell /><ns0:cell>molecule</ns0:cell><ns0:cell>mean</ns0:cell><ns0:cell cols='2'>mean distance</ns0:cell><ns0:cell>p-value</ns0:cell><ns0:cell>mean</ns0:cell><ns0:cell>distance</ns0:cell><ns0:cell>p-value</ns0:cell></ns0:row><ns0:row><ns0:cell>HL-46</ns0:cell><ns0:cell>EI34DRAFT_7210</ns0:cell><ns0:cell cols='3'>64.42 63.96±1.94 1.55±1.25</ns0:cell><ns0:cell>0.997</ns0:cell><ns0:cell>65.12±2.13</ns0:cell><ns0:cell>1.61±1.56</ns0:cell><ns0:cell>0..263</ns0:cell></ns0:row><ns0:row><ns0:cell>HL-46</ns0:cell><ns0:cell>EI34DRAFT_6181</ns0:cell><ns0:cell cols='3'>59.94 60.78±2.27 1.97±1.41</ns0:cell><ns0:cell>0.856</ns0:cell><ns0:cell>60.97±1.78</ns0:cell><ns0:cell>1.92±0.75</ns0:cell><ns0:cell>0.605</ns0:cell></ns0:row><ns0:row><ns0:cell>HL-48</ns0:cell><ns0:cell>CY41DRAFT</ns0:cell><ns0:cell cols='3'>58.98 59.00±1.52 1.01±1.13</ns0:cell><ns0:cell cols='4'>0.996 45.76±19.69 13.22±19.69 <0.001 a</ns0:cell></ns0:row><ns0:row><ns0:cell>HL-49</ns0:cell><ns0:cell>K302DRAFT</ns0:cell><ns0:cell cols='3'>42.22 42.24±1.71 1.15±1.27</ns0:cell><ns0:cell>0.434</ns0:cell><ns0:cell>42.73±3.37</ns0:cell><ns0:cell>2.44±2.38</ns0:cell><ns0:cell>0.001</ns0:cell></ns0:row><ns0:row><ns0:cell>HL-53</ns0:cell><ns0:cell>Ga0003345</ns0:cell><ns0:cell cols='3'>47.50 46.95±1.61 0.96±1.40</ns0:cell><ns0:cell>0.031</ns0:cell><ns0:cell>48.83±3.55</ns0:cell><ns0:cell>3.70±0.82</ns0:cell><ns0:cell><0.001</ns0:cell></ns0:row><ns0:row><ns0:cell>HL-55</ns0:cell><ns0:cell>K417DRAFT</ns0:cell><ns0:cell cols='3'>56.26 55.87±1.97 1.42±1.41</ns0:cell><ns0:cell>0.025</ns0:cell><ns0:cell>55.44±3.30</ns0:cell><ns0:cell>3.00±1.59</ns0:cell><ns0:cell>0.001</ns0:cell></ns0:row><ns0:row><ns0:cell>HL-58</ns0:cell><ns0:cell>CD01DRAFT</ns0:cell><ns0:cell cols='3'>57.56 56.83±2.61 1.69±2.12</ns0:cell><ns0:cell>0.047</ns0:cell><ns0:cell>56.11±3.69</ns0:cell><ns0:cell>3.93±0.51</ns0:cell><ns0:cell>0.016</ns0:cell></ns0:row><ns0:row><ns0:cell>HL-91</ns0:cell><ns0:cell>Ga0058931_11</ns0:cell><ns0:cell cols='3'>61.75 62.05±0.25 0.31±0.23</ns0:cell><ns0:cell>0.954</ns0:cell><ns0:cell>60.39±3.17</ns0:cell><ns0:cell>2.79±2.02</ns0:cell><ns0:cell>0.053</ns0:cell></ns0:row><ns0:row><ns0:cell>HL-91</ns0:cell><ns0:cell>Ga0058931_12</ns0:cell><ns0:cell>60.37</ns0:cell><ns0:cell>nd b</ns0:cell><ns0:cell>nd</ns0:cell><ns0:cell>nd</ns0:cell><ns0:cell>nd</ns0:cell><ns0:cell>nd</ns0:cell><ns0:cell>nd</ns0:cell></ns0:row><ns0:row><ns0:cell>HL-91</ns0:cell><ns0:cell>Ga0058931_13</ns0:cell><ns0:cell>61.77</ns0:cell><ns0:cell>nd</ns0:cell><ns0:cell>nd</ns0:cell><ns0:cell>nd</ns0:cell><ns0:cell>nd</ns0:cell><ns0:cell>nd</ns0:cell><ns0:cell>nd</ns0:cell></ns0:row><ns0:row><ns0:cell>HL-91</ns0:cell><ns0:cell>Ga0058931_14</ns0:cell><ns0:cell cols='3'>61.84 60.99±1.90 1.33±1.60</ns0:cell><ns0:cell>0.030</ns0:cell><ns0:cell>59.11±2.96</ns0:cell><ns0:cell>3.52±1.96</ns0:cell><ns0:cell>0.005</ns0:cell></ns0:row><ns0:row><ns0:cell>HL-93</ns0:cell><ns0:cell>Ga0071314_11</ns0:cell><ns0:cell cols='3'>55.88 56.75±2.20 1.75±1.59</ns0:cell><ns0:cell>1.000</ns0:cell><ns0:cell>56.08±4.42</ns0:cell><ns0:cell>3.6±2.57</ns0:cell><ns0:cell><0.001</ns0:cell></ns0:row><ns0:row><ns0:cell cols='2'>HL-109 Ga0071312_11</ns0:cell><ns0:cell cols='3'>64.09 64.55±1.46 1.12±1.05</ns0:cell><ns0:cell>0.715</ns0:cell><ns0:cell>60.96±3.02</ns0:cell><ns0:cell>3.28±2.85</ns0:cell><ns0:cell>0.073</ns0:cell></ns0:row><ns0:row><ns0:cell cols='2'>HL-109 Ga0071312_12</ns0:cell><ns0:cell cols='3'>64.07 63.89±1.41 0.92±1.09</ns0:cell><ns0:cell>0.169</ns0:cell><ns0:cell>63.11±2.21</ns0:cell><ns0:cell>1.94±1.43</ns0:cell><ns0:cell>0.593</ns0:cell></ns0:row><ns0:row><ns0:cell cols='2'>HL-109 Ga0071312_13</ns0:cell><ns0:cell cols='3'>65.34 65.47±0.07 0.13±0.07</ns0:cell><ns0:cell>0.778</ns0:cell><ns0:cell>61.68±2.24</ns0:cell><ns0:cell>3.66±2.24</ns0:cell><ns0:cell>0.009</ns0:cell></ns0:row><ns0:row><ns0:cell cols='2'>HL-111 Ga0071316_11</ns0:cell><ns0:cell cols='3'>68.12 68.20±1.44 0.99±1.05</ns0:cell><ns0:cell>0.465</ns0:cell><ns0:cell>64.26±1.39</ns0:cell><ns0:cell>3.86±1.39</ns0:cell><ns0:cell><0.001</ns0:cell></ns0:row></ns0:table><ns0:note>a Bold type indicates significant results (P 0.005).≤ b Not determined because the entire molecule was missing from the reconstructed genome.</ns0:note></ns0:figure>
<ns0:figure type='table' xml:id='tab_7'><ns0:head>Table 3 . Tetranucleotide frequency  2 analysis.</ns0:head><ns0:label>3</ns0:label><ns0:figDesc /><ns0:table><ns0:row><ns0:cell /><ns0:cell /><ns0:cell>CR</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell>NR</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>molecule</ns0:cell><ns0:cell>mean</ns0:cell><ns0:cell>sd</ns0:cell><ns0:cell cols='2'>p-value mean</ns0:cell><ns0:cell>sd</ns0:cell><ns0:cell>p-value</ns0:cell></ns0:row><ns0:row><ns0:cell cols='3'>HL-46 EI34DRAFT_6181 0.2323 0.1883</ns0:cell><ns0:cell cols='3'>0.154 0.1518 0.1429</ns0:cell><ns0:cell>0.983</ns0:cell></ns0:row><ns0:row><ns0:cell cols='3'>HL-46 EI34DRAFT_7210 0.2042 0.0696</ns0:cell><ns0:cell cols='3'>0.896 0.1701 0.1332</ns0:cell><ns0:cell>0.975</ns0:cell></ns0:row><ns0:row><ns0:cell>HL-48 CY41DRAFT</ns0:cell><ns0:cell cols='2'>0.0276 0.0577</ns0:cell><ns0:cell cols='4'>0.387 0.4425 0.2689 <0.001 a</ns0:cell></ns0:row><ns0:row><ns0:cell>HL-49 K302DRAFT</ns0:cell><ns0:cell cols='2'>0.0522 0.0431</ns0:cell><ns0:cell cols='4'>0.757 0.2340 0.2164 <0.001</ns0:cell></ns0:row><ns0:row><ns0:cell>HL-53 Ga0003345</ns0:cell><ns0:cell cols='2'>0.0261 0.0451</ns0:cell><ns0:cell cols='4'>0.001 0.3851 0.1525 <0.001</ns0:cell></ns0:row><ns0:row><ns0:cell>HL-55 K417DRAFT</ns0:cell><ns0:cell cols='2'>0.0458 0.0726</ns0:cell><ns0:cell cols='3'>0.086 0.2774 0.2168</ns0:cell><ns0:cell>0.004</ns0:cell></ns0:row><ns0:row><ns0:cell>HL-58 CD01DRAFT</ns0:cell><ns0:cell cols='2'>0.0761 0.1451</ns0:cell><ns0:cell cols='3'>0.008 0.2974 0.969</ns0:cell><ns0:cell>0.004</ns0:cell></ns0:row><ns0:row><ns0:cell>HL-91 Ga0058931_11</ns0:cell><ns0:cell cols='2'>0.0266 0.0213</ns0:cell><ns0:cell cols='3'>0.313 0.3043 0.1416</ns0:cell><ns0:cell>0.011</ns0:cell></ns0:row><ns0:row><ns0:cell>HL-91 Ga0058931_12</ns0:cell><ns0:cell>nd b</ns0:cell><ns0:cell>nd</ns0:cell><ns0:cell>nd</ns0:cell><ns0:cell>nd</ns0:cell><ns0:cell>nd</ns0:cell><ns0:cell>nd</ns0:cell></ns0:row><ns0:row><ns0:cell>HL-91 Ga0058931_13</ns0:cell><ns0:cell>nd</ns0:cell><ns0:cell>nd</ns0:cell><ns0:cell>nd</ns0:cell><ns0:cell>nd</ns0:cell><ns0:cell>nd</ns0:cell><ns0:cell>nd</ns0:cell></ns0:row><ns0:row><ns0:cell>HL-91 Ga0058931_14</ns0:cell><ns0:cell cols='2'>0.0557 0.0647</ns0:cell><ns0:cell cols='4'>0.004 0.3614 0.2052 <0.001</ns0:cell></ns0:row><ns0:row><ns0:cell>HL-93 Ga0071314_11</ns0:cell><ns0:cell cols='2'>0.0925 0.0738</ns0:cell><ns0:cell cols='3'>0.993 0.2254 0.1595</ns0:cell><ns0:cell>0.062</ns0:cell></ns0:row><ns0:row><ns0:cell>HL-109 Ga0071312_11</ns0:cell><ns0:cell cols='2'>0.0262 0.0401</ns0:cell><ns0:cell cols='3'>0.396 0.3148 0.1842</ns0:cell><ns0:cell>0.087</ns0:cell></ns0:row><ns0:row><ns0:cell>HL-109 Ga0071312_12</ns0:cell><ns0:cell cols='2'>0.0216 0.0281</ns0:cell><ns0:cell cols='3'>0.076 0.2907 0.1913</ns0:cell><ns0:cell>0.231</ns0:cell></ns0:row><ns0:row><ns0:cell>HL-109 Ga0071312_13</ns0:cell><ns0:cell cols='2'>0.0048 0.0019</ns0:cell><ns0:cell cols='3'>0.538 0.3651 0.2299</ns0:cell><ns0:cell>0.016</ns0:cell></ns0:row><ns0:row><ns0:cell>HL-111 Ga0071316_11</ns0:cell><ns0:cell cols='2'>0.0396 0.0561</ns0:cell><ns0:cell cols='4'>0.322 0.4504 0.1640 <0.001</ns0:cell></ns0:row></ns0:table><ns0:note>a Bold text indicates significant result b Not determined because the entire molecule was missing from the reconstructed genome.</ns0:note></ns0:figure>
<ns0:figure type='table' xml:id='tab_8'><ns0:head>Table 4 . Metagenomic redundancy.</ns0:head><ns0:label>4</ns0:label><ns0:figDesc /><ns0:table><ns0:row><ns0:cell /><ns0:cell /><ns0:cell>Genome</ns0:cell><ns0:cell /><ns0:cell>CR</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell>NR</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell /><ns0:cell>molecule</ns0:cell><ns0:cell>mean</ns0:cell><ns0:cell>mean</ns0:cell><ns0:cell>distance</ns0:cell><ns0:cell>p-value</ns0:cell><ns0:cell>mean</ns0:cell><ns0:cell>distance</ns0:cell><ns0:cell>p-value</ns0:cell></ns0:row><ns0:row><ns0:cell>HL-46</ns0:cell><ns0:cell>EI34DRAFT_6181</ns0:cell><ns0:cell>2.76</ns0:cell><ns0:cell>2.78</ns0:cell><ns0:cell>0.26±0.15</ns0:cell><ns0:cell>0.992</ns0:cell><ns0:cell>2.42</ns0:cell><ns0:cell>0.43±0.31</ns0:cell><ns0:cell>0.264</ns0:cell></ns0:row><ns0:row><ns0:cell>HL-46</ns0:cell><ns0:cell>EI34DRAFT_7210</ns0:cell><ns0:cell>5.98</ns0:cell><ns0:cell>4.43</ns0:cell><ns0:cell>2.95±2.34</ns0:cell><ns0:cell>0.978</ns0:cell><ns0:cell>4.99</ns0:cell><ns0:cell>4.01±13.35</ns0:cell><ns0:cell>0.649</ns0:cell></ns0:row><ns0:row><ns0:cell>HL-48</ns0:cell><ns0:cell>CY41DRAFT</ns0:cell><ns0:cell>72.40</ns0:cell><ns0:cell>69.29</ns0:cell><ns0:cell>3.65±2.45</ns0:cell><ns0:cell>1.000</ns0:cell><ns0:cell cols='3'>140.93 100.97±153.33 <0.001 a</ns0:cell></ns0:row><ns0:row><ns0:cell>HL-49</ns0:cell><ns0:cell>K302DRAFT</ns0:cell><ns0:cell>8.97</ns0:cell><ns0:cell>8.67</ns0:cell><ns0:cell>0.51±0.58</ns0:cell><ns0:cell>0.999</ns0:cell><ns0:cell>11.38</ns0:cell><ns0:cell>4.16±17.52</ns0:cell><ns0:cell>0.002</ns0:cell></ns0:row><ns0:row><ns0:cell>HL-53</ns0:cell><ns0:cell>Ga0003345</ns0:cell><ns0:cell>441.81</ns0:cell><ns0:cell cols='2'>446.29 24.51±18.21</ns0:cell><ns0:cell>0.073</ns0:cell><ns0:cell cols='2'>517.07 115.56±59.71</ns0:cell><ns0:cell><0.001</ns0:cell></ns0:row><ns0:row><ns0:cell>HL-55</ns0:cell><ns0:cell>K417DRAFT</ns0:cell><ns0:cell>16.76</ns0:cell><ns0:cell>15.35</ns0:cell><ns0:cell>7.06±10.45</ns0:cell><ns0:cell>0.679</ns0:cell><ns0:cell cols='2'>117.37 110.35±333.81</ns0:cell><ns0:cell><0.001</ns0:cell></ns0:row><ns0:row><ns0:cell>HL-58</ns0:cell><ns0:cell>CD01DRAFT</ns0:cell><ns0:cell>128.28</ns0:cell><ns0:cell>127.85</ns0:cell><ns0:cell>9.10±15.71</ns0:cell><ns0:cell>1.000</ns0:cell><ns0:cell>180.14</ns0:cell><ns0:cell>60.44±27.54</ns0:cell><ns0:cell><0.001</ns0:cell></ns0:row><ns0:row><ns0:cell>HL-91</ns0:cell><ns0:cell>Ga0058931_11</ns0:cell><ns0:cell>231.39</ns0:cell><ns0:cell>228.46</ns0:cell><ns0:cell>3.64±2.25</ns0:cell><ns0:cell>0.786</ns0:cell><ns0:cell>311.6</ns0:cell><ns0:cell>91.27±97.44</ns0:cell><ns0:cell>0.001</ns0:cell></ns0:row><ns0:row><ns0:cell>HL-91</ns0:cell><ns0:cell>Ga0058931_12</ns0:cell><ns0:cell>163.24</ns0:cell><ns0:cell>nd b</ns0:cell><ns0:cell>nd</ns0:cell><ns0:cell>nd</ns0:cell><ns0:cell>nd</ns0:cell><ns0:cell>nd</ns0:cell><ns0:cell>nd</ns0:cell></ns0:row><ns0:row><ns0:cell>HL-91</ns0:cell><ns0:cell>Ga0058931_13</ns0:cell><ns0:cell>168.27</ns0:cell><ns0:cell>nd</ns0:cell><ns0:cell>nd</ns0:cell><ns0:cell>nd</ns0:cell><ns0:cell>nd</ns0:cell><ns0:cell>nd</ns0:cell><ns0:cell>nd</ns0:cell></ns0:row><ns0:row><ns0:cell>HL-91</ns0:cell><ns0:cell>Ga0058931_14</ns0:cell><ns0:cell>227.56</ns0:cell><ns0:cell>231.77</ns0:cell><ns0:cell>8.18±6.59</ns0:cell><ns0:cell>0.220</ns0:cell><ns0:cell cols='2'>273.03 97.82±117.47</ns0:cell><ns0:cell><0.001</ns0:cell></ns0:row><ns0:row><ns0:cell>HL-93</ns0:cell><ns0:cell>Ga0071314_11</ns0:cell><ns0:cell>50.87</ns0:cell><ns0:cell>50.03</ns0:cell><ns0:cell>4.04±2.92</ns0:cell><ns0:cell>1.000</ns0:cell><ns0:cell>65.73</ns0:cell><ns0:cell>16.16±35.87</ns0:cell><ns0:cell><0.001</ns0:cell></ns0:row><ns0:row><ns0:cell cols='2'>HL-109 Ga0071312_11</ns0:cell><ns0:cell cols='3'>3103.11 3098.73 97.47±72.15</ns0:cell><ns0:cell cols='3'>0.748 3072.59 323.24±240.86</ns0:cell><ns0:cell>0.005</ns0:cell></ns0:row><ns0:row><ns0:cell cols='2'>HL-109 Ga0071312_12</ns0:cell><ns0:cell cols='3'>2821.18 2822.26 113.08±78.18</ns0:cell><ns0:cell cols='3'>0.124 2778.03 352.81±436.28</ns0:cell><ns0:cell>0.003</ns0:cell></ns0:row><ns0:row><ns0:cell cols='2'>HL-109 Ga0071312_13</ns0:cell><ns0:cell cols='2'>2853.84 2901.40</ns0:cell><ns0:cell>47.56±9.73</ns0:cell><ns0:cell cols='3'>0.179 2097.01 756.83±256.91</ns0:cell><ns0:cell>0.018</ns0:cell></ns0:row><ns0:row><ns0:cell cols='2'>HL-111 Ga0071316_11</ns0:cell><ns0:cell>90.14</ns0:cell><ns0:cell>88.03</ns0:cell><ns0:cell>3.98±4.31</ns0:cell><ns0:cell>0.993</ns0:cell><ns0:cell cols='2'>98.25 38.42±104.87</ns0:cell><ns0:cell>0.027</ns0:cell></ns0:row></ns0:table><ns0:note>a Bold text indicates significant result b Not determined because the entire molecule was missing from the reconstructed genome.</ns0:note></ns0:figure>
<ns0:figure type='table' xml:id='tab_9'><ns0:head /><ns0:label /><ns0:figDesc>The gene features of each genome region were assigned to functional COG categories or as non-coding genes (rRNA; tRNA; ncRNA). Organisms' gene sets were compared using Principal Component Analysis. Organisms are represented by colors (HL-46, yellow; HL-48, purple; HL-49, blue; HL-53, light blue; HL-55, gray; HL-58, orange; HL-91, black; HL-93, pink; HL-109, red; HL-111, green). The genome region categories are represented by shapes (whole isolate genomes, circles; CDRs, squares; MDRs, triangles; extrachromosomal elements, diamonds).</ns0:figDesc><ns0:table /></ns0:figure>
<ns0:note place='foot'>PeerJ reviewing PDF | (2017:03:17041:1:1:NEW 4 Sep 2020)Manuscript to be reviewed</ns0:note>
</ns0:body>
" | "August 26, 2020
Dear Editor,
Enclosed is a revised draft of our manuscript “Biases in reconstruction of genomes from metagenomic data.” We have addressed the major criticisms of the reviewers – that the analysis was weak because it only evaluated one binning protocol and one simple community – by performing an additional analysis on an environmental metagenomic data set, the Tara Oceans data, which is a complex environmental metagenome and also has been analyzed by three different groups using different protocols, allowing a further evaluation of standard binning practices for evidences of the biases we identified in our model data set. We thank the reviewers for their time and the quality of their comments, and we thank the editor and journal for their patience in our turn-around time. We hope that the extensive edits made to the manuscript have improved it to meet the standards for publication.
Direct responses to the reviewer comments are interspersed below in blue text.
On behalf of all authors,
William C. Nelson, Ph.D.
william.nelson@pnnl.gov
Reviewer 1
Basic reporting
This manuscript described the biases in the genome reconstruction process from metagenomic data. The authors explored several factors that might affect the genome reconstruction by comparing their unicyanobacterial consortia paralleled genomic and metagenomic datasets. They reported that repeated sequences and genomic regions with variant nucleotide compositions are more likely to be missing, and that missing regions are strong biased toward rRNA, tRNA, and mobile elements.
Although most of the conclusions drawn from this study are not very surprising, the study itself is plausible in that it dig through the elements that might affect the genome reconstruction process. There are however still room to improve this study. I give my review as follows.
1. The dataset that the authors used for probing the problem is the unicyanobacterial consortia that they developed as model systems, which however represents just one system. This makes their conclusions a bit weaker since their results were only drawn from that very system. I would appreciate if the authors can extend their study to one or two more systems, be them simulated or real ones with genuine answers.
To address this issue, we have added an additional analysis of the Tara Ocean metagenome, which has been processed for MAGs by three different groups using distinct approaches. The results of this analysis are in agreement with our findings from the model system.
2. The microbial population complexity may also affect the genome reconstruction results. Looks like the unicyanobacterial consortia is a simple system since not too many genomes can be recovered and reconstructed. Perhaps the study can be extended to a more complex microbial community. The authors may also be able to say something about the relationships between microbial community complexity and the reconstruction process.
This suggestion is also addressed by the addition of the Tara Ocean dataset, as it derives from a complex microbial community.
3. I wonder when the authors term MDR (missed detection region), do they distinguish contigs that were not binned correctly between contigs that were filtered out due to short contig sizes? For example, 16S rRNA genes cannot usually be assembled very well and were usually split into pieces. Hence these fragmented were usually filtered out in the very first step of binning due to the contig length requirements of most binning tools. This also applies to most repetitive elements since they confuse assemblers. I guess the analysis of MDR is focused on those long enough to be included in the binning process and would very much appreciate if the short contigs can also be included in the analysis as well.
For clarity, we have changed our terminology from ‘missed detection region’ (MDR) to ‘not-binned region’ (NR). NRs were identified solely on the basis of their absence from the MAGs. It is difficult, however, to get reliable compositional data on short sequences. For this reason, we chose to only analyze the composition of NRs 1000 bp or longer. Since most binning processes tend to use at least 1000 bp as a minimum length for inclusion, this also meant that the NRs we were examining represented NRs that were missing from the MAG for a reason other than they were simply too short. For the genome and metagenome redundancy, all NRs were evaluated since the redundancy values were calculated on a per-base basis and then averaged across the length of the region.
4. (line 109) how come tetranucleotide frequencies need to be extracted across all six reading frames? Tetranucleotide need to be considered for only forward and reverse-complement strains. Please explain or revise this part.
The analysis has been repeated considering only forward and reverse, and the manuscript reports the updated data, which closely aligns with our previously reported figures.
5. In table 3 there is a “mean tetranucleotide frequency,” however I do not know what does that mean. Are all tetranuclotide frequencies for the CDR or MDR regions of organisms be calculated and averaged? Perhaps the authors can explain more in both table 3 and the text.
We apologize for the table being unclear. What is being reported is the mean and standard deviation of the chi-squared values, calculated across all CRs or NRs.
6. Two words are used interchangeably in this manuscript: repeat and redundancy. I however feel that the word “redundancy” is a bit misleading and was not used very frequently throughout genomic analysis. Alternatively the term “per-base coverage” may be more informative. Please consider using terms that are more widely used and comprehensible.
We are sensitive to the difficulty of the semantics related to this work. We agree that perhaps “redundancy” is not the optimal term to use and suggest in its place “repetitiveness”. We feel that, while “per-base coverage” is not an inaccurate description of what we calculated, describing repeat sequences as having deeper coverage than non-repeat sequence could be equally confusing to a reader.
7. Figure 2 is quite clear in delivering the message to readers. I wonder if the values such as TNF, GC%, or “redundancy” is derived by a sliding window?
The figure legend now states that “values were calculated across 2000 nt windows with a step size of 1000 nt.”
Experimental design
no comment
Validity of the findings
no comment
Comments for the author
no comment
Reviewer: Tom Delmont
Basic reporting
In the manuscript entitled ‘Biases in genome reconstruction from metagenomic data’, authors describe genomic regions associated with ten microbial populations that were missing in metagenome-assembled genomes, yet present in culture representatives. Authors found that the missing regions exhibit distinct traits compared to the rest of the genomes studies, and conclude that high-completion metagenome-assembled genomes are generally good representatives of the metabolic potential of the microbial populations they represent.
I note that links to the genomic and metagenomic raw reads used for mapping are not available and need to be added in the material and methods section.
We apologize for the oversight. Genbank accessions for the UCC MAGs, genomes and metagenomic data sets are now all listed in the Data and Code Availability section of the Materials and Methods.
Experimental design
The experimental design used in this manuscript is valid. Authors applied the same bioinformatics methodology to genomic regions present and missing in metagenome-assembled genomes (GC%, sequence composition, coverage, and functional potential). Methods are described with enough details.
I would like to mention one limit of the experimental design. There are many ways to characterize metagenome-assembled genomes, and it would have strengthened to study to test the effect of assembly (using different software) and binning (using both manual and automatic tools) on downstream results. Are the missing regions due to fundamental limitations of assembly and/or binning, or specific to the tools used to characterize the metagenome-assembled genomes? To me, this is the main limitation of the study. However, it does not impact the described observations. In my opinion, it merely limits the extent of the conclusions and does not prevent publication in its current form.
We agree with the reviewer that focusing on such a highly curated data set limits the applicability of our results. To address this issue, we have examined the Tara Oceans metagenomic data set, which is derived from a much more complex microbial community, and which has been separately processed by three groups using different protocols to generate MAG sets. This extension of our work demonstrates a practical presentation of the biases we observed in our model system.
Validity of the findings
The findings of this study are valid, and in line with the field’s understanding of assembly-based metagenomics.
Comments for the author
The overall aim of the study is to investigate the genomic regions of microbial populations systematically missing when using assembly-based metagenomics, as compared to cultivation. This is an important topic, as metagenome-assembled genomes largely contribute to our understanding of the microbial tree of life. Authors demonstrated that in the studied microbial community, nucleotide composition of genomic regions missing in reconstructed “genomes from metagenomes” frequently differ from the genome average. This contributes to our understanding of the limits of metagenomic assemblies, and/or binning. The manuscript is well written, and general trends appear to emerge from the analysis of 10 microbial populations.
The manuscript is of interest, and I could not see any flaw in the methodology. However, as far as I could see authors used only one metagenomic assembly software, and only one metagenomic binning workflow. It is unclear how variations in the bioinformatics workflow for assembly and binning impacts the genomic regions determined as missing. As a result, authors can only compare culture representatives with one single metagenomic workflow. I would appreciate an extensive answer of the authors regarding this matter. Would they consider expending their experimental design for a more comprehensive study?
We believe our addition of the Tara Ocean data addresses this critique.
Specific notes:
Introduction:
Ln 42-43: Key references of the pioneer publications supporting the sentence are missing. Please consider introducing a more relevant history of high-throughput sequencing in the context of culture-independent surveys.
We have supplemented the existing references with those of important work in the field.
Ln 58-59: Please consider reformulating the sentence, as single copy core genes are used to determine the completion, as well as the redundancy, of bins without the need for any reference genomes.
This section of the Introduction has been edited to more deftly relay our point that techniques such as single-copy gene analysis are estimations that cannot provide definitive answers regarding completeness and correctness. See lines 50-56 in the resubmission.
Ln 81-82: what workflow did the authors used, specifically? Many automatic, and a few manual-binning tools are available. They all provide different results.
The assembly and binning process for the MAGs was highly manual and is described in our prior publication, which is cited. This work was done prior to the release of any of the automated tools available today, although we did use an earlier version of MetaBAT as a quality check on our results. A brief summary of our process has been added to the manuscript (L 186-196)
M&M:
Please provide links to the genomic and metagenomic raw reads used for mapping. The study uses this data to assess coverage values and as a result it is important to make it easily available to the reader.
As noted above, the relevant accessions have been added to the Data and Code Availability section of the Materials and Methods.
Results and discussion:
Ln 197: does “save” corresponds to “except”? Please consider using a more commonly used term for an optimal reading experience.
The clarification has been made. (L 251)
Ln 207: please consider using a different term. Genome reconstruction could refer to the assembly, or the binning. A more specific term would be appreciated.
We have edited this to read “binning” (L 261)
Ln 221-223: This is incorrect. Duplicated regions will exhibit 2-fold coverage increase across all samples, and thus be clustered with the associated genome when binned using differential coverage. On the other hand, multi-copy plasmids for instance can create problematic situations, if the regulation of copy-numbers changes across samples. Please reformulate the sentence.
Thank you for pointing out this error. We have re-written the section (L 274-282) to emphasize how coverage issues differ from compositional issues.
Ln 229-230: sequence coverage and genome coverage of what? I understand the metagenomic coverage variations described later on, but not this one. Is it based on reads recruitment from the pure culture? Please explain.
The section has been edited to clarify that intragenome repetitiveness was determined by a self-search. (L 283-285)
Figure 3: legend does not match the figure. Was the wrong figure uploaded?
We regret the error. The correct figure has been attached to the submission.
Ln 296-298: Unclear. How adding controls not related to the studied environment (generally a black box) will help understand assembly accuracy?
Perhaps we don’t understand the comment, but our suggestion is not to add irrelevant controls, but just to encourage researchers to employ replication and sequencing controls to enable analysis of the correctness of their results.
Ln 304: Why 2000nt?
As is mentioned earlier in the paragraph, insertion sequence elements appear to be a major source of the repeats that break assemblies and assort into NR regions and they tend to be between 1kb and 2kb in length. Thus, if metagenome read length is extended to 2kb or longer (which is becoming ever more common with current long-read technology), significant improvements in genome reconstruction (from both an assembly and binning standpoint) should be achieved.
Ln 306: The “enhanced” binning results of this reference are contaminated due to a lack of proper curation step. Please consider using a better reference.
We thank the reviewer for the suggestion and have added additional relevant references.
Conclusion:
Ln 325: Please fix typo.
We did not identify a typo on that line, nor did our spell-checker or grammar-checker.
Reviewer: Johannes Dröge
Basic reporting
## Minor
* the central figures (Fig. 2, Fig. 3) need better annotation and a legend, otherwise the reader has to spend quite some to switching between the caption text and figure to understand them
Figures 2 and 3 have been altered to incorporate descriptive information.
l. 59-63: other comparisons to sequenced isolates and controlled simulation benchmarks with complex communities have been made to assess properties such as quality and completeness of reconstructed genomes. The authors should mention and cite some of them.
A brief description of simulation efforts has been added with references. (L57-69)
l. 72: explain axenic, this is a special term mostly unknown to non-biologist readers
In hindsight, the term is redundant with ‘isolate’, and thus has been eliminated.
l. 94-95: data set vs. dataset (should be consistent throughout the manuscript)
The manuscript has been edited to have a consistent and intentional use of the terms ‘data’ and ‘data set’. ‘Dataset’ has been eliminated.
l. 107: 'tetranucleotide frequency distance chi squared analysis': why not TNF chi square analysis/test and where is the distance? It's a little confusing because there is also a term called chi2 distance/divergence.
The text has been edited to read “tetranucleotide frequency (TNF) chi-square test” (L 121-122).
l. 108: typo in 'a custom perl scripts'
The typo has been corrected.
l. 113: 'absolute distance' == Euclidean distance? If needed, provide formula
The text has been edited to read “absolute difference”.
l. 119: explain shortly what 'per-base redundancy' means
This section has been edited to clarify how our analysis was performed. L 145-156
l. 121: what is the 'arithmetic distance' here? If needed, provide formula
This has been clarified in the edited section mentioned above.
Experimental design
## Major
l. 81-82: The genome reconstruction alias binning process is treated as a black box, but there are pronounced difference between different binning procedures (see e.g. http://biorxiv.org/content/early/2017/01/09/099127). The process described as 'current standard genome reconstruction techniques' is very nebulous and hides the complexity of genome reconstruction. In general, the details of the metagenome sequencing and in-silico processing (assembly, binning) are nowhere described although they represent the actual subject of the study. The choices made here have direct implications on the validity of the findings. Therefore, the authors must disclose which technology (sequencing platform, read lengths, insert size etc.) and algorithms (assemblers, contigs lengths, manual inspections etc., binning program, binning procedure, binning features) their results will relate to.
Assembly of the genomes and assembly and binning of the MAGs has previously been published and is cited. To address this criticism, we have added text briefly describing our process L 186-196
* Please provide scripts used in the calculations for the distances and p-values together with a minimal documentation and a usage/license statement.
The scripts are now available on github as described on L 114.
# Minor
* When the authors generate empirical null distributions for the p-value calculation using random draws (for instance l. 112,129, ), it is not entire clear to me what pool they sample from. Is this the same genome including MDRs and CDRs?
Bootstrapping datasets were drawn from the same genome as the CR/NR data sets being tested. This has been specified on L 130
Validity of the findings
## Major
In general, the overall findings confirm what is generally known for genome binning and add additional facts on the functional level. However, results are presented being universal although this extrapolation cannot be made without looking at different data and algorithms.
We feel that our results confirm what has generally been assumed, since there are no other instances of direct comparison of a MAG set to a collection of genomes all isolated together from an active natural community, which, although reduced in complexity, does contain various degrees of biodiversity. To address the different data and algorithms criticism, we have added an analysis of the Tara Ocean data, which has been assembled and binned by three different groups using different techniques. The results of this analysis demonstrate the types of bias we describe in our simplified system.
* It remains unresolved whether the missing regions are a result of an incomplete assembly or the binning. For both steps, there exists a multitude of different algorithms which lead to different output. In l. 151 the authors write that the assembly contained no errors, but it is not clear whether this also relates to missing regions. In l. 188 they write that that differing GC content in MDRs and CDRs are due to the assembly but unless verified, it could also be an artifact of the sequencing itself.
The reviewer is correct that our evaluation of assembly correctness was limited to the CRs. We have more closely examined the metagenome contig coverage of the NRs. Many of the short NRs (<500nt) have no contigs that map to them, reflecting the minimum contig size requested from the assembler. Most NRs longer than 500bp did have at least one contig mapping to them, with a strong positive correlation between NR length and number of contigs mapping. The coverage of the NR region by the contigs ranged from < 50% to over 400%, revealing that some NR sequence is repeat sequence and represented many times within the metagenomic contig set, and some NR sequence is missing entirely from the metagenomic contig set. There was evidence of indel differences between contigs and the NR regions, but no clear evidence of mis-assembly. This analysis is represented in the text and Supplemental Figures 1 and 2.
* In l. 212 and following the authors elaborate on the effects of differential coverage binning. Basically, if absolute counts are used for binning, collapsed regions are missed because they have higher coverage. However, many binners also group such contigs by using information which is not dependent on absolute counts, such as covariance/correlation distances. For instance, the program MetaBAT recruits shorter contigs to existing bins at the end of the binning procedure. The authors must disclose the details of genome binning, otherwise the results are hard to interpret or validate. The authors should also show that their findings are not an artifact of a specific binning program or setting.
We have added a description of our binning process (L 186-196).
" | Here is a paper. Please give your review comments after reading it. |
9,757 | "<ns0:abstract xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:p>Intervertebral disc degeneration (IDD), a major cause of lower back pain, has multiple contributing factors including genetics, environment, age, and loading history.</ns0:p><ns0:p>Bioinformatics analysis has been extensively used to identify diagnostic biomarkers and therapeutic targets for IDD diagnosis and treatment. However, multiple microarray dataset analysis and machine learning methods have not been integrated. In this study, we downloaded the mRNA, microRNA (miRNA), long noncoding RNA (lncRNA), and circular RNA (circRNA) expression profiles (GSE34095, GSE15227, GSE63492, GSE116726, GSE56081, and GSE67566) associated with IDD from the GEO database. Using differential expression analysis and recursive feature elimination, we extracted four optimal feature genes. We then used the support vector machine (SVM) to make a classification model with the four optimal feature genes. The ROC curve was used to evaluate the model's performance, and the expression profiles (GSE63492, GSE116726, GSE56081, and GSE67566) were used to construct a competitive endogenous RNA (ceRNA) regulatory network and explore the underlying mechanisms of the feature genes. We found that three miRNAs (hsa-miR-4728-5p, hsa-miR-5196-5p, and hsa-miR-185-5p) and three circRNAs (hsa_circRNA_100723, hsa_circRNA_104471, and hsa_circRNA_100750) were important regulators with more interactions than the other RNAs across the whole network. The expression level analysis of the three datasets revealed that BCAS4 and SCRG1 were key genes involved in IDD development. Ultimately, our study proposes a novel approach to determining reliable and effective targets in IDD diagnosis and treatment.</ns0:p></ns0:div>
</ns0:abstract>
<ns0:body xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:head>Introduction</ns0:head><ns0:p>Lower back pain, commonly caused by intervertebral disc degeneration (IDD), can be a significant socioeconomic burden on patients <ns0:ref type='bibr' target='#b46'>(Vergroesen et al. 2015)</ns0:ref>. IDD is characterized by the apoptosis of nucleus pulposus (NP) cells, the degradation of extracellular matrix (ECM) components, and several contributing factors including genetics and environment <ns0:ref type='bibr' target='#b5'>(Battie et al. 2008;</ns0:ref><ns0:ref type='bibr' target='#b14'>Feng et al. 2016)</ns0:ref>. However, the precise etiology of IDD remains largely unknown. Diagnosing degenerative disc disease is difficult, and common IDD treatment and management strategies primarily consist of conservative management or surgical treatment to relieve pain, without resolving the underlying tissue pathology <ns0:ref type='bibr' target='#b1'>(An et al. 2003;</ns0:ref><ns0:ref type='bibr' target='#b55'>Zaina et al. 2016</ns0:ref>). To identify potential biomarkers and specific therapeutic targets, a more detailed understanding of the molecular and cellular events underlying IDD formation is needed. The etiology of IDD is complex, but over the past several decades, it has become clear that genetic factors are most dominant. Multiple candidate genes, including thrombospondin-2, vitamin D receptor, COL2A1, <ns0:ref type='bibr'>ACAN,</ns0:ref><ns0:ref type='bibr'>interleukins (IL1α,</ns0:ref><ns0:ref type='bibr'>IL 1β,</ns0:ref><ns0:ref type='bibr'>and IL6)</ns0:ref>, matrix metalloproteinases (MMP-3 and MMP-9), and growth/differentiation factor 5, have been associated with the pathophysiological process of IDD development <ns0:ref type='bibr' target='#b14'>(Feng et al. 2016;</ns0:ref><ns0:ref type='bibr' target='#b23'>Kalb et al. 2012;</ns0:ref><ns0:ref type='bibr' target='#b51'>Wang et al. 2018b;</ns0:ref><ns0:ref type='bibr' target='#b54'>Yuan et al. 2018</ns0:ref>). Genome-wide association studies (GWAS) have been used to help identify novel variants. Several GWAS found that the genetic polymorphisms of PARK2 and CHST3 relevant to IDD etiology <ns0:ref type='bibr'>(Song et al. 2013;</ns0:ref><ns0:ref type='bibr' target='#b52'>Williams et al. 2013)</ns0:ref>. Additionally, some bioinformatics analyses based on gene expression profiles revealed that FYN, PRKCD, YWHAB, YWHAZ, AR, Fibronectin 1, COL2A1, β-catenin, COL6A2, IBSP, RAP1A, and FOXF2 genes may play key roles in IDD development <ns0:ref type='bibr' target='#b7'>(Chen et al. 2013;</ns0:ref><ns0:ref type='bibr' target='#b16'>Guo et al. 2017;</ns0:ref><ns0:ref type='bibr' target='#b21'>Ji et al. 2015)</ns0:ref>. Although a considerable number of genes associated with IDD development have been found, early IDD diagnosis and precise treatment remain difficult and require further study. Integrated analyses using combined multiple microarray data can provide a more accurate understanding of the interplay across multi-level genomic features and the molecular mechanisms that cause complex diseases <ns0:ref type='bibr' target='#b31'>(Momtaz et al. 2018)</ns0:ref>. Machine learning is a type of artificial intelligence that can 'learn' a model using past data in order to predict future data. Machine learning algorithms have been used in key feature training, recognition, and group classification <ns0:ref type='bibr' target='#b19'>(Huang et al. 2018)</ns0:ref>. Modern researchers have unprecedented access to machine learning methods that can elucidate complex molecular mechanisms and predict disease genes from large biomedical datasets <ns0:ref type='bibr' target='#b28'>(Libbrecht & Noble 2015;</ns0:ref><ns0:ref type='bibr' target='#b32'>Obermeyer & Emanuel 2016)</ns0:ref>. To the best of our knowledge, no investigations have used integrated analysis and machine learning methods to identify key IDD-associated genes. Accumulating evidence has indicated that noncoding RNAs, including microRNAs (miRNAs), long noncoding RNAs (lncRNAs), and circular RNAs (circRNAs), are important gene expression regulators that influence cellular function and disease states <ns0:ref type='bibr' target='#b0'>(Adams et al. 2017)</ns0:ref>. lncRNAs and circRNAs act as competitive endogenous RNAs (ceRNAs) and competitively bind miRNA response elements (MREs) to construct a regulatory network involved in IDD progression. <ns0:ref type='bibr' target='#b56'>Zhao et al. (2016)</ns0:ref> used RNA sequencing to identify 1,854 lncRNAs and 2,804 protein-coding genes that were differentially expressed in the IDD group. <ns0:ref type='bibr'>Tan et al. (2018)</ns0:ref> found that LncSNHG1 promoted NP cell proliferation by suppressing miR-326 expression and upregulating CCND1 expression. Recent studies have focused on the functional roles of circRNAs during IDD development and found that circ-4099, circ-GRB10, circVMA21, and circ_001653 play pivotal roles during NP cell proliferation, apoptosis, and extracellular matrix synthesis/degradation <ns0:ref type='bibr' target='#b8'>(Cheng et al. 2018;</ns0:ref><ns0:ref type='bibr'>Cui & Zhang 2020;</ns0:ref><ns0:ref type='bibr' target='#b17'>Guo et al. 2018;</ns0:ref><ns0:ref type='bibr' target='#b49'>Wang et al. 2018a</ns0:ref>). The diverse biological functions of ceRNAs deserve further exploration. In this study, we aimed to identify the key genes and underlying mechanisms of IDD development by constructing an lncRNA/circRNA-miRNA-mRNA network using multiple microarray datasets and machine learning methods. Fig. <ns0:ref type='figure' target='#fig_6'>S1</ns0:ref> shows the flow chart for this study. Our results present novel biomarkers and therapeutic targets that can be used for IDD diagnosis and treatment.</ns0:p></ns0:div>
<ns0:div><ns0:head>Materials & Methods</ns0:head></ns0:div>
<ns0:div><ns0:head>Microarray datasets</ns0:head><ns0:p>We retrieved a total of six expression profiles from the GEO database (www.ncbi.Nlm.nih.gov/geo): two mRNA expression profiles (GSE34095 <ns0:ref type='bibr' target='#b45'>(Tsai et al. 2013</ns0:ref>) and GSE15227 <ns0:ref type='bibr' target='#b15'>(Gruber et al. 2009</ns0:ref>)), two miRNA expression profiles (GSE63492 <ns0:ref type='bibr' target='#b26'>(Lan et al. 2016</ns0:ref>) and GSE116726 <ns0:ref type='bibr' target='#b20'>(Ji et al. 2018</ns0:ref>)), one mRNA-lncRNA expression profile (GSE56081 <ns0:ref type='bibr' target='#b26'>(Lan et al. 2016;</ns0:ref><ns0:ref type='bibr' target='#b48'>Wan et al. 2014</ns0:ref>)), and one circRNA expression profile (GSE67566 <ns0:ref type='bibr' target='#b26'>(Lan et al. 2016;</ns0:ref><ns0:ref type='bibr' target='#b30'>Liu et al. 2015)</ns0:ref>). Table <ns0:ref type='table' target='#tab_0'>1</ns0:ref> contains the basic information for these expression profiles. Manuscript to be reviewed</ns0:p></ns0:div>
<ns0:div><ns0:head>Differential expression analysis</ns0:head><ns0:p>The raw data were annotated, normalized, log 2 transformed, and screened for differentially expressed genes (DEGs), differentially expressed miRNAs (DEMs), differentially expressed lncRNAs (DELs), and differentially expressed circRNAs (DECs) in IDD and normal disc tissue using the 'Limma' R package <ns0:ref type='bibr' target='#b37'>(Ritchie et al. 2015)</ns0:ref>. Since each dataset came from a different experiment and microarray platform, the way to get the data may be different. When filtering using consistent thresholds, some datasets did not yield valid differential genes. In order to obtain effective differentially expressed genes for subsequent analysis, we used different thresholds according to each dataset's conditions. The specific DEG screening thresholds were as follows: p-value < 0.05 for the GSE34095 dataset, p-value < 0.01 and |log2FC| ≥ 2 for the GSE15227 dataset, and p-value < 0.01 and |log2 FC| > 2 for the GSE116726 dataset. We obtained the GSE67566, GSE63492, and GSE56081 datasets from the same tissue samples, and provided the differential expression analysis results and thresholds as Supplemental Materials (Tables <ns0:ref type='table' target='#tab_1'>S1 and S2</ns0:ref>, Documents S1and S2). We visualized the differential expression analysis results using a volcano plot and heatmap with hierarchical clustering (Fig. <ns0:ref type='figure' target='#fig_4'>S2</ns0:ref>).</ns0:p></ns0:div>
<ns0:div><ns0:head>Feature gene extraction</ns0:head><ns0:p>Using VENNY 2.1 software (http://bioinfogp.cnb.csic.es/tools/venny/index.html), we extracted the differential gene intersections of the GSE34095 and GSE15227 datasets to use as IDD-related DEGs. We used the GSE15227 dataset as the training set to screen for important feature genes. Using the R caret package, random forest, and neural network methods, we constructed the model and obtained the important DEG features <ns0:ref type='bibr' target='#b25'>(Kuhn 2015)</ns0:ref>. We used recursive feature elimination (RFE), a machine learning method, to extract the optimum feature genes to identify the functional biomarkers involved in IDD progression <ns0:ref type='bibr' target='#b18'>(Guyon et al. 2002)</ns0:ref>.</ns0:p></ns0:div>
<ns0:div><ns0:head>SVM model verification</ns0:head><ns0:p>The support vector machine (SVM) is particularly effective for binary classification in a supervised learning manner, and is better than other machine learning methods at identifying subtle patterns in complex datasets <ns0:ref type='bibr' target='#b3'>(Aruna & Dr 2011)</ns0:ref>. Radial basis function kernel (RBF kernel) is commonly used in nonlinear support vector machine classification since it can enable data to operate in a high-dimensional and implicit feature space <ns0:ref type='bibr' target='#b22'>(Jiao et al. 2017)</ns0:ref>. In this study, we used the R e1071 package (https://CRAN.R-project.org/package=e1071) to build an RBF kernel SVM model to identify the optimal feature genes <ns0:ref type='bibr' target='#b11'>(Dimitriadou et al. 2012</ns0:ref>), and we selected the GSE15227 dataset to train the machine. We then evaluated its performance in the GSE15227 training set using the R pROC package (http://www.biomedcentral.com/1471-2105/12/77/) <ns0:ref type='bibr' target='#b38'>(Robin et al. 2011)</ns0:ref>. The optimal feature genes were taken from the training set, and over-fitting would occur whenever ROC verification was performed. Therefore, we used two validation sets (GSE34095 and GSE56081) to further verify the model's performance.</ns0:p></ns0:div>
<ns0:div><ns0:head>Identifying target miRNAs of the optimal feature genes</ns0:head><ns0:p>We used miRWalk 2.0 and 12 prediction programs (MIMATid, Microt4, miRanda, mirbridge, miRDB, miRMap, miRNAMap, Pictar2, PITA, RNA22, RNAhybrid, and Targetscan) to predict the optimal feature genes' target miRNAs <ns0:ref type='bibr' target='#b13'>(Dweep & Gretz 2015)</ns0:ref>. miRNAs present in more than five of the 12 prediction programs were considered target miRNAs. We selected overlapping DEMs in the GSE116726 and GSE63492 datasets as candidate DEMs. Intersecting target miRNAs and candidate DEMs were selected as optimal feature gene interaction pairs. ceRNA network construction We obtained the miRNA sequences interacting with optimal feature genes from the miRbase (Kozomara & Griffiths-Jones 2011) and extracted mature sequences using the Perl program. The DEL sequences were downloaded from NCBI. For DELs with several transcripts, we selected the longest transcripts for subsequent analysis. We used the BEDTOOLS command <ns0:ref type='bibr' target='#b36'>(Quinlan & Hall 2010)</ns0:ref> and genomic coordinates to obtain the DEC sequences, and converted the gene name to its circRNA symbol using the Perl program. We then used the miRanda tool to analyze the combinations of miRNAs, DELs, and DECs. We set the analysis parameters for the miRNA-DELs and miRNA-DECs as sc:120, en:-20 and sc:150, en:-7, respectively, and processed the results using python script (Documents S3 and S4). The DELs and DECs with ≥5 miRNA binding sites were identified as reliable miRNA-lncRNA and miRNA-circRNA interaction pairs. The ceRNA (DELs/DECs-miRNA-optimal feature gene) regulatory network was constructed using a combination of miRNA-DEL pairs, miRNA-DEC pairs, and miRNA-optimal feature gene pairs. We visualized the network using Cytoscape 3.6.0 (http://www.cytoscape.org/).</ns0:p></ns0:div>
<ns0:div><ns0:head>Statistical analysis</ns0:head><ns0:p>All data were presented as the mean±SEM. Differential expression levels were compared using the Student's ttest in GraphPad Prism 7.0 (GraphPad Software Inc., La Jolla, CA, USA). P values < 0.05 were considered statistically significant.</ns0:p></ns0:div>
<ns0:div><ns0:head>Results</ns0:head></ns0:div>
<ns0:div><ns0:head>Differentially expressed genes in IDD</ns0:head><ns0:p>We obtained a total of 334 DEGs (199 up-regulated and 135 down-regulated) in the GSE34095 dataset (Fig. <ns0:ref type='figure' target='#fig_6'>1A</ns0:ref>) and 188 DEGs (141 up-regulated and 47 down-regulated) in the GSE15227 dataset (Fig. <ns0:ref type='figure' target='#fig_6'>1B</ns0:ref>). Hierarchical cluster heatmaps showed that these DEGs could distinguish between the degenerative disc samples and the control disc samples (Fig. <ns0:ref type='figure' target='#fig_6'>1C and 1D</ns0:ref>). We obtained a total of 13 overlapping DEGs (COL3A1, SCRG1, HTRA1, BCAS4, C11orf80, CRNKL1, GREM1, FGFR3, BDKRB1, WDR46, FN1, LMF2, and GDI2) via the intersection between the two datasets, and considered these as IDD-related DEGs (Fig. <ns0:ref type='figure' target='#fig_6'>1E</ns0:ref>).</ns0:p></ns0:div>
<ns0:div><ns0:head>Optimal feature gene selection</ns0:head><ns0:p>We used random forest and neural network models to evaluate the importance of the 13 overlapping DEGs (Documents S5 and S6). The results showed that most of the DEGs from the two models had minimal differences in terms of gene importance rankings (Fig. <ns0:ref type='figure'>2A and 2B</ns0:ref>). Using the RFE method, we identified four important genes (WDR46, BCAS4, CRNKL1, and SCRG1) as optimal feature genes associated with IDD (Fig. <ns0:ref type='figure'>2C</ns0:ref>). A classification model was constructed using the RBF kernel SVM, the four genes as features, and the GSE15227 dataset to train the machine. The parameters were set as: SVM-Kernel: radial, cost:1, gamma:0.25, and epsilon:0.1. Additionally, their performance was assessed using the R pROC package. In the GSE15227 training set, the area under the ROC curve (AUC) of the SVM model was 100 percent, suggesting that the model could accurately distinguish between IDD and normal samples (Fig. <ns0:ref type='figure'>2D</ns0:ref>). We used the GSE34095 and GSE56081 datasets as validation sets to further evaluate the model's performance and to help avoid over-fitting in the training set. The SVM model in the GSE34095 validation set had an AUC of 55.6 percent, which may be correlated with the small sample size of the microarray dataset (Fig. <ns0:ref type='figure'>2E</ns0:ref>). However, the AUC for the GSE56081 validation set was 100 percent (Fig. <ns0:ref type='figure'>2F</ns0:ref>). These results indicated that the optimal feature genes (WDR46, BCAS4, CRNKL1, and SCRG1) could be used as effective and accurate IDD diagnostic biomarkers. Identifying target miRNAs of the optimal feature genes We predicted a total of 467 target miRNA optimal feature genes using miRWalk 2.0. Compared to the control disc samples, we identified 724 DEMs (527 up-regulated and 197 down-regulated) from the GSE116726 dataset (Fig. <ns0:ref type='figure' target='#fig_8'>3A and 3B</ns0:ref>) and 149 DEMs from the GSE63492 dataset. Fig. <ns0:ref type='figure' target='#fig_8'>3C</ns0:ref> shows a Venn diagram of the 46 common DEMs that were found. Based on the intersections between the target miRNAs and common DEMs, we further analyzed 12 overlapping miRNAs which we considered to be the optimal feature genes' interaction miRNAs (Fig. <ns0:ref type='figure' target='#fig_8'>3D</ns0:ref>).</ns0:p></ns0:div>
<ns0:div><ns0:head>ceRNA network construction</ns0:head><ns0:p>The differentially expressed lncRNAs and circRNAs between the degenerated disc samples and the control disc samples were downloaded from the GEO database. We analyzed the miRNA-DEL interactions and the miRNA-DEC interactions using the miRanda tool with ≥5 miRNA binding sites. The miRNA-DEL pairs, miRNA-DEC pairs, and miRNA-optimal feature gene pairs were combined to build a DELs/DECs-DEMs-optimal feature gene regulatory network, which included four mRNA nodes, 12 miRNA nodes, 10 lncRNA nodes, and 75 circRNA nodes (Fig. <ns0:ref type='figure' target='#fig_2'>4</ns0:ref>, Table <ns0:ref type='table'>S3</ns0:ref>). Out of these, three miRNAs (hsa-miR-4728-5p, hsa-miR-5196-5p, and hsa-miR-185-5p) and three circRNAs (hsa_circRNA_100723, hsa_circRNA_104471, and hsa_circRNA_100750) were key regulators, based on the optimal feature genes and connective degrees of the whole network (Table <ns0:ref type='table' target='#tab_1'>2</ns0:ref>). However, the LncRNA connective degree was very low.</ns0:p></ns0:div>
<ns0:div><ns0:head>Validating optimal feature genes</ns0:head><ns0:p>The expression of the four optimal feature genes across the three datasets was visualized using box plots. In the GSE15227 and GSE56081 datasets, BCAS4 expression levels were significantly down-regulated (Fig. <ns0:ref type='figure' target='#fig_10'>5A and 5C</ns0:ref>), and SCRG1 was significantly up-regulated (Fig. <ns0:ref type='figure' target='#fig_10'>5G and 5I</ns0:ref>). CRNKL1 was only up-regulated in the GSE34095 dataset (Fig. <ns0:ref type='figure' target='#fig_10'>5E</ns0:ref>). Unfortunately, the expression levels of WDR46 showed no significant difference across the three datasets (Fig. <ns0:ref type='figure' target='#fig_10'>5J</ns0:ref>,5K and 5L). These results indicated that BCAS4 and SCRG1 are key genes involved in IDD development.</ns0:p></ns0:div>
<ns0:div><ns0:head>Discussion</ns0:head><ns0:p>IDD is linked to lower back pain and spine-related diseases, and although IDD's underlying mechanisms have been studied for many years, they still remain unclear. Insufficient early IDD diagnosis and treatment methods affect the quality of life for patients and impose a heavy economic burden on society <ns0:ref type='bibr' target='#b46'>(Vergroesen et al. 2015)</ns0:ref>. NP cells play an important role in maintaining intervertebral disc homeostasis by synthesizing ECM, which includes aggrecan and type II collagen <ns0:ref type='bibr' target='#b56'>(Zhang et al. 2016)</ns0:ref>. Recent studies show that targeting gene therapy can inhibit NP cell senescence and apoptosis, and can ultimately ameliorate IDD <ns0:ref type='bibr' target='#b6'>(Chen et al. 2018)</ns0:ref>. Therefore, reliable and specific gene targets are essential for IDD diagnosis and treatment. Recent developments in bioinformatics and computational biology have led to the identification of several key genetic targets related to IDD and the prediction of potential molecular mechanisms <ns0:ref type='bibr' target='#b35'>(Petryszak et al. 2014)</ns0:ref>. In this study, we downloaded multiple microarray datasets associated with IDD from the GEO database, including two mRNA expression profiles, two miRNA expression profiles, one mRNA-lncRNA expression profile, and one circRNA expression profile. A total of four optimal feature genes (WDR46, BCAS4, CRNKL1, and SCRG1) were identified using machine-learning methods. The construction of a DELs/DECs-miRNA-optimal feature genes network revealed that three miRNAs (hsa-miR-4728-5p, hsa-miR-5196-5p, and hsa-miR-185-5p) and three circRNAs (hsa_circRNA_100723, hsa_circRNA_104471, and hsa_circRNA_100750) may be important mediators for optimal feature genes. To further explore the different optimal feature genes in normal and degenerative NP tissues, we also investigated their expression levels across three datasets. The results indicated that BCAS4 and SCRG1 were key genes related to IDD. Breast carcinoma amplified sequence 4 (BCAS4), a novel gene cloned from breast cancer cells, encodes a 211amino acid cytoplasmic protein with no significant homologies to any known protein <ns0:ref type='bibr' target='#b4'>(Barlund et al. 2002)</ns0:ref>. Previous studies have demonstrated that the specific DNA methylation of BCAS4 acts as an epigenetic marker and can be used to distinguish saliva from other body fluids. It is also widely used in forensic investigations <ns0:ref type='bibr' target='#b39'>(Silva et al. 2016;</ns0:ref><ns0:ref type='bibr' target='#b42'>Taki & Kibayashi 2015)</ns0:ref>. However, BCAS4's biological role in disease requires further investigation. Stimulator of chondrogenesis 1 (SCRG1) was first found in the genes associated with, or responsible for, the neurodegenerative changes observed in transmissible spongiform encephalopathies <ns0:ref type='bibr' target='#b10'>(Dandoy-Dron et al. 1998</ns0:ref>). SCRG1 transcript is found in the brain, heart, and spinal cord, and its sequence is highly conserved in humans, mice, and rats. SCRG1 has also been observed to be specifically expressed in human articular cartilage, and is involved in human mesenchymal stem cell (hMSC) growth suppression and differentiation during dexamethasone-dependent chondrogenesis <ns0:ref type='bibr' target='#b33'>(Ochi et al. 2006)</ns0:ref>. Recent studies have shown that SCRG1 is an important regulator during hMSC self-renewal, migration, and osteogenic differentiation along with its receptor BST1 <ns0:ref type='bibr' target='#b2'>(Aomatsu et al. 2014</ns0:ref>). However, the function of SCRG1 in IDD development has not yet been explored. circRNAs and lncRNAs may act as ceRNAs by competitively binding to miRNA and suppressing mRNA expression <ns0:ref type='bibr' target='#b0'>(Adams et al. 2017;</ns0:ref><ns0:ref type='bibr' target='#b43'>Tay et al. 2014)</ns0:ref>. This ceRNA hypothesis suggests that there is a novel mechanism for RNA interactions. In this study's ceRNA analysis, hsa-miR-4728-5p, hsa-miR-5196-5p, hsa-miR-185-5p, hsa_circRNA_100723, hsa_circRNA_104471, and hsa_circRNA_100750 showed more interactions compared to the other RNAs in the whole network. Additionally, miR-155, miR-21, and miR-133a were shown to be differentially expressed in degenerative NP cells, indicating that they may be potential biomarkers for early IDD diagnosis <ns0:ref type='bibr' target='#b29'>(Liu et al. 2014;</ns0:ref><ns0:ref type='bibr' target='#b50'>Wang et al. 2011;</ns0:ref><ns0:ref type='bibr' target='#b53'>Xu et al. 2016)</ns0:ref>. The dysregulation of these miRNAs is closely associated with NP cell apoptosis which affects IDD progression. miR-185-5p has been reported as a critical regulator, but miR-4728-5p and miR-5196-5p have rarely been reported. <ns0:ref type='bibr'>Chang et al. (2017)</ns0:ref> reported that miR-185-5p induced by Runx2 could directly target Dlx2 to inhibit amylogenesis and osteogenesis, providing a new treatment option for cleidocranial dysplasia. Multiple lncRNAs bind to miR-185-5p in order to modulate the progression of different types of human cancer, including prostate cancer <ns0:ref type='bibr' target='#b44'>(Tian et al. 2018)</ns0:ref>, colorectal cancer <ns0:ref type='bibr' target='#b57'>(Zhu et al. 2018)</ns0:ref>, and glioblastoma <ns0:ref type='bibr' target='#b51'>(Wang et al. 2018b</ns0:ref>). However, the regulatory effects of miR-185-5p on IDD require further investigation. CircRNAs, a new type of ncRNAs formed by special loop splicing, are thought to be potential diagnostic biomarkers and therapeutic targets because they are more stable and conserved than other RNAs <ns0:ref type='bibr' target='#b27'>(Lee et al. 2019</ns0:ref>). However, very few studies have focused on the role of circRNAs in IDD development. <ns0:ref type='bibr' target='#b17'>Guo et al. (2018)</ns0:ref> found that circ-GRB10 was downregulated in IDD cells, and circ-GRB10 overexpression inhibited NP cell apoptosis by sequestering miR-328-5p and upregulating target genes involved in cell proliferation through the ErbB pathway. <ns0:ref type='bibr' target='#b8'>Cheng et al. (2018)</ns0:ref> reported that circVMA21 acted as a sponge for miR-200c and regulated the activity and function of NP cells by targeting miRNA-200c and XIAP, providing a new IDD intervention and treatment strategy. <ns0:ref type='bibr' target='#b49'>Wang et al. (2018a)</ns0:ref> found that circRNA_4099 could act as a sponge for miR-616-5p and eliminate Sox9 inhibition, increasing ECM secretion. Similarly, <ns0:ref type='bibr'>Cui & Zhang (2020)</ns0:ref> reported that circ_001653 silencing may bind to miR-486-3p in order to inhibit CEMIP expression, thus attenuating NP cell apoptosis and ECM degradation. In this study, we used integrated analysis to identify that hsa_circRNA_100723, hsa_circRNA_104471, and hsa_circRNA_100750, which have not been previously reported, are more likely to be important molecules involved in IDD regulation, and require further investigation.</ns0:p></ns0:div>
<ns0:div><ns0:head>Conclusion</ns0:head><ns0:p>In this study, we used integrated bioinformatics analysis and machine learning methods to identify BCAS4 and SCRG1 as key genes associated with IDD development. Additionally, after constructing the ceRNA network, we found three miRNAs and three circRNAs that may act as important regulators during IDD development by targeting key genes. This novel study may provide new insights into IDD pathogenesis and therapy. Further experiments should be conducted to verify this study's results.</ns0:p><ns0:p>Ranking of the top 13 IDD-related genes using neural networks (A) and random forest (B). Extraction of the optimum feature genes from the 13 IDD-related genes was carried out using recursive feature elimination (C). Classification efficiency of the optimum feature genes in the model as evaluated using the ROC curve in the GSE15227 (D), GSE34095 (E), and GSE56081 (F) datasets, respectively. Table <ns0:ref type='table' target='#tab_0'>1</ns0:ref>. Basic information of the expression profiles Table <ns0:ref type='table' target='#tab_1'>2</ns0:ref>. The top three miRNAs and circRNAs related to optimal feature genes in the network Supplemental Files Figure <ns0:ref type='figure' target='#fig_6'>S1：Flow</ns0:ref> chart presentation of the study IDD, intervertebral disc degeneration; DEG, differential expression gene; RFE, recursive feature elimination; SVM, support vector machine; DEM, differential expression miRNA; DEL, differential expression lncRNA; DEC, differential expression circRNA; ceRNA, competing endogenous RNA. Blue represents the downregulated and red represents the upregulated. Table <ns0:ref type='table' target='#tab_0'>S1</ns0:ref>: All differentially expressed circRNAs in the GSE67566 dataset. Table <ns0:ref type='table' target='#tab_1'>S2</ns0:ref>: All differentially expressed miRNAs in the GSE63492 dataset. Table <ns0:ref type='table'>S3</ns0:ref>: The nodes of optimal feature genes, miRNAs, lncRNAs, and circRNAs in the ceRNA network. Figure Ranking of the top 13 IDD-related genes using neural networks (A) and random forest (B).</ns0:p><ns0:note type='other'>1</ns0:note><ns0:p>Extraction of the optimum feature genes from the 13 IDD-related genes was carried out using recursive feature elimination (C). Classification efficiency of the optimum feature genes in the model as evaluated using the ROC curve in the GSE15227 (D), GSE34095 (E), and GSE56081 (F) datasets, respectively. Manuscript to be reviewed </ns0:p><ns0:note type='other'>Figure 3</ns0:note><ns0:note type='other'>Figure 5</ns0:note></ns0:div><ns0:figure xml:id='fig_0'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:05:48863:1:2:NEW 20 Aug 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_1'><ns0:head>Figure3:</ns0:head><ns0:label /><ns0:figDesc>Identification of target miRNAs of the optimal feature genes (A)Volcano plots represent the DEMs of the degenerative disc samples and control disc samples in the GSE116726 dataset. (B) Hierarchical cluster heatmaps of the GSE116726 dataset display the DEMs to compare degenerative disc samples and control disc samples. Blue represents the degenerative samples and red represents the control samples. (C) Venn diagram of DEMs in the GSE116726 and GSE63492 datasets. The common area represents the overlapping DEMs. (D)Venn diagram of miRNAs in overlapping DEMs and target miRNAs of optimum feature genes. The common area represents the overlapping miRNAs. DEM, differentially expressed miRNA.</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_2'><ns0:head>Figure 4 :</ns0:head><ns0:label>4</ns0:label><ns0:figDesc>ceRNA network construction ceRNA network of optimum IDD feature genes with DEMs, DELs, and DECs. The blue circles represent optimum IDD feature genes, the red triangles represent DECs, the green diamonds represent DELs, and the yellow arrows represent DEMs. ceRNA, competing endogenous RNA; DEC, differentially expressed circRNA; DEL, differentially expressed lncRNA; and DEM, differentially expressed miRNA.</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_3'><ns0:head>Figure 5 :</ns0:head><ns0:label>5</ns0:label><ns0:figDesc>Validation of optimal feature genes The expression levels of the four optimum feature genes in the GSE15227, GSE34095, and GSE56081 datasets, respectively. (A, B and C) The expression level of BCSA4. (D, E and F) The expression level of CRNKL1. (G, H and I) The expression level of SCRG1. (J, K and L) The expression level of WDR46. * represents P value < 0.05, *** represents P value < 0.001, **** represents P value < 0.0001, and NS represents not significant.</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_4'><ns0:head>Figure S2 :</ns0:head><ns0:label>S2</ns0:label><ns0:figDesc>Differential expression analysis of the GSE56081, GSE63492, and GSE67566 datasets. Volcano plots represent the DEGs (A) and DELs (C) of the degenerative disc samples and control disc samples, respectively, in the GSE56081 dataset. Hierarchical cluster heatmaps represent the DEGs (B) and DELs (D) of the degenerative disc samples and control disc samples, respectively, in the GSE56081 dataset. Hierarchical cluster heatmaps represent the DEMs in the GSE63492 dataset (E) and the DECs (F) in the GSE67566 dataset.</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_5'><ns0:head /><ns0:label /><ns0:figDesc>Document S1: Differentially expressed mRNAs in the GSE56081 dataset. Document S2: Differentially expressed lncRNAs in the GSE56081 dataset. Document S3: The target miRNAs of DECs analyzed by miRanda tool. Document S4: The target miRNAs of DELs analyzed by miRanda tool. Document S5: Importance rankings of the 13 overlapping DEGs evaluated by random forest. Document S6: Importance rankings of the 13 overlapping DEGs evaluated by neural network.</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_6'><ns0:head>Figure 1 :</ns0:head><ns0:label>1</ns0:label><ns0:figDesc>Figure 1: Differentially expressed IDD genes</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_7'><ns0:head>Figure 2 Figure 2 :</ns0:head><ns0:label>22</ns0:label><ns0:figDesc>Figure 2</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_8'><ns0:head>Figure 3 :</ns0:head><ns0:label>3</ns0:label><ns0:figDesc>Figure 3: Identification of target miRNAs of optimal feature genes</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_9'><ns0:head>Figure 4 Figure 4 :</ns0:head><ns0:label>44</ns0:label><ns0:figDesc>Figure 4</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_10'><ns0:head>Figure 5 :</ns0:head><ns0:label>5</ns0:label><ns0:figDesc>Figure 5: Validation of optimal feature genes</ns0:figDesc></ns0:figure>
<ns0:figure type='table' xml:id='tab_0'><ns0:head>Table 1 . Basic information of expression profiles included in the study</ns0:head><ns0:label>1</ns0:label><ns0:figDesc /><ns0:table><ns0:row><ns0:cell>Type</ns0:cell><ns0:cell>Series</ns0:cell><ns0:cell>Platform</ns0:cell><ns0:cell>Source</ns0:cell><ns0:cell>Number of samples</ns0:cell><ns0:cell>Publication</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell>name</ns0:cell><ns0:cell>(control/degenerative)</ns0:cell><ns0:cell>year</ns0:cell></ns0:row><ns0:row><ns0:cell>mRNA</ns0:cell><ns0:cell>GSE34095</ns0:cell><ns0:cell>GPL96</ns0:cell><ns0:cell>Disc</ns0:cell><ns0:cell>6(3/3)</ns0:cell><ns0:cell>2012</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell>tissue</ns0:cell><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>mRNA</ns0:cell><ns0:cell>GSE15227</ns0:cell><ns0:cell>GPL1352</ns0:cell><ns0:cell>Disc</ns0:cell><ns0:cell>15 (12/3)</ns0:cell><ns0:cell>2009</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell>tissue</ns0:cell><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>miRNA</ns0:cell><ns0:cell>GSE63492</ns0:cell><ns0:cell>GPL19449</ns0:cell><ns0:cell>Nucleus</ns0:cell><ns0:cell>10 (5/5)</ns0:cell><ns0:cell>2016</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell>pulposus</ns0:cell><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>miRNA</ns0:cell><ns0:cell cols='2'>GSE116726 GPL20712</ns0:cell><ns0:cell>Nucleus</ns0:cell><ns0:cell>6(3/3)</ns0:cell><ns0:cell>2018</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell>pulposus</ns0:cell><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>mRNA-l</ns0:cell><ns0:cell>GSE56081</ns0:cell><ns0:cell>GPL15314</ns0:cell><ns0:cell>Nucleus</ns0:cell><ns0:cell>10(5/5)</ns0:cell><ns0:cell>2014</ns0:cell></ns0:row><ns0:row><ns0:cell>ncRNA</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell>pulposus</ns0:cell><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell cols='2'>circRNA GSE67566</ns0:cell><ns0:cell>GPL19978</ns0:cell><ns0:cell>Nucleus</ns0:cell><ns0:cell>10(5/5)</ns0:cell><ns0:cell>2016</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell>pulposus</ns0:cell><ns0:cell /><ns0:cell /></ns0:row></ns0:table></ns0:figure>
<ns0:figure type='table' xml:id='tab_1'><ns0:head>Table 2 . The top 3 miRNAs and circRNAs related to optimal feature genes in the network</ns0:head><ns0:label>2</ns0:label><ns0:figDesc /><ns0:table><ns0:row><ns0:cell>Type</ns0:cell><ns0:cell>name</ns0:cell><ns0:cell>Number of directed edges</ns0:cell></ns0:row><ns0:row><ns0:cell>miRNA</ns0:cell><ns0:cell>hsa-miR-4728-5p</ns0:cell><ns0:cell>58</ns0:cell></ns0:row><ns0:row><ns0:cell>miRNA</ns0:cell><ns0:cell>hsa-miR-5196-5p</ns0:cell><ns0:cell>41</ns0:cell></ns0:row><ns0:row><ns0:cell>miRNA</ns0:cell><ns0:cell>hsa-miR-185-5p</ns0:cell><ns0:cell>34</ns0:cell></ns0:row><ns0:row><ns0:cell>circRNA</ns0:cell><ns0:cell>hsa_circRNA_100723</ns0:cell><ns0:cell>12</ns0:cell></ns0:row><ns0:row><ns0:cell>circRNA</ns0:cell><ns0:cell>hsa_circRNA_104471</ns0:cell><ns0:cell>11</ns0:cell></ns0:row><ns0:row><ns0:cell>circRNA</ns0:cell><ns0:cell>hsa_circRNA_100750</ns0:cell><ns0:cell>9</ns0:cell></ns0:row></ns0:table></ns0:figure>
<ns0:note place='foot'>PeerJ reviewing PDF | (2020:05:48863:1:2:NEW 20 Aug 2020)</ns0:note>
<ns0:note place='foot'>PeerJ reviewing PDF | (2020:05:48863:1:2:NEW 20 Aug 2020)Manuscript to be reviewed</ns0:note>
</ns0:body>
" | "
Yangpu Hospital
Tongji University School of Medicine
450 Tengyue Road
Shanghai 200090, China July 28th,2020
Dear editors and reviewers,
We deeply appreciate the time and effort you’ve spent in reviewing our paper Integrated analysis of multi-microarray datasets and machine learning methods reveal key genes and regulatory mechanisms underlying human intervertebral disc degeneration.
We have revised the manuscript, according to the comments and suggestions of reviewers, and responded, point by point to, which marked in blue in the paper. We attached a copy of the revised manuscript at the end of this document.
The revised manuscript has been edited and proofread by Peer J language editing, which marked in red in the paper.
I would like to re-submit this revised manuscript to Peer J, and hope it is acceptable for publication in the journal.
Looking forward to hearing from you soon.
With kindest regards,
Yours Sincerely,
Xiaodong Liu
On behalf of all authors.
Responses to Reviewers
First of all, we thank reviewers for their positive and constructive comments. Those comments are all valuable and very helpful for revising and improving our paper, as well as the important guiding significance to our researches. The responds to the reviewer’s comments are as follows:
Replies to Reviewer #1:
Basic reporting
Question: 1. Review by a primary English writer for grammar, English composition. Several mistakes should be corrected in the manuscript. Such as:
Page 6: 69-72 “Diagnosis of degenerative disc disease is difficult, and currently available treatment and management strategies for IDD primarily involve conservative or surgical treatment to relieve pain, without resolving underlying tissue pathology (An et al. 2003; Zaina et al. 2016).” “involve” maybe not suit in the present context, “consist of” may be more appropriate.
Answer: Agreed. We have changed “involve” to “consist of” in line 83.
Page 6: 72-73 “Therefore, a detailed understanding of the molecular and cellular events that underlie the formation of IDD is needed to identify diagnostic markers and design new therapeutic targets.” This sentence needs to be rewritten. “to” is not suited in the sentence.
Answer: Agreed. We have rewritten this sentence as “To identify potential biomarkers and specific therapeutic targets, a more detailed understanding of the molecular and cellular events underlying IDD formation is needed.” in line 86-88.
Page 10: 239-243 “Recent developments in bioinformatics and computational Biology has led to the identification of several key genetic targets related to IDD and potential molecular mechanisms have also been predicted (Petryszak et al.2014). In this study, integrated analysis of multi-microarray datasets including two mRNA expression profiles, two miRNA expression profiles, one mRNA-lncRNA expression profile, and a circRNA expression profile, associated with IDD was downloaded from the GEO database.” “Biology” should be corrected as biology. “multi-microarray datasets” is incorrect in English. The sentence, “In this study, integrated analysis of multi-microarray datasets including two mRNA expression profiles, two miRNA expression profiles, one mRNA-lncRNA expression profile, and a circRNA expression profile, associated with IDD was downloaded from the GEO database.” should be rewritten.
Answer: Thank you very much for the suggestion. We have corrected “Biology” as “biology” in line 344, and corrected “multi-microarray datasets” as “multiple microarray datasets” in the title and main body. Moreover, we have rewritten the sentence “In this study, integrated analysis of multi-microarray datasets including two mRNA expression profiles, two miRNA expression profiles, one mRNA-lncRNA expression profile, and a circRNA expression profile, associated with IDD was downloaded from the GEO database” as “In this study, we downloaded multiple microarray datasets associated with IDD from the GEO database, including two mRNA expression profiles, two miRNA expression profiles, one mRNA-lncRNA expression profile, and one circRNA expression profile” in line 275-277.
Page 11: 266-267 The sentence, “The ceRNA hypothesis is a new gene regulatory model that plays an important role in various diseases.”, should be rewritten.
Answer: Agreed. We have rewritten this sentence as “The ceRNA hypothesis reveals a new mechanism for RNA interactions” in line 386-387.
Page 12: 294-295 “In conclusion, in this study, BCAS4 and SCRG1as were identified as key genes associated with the progression of IDD using bioinformatics analysis and machine-learning methods.” “as” should be deleted.
Answer: Agreed. We have deleted the word “as” in line 426.
Page 12: 295-296 The sentence, “Besides, 3 miRNAs and 3 circRNAs were shown to play a vital role in IDD through interaction with key genes after constructing a ceRNA network.”, should be rewritten.
Answer: Agreed. We have rewritten this sentence as “Additionally, after constructing the ceRNA network, we found three miRNAs and three circRNAs that may act as important regulators during IDD development by targeting key genes” in line 429-431.
2. Page 6: 67-69 “Previous studies have indicated that nucleus pulposus (NP) cell apoptosis (Jiang et al. 2013), and several other factors including genetic and environmental factors are the main causes of IDD (Battie et al. 2008)”.
Genetic and environmental factors are the primary influencing factors of IDD. Nucleus pulposus cell apoptosis is a phenomenon and process observed in the degeneration, not a cause of IDD. This statement is not accurate.
Answer: Agreed. We have rewritten this sentence as “IDD is characterized by the apoptosis of nucleus pulposus (NP) cells, the degradation of extracellular matrix (ECM) components, and several contributing factors including genetics and environment” in line 77-80. Meanwhile, I delete reference “Jiang et al. 2013” and add reference “Feng et al. 2016” in the revised manuscript.
3. Page 7: 109-110 “DEGs, miRNA (DEMs), lncRNAs (DELs), and circRNAs (DECs) in IDD compared to normal discs tissues were identified using the “Limma” R package (Ritchie et al. 2015).”
The abbreviations in the sentence should be normalized.
Answer: Agreed. The abbreviations are corrected as “differentially expressed genes (DEGs), differentially expressed miRNAs (DEMs), differentially expressed lncRNAs (DELs), and differentially expressed circRNAs (DECs)” in line 167-168.
4. Page 10: 234-235 “NP plays an important role in maintaining the intervertebral disc homeostasis by synthesizing aggrecan, type II collagen, and extracellular matrix (ECM).”
Aggrecan and type II collagen are the components of ECM. This statement is not accurate.
Answer: Thanks for your suggestion. We have corrected this statement as “NP cells play an important role in maintaining intervertebral disc homeostasis by synthesizing ECM, which includes aggrecan and type II collagen” in line 337-338.
5. Page 6: 75-97 “Genetic factors play a significant role in the development of IDD, and variants in several genes have been identified (Battie et al. 2008). Studies have reported that vitamin D receptor (VDR), COL1A1, and COL9a3 genes are associated with the degeneration of the lumbar disc, and this is significantly more pronounced in individuals with multiple mutations (Toktas et al. 2015). In another study, bioinformatics analyses revealed that five genes (FYN, PRKCD, YWHAB, YWHAZ, and AR) were associated with IDD (Ji et al. 2015). However, the relationship between genes and IDD remains controversial and there is a need for further studies.”
Page 6: 88-92 “Tan et al. suggested that LncSNHG1 promotes NP cell proliferation by suppressing miR-326 expression and upregulating CCND1 expression (Tan et al. 2018). Wang et al. demonstrated that circRNA_4099 can sponge to miR-616-5p and eliminate the inhibition of Sox9 by miR-616-5p hence increasing the secretion of the extracellular matrix (Wang et al. 2018). However, the regulatory mechanism of the lncRNA/circRNA-miRNA-mRNA network in the IDD has not been explored.”
In the introduction segment, the authors should provide a greater perspective on the number of studies that have addressed this same research question, analyze the unsolved question exist in this aspect, and expound why this research question is urgent and meaningful.
Answer: Thanks for your valuable comments. We have rewritten the introduction segment, and cited GWAS, gene expression profiling and RNA-seq studies to clarify the importance of my research question:
The etiology of IDD is complex, but over the past several decades, it has become clear that genetic factors are most dominant. Multiple candidate genes, including thrombospondin-2, vitamin D receptor, COL2A1, ACAN, interleukins (IL1α, IL 1β, and IL6), matrix metalloproteinases (MMP-3 and MMP-9), and growth/differentiation factor 5, have been associated with the pathophysiological process of IDD development (Feng et al. 2016; Kalb et al. 2012; Wang et al. 2018b; Yuan et al. 2018). Genome-wide association studies (GWAS) have been used to help identify novel variants. Several GWAS found that the genetic polymorphisms of PARK2 and CHST3 were relevant to IDD etiology (Song et al. 2013; Williams et al. 2013). Additionally, some bioinformatics analyses based on gene expression profiles revealed that FYN, PRKCD, YWHAB, YWHAZ, AR, Fibronectin 1, COL2A1, β-catenin, COL6A2, IBSP, RAP1A, and FOXF2 genes may play key roles in IDD development (Chen et al. 2013; Guo et al. 2017; Ji et al. 2015). Although a considerable number of genes associated with IDD development have been found, early IDD diagnosis and precise treatment remain difficult and require further study.
Accumulating evidence has indicated that noncoding RNAs, including microRNAs (miRNAs), long noncoding RNAs (lncRNAs), and circular RNAs (circRNAs), are important gene expression regulators that influence cellular function and disease states (Adams et al. 2017). lncRNAs and circRNAs act as competitive endogenous RNAs (ceRNAs) and competitively bind miRNA response elements (MREs) to construct a regulatory network involved in IDD progression. Zhao et al. (2016) used RNA sequencing to identify 1,854 lncRNAs and 2,804 protein-coding genes that were differentially expressed in the IDD group. Tan et al. (2018) found that LncSNHG1 promoted NP cell proliferation by suppressing miR-326 expression and upregulating CCND1 expression. Recent studies have focused on the functional roles of circRNAs during IDD development and found that circ-4099, circ‐GRB10, circVMA21, and circ_001653 play pivotal roles during NP cell proliferation, apoptosis, and extracellular matrix synthesis/degradation (Cheng et al. 2018; Cui & Zhang 2020; Guo et al. 2018; Wang et al. 2018a). The diverse biological functions of ceRNAs deserve further exploration.
Reference:
Chen K, Wu D, Zhu X, Ni H, Wei X, Mao N, Xie Y, Niu Y, and Li M. 2013. Gene expression profile analysis of human intervertebral disc degeneration. Genet Mol Biol 36:448-454. 10.1590/s1415-47572013000300021.
Cheng X, Zhang L, Zhang K, Zhang G, Hu Y, Sun X, Zhao C, Li H, Li YM, and Zhao J. 2018. Circular RNA VMA21 protects against intervertebral disc degeneration through targeting miR-200c and X linked inhibitor-of-apoptosis protein. Ann Rheum Dis 77:770-779. 10.1136/annrheumdis-2017-212056.
Cui S, and Zhang L. 2020. circ_001653 Silencing Promotes the Proliferation and ECM Synthesis of NPCs in IDD by Downregulating miR-486-3p-Mediated CEMIP. Mol Ther Nucleic Acids 20:385-399. 10.1016/j.omtn.2020.01.026.
Feng Y, Egan B, and Wang J. 2016. Genetic Factors in Intervertebral Disc Degeneration. Genes Dis 3:178-185. 10.1016/j.gendis.2016.04.005.
Guo W, Zhang B, Li Y, Duan HQ, Sun C, Xu YQ, and Feng SQ. 2017. Gene expression profile identifies potential biomarkers for human intervertebral disc degeneration. Mol Med Rep 16:8665-8672. 10.3892/mmr.2017.7741.
Guo W, Zhang B, Mu K, Feng SQ, Dong ZY, Ning GZ, Li HR, Liu S, Zhao L, Li Y, Yu BB, Duan HQ, Sun C, and Li YJ. 2018. Circular RNA GRB10 as a competitive endogenous RNA regulating nucleus pulposus cells death in degenerative intervertebral disk. Cell Death Dis 9:319. 10.1038/s41419-017-0232-z.
Kalb S, Martirosyan NL, Kalani MY, Broc GG, and Theodore N. 2012. Genetics of the degenerated intervertebral disc. World Neurosurg 77:491-501. 10.1016/j.wneu.2011.07.014.
Song YQ, Karasugi T, Cheung KM, Chiba K, Ho DW, Miyake A, Kao PY, Sze KL, Yee A, Takahashi A, Kawaguchi Y, Mikami Y, Matsumoto M, Togawa D, Kanayama M, Shi D, Dai J, Jiang Q, Wu C, Tian W, Wang N, Leong JC, Luk KD, Yip SP, Cherny SS, Wang J, Mundlos S, Kelempisioti A, Eskola PJ, Männikkö M, Mäkelä P, Karppinen J, Järvelin MR, O'Reilly PF, Kubo M, Kimura T, Kubo T, Toyama Y, Mizuta H, Cheah KS, Tsunoda T, Sham PC, Ikegawa S, and Chan D. 2013. Lumbar disc degeneration is linked to a carbohydrate sulfotransferase 3 variant. J Clin Invest 123:4909-4917. 10.1172/jci69277.
Wang Z, Li Y, Wang Y, Wang X, Zhang J, and Tian J. 2018. Association between GDF5 single nucleotide polymorphism rs143383 and lumbar disc degeneration. Exp Ther Med 16:1900-1904. 10.3892/etm.2018.6382.
Williams FM, Bansal AT, van Meurs JB, Bell JT, Meulenbelt I, Suri P, Rivadeneira F, Sambrook PN, Hofman A, Bierma-Zeinstra S, Menni C, Kloppenburg M, Slagboom PE, Hunter DJ, MacGregor AJ, Uitterlinden AG, and Spector TD. 2013. Novel genetic variants associated with lumbar disc degeneration in northern Europeans: a meta-analysis of 4600 subjects. Ann Rheum Dis 72:1141-1148. 10.1136/annrheumdis-2012-201551.
Yuan B, Ji W, Fan B, Zhang B, Zhao Y, and Li J. 2018. Association analysis between thrombospondin-2 gene polymorphisms and intervertebral disc degeneration in a Chinese Han population. Medicine (Baltimore) 97:e9586. 10.1097/md.0000000000009586.
Zhao B, Lu M, Wang D, Li H, and He X. 2016. Genome-Wide Identification of Long Noncoding RNAs in Human Intervertebral Disc Degeneration by RNA Sequencing. BioMed Research International 2016:3684875. 10.1155/2016/3684875.
Experimental design
Question:6. Page 7: 110-115 “Different thresholds were used according to the conditions of each dataset. The specific screening thresholds for DEGs were as follows: GSE34095 dataset was p-value< 0.05; GSE15227 dataset was p-value< 0.01 & |log2FC|≥2; GSE116726 dataset was p-value< 0.01& |log2 FC|>2. The GSE67566, GSE63492, and GSE56081 datasets were obtained from the same tissue samples, and the differential expression analysis results and thresholds were uploaded in the raw data (Table S1,S2;Document S1,S2）.”
P-value< 0.05 and |log2FC|≥2 is the generally accepted screening thresholds for the differential expressed genes. In the authors’ study, different thresholds were set for several data sets. Context should be added to further explain the rationale for the setting of different thresholds.
Answer: The reviewer’s comment is helpful. Because each dataset comes from a different experiment and microarray platform, the way to get the data may be different. If the screening is performed according to a consistent threshold, some datasets may not get effective differentially expressed genes, and subsequent analysis cannot be performed. Therefore, in the case of statistical significance, this analysis adjusts the threshold to obtain reasonable differentially expressed genes for subsequent analysis.
Validity of the findings
Question:7. Page 11: 257-262 “SCRG1 (scrapie responsive gene 1) was first discovered in the brain of scrapie-infected mice (Dandoy-Dron et al. 1998). Recent studies have shown that SCRG1 is closely related to neurodegeneration, and is an important regulator of human mesenchymal stem cells (hMSCs) self-renewal, migration and osteogenic differentiation with its receptor BST1 (Aomatsu et al. 2014; Dron et al. 2006). SCRG1 gene encodes a 98-amino acid, a cytokine-like protein that is highly conserved in mammals and has no significant homology to any other known protein (Dandoy-Dron et al. 2003; Dron et al. 2000).”
SCRG1 is a member of human genes, the Ensemble Gene ID of which is ENSG00000164106. The description of SCRG1 in Ensemble is “stimulator of chondrogenesis 1”. Scrg1 is a member of the mouse genetic base, the description of which in Ensemble is “scrapie responsive gene 1”, with the gene ID of ENSMUSG00000031610. The discussion should be expanded about the SCRG1, not Scrg1.
Answer: Thanks for your constructive comments. We have corrected the mistake and expanded the discussion of SCRG1 as follows:
SCRG1 transcript is found in the brain, heart, and spinal cord, and its sequence is highly conserved in humans, mice, and rats. SCRG1 has also been observed to be specifically expressed in human articular cartilage, and is involved in human mesenchymal stem cell (hMSC) growth suppression and differentiation during dexamethasone-dependent chondrogenesis (Ochi et al. 2006). Recent studies have shown that SCRG1 is an important regulator during hMSC self-renewal, migration, and osteogenic differentiation along with its receptor BST1 (Aomatsu et al. 2014).
Reference:
Ochi K, Derfoul A, and Tuan RS. 2006. A predominantly articular cartilage-associated gene, SCRG1, is induced by glucocorticoid and stimulates chondrogenesis in vitro. Osteoarthritis and Cartilage 14:30-38. https://doi.org/10.1016/j.joca.2005.07.015.
8. There are several statements and recommendations in the discussion section that are not supported by this study. The authors are encouraged to remove biased language and hyperbole throughout the manuscript. such as:
page 11: 262-263 “The high conservation suggests that BCAS4 and SCRG1 can serve as good diagnostic markers of IDD.”
page 11: 278-279 “Therefore, based on these previous findings, this study presents new insights into the molecular mechanism of IDD development.”
page 12: 296-297 “These are reliable and effective molecular targets for the diagnosis and treatment of IDD.”
Answer: Thank you for your suggestions. We have removed biased language and hyperbole throughout the manuscript.
9. The tissue samples used in the six data sets were different. While GSE34095 and GSE15227 are the disc tissue, the other four are the NP tissue. The analysis across the six data sets may led to a violated validation of results.
The authors are encouraged to consider discussing the limitations of the study, including but not limited to the item mentioned above.
Answer: Thank you for your suggestions. In this analysis, both disc tissue and NP tissue are treated as tissue of IDD, instead of being treated as two tissues for analysis, respectively. Thereby, we can identify the common feature genes in the entire process of IDD rather than a certain part of the intervertebral disc.
Comments for the author
The authors presented an integrated analysis of several microarray datasets. Bioinformatics and machine learning methods were employed in the analysis. Several genes, miRNAs, and circRNAs were revealed as the key regulators involved in the degenerated progress. Crosstalk among mRNAs, miRNAs, lncRNAs, and circRNAs in IDD was displayed in the paper. The research perspective is very novel. I appreciate their eagerness to provide insights into this area.
Answer: We greatly appreciate the recognition of our work by the reviewer, and we express our sincere gratitude here.
Replies to Reviewer #2:
Basic reporting
Question: The paper is devoted to a meta-analysis of publicly available expression datasets for intervertebral disc degeneration by machine learning techniques. The paper is mostly well written, well referenced and properly organised. Figures and tables are of good quality. All the results are clearly presented and discussed. There are some minor inaccuracies in the typing, namely spaces between end of sentences and full stops, e.g. '(Dandoy-Dron et al. 1998) .'
Answer: We are very grateful for your recognition of our work. We have checked the entire manuscript for minor inaccuracies in the typing.
Experimental design
The design of the study is sound, the study pipeline is clear, the methods are adequate. The study hypothesis is well defined. Methods are fully described and the study seems reproducible. A set of various ML methods was applied with cross-validation between them, thus increasing validity of the findings. This type of studies in rare in the field, therefore, must be encouraged.
Answer: We are very grateful for your recognition of our work.
Validity of the findings
Findings seem to be meaningful, some of the genes identified are known to be involved in IDD from other studies, thus proving the validity of the study. Any possible bias may be due to the use of publicly available datasets that can be biased one way or another. Discussion is relevant. All the raw and summary data is provided.
Answer: We thank you again for your recognition of our work.
Comments for the author
I have no specific issues with this study, it seems to be well thought and implemented. There are some minor issues to be addressed:
Question1. Make sure all gene names are in italic.
Answer: Thanks for your suggestion. I have checked the entire manuscript and italicized all gene names.
Question 2. In introduction, you may want to mention the results of GWASs on IDD.
Answer: The reviewer’s comment is helpful. I have cited the results of GWAS on IDD in line 102-106:
Genome-wide association studies (GWAS) have been used to help identify novel variants. Several GWAS found that the genetic polymorphisms of PARK2 and CHST3 were relevant to IDD etiology (Song et al. 2013; Williams et al. 2013).
Reference:
Song YQ, Karasugi T, Cheung KM, Chiba K, Ho DW, Miyake A, Kao PY, Sze KL, Yee A, Takahashi A, Kawaguchi Y, Mikami Y, Matsumoto M, Togawa D, Kanayama M, Shi D, Dai J, Jiang Q, Wu C, Tian W, Wang N, Leong JC, Luk KD, Yip SP, Cherny SS, Wang J, Mundlos S, Kelempisioti A, Eskola PJ, Männikkö M, Mäkelä P, Karppinen J, Järvelin MR, O'Reilly PF, Kubo M, Kimura T, Kubo T, Toyama Y, Mizuta H, Cheah KS, Tsunoda T, Sham PC, Ikegawa S, and Chan D. 2013. Lumbar disc degeneration is linked to a carbohydrate sulfotransferase 3 variant. J Clin Invest 123:4909-4917. 10.1172/jci69277.
Williams FM, Bansal AT, van Meurs JB, Bell JT, Meulenbelt I, Suri P, Rivadeneira F, Sambrook PN, Hofman A, Bierma-Zeinstra S, Menni C, Kloppenburg M, Slagboom PE, Hunter DJ, MacGregor AJ, Uitterlinden AG, and Spector TD. 2013. Novel genetic variants associated with lumbar disc degeneration in northern Europeans: a meta-analysis of 4600 subjects. Ann Rheum Dis 72:1141-1148. 10.1136/annrheumdis-2012-201551.
Question 3. line 123: of the DEGs (Kuhn 2015). Recursive feature elimination (RFE), which is a machine-learning method, was used - which is a machine-learning method phrase is obsolete here.
Answer: Because the number of genes obtained from intersection is small, the RFE method can continuously reduce the size of the feature set to select the desired features by the recursion method. Although it is a relatively early machine learning algorithm, RFE can efficiently get the results we expected, and the results have also been verified. Of course, in the future analysis, we will continue to try new machine learning models.
Question 4. line 187: The parameters of the SVM model are shown in Fig.S3. In - just list the parameters in the text, no need for this figure.
Answer: Thanks for your suggestion. We have listed the parameters of the SVM model in line 276-277 and removed the Fig.S3.
Question 5. lines 201-202: including physical interactions, co-expression, predicted, co-localization, pathway, genetic interactions and shared protein domains, accounted for 67.64%, 13.50%, 6.35%, 6.17%, 4.35%, 1.40% and 0.59%, respectively. - it is unclear what this percentages represent, account for what? This needs to be rephrased.
Answer: Given the key genes we finally screened are BCAS4 and SCRG1, the gene function prediction of CRNKL1 is meaningless. We removed the relevant paragraphs from the revised manuscript.
Question 6. Line 239: Biology -> biology.
Answer: Thanks for your suggestion. The “Biology” has changed to “biology” in line 344.
Question 7. line 294: 'In conclusion, in this study, BCAS4 and SCRG1as were identified as key genes associated with the progression' - you cannot claim these genes are associated with the progression of the IDD as you analysed cross-sectional data; to talk about progression you need to analysed longitudinal data. I would rephrased 'key genes associated with the development'.
Answer: Thanks for your suggestion. We have rewritten the sentence as “In this study, we used integrated bioinformatics analysis and machine learning methods to identify BCAS4 and SCRG1 as key genes associated with IDD development” in line 425-427.
Replies to Reviewer #3:
Basic reporting
Question: The manuscript presents a valuable analysis of biomarker genes relevant to intervertebral disc degeneration (IDD). The topic itself is of importance since the gene signatures shed light on diagnosis and treatment of IDD. The paper is in general well organized. It is however not easy to follow – this is to some degree expected, given that many datasets and many bioinformatics approaches were introduced. I would suggest add more detailed and concrete explanations, especially detailed description of relationship between non-coding RNA and genes. The authors didn’t cite enough sources to provide background about this study: 1) several studies have been done to identify potential biomarkers for IDD using microarray data (DOI: 10.3892/mmr.2017.7741, https://www.scielo.br/pdf/gmb/v36n3/a21v36n3.pdf); 2) Since RNA-Seq has emerged as an alternative method for gene expression profiling, the results from RNA-Seq can be cited (i.e. https://doi.org/10.1155/2016/3684875). The workflow in Fig. S1 is very helpful for understanding experimental design in this study, but it should be well organized (i.e. all GEO dataset should be placed on the top; and arrow line shouldn’t cross boxes).
Answer: Thanks for your constructive comments. The studies mentioned by the reviewer have been supplemented in the second and fourth paragraph of introduction, as follow:
Additionally, some bioinformatics analyses based on gene expression profiles revealed that FYN, PRKCD, YWHAB, YWHAZ, AR, Fibronectin 1, COL2A1, β-catenin, COL6A2, IBSP, RAP1A, and FOXF2 genes may play key roles in IDD development (Chen et al. 2013; Guo et al. 2017; Ji et al. 2015).
Zhao et al. (2016) used RNA sequencing to identify 1,854 lncRNAs and 2,804 protein-coding genes that were differentially expressed in the IDD group.
Reference:
Chen K, Wu D, Zhu X, Ni H, Wei X, Mao N, Xie Y, Niu Y, and Li M. 2013. Gene expression profile analysis of human intervertebral disc degeneration. Genet Mol Biol 36:448-454. 10.1590/s1415-47572013000300021.
Guo W, Zhang B, Li Y, Duan HQ, Sun C, Xu YQ, and Feng SQ. 2017. Gene expression profile identifies potential biomarkers for human intervertebral disc degeneration. Mol Med Rep 16:8665-8672. 10.3892/mmr.2017.7741.
Zhao B, Lu M, Wang D, Li H, and He X. 2016. Genome-Wide Identification of Long Noncoding RNAs in Human Intervertebral Disc Degeneration by RNA Sequencing. BioMed Research International 2016:3684875. 10.1155/2016/3684875.
In addition, we have redrawn the workflow as follow:
Experimental design
Question: (1) The topic itself fall within the scope of PeerJ and will be of interests to the readers of the journal. However, the aim and purpose of the study was not clearly presented: 1) why integrating multiple microarray datasets can improve identification of biomarkers; 2) what roles machine learning methods play in this study? (2) In Line 112, the different thresholds (p-value < 0.05 & p-value < 0.01) were used and the authors didn’t explain the reason why using different p-values. (3) In “SVM model verification”, the authors should include more details about SVM models, such as kernel (it is very important, so don’t just put R output in Fig S3). (4) I am wondering how the training dataset(s) and validation dataset(s) were decided. Also, there are different machine learning methods (random forest, K-nearest neighbors, decision tree etc.) which can be used for verification. So, the authors can apply these methods and do comparison.
Answer: The reviewer’s comment is helpful. (1) We have added a paragraph in the introduction to clarify the advantages of integrated analysis and machine learning in identifying key genes, as follow:
Integrated analyses using combined multiple microarray data can provide a more accurate understanding of the interplay across multi-level genomic features and the molecular mechanisms that cause complex diseases (Momtaz et al. 2018). Machine learning is a type of artificial intelligence that can 'learn' a model using past data in order to predict future data. Machine learning algorithms have been used in key feature training, recognition, and group classification (Huang et al. 2018). Modern researchers have unprecedented access to machine learning methods that can elucidate complex molecular mechanisms and predict disease genes from large biomedical datasets (Libbrecht & Noble 2015; Obermeyer & Emanuel 2016). To the best of our knowledge, no investigations have used integrated analysis and machine learning methods to identify key IDD-associated genes.
Reference:
Momtaz R, Ghanem NM, El-Makky NM, and Ismail MA. 2018. Integrated analysis of SNP, CNV and gene expression data in genetic association studies. Clin Genet 93:557-566. 10.1111/cge.13092.
Huang S, Cai N, Pacheco PP, Narrandes S, Wang Y, and Xu W. 2018. Applications of Support Vector Machine (SVM) Learning in Cancer Genomics. Cancer Genomics Proteomics 15:41-51. 10.21873/cgp.20063.
Libbrecht MW, and Noble WS. 2015. Machine learning applications in genetics and genomics. Nat Rev Genet 16:321-332. 10.1038/nrg3920.
Obermeyer Z, and Emanuel EJ. 2016. Predicting the Future — Big Data, Machine Learning, and Clinical Medicine. New England Journal of Medicine 375:1216-1219. 10.1056/NEJMp1606181.
(2) We have explained the reason why using different p-values in line 170-174, as follow:
Since each dataset came from a different experiment and microarray platform, the way to get the data may be different. When filtering using consistent thresholds, some datasets did not yield valid differential genes. In order to obtain effective differentially expressed genes for subsequent analysis, we used different thresholds according to each dataset’s conditions.
(3) We have added more details about SVM models in the method and results segment, as follows:
The support vector machine (SVM) is particularly effective for binary classification in a supervised learning manner, and is better than other machine learning methods at identifying subtle patterns in complex datasets (Aruna & Dr 2011). Radial basis function kernel (RBF kernel) is commonly used in nonlinear support vector machine classification since it can enable data to operate in a high-dimensional and implicit feature space (Jiao et al. 2017). In this study, we used the R e1071 package (https://CRAN.R-project.org/package=e1071) to build an RBF kernel SVM model to identify the optimal feature genes (Dimitriadou et al. 2012), and we selected the GSE15227 dataset to train the machine. We then evaluated its performance in the GSE15227 training set using the R pROC package (http://www.biomedcentral.com/1471-2105/12/77/) (Robin et al. 2011).
A classification model was constructed using the RBF kernel SVM, the four genes as features, and the GSE15227 dataset to train the machine. The parameters were set as: SVM-Kernel: radial, cost:1, gamma:0.25, and epsilon:0.1. Additionally, their performance was assessed using the R pROC package.
Reference:
Aruna S, and Dr S. 2011. A Novel SVM based CSSFFS Feature Selection Algorithm for Detecting Breast Cancer. International Journal of Computer Applications 31:14-20.
Jiao P, Cai F, Feng Y, and Wang W. 2017. Link predication based on matrix factorization by fusion of multi class organizations of the network. Scientific reports 7:8937-8937. 10.1038/s41598-017-09081-9
(4) In essence, there is no difference between training set and validation set. When training a supervised machine learning model, the data will be divided into training set and validation set to select the model with the best accuracy and generalization ability. In addition, we have used four machine learning methods (random forest, neural network, recursive feature elimination and support vector machine) to identify the optimal feature genes in the development of IDD in this study. Four optimal feature genes have been identified, and intervertebral disc degenerated samples and normal samples can be accurately distinguished.
Validity of the findings
Question：（1）Since several studies have been done for identification of biomarkers for IDD, the authors need to claim the novelty of this study. Since only 13 genes were found overlap between DEGs from GSE34095 and GSE15227, they could be overlapped by chance (False positives). So, the authors should implement overlap test (i.e. hypergeometric test). （2）In Line 192, I don’t think the sample size is reason of low AUC for GSE34095.（3） In this study, BCAS4 and SCRG1 were identified as final key genes, however CRNKL1 was used to predict the relevant genes in Line 198 and 199. It doesn’t make sense.
Answer: Thanks for your suggestions. (1) Differential expression genes obtained from the two datasets are inherently statistically significant and the samples from the original dataset can be accurately clustered through bidirectional clustering. Moreover, many studies have used the method of intersecting multiple datasets to further determine the differential expression genes. We think that there is no need for perform overlapping tests such as hypergeometric tests.
(2) The principal components analysis (PCA) result showed that the features between the normal and disease samples of GSE34095 were not very evident (data is not uploaded), and with the small sample size of GSE34095 dataset (only 6 samples), a typing errors of one sample would have greatly reduced the efficacy of the model, so we think that the small sample size may cause low AUC.
The principal components analysis of GSE34095
(3) We have removed the CRNKL1 interaction analysis and related results and Figure 3 in the revised manuscript.
Comments for the author
The manuscript presents a valuable analysis of biomarker genes relevant to intervertebral disc degeneration (IDD). The topic itself is of importance since the gene signatures shed light on diagnosis and treatment of IDD. I would suggest add more detailed and concrete explanations.
Answer: Special thanks to you for your good comments. We have added more detailed and concrete explanations according to your comments.
" | Here is a paper. Please give your review comments after reading it. |
9,758 | "<ns0:abstract xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:p>Intervertebral disc degeneration (IDD), a major cause of lower back pain, has multiple contributing factors including genetics, environment, age, and loading history.</ns0:p><ns0:p>Bioinformatics analysis has been extensively used to identify diagnostic biomarkers and therapeutic targets for IDD diagnosis and treatment. However, multiple microarray dataset analysis and machine learning methods have not been integrated. In this study, we downloaded the mRNA, microRNA (miRNA), long noncoding RNA (lncRNA), and circular RNA (circRNA) expression profiles (GSE34095, GSE15227, GSE63492, GSE116726, GSE56081, and GSE67566) associated with IDD from the GEO database. Using differential expression analysis and recursive feature elimination, we extracted four optimal feature genes. We then used the support vector machine (SVM) to make a classification model with the four optimal feature genes. The ROC curve was used to evaluate the model's performance, and the expression profiles (GSE63492, GSE116726, GSE56081, and GSE67566) were used to construct a competitive endogenous RNA (ceRNA) regulatory network and explore the underlying mechanisms of the feature genes. We found that three miRNAs (hsa-miR-4728-5p, hsa-miR-5196-5p, and hsa-miR-185-5p) and three circRNAs (hsa_circRNA_100723, hsa_circRNA_104471, and hsa_circRNA_100750) were important regulators with more interactions than the other RNAs across the whole network. The expression level analysis of the three datasets revealed that BCAS4 and SCRG1 were key genes involved in IDD development. Ultimately, our study proposes a novel approach to determining reliable and effective targets in IDD diagnosis and treatment.</ns0:p></ns0:div>
</ns0:abstract>
<ns0:body xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:head>Introduction</ns0:head><ns0:p>Lower back pain, commonly caused by intervertebral disc degeneration (IDD), can be a significant socioeconomic burden on patients <ns0:ref type='bibr' target='#b47'>(Vergroesen et al. 2015)</ns0:ref>. IDD is characterized by the apoptosis of nucleus pulposus (NP) cells, the degradation of extracellular matrix (ECM) components, and several contributing factors including genetics and environment <ns0:ref type='bibr' target='#b4'>(Battie et al. 2008;</ns0:ref><ns0:ref type='bibr' target='#b12'>Feng et al. 2016)</ns0:ref>. However, the precise etiology of IDD remains largely unknown. Diagnosing degenerative disc disease is difficult, and common IDD treatment and management strategies primarily consist of conservative management or surgical treatment to relieve pain, without resolving the underlying tissue pathology <ns0:ref type='bibr' target='#b0'>(An et al. 2003;</ns0:ref><ns0:ref type='bibr' target='#b55'>Zaina et al. 2016</ns0:ref>). To identify potential biomarkers and specific therapeutic targets, a more detailed understanding of the molecular and cellular events underlying IDD formation is needed. The etiology of IDD is complex, but over the past several decades, it has become clear that genetic factors are most dominant. Multiple candidate genes, including thrombospondin-2, vitamin D receptor, COL2A1, <ns0:ref type='bibr'>ACAN,</ns0:ref><ns0:ref type='bibr'>interleukins (IL1α,</ns0:ref><ns0:ref type='bibr'>IL 1β,</ns0:ref><ns0:ref type='bibr'>and IL6)</ns0:ref>, matrix metalloproteinases (MMP-3 and MMP-9), and growth/differentiation factor 5, have been associated with the pathophysiological process of IDD development <ns0:ref type='bibr' target='#b12'>(Feng et al. 2016;</ns0:ref><ns0:ref type='bibr' target='#b21'>Kalb et al. 2012;</ns0:ref><ns0:ref type='bibr' target='#b51'>Wang et al. 2018b;</ns0:ref><ns0:ref type='bibr' target='#b54'>Yuan et al. 2018</ns0:ref>). Genome-wide association studies (GWAS) have been used to help identify novel variants. Several GWAS found that the genetic polymorphisms of PARK2 and CHST3 relevant to IDD etiology <ns0:ref type='bibr'>(Song et al. 2013;</ns0:ref><ns0:ref type='bibr' target='#b52'>Williams et al. 2013)</ns0:ref>. Additionally, some bioinformatics analyses based on gene expression profiles revealed that FYN, PRKCD, YWHAB, YWHAZ, AR, Fibronectin 1, COL2A1, β-catenin, COL6A2, IBSP, RAP1A, and FOXF2 genes may play key roles in IDD development <ns0:ref type='bibr' target='#b6'>(Chen et al. 2013;</ns0:ref><ns0:ref type='bibr' target='#b14'>Guo et al. 2017;</ns0:ref><ns0:ref type='bibr' target='#b19'>Ji et al. 2015)</ns0:ref>. Although a considerable number of genes associated with IDD development have been found, early IDD diagnosis and precise treatment remain difficult and require further study. Integrated analyses using combined multiple microarray data can provide a more accurate understanding of the interplay across multi-level genomic features and the molecular mechanisms that cause complex diseases <ns0:ref type='bibr' target='#b31'>(Momtaz et al. 2018)</ns0:ref>. Machine learning is a type of artificial intelligence that can 'learn' a model using past data in order to predict future data. Machine learning algorithms have been used in key feature training, recognition, and group classification <ns0:ref type='bibr' target='#b17'>(Huang et al. 2018)</ns0:ref>. Modern researchers have unprecedented access to machine learning methods that can elucidate complex molecular mechanisms and predict disease genes from large biomedical datasets <ns0:ref type='bibr' target='#b27'>(Libbrecht & Noble 2015;</ns0:ref><ns0:ref type='bibr' target='#b32'>Obermeyer & Emanuel 2016)</ns0:ref>. To the best of our knowledge, no investigations have used integrated analysis and machine learning methods to identify key IDD-associated genes. Accumulating evidence has indicated that noncoding RNAs, including microRNAs (miRNAs), long noncoding RNAs (lncRNAs), and circular RNAs (circRNAs), are important gene expression regulators that influence cellular function and disease states <ns0:ref type='bibr'>(Adams et al. 2017)</ns0:ref>. lncRNAs and circRNAs act as competitive endogenous RNAs (ceRNAs) and competitively bind miRNA response elements (MREs) to construct a regulatory network involved in IDD progression. <ns0:ref type='bibr' target='#b56'>Zhao et al. (2016)</ns0:ref> used RNA sequencing to identify 1,854 lncRNAs and 2,804 protein-coding genes that were differentially expressed in the IDD group. <ns0:ref type='bibr'>Tan et al. (2018)</ns0:ref> found that LncSNHG1 promoted NP cell proliferation by suppressing miR-326 expression and upregulating CCND1 expression. Recent studies have focused on the functional roles of circRNAs during IDD development and found that circ-4099, circ-GRB10, circVMA21, and circ_001653 play pivotal roles during NP cell proliferation, apoptosis, and extracellular matrix synthesis/degradation <ns0:ref type='bibr' target='#b7'>(Cheng et al. 2018;</ns0:ref><ns0:ref type='bibr'>Cui & Zhang 2020;</ns0:ref><ns0:ref type='bibr' target='#b15'>Guo et al. 2018;</ns0:ref><ns0:ref type='bibr' target='#b49'>Wang et al. 2018a</ns0:ref>). The diverse biological functions of ceRNAs deserve further exploration. In this study, we aimed to identify the key genes and underlying mechanisms of IDD development by constructing an lncRNA/circRNA-miRNA-mRNA network using multiple microarray datasets and machine learning methods. Fig. <ns0:ref type='figure' target='#fig_4'>S1</ns0:ref> shows the flow chart for this study. Our results present novel biomarkers and therapeutic targets that can be used for IDD diagnosis and treatment.</ns0:p></ns0:div>
<ns0:div><ns0:head>Materials & Methods</ns0:head></ns0:div>
<ns0:div><ns0:head>Microarray datasets</ns0:head><ns0:p>We retrieved a total of six expression profiles from the GEO database (www.ncbi.Nlm.nih.gov/geo): two mRNA expression profiles (GSE34095 <ns0:ref type='bibr' target='#b45'>(Tsai et al. 2013</ns0:ref>) and GSE15227 <ns0:ref type='bibr' target='#b13'>(Gruber et al. 2009</ns0:ref>)), two miRNA expression profiles (GSE63492 <ns0:ref type='bibr' target='#b24'>(Lan et al. 2016</ns0:ref>) and GSE116726 <ns0:ref type='bibr' target='#b18'>(Ji et al. 2018</ns0:ref>)), one mRNA-lncRNA expression profile (GSE56081 <ns0:ref type='bibr' target='#b24'>(Lan et al. 2016;</ns0:ref><ns0:ref type='bibr' target='#b48'>Wan et al. 2014</ns0:ref>)), and one circRNA expression profile (GSE67566 <ns0:ref type='bibr' target='#b24'>(Lan et al. 2016;</ns0:ref><ns0:ref type='bibr' target='#b30'>Liu et al. 2015)</ns0:ref>). Table <ns0:ref type='table' target='#tab_0'>1</ns0:ref> contains the basic information for these expression profiles. Manuscript to be reviewed</ns0:p></ns0:div>
<ns0:div><ns0:head>Differential expression analysis</ns0:head><ns0:p>The raw data were annotated, normalized, log 2 transformed, and screened for differentially expressed genes (DEGs), differentially expressed miRNAs (DEMs), differentially expressed lncRNAs (DELs), and differentially expressed circRNAs (DECs) in IDD and normal disc tissue using the 'Limma' R package <ns0:ref type='bibr' target='#b37'>(Ritchie et al. 2015)</ns0:ref>. Since each dataset came from a different experiment and microarray platform, the way to get the data may be different. When filtering using consistent thresholds, some datasets did not yield valid differential genes. In order to obtain effective differentially expressed genes for subsequent analysis, we used different thresholds according to each dataset's conditions. The specific DEG screening thresholds were as follows: p-value < 0.05 for the GSE34095 dataset, p-value < 0.01 and |log2FC| ≥ 2 for the GSE15227 dataset, and p-value < 0.01 and |log2 FC| > 2 for the GSE116726 dataset. We obtained the GSE67566, GSE63492, and GSE56081 datasets from the same tissue samples, and provided the differential expression analysis results and thresholds as Supplemental Materials (Tables <ns0:ref type='table' target='#tab_2'>S1 and S2</ns0:ref>, Documents S1and S2). We visualized the differential expression analysis results using a volcano plot and heatmap with hierarchical clustering (Fig. <ns0:ref type='figure'>S2</ns0:ref>).</ns0:p></ns0:div>
<ns0:div><ns0:head>Feature gene extraction</ns0:head><ns0:p>Using VENNY 2.1 software (http://bioinfogp.cnb.csic.es/tools/venny/index.html), we extracted the differential gene intersections of the GSE34095 and GSE15227 datasets to use as IDD-related DEGs <ns0:ref type='bibr' target='#b26'>(Li et al. 2019;</ns0:ref><ns0:ref type='bibr' target='#b57'>Zhang et al. 2018)</ns0:ref>. We used the GSE15227 dataset as the training set to screen for important feature genes. Using the R caret package, random forest, and neural network methods, we constructed the model and obtained the important DEG features <ns0:ref type='bibr' target='#b23'>(Kuhn 2015)</ns0:ref>. We used recursive feature elimination (RFE), a machine learning method, to extract the optimum feature genes to identify the functional biomarkers involved in IDD progression <ns0:ref type='bibr' target='#b16'>(Guyon et al. 2002)</ns0:ref>.</ns0:p></ns0:div>
<ns0:div><ns0:head>SVM model verification</ns0:head><ns0:p>The support vector machine (SVM) is particularly effective for binary classification in a supervised learning manner, and is better than other machine learning methods at identifying subtle patterns in complex datasets <ns0:ref type='bibr' target='#b2'>(Aruna & Dr 2011)</ns0:ref>. Radial basis function kernel (RBF kernel) is commonly used in nonlinear support vector machine classification since it can enable data to operate in a high-dimensional and implicit feature space <ns0:ref type='bibr' target='#b20'>(Jiao et al. 2017)</ns0:ref>. In this study, we used the R e1071 package (https://CRAN.R-project.org/package=e1071) to build an RBF kernel SVM model to identify the optimal feature genes <ns0:ref type='bibr' target='#b10'>(Dimitriadou et al. 2012</ns0:ref>), and we selected the GSE15227 dataset to train the machine. We then evaluated its performance in the GSE15227 training set using the R pROC package (http://www.biomedcentral.com/1471-2105/12/77/) <ns0:ref type='bibr' target='#b38'>(Robin et al. 2011)</ns0:ref>. The optimal feature genes were taken from the training set, and over-fitting would occur whenever ROC verification was performed. Therefore, we used two validation sets (GSE34095 and GSE56081) to further verify the model's performance.</ns0:p></ns0:div>
<ns0:div><ns0:head>Identifying target miRNAs of the optimal feature genes</ns0:head><ns0:p>We used miRWalk 2.0 with 12 prediction programs (MIMATid, Microt4, miRanda, mirbridge, miRDB, miRMap, miRNAMap, Pictar2, PITA, RNA22, RNAhybrid, and Targetscan) to predict the optimal feature genes' target miRNAs <ns0:ref type='bibr' target='#b11'>(Dweep & Gretz 2015)</ns0:ref>. miRNAs present in more than five of the 12 prediction programs were considered as target miRNAs. We selected overlapping DEMs in the GSE116726 and GSE63492 datasets as candidate DEMs. Intersecting target miRNAs and candidate DEMs were selected as optimal feature gene interaction pairs. ceRNA network construction We obtained the miRNA sequences interacting with optimal feature genes from the miRbase (Kozomara & Griffiths-Jones 2011) and extracted mature sequences using the Perl program. The DEL sequences were downloaded from NCBI. For DELs with several transcripts, we selected the longest transcripts for subsequent analysis. We used the BEDTOOLS command <ns0:ref type='bibr' target='#b36'>(Quinlan & Hall 2010)</ns0:ref> and genomic coordinates to obtain the DEC sequences, and converted the gene name to its circRNA symbol using the Perl program. We then used the miRanda tool to analyze the combinations of miRNAs, DELs, and DECs. We set the analysis parameters for the miRNA-DELs and miRNA-DECs as sc:120, en:-20 and sc:150, en:-7, respectively, and processed the results using python script (Documents S3 and S4). The DELs and DECs with ≥5 miRNA binding sites were identified as reliable miRNA-lncRNA and miRNA-circRNA interaction pairs. The ceRNA (DELs/DECs-miRNA-optimal feature gene) regulatory network was constructed using a combination of miRNA-DEL pairs, miRNA-DEC pairs, and miRNA-optimal feature gene pairs. We visualized the network using Cytoscape 3.6.0 (http://www.cytoscape.org/).</ns0:p></ns0:div>
<ns0:div><ns0:head>Statistical analysis</ns0:head><ns0:p>All data were presented as the mean±SEM. Differential expression levels were compared using the Student's ttest in GraphPad Prism 7.0 (GraphPad Software Inc., La Jolla, CA, USA). P values < 0.05 were considered statistically significant.</ns0:p></ns0:div>
<ns0:div><ns0:head>Results</ns0:head></ns0:div>
<ns0:div><ns0:head>Differentially expressed genes in IDD</ns0:head><ns0:p>We obtained a total of 334 DEGs (199 up-regulated and 135 down-regulated) in the GSE34095 dataset (Fig. <ns0:ref type='figure' target='#fig_4'>1A</ns0:ref>) and 188 DEGs (141 up-regulated and 47 down-regulated) in the GSE15227 dataset (Fig. <ns0:ref type='figure' target='#fig_4'>1B</ns0:ref>). Hierarchical cluster heatmaps showed that these DEGs could distinguish between the degenerative disc samples and the control disc samples (Fig. <ns0:ref type='figure' target='#fig_4'>1C and 1D</ns0:ref>). We obtained a total of 13 overlapping DEGs (COL3A1, SCRG1, HTRA1, BCAS4, C11orf80, CRNKL1, GREM1, FGFR3, BDKRB1, WDR46, FN1, LMF2, and GDI2) via the intersection between the two datasets, and considered these as IDD-related DEGs (Fig. <ns0:ref type='figure' target='#fig_4'>1E</ns0:ref>).</ns0:p></ns0:div>
<ns0:div><ns0:head>Optimal feature gene selection</ns0:head><ns0:p>We used random forest and neural network models to evaluate the importance of the 13 overlapping DEGs (Documents S5 and S6). The results showed that most of the DEGs from the two models had minimal differences in terms of gene importance rankings (Fig. <ns0:ref type='figure'>2A and 2B</ns0:ref>). Using the RFE method, we identified four important genes (WDR46, BCAS4, CRNKL1, and SCRG1) as optimal feature genes associated with IDD (Fig. <ns0:ref type='figure'>2C</ns0:ref>).</ns0:p><ns0:p>A classification model was constructed using the RBF kernel SVM, the four genes as features, and the GSE15227 dataset to train the machine. The parameters were set as: SVM-Kernel: radial, cost:1, gamma:0.25, and epsilon:0.1. Additionally, their performance was assessed using the R pROC package. In the GSE15227 training set, the area under the ROC curve (AUC) of the SVM model was 100 percent, suggesting that the model could accurately distinguish between IDD and normal samples (Fig. <ns0:ref type='figure'>2D</ns0:ref>). We used the GSE34095 and GSE56081 datasets as validation sets to further evaluate the model's performance and to help avoid over-fitting in the training set. The SVM model in the GSE34095 validation set had an AUC of 55.6 percent, which may be correlated with the small sample size of the microarray dataset (Fig. <ns0:ref type='figure'>2E</ns0:ref>). However, the AUC for the GSE56081 validation set was 100 percent (Fig. <ns0:ref type='figure'>2F</ns0:ref>). These results indicated that the optimal feature genes (WDR46, BCAS4, CRNKL1, and SCRG1) could be used as effective and accurate IDD diagnostic biomarkers. Identifying target miRNAs of the optimal feature genes We predicted a total of 467 target miRNA optimal feature genes using miRWalk 2.0. Compared to the control disc samples, we identified 724 DEMs (527 up-regulated and 197 down-regulated) from the GSE116726 dataset (Fig. <ns0:ref type='figure' target='#fig_6'>3A and 3B</ns0:ref>) and 149 DEMs from the GSE63492 dataset. Fig. <ns0:ref type='figure' target='#fig_6'>3C</ns0:ref> shows a Venn diagram of the 46 common DEMs that were found. Based on the intersections between the target miRNAs and common DEMs, we further analyzed 12 overlapping miRNAs which we considered to be the optimal feature genes' interaction miRNAs (Fig. <ns0:ref type='figure' target='#fig_6'>3D</ns0:ref>).</ns0:p></ns0:div>
<ns0:div><ns0:head>ceRNA network construction</ns0:head><ns0:p>The differentially expressed lncRNAs and circRNAs between the degenerated disc samples and the control disc samples were downloaded from the GEO database. We analyzed the miRNA-DEL interactions and the miRNA-DEC interactions using the miRanda tool with ≥5 miRNA binding sites. The miRNA-DEL pairs, miRNA-DEC pairs, and miRNA-optimal feature gene pairs were combined to build a DELs/DECs-DEMs-optimal feature gene regulatory network, which included four mRNA nodes, 12 miRNA nodes, 10 lncRNA nodes, and 75 circRNA nodes (Fig. <ns0:ref type='figure' target='#fig_2'>4</ns0:ref>, Table <ns0:ref type='table'>S3</ns0:ref>). Out of these, three miRNAs (hsa-miR-4728-5p, hsa-miR-5196-5p, and hsa-miR-185-5p) and three circRNAs (hsa_circRNA_100723, hsa_circRNA_104471, and hsa_circRNA_100750) were key regulators, based on the optimal feature genes and connective degrees of the whole network (Table <ns0:ref type='table' target='#tab_2'>2</ns0:ref>). However, the LncRNA connective degree was very low.</ns0:p></ns0:div>
<ns0:div><ns0:head>Validating optimal feature genes</ns0:head><ns0:p>The expression of the four optimal feature genes across the three datasets was visualized using box plots. In the GSE15227 and GSE56081 datasets, BCAS4 expression levels were significantly down-regulated (Fig. <ns0:ref type='figure' target='#fig_8'>5A and 5C</ns0:ref>), and SCRG1 was significantly up-regulated (Fig. <ns0:ref type='figure' target='#fig_8'>5G and 5I</ns0:ref>). CRNKL1 was only up-regulated in the GSE34095 dataset (Fig. <ns0:ref type='figure' target='#fig_8'>5E</ns0:ref>). Unfortunately, the expression levels of WDR46 showed no significant difference across the three datasets (Fig. <ns0:ref type='figure' target='#fig_8'>5J</ns0:ref>,5K and 5L). These results indicated that BCAS4 and SCRG1 are key genes involved in IDD development.</ns0:p></ns0:div>
<ns0:div><ns0:head>Discussion</ns0:head><ns0:p>IDD is linked to lower back pain and spine-related diseases, and although IDD's underlying mechanisms have been studied for many years, they still remain unclear. Insufficient early IDD diagnosis and treatment methods affect the quality of life for patients and impose a heavy economic burden on society <ns0:ref type='bibr' target='#b47'>(Vergroesen et al. 2015)</ns0:ref>. NP cells play an important role in maintaining intervertebral disc homeostasis by synthesizing ECM, which includes aggrecan and type II collagen <ns0:ref type='bibr' target='#b56'>(Zhang et al. 2016)</ns0:ref>. Recent studies show that targeting gene therapy can inhibit NP cell senescence and apoptosis, and can ultimately ameliorate IDD <ns0:ref type='bibr' target='#b5'>(Chen et al. 2018)</ns0:ref>. Therefore, reliable and specific gene targets are essential for IDD diagnosis and treatment. Recent developments in bioinformatics and computational biology have led to the identification of several key genetic targets related to IDD and the prediction of potential molecular mechanisms <ns0:ref type='bibr' target='#b35'>(Petryszak et al. 2014)</ns0:ref>. In this study, we downloaded multiple microarray datasets associated with IDD from the GEO database, including two mRNA expression profiles, two miRNA expression profiles, one mRNA-lncRNA expression profile, and one circRNA expression profile. A total of four optimal feature genes (WDR46, BCAS4, CRNKL1, and SCRG1) were identified using machine-learning methods. The construction of a DELs/DECs-miRNA-optimal feature genes network revealed that three miRNAs (hsa-miR-4728-5p, hsa-miR-5196-5p, and hsa-miR-185-5p) and three circRNAs (hsa_circRNA_100723, hsa_circRNA_104471, and hsa_circRNA_100750) may be important mediators for optimal feature genes. To further explore the different optimal feature genes in normal and degenerative NP tissues, we also investigated their expression levels across three datasets. The results indicated that BCAS4 and SCRG1 were key genes related to IDD. Breast carcinoma amplified sequence 4 (BCAS4), a novel gene cloned from breast cancer cells, encodes a 211amino acid cytoplasmic protein with no significant homologies to any known protein <ns0:ref type='bibr' target='#b3'>(Barlund et al. 2002)</ns0:ref>. Previous studies have demonstrated that the specific DNA methylation of BCAS4 acts as an epigenetic marker and can be used to distinguish saliva from other body fluids. It is also widely used in forensic investigations <ns0:ref type='bibr' target='#b39'>(Silva et al. 2016;</ns0:ref><ns0:ref type='bibr' target='#b42'>Taki & Kibayashi 2015)</ns0:ref>. However, BCAS4's biological role in disease requires further investigation. Stimulator of chondrogenesis 1 (SCRG1) was first found in the genes associated with, or responsible for, the neurodegenerative changes observed in transmissible spongiform encephalopathies <ns0:ref type='bibr' target='#b9'>(Dandoy-Dron et al. 1998</ns0:ref>). SCRG1 transcript is found in the brain, heart, and spinal cord, and its sequence is highly conserved in humans, mice, and rats. SCRG1 has also been observed to be specifically expressed in human articular cartilage, and is involved in human mesenchymal stem cell (hMSC) growth suppression and differentiation during dexamethasone-dependent chondrogenesis <ns0:ref type='bibr' target='#b33'>(Ochi et al. 2006)</ns0:ref>. Recent studies have shown that SCRG1 is an important regulator during hMSC self-renewal, migration, and osteogenic differentiation along with its receptor BST1 <ns0:ref type='bibr' target='#b1'>(Aomatsu et al. 2014</ns0:ref>). However, the function of SCRG1 in IDD development has not yet been explored. circRNAs and lncRNAs may act as ceRNAs by competitively binding to miRNA and suppressing mRNA expression <ns0:ref type='bibr'>(Adams et al. 2017;</ns0:ref><ns0:ref type='bibr' target='#b43'>Tay et al. 2014)</ns0:ref>. This ceRNA hypothesis suggests that there is a novel mechanism for RNA interactions. In this study's ceRNA analysis, hsa-miR-4728-5p, hsa-miR-5196-5p, hsa-miR-185-5p, hsa_circRNA_100723, hsa_circRNA_104471, and hsa_circRNA_100750 showed more interactions compared to the other RNAs in the whole network. Additionally, miR-155, miR-21, and miR-133a were shown to be differentially expressed in degenerative NP cells, indicating that they may be potential biomarkers for early IDD diagnosis <ns0:ref type='bibr' target='#b29'>(Liu et al. 2014;</ns0:ref><ns0:ref type='bibr' target='#b50'>Wang et al. 2011;</ns0:ref><ns0:ref type='bibr' target='#b53'>Xu et al. 2016)</ns0:ref>. The dysregulation of these miRNAs is closely associated with NP cell apoptosis which affects IDD progression. miR-185-5p has been reported as a critical regulator, but miR-4728-5p and miR-5196-5p have rarely been reported. <ns0:ref type='bibr'>Chang et al. (2017)</ns0:ref> reported that miR-185-5p induced by Runx2 could directly target Dlx2 to inhibit amylogenesis and osteogenesis, providing a new treatment option for cleidocranial dysplasia. Multiple lncRNAs bind to miR-185-5p in order to modulate the progression of different types of human cancer, including prostate cancer <ns0:ref type='bibr' target='#b44'>(Tian et al. 2018)</ns0:ref>, colorectal cancer <ns0:ref type='bibr' target='#b58'>(Zhu et al. 2018)</ns0:ref>, and glioblastoma <ns0:ref type='bibr' target='#b51'>(Wang et al. 2018b</ns0:ref>). However, the regulatory effects of miR-185-5p on IDD require further investigation. CircRNAs, a new type of ncRNAs formed by special loop splicing, are thought to be potential diagnostic biomarkers and therapeutic targets because they are more stable and conserved than other RNAs <ns0:ref type='bibr' target='#b25'>(Lee et al. 2019</ns0:ref>). However, very few studies have focused on the role of circRNAs in IDD development. <ns0:ref type='bibr' target='#b15'>Guo et al. (2018)</ns0:ref> found that circ-GRB10 was downregulated in IDD cells, and circ-GRB10 overexpression inhibited NP cell apoptosis by sequestering miR-328-5p and upregulating target genes involved in cell proliferation through the ErbB pathway. <ns0:ref type='bibr' target='#b7'>Cheng et al. (2018)</ns0:ref> reported that circVMA21 acted as a sponge for miR-200c and regulated the activity and function of NP cells by targeting miRNA-200c and XIAP, providing a new IDD intervention and treatment strategy. <ns0:ref type='bibr' target='#b49'>Wang et al. (2018a)</ns0:ref> found that circRNA_4099 could act as a sponge for miR-616-5p and eliminate Sox9 inhibition, increasing ECM secretion. Similarly, <ns0:ref type='bibr'>Cui & Zhang (2020)</ns0:ref> reported that circ_001653 silencing may bind to miR-486-3p in order to inhibit CEMIP expression, thus attenuating NP cell apoptosis and ECM degradation. In this study, we used integrated analysis to identify that hsa_circRNA_100723, hsa_circRNA_104471, and hsa_circRNA_100750, which have not been previously reported, are more likely to be important molecules involved in IDD regulation, and require further investigation.</ns0:p></ns0:div>
<ns0:div><ns0:head>Conclusion</ns0:head><ns0:p>In this study, we used integrated bioinformatics analysis and machine learning methods to identify BCAS4 and SCRG1 as key genes associated with IDD development. Additionally, after constructing the ceRNA network, we found three miRNAs and three circRNAs that may act as important regulators during IDD development by targeting key genes. This novel study may provide new insights into IDD pathogenesis and therapy. Further experiments should be conducted to verify this study's results. GSE15227 and GSE34095 datasets, respectively. (C and D) Hierarchical cluster heatmaps of the GSE15227 and GSE34095 datasets displaying the DEGs in the degenerative disc samples and the control disc samples. Blue represents the downregulated and red represents the upregulated. (E) Venn diagram of DEGs in the GSE15227 and GSE34095 datasets. The common area represents the overlapping genes. DEG, differentially expressed genes. Figure <ns0:ref type='figure'>2</ns0:ref>: Selection of optimal feature genes Ranking of the top 13 IDD-related genes using neural networks (A) and random forest (B). Extraction of the optimum feature genes from the 13 IDD-related genes was carried out using recursive feature elimination (C). Classification efficiency of the optimum feature genes in the model as evaluated using the ROC curve in the GSE15227 (D), GSE34095 (E), and GSE56081 (F) datasets, respectively. Figure Ranking of the top 13 IDD-related genes using neural networks (A) and random forest (B).</ns0:p><ns0:note type='other'>1</ns0:note><ns0:p>Extraction of the optimum feature genes from the 13 IDD-related genes was carried out using recursive feature elimination (C). Classification efficiency of the optimum feature genes in the model as evaluated using the ROC curve in the GSE15227 (D), GSE34095 (E), and GSE56081 (F) datasets, respectively. Manuscript to be reviewed </ns0:p><ns0:note type='other'>Figure 3</ns0:note><ns0:note type='other'>Figure 5</ns0:note></ns0:div><ns0:figure xml:id='fig_0'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:05:48863:2:0:NEW 11 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_1'><ns0:head>Figure3:</ns0:head><ns0:label /><ns0:figDesc>Identification of target miRNAs of the optimal feature genes (A)Volcano plots represent the DEMs of the degenerative disc samples and control disc samples in the GSE116726 dataset. (B) Hierarchical cluster heatmaps of the GSE116726 dataset display the DEMs to compare degenerative disc samples and control disc samples. Blue represents the degenerative samples and red represents the control samples. (C) Venn diagram of DEMs in the GSE116726 and GSE63492 datasets. The common area represents the overlapping DEMs. (D)Venn diagram of miRNAs in overlapping DEMs and target miRNAs of optimum feature genes. The common area represents the overlapping miRNAs. DEM, differentially expressed miRNA.</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_2'><ns0:head>Figure 4 :</ns0:head><ns0:label>4</ns0:label><ns0:figDesc>ceRNA network construction ceRNA network of optimum IDD feature genes with DEMs, DELs, and DECs. The blue circles represent optimum IDD feature genes, the red triangles represent DECs, the green diamonds represent DELs, and the yellow arrows represent DEMs. ceRNA, competing endogenous RNA; DEC, differentially expressed circRNA; DEL, differentially expressed lncRNA; and DEM, differentially expressed miRNA.</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_3'><ns0:head>Figure 5 :</ns0:head><ns0:label>5</ns0:label><ns0:figDesc>Validation of optimal feature genes The expression levels of the four optimum feature genes in the GSE15227, GSE34095, and GSE56081 datasets, respectively. (A, B and C) The expression level of BCSA4. (D, E and F) The expression level of CRNKL1. (G, H and I) The expression level of SCRG1. (J, K and L) The expression level of WDR46. * represents P value < 0.05, *** represents P value < 0.001, **** represents P value < 0.0001, and NS represents not significant.</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_4'><ns0:head>Figure 1 :</ns0:head><ns0:label>1</ns0:label><ns0:figDesc>Figure 1: Differentially expressed IDD genes</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_5'><ns0:head>Figure 2 Figure 2 :</ns0:head><ns0:label>22</ns0:label><ns0:figDesc>Figure 2</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_6'><ns0:head>Figure 3 :</ns0:head><ns0:label>3</ns0:label><ns0:figDesc>Figure 3: Identification of target miRNAs of optimal feature genes</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_7'><ns0:head>Figure 4 Figure 4 :</ns0:head><ns0:label>44</ns0:label><ns0:figDesc>Figure 4</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_8'><ns0:head>Figure 5 :</ns0:head><ns0:label>5</ns0:label><ns0:figDesc>Figure 5: Validation of optimal feature genes</ns0:figDesc></ns0:figure>
<ns0:figure type='table' xml:id='tab_0'><ns0:head>Table 1 . Basic information of the expression profiles Table 2. The top three miRNAs and circRNAs related to optimal feature genes in the network Supplemental Files Figure S1：Flow chart presentation of the study IDD</ns0:head><ns0:label>1</ns0:label><ns0:figDesc /><ns0:table /><ns0:note>, intervertebral disc degeneration; DEG, differential expression gene; RFE, recursive feature elimination;</ns0:note></ns0:figure>
<ns0:figure type='table' xml:id='tab_1'><ns0:head>Table 1 . Basic information of expression profiles included in the study</ns0:head><ns0:label>1</ns0:label><ns0:figDesc /><ns0:table><ns0:row><ns0:cell>Type</ns0:cell><ns0:cell>Series</ns0:cell><ns0:cell>Platform</ns0:cell><ns0:cell>Source</ns0:cell><ns0:cell>Number of samples</ns0:cell><ns0:cell>Publication</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell>name</ns0:cell><ns0:cell>(control/degenerative)</ns0:cell><ns0:cell>year</ns0:cell></ns0:row><ns0:row><ns0:cell>mRNA</ns0:cell><ns0:cell>GSE34095</ns0:cell><ns0:cell>GPL96</ns0:cell><ns0:cell>Disc</ns0:cell><ns0:cell>6(3/3)</ns0:cell><ns0:cell>2012</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell>tissue</ns0:cell><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>mRNA</ns0:cell><ns0:cell>GSE15227</ns0:cell><ns0:cell>GPL1352</ns0:cell><ns0:cell>Disc</ns0:cell><ns0:cell>15 (12/3)</ns0:cell><ns0:cell>2009</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell>tissue</ns0:cell><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>miRNA</ns0:cell><ns0:cell>GSE63492</ns0:cell><ns0:cell>GPL19449</ns0:cell><ns0:cell>Nucleus</ns0:cell><ns0:cell>10 (5/5)</ns0:cell><ns0:cell>2016</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell>pulposus</ns0:cell><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>miRNA</ns0:cell><ns0:cell cols='2'>GSE116726 GPL20712</ns0:cell><ns0:cell>Nucleus</ns0:cell><ns0:cell>6(3/3)</ns0:cell><ns0:cell>2018</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell>pulposus</ns0:cell><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>mRNA-l</ns0:cell><ns0:cell>GSE56081</ns0:cell><ns0:cell>GPL15314</ns0:cell><ns0:cell>Nucleus</ns0:cell><ns0:cell>10(5/5)</ns0:cell><ns0:cell>2014</ns0:cell></ns0:row><ns0:row><ns0:cell>ncRNA</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell>pulposus</ns0:cell><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell cols='2'>circRNA GSE67566</ns0:cell><ns0:cell>GPL19978</ns0:cell><ns0:cell>Nucleus</ns0:cell><ns0:cell>10(5/5)</ns0:cell><ns0:cell>2016</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell>pulposus</ns0:cell><ns0:cell /><ns0:cell /></ns0:row></ns0:table></ns0:figure>
<ns0:figure type='table' xml:id='tab_2'><ns0:head>Table 2 . The top 3 miRNAs and circRNAs related to optimal feature genes in the network</ns0:head><ns0:label>2</ns0:label><ns0:figDesc /><ns0:table><ns0:row><ns0:cell>Type</ns0:cell><ns0:cell>name</ns0:cell><ns0:cell>Number of directed edges</ns0:cell></ns0:row><ns0:row><ns0:cell>miRNA</ns0:cell><ns0:cell>hsa-miR-4728-5p</ns0:cell><ns0:cell>58</ns0:cell></ns0:row><ns0:row><ns0:cell>miRNA</ns0:cell><ns0:cell>hsa-miR-5196-5p</ns0:cell><ns0:cell>41</ns0:cell></ns0:row><ns0:row><ns0:cell>miRNA</ns0:cell><ns0:cell>hsa-miR-185-5p</ns0:cell><ns0:cell>34</ns0:cell></ns0:row><ns0:row><ns0:cell>circRNA</ns0:cell><ns0:cell>hsa_circRNA_100723</ns0:cell><ns0:cell>12</ns0:cell></ns0:row><ns0:row><ns0:cell>circRNA</ns0:cell><ns0:cell>hsa_circRNA_104471</ns0:cell><ns0:cell>11</ns0:cell></ns0:row><ns0:row><ns0:cell>circRNA</ns0:cell><ns0:cell>hsa_circRNA_100750</ns0:cell><ns0:cell>9</ns0:cell></ns0:row></ns0:table></ns0:figure>
</ns0:body>
" | "Dear editors and reviewers,
We deeply appreciate the time and effort you’ve spent in reviewing our paper Integrated analysis of multi-microarray datasets and machine learning methods reveal key genes and regulatory mechanisms underlying human intervertebral disc degeneration.
We are very grateful to the reviewer Ling Zhang for helpful comments on our manuscript. Following the suggested comments, the manuscript has been correspondingly altered. In our revised version of the manuscript, all changes are highlighted in red. Responses to reviewer Ling Zhang comments are listed as follow:
Basic reporting
The authors have addressed all of concerns from my previous review.
Experimental design
Question: The authors have revised the manuscript substantially and addressed most of issues raised. The explanation about using different p-values for different experiment and microarray platform are not clear to me. I am wondering if the similar numbers of DEGs were generated from different experiments when using different p-values? I recommend that the authors add one or two more sentences to explain it.
Answer: Thanks for your suggestion. As we all know, P values < 0.05 were considered statistically significant. Although we set different thresholds in this manuscript, all of them less than 0.05. If there are too many differentially expressed genes screened at P<0.05, we will set the threshold to P<0.01 according to the specific situation, which helps us to screen out genes with more significant expression differences. We think there is no need to explain any more.
Validity of the findings
Question: The authors have thoroughly addressed my comments. The authors mentioned ' Moreover, many studies have used the method of intersecting multiple datasets to further determine the differential expression genes.' I recommend that the authors cite some references to them.
Answer: Thank you for your valuable comments on our work. We have cited two references which used the method of intersecting multiple datasets to identify differential expression genes in line 152. References as follow:
Zhang Y, Zhang R, Ding X, and Ai K. 2018. EFNB2 acts as the target of miR-557 to facilitate cell proliferation, migration and invasion in pancreatic ductal adenocarcinoma by bioinformatics analysis and verification. Am J Transl Res 10:3514-3528.
Li CY, Zhang WW, Xiang JL, Wang XH, Wang JL, and Li J. 2019. Integrated analysis highlights multiple long non‑coding RNAs and their potential roles in the progression of human esophageal squamous cell carcinoma. Oncol Rep 42:2583-2599. 10.3892/or.2019.7377
Comments for the author
The manuscript is, in my opinion, ready to be published in PeerJ with minor corrections and clarifications.
Answer: We are very grateful for your recognition of our work.
" | Here is a paper. Please give your review comments after reading it. |
9,759 | "<ns0:abstract xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:p>Next generation sequencing technologies are becoming more accessible and affordable over the years, with entire genome sequences of several pathogens being deciphered in few hours. However, there is the need to analyze multiple genomes within a short time, in order to provide critical information about a pathogen of interest such as drug resistance, mutations and genetic relationship of isolates in an outbreak setting. Many pipelines that currently do this are stand-alone workflows and require huge computational requirements to analyze multiple genomes. We present an automated and scalable pipeline called BAGEP for monomorphic bacteria that performs quality control on FASTQ paired end files, scan reads for contaminants using a taxonomic classifier, maps reads to a reference genome of choice for variant detection, detects antimicrobial resistant (AMR) genes, constructs a phylogenetic tree from core genome alignments and provide interactive short nucleotide polymorphism (SNP) visualization across core genomes in the data set. The objective of our research was to create an easy to use pipeline from existing bioinformatics tools that can be deployed on a personal computer. The pipeline was built on Snakemake framework and utilizes existing tools for each processing step: fastp for quality trimming, snippy for variant calling, Centrifuge for taxonomic classification, Abricate for AMR gene detection, snippy-core for generating whole and core genome alignments, IQ-TREE for phylogenetic tree construction and vcfR for an interactive heatmap visualization which shows SNPs at specific locations across the genomes. BAGEP was successfully tested and validated with Mycobacterium tuberculosis (n=20) and Salmonella enterica serovar Typhi (n=20) genomes which are about 4.4 million and 4.8 million base pairs, respectively. Running these test data on a 8 GB RAM, 2.5 GHz quad core laptop took 122 and 61 minutes on respective data sets to complete the analysis. BAGEP is a fast, calls accurate SNPs and an easy to run pipeline that can be executed on a mid-range laptop; it is freely available on: https://github.com/idolawoye/BAGEP .</ns0:p></ns0:div>
</ns0:abstract>
<ns0:body xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:head>Introduction</ns0:head><ns0:p>Over the years, as next generation sequencing has rapidly become popular, molecular biology has taken a giant leap in the way genomes of various organisms are deciphered. Genomics have grown expeditiously with high throughput sequencing technologies and paving the way for novel biological analytical approaches <ns0:ref type='bibr' target='#b39'>(Schuster, 2008)</ns0:ref>.</ns0:p><ns0:p>Genome sequencing of pathogens such as bacteria and viruses generate FASTQ files which contains quality cores of the nucleotide bases in raw format. These files usually need various technical processing steps such as adapter removal, sequence quality filtering and quality control. This is crucial for other downstream analysis, as seen in a work where choosing the appropriate FASTQ pre-processor improved downstream analysis significantly in comparison to other software <ns0:ref type='bibr' target='#b6'>(Chen et al., 2018)</ns0:ref>.</ns0:p><ns0:p>In addition, reconstructing a sequenced genome is usually achieved through de novo assembly (genome assembly using overlapping regions in the genome) or through a reference sequence guided approach (mapping sequence reads to a reference genome) <ns0:ref type='bibr'>(Korbel et al., 2007;</ns0:ref><ns0:ref type='bibr' target='#b45'>Zhang et al., 2011)</ns0:ref>. There are numerous tools that can perform these tasks at various efficiency and for different lengths of sequence reads. A few genome assemblers such as SPAdes, Burrows-Wheeler Alignment tool (BWA), minimap2 and Bowtie are widely used in bioinformatics <ns0:ref type='bibr' target='#b5'>(Bankevich et al., 2012;</ns0:ref><ns0:ref type='bibr' target='#b28'>Langmead and Salzberg, 2012;</ns0:ref><ns0:ref type='bibr' target='#b29'>Li, 2018;</ns0:ref><ns0:ref type='bibr'>Li and Durbin, 2009)</ns0:ref>.</ns0:p><ns0:p>Variants in microbes occur as short nucleotide polymorphisms (SNPs), insertions and deletions (indels) and or translocations. Mapping the reads to a reference genome or comparing the assembled genome to a reference genome are the ways in detecting variants <ns0:ref type='bibr' target='#b35'>(Olson et al., 2015)</ns0:ref>. SNP detection is important for comparative genomics in bacterial isolates as they have been used to characterized different bacteria species sharing the same genus <ns0:ref type='bibr'>(Ledwaba et al., 2019)</ns0:ref>. SNP analysis is also an integral part of understanding evolution in bacterial genomes, detecting the cause or source of an outbreak, phylogeography and genome-wide association studies (GWAS) <ns0:ref type='bibr'>(Farhat et al., 2019;</ns0:ref><ns0:ref type='bibr'>O'Neill et al., 2019;</ns0:ref><ns0:ref type='bibr'>Stimson et al., 2019)</ns0:ref>.</ns0:p><ns0:p>At the moment, there is no generally accepted standard operating procedure for evaluating and calling SNPs and or indels which has led to a wide variety of tools and methods for variant detection. Furthermore, best practices for variant identification in microbial genomes have been proposed in literature and it has been adopted to a large extent <ns0:ref type='bibr' target='#b35'>(Olson et al., 2015;</ns0:ref><ns0:ref type='bibr' target='#b44'>Van der Auwera et al., 2013)</ns0:ref>. Downstream analysis of numerous bacteria genomes using whole genome sequencing (WGS) pipelines require high performance computing servers or a cloud-based support system. In addition, while a number of WGS pipeline exist such as UVP <ns0:ref type='bibr' target='#b9'>(Ezewudo et al., 2018)</ns0:ref> and MTBseq <ns0:ref type='bibr' target='#b21'>(Kohl et al., 2018)</ns0:ref> for Mycobacterium tuberculosis, these pipelines require huge computing infrastructure. Moreover, the pipelines usually have many dependencies to be installed especially if the analysis requires multiple tasks to be performed such as phylogenetic tree construction, drug resistance prediction, clustering, and so on. The huge amount of bioinformatics tools available makes it a daunting task in picking the suitable software for a certain analysis. In low-to-middle countries who are becoming exposed to WGS of pathogens, it is important to also provide ready-to-use, quick and easily deployable pipelines for WGS data analysis. These pipelines should also be a stand-alone workflow on a local server or even a personal computer. For example, UVP and MTBseq needs at least 100 GB and 25 GB of RAM respectively to run locally which is capitally intensive for many researchers in low-to-middle countries (LMICs) who need to analyze WGS data.</ns0:p><ns0:p>To address the challenges identified above, we embarked on the development of BAGEP (Bacteria Genome Pipeline, available online at https://github.com/idolawoye/BAGEP#). It uses existing pipelines and bioinformatics tools and an advanced workflow management system called Snakemake <ns0:ref type='bibr' target='#b26'>(Koster and Rahmann, 2012)</ns0:ref>. This pipeline combines many tools used in downstream analysis of paired-end FASTQ raw reads such as quality control, read mapping, variant calling, core genome and/or full genome alignments, SNP visualization and phylogeny into a single, fast, easy to use workflow.</ns0:p></ns0:div>
<ns0:div><ns0:head>Materials & Methods</ns0:head></ns0:div>
<ns0:div><ns0:head>Sampling</ns0:head><ns0:p>To display the versatility of BAGEP, we applied it to a set of multi-drug resistant (MDR) M. tuberculosis genomes (n=20) from Southwest Nigeria <ns0:ref type='bibr' target='#b42'>(Senghore et al., 2017)</ns0:ref> and S. enterica serovar Typhi genomes (n=20) from the United Kingdom <ns0:ref type='bibr' target='#b2'>(Ashton et al., 2017)</ns0:ref>. These publicly available paired-end FASTQ reads were extracted from National Centre for Biotechnology Information (NCBI) Sequence Read Archive (SRA). Genomic DNA of isolates were extracted from liquid cultures and sequenced on Illumina MiSeq and HiSeq platforms. These raw reads are deposited under BioProjects PRJEB15857 and PRJNA248792.</ns0:p></ns0:div>
<ns0:div><ns0:head>Implementation</ns0:head><ns0:p>Installation of BAGEP requires the pipeline to be downloaded on to a personal computer and installed by creating a conda environment to set up all dependencies. Centrifuge database should be downloaded separately in order to detect contaminants in the samples. Complete installation steps are stated in the github README file: https://github.com/idolawoye/BAGEP/blob/master/README.md.</ns0:p><ns0:p>The analysis of BAGEP is segmented into set of 'rules' that connects the output file of an analysis into the input of the next task in the general workflow (Figure <ns0:ref type='figure'>1</ns0:ref>). The dependencies are fastp for quality control of raw reads, Centrifuge for taxonomic classification, Snippy for variant detection, snippy-core for core and whole genome alignments, Abricate for AMR detection, IQ-TREE for phylogenetics, Krona for taxonomic visualization, vcfR and heatmaply for SNP visualization.</ns0:p><ns0:p>These tools can be installed in a containerized manner using a bioconda channel <ns0:ref type='bibr' target='#b8'>(Dale et al., 2018)</ns0:ref> which can be activated and deactivated easily, whereas the SNP visualization is provided by R libraries included in the dependencies. The input files for BAGEP are paired-end FASTQ files and a reference genome in FASTA or GenBank format, only the latter is provided in the single line command to run, whilst the former should be saved in a designated local directory.</ns0:p><ns0:p>BAGEP also allows full customization of the pipeline, such that users can modify the parameters used in running their samples. For example, the pipeline does not mask repeat regions when aligning the core genomes by default, but this can be done by adding the option to the snakefile rule handling that process. It is possible to modify every step of the workflow to suit the samples being processed, even the model for phylogenetics.</ns0:p><ns0:p>Upon execution of the pipeline, Snakemake organizes the correct combination of tasks in order to generate the desired output and also leverages on the wildcard feature to analyze multiple genomes that have similar format allowed in the rule <ns0:ref type='bibr' target='#b26'>(Koster and Rahmann, 2012)</ns0:ref>. Preliminary results such as quality trimmed FASTQ files are saved in a new directory, while primary results like detected variants, core genome alignments, phylogeny and SNP visualization are saved in a Manuscript to be reviewed different directory. The pipeline runs on Linux/Unix, MacOS terminal or Windows Subsystem for Linux. However, no prior skill in programming is required to run and interpret the results as only one stage of input is needed and thereafter the pipeline runs automatically. Another significant advantage of BAGEP is that it can re-execute steps that have not been ran if errors occur and resume from where it stopped rather than restarting the entire process which another valuable feature of the underlying framework.</ns0:p></ns0:div>
<ns0:div><ns0:head>Results</ns0:head><ns0:p>FASTQ Pre-processing BAGEP performs several quality control tasks on raw FASTQ reads using fastp <ns0:ref type='bibr' target='#b6'>(Chen et al., 2018)</ns0:ref>. Fastp capitalizes on speed, efficiency and an all-in-one pre-processor. It carries out sliding window quality filtering, adapter trimming, base correction, polyG and polyX tail trimming for Illumina NextSeq and NovaSeq series, UMI pre-processing to reduce background noise and improve sensitivity when detecting ultra-low frequency mutations in deep sequencing, duplication analysis, quality control and reporting (Figure <ns0:ref type='figure'>2</ns0:ref>).</ns0:p><ns0:p>Fastp was chosen as the suitable tool for this task as it is seen to outperform other quality control FASTQ software and it also allows multi-threading to improve its performance when dealing with numerous samples <ns0:ref type='bibr' target='#b6'>(Chen et al., 2018)</ns0:ref>.</ns0:p></ns0:div>
<ns0:div><ns0:head>Taxonomic classification</ns0:head><ns0:p>This step is crucial in the pipeline as contaminants from the environment such as host cells or other bacteria can negatively impact the accuracy of variants detected. We utilized Centrifuge for this step due to its lower memory requirements compared to other similar tools like Kraken <ns0:ref type='bibr' target='#b17'>(Kim et al., 2016)</ns0:ref>. To visualize the results, Krona was employed such that one can interact with the output and also export it as a static image file (Figure <ns0:ref type='figure'>3</ns0:ref>) <ns0:ref type='bibr' target='#b36'>(Ondov et al., 2011)</ns0:ref>.</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:04:48226:1:0:NEW 4 Aug 2020)</ns0:p><ns0:p>Manuscript to be reviewed</ns0:p></ns0:div>
<ns0:div><ns0:head>Read mapping and variant discovery</ns0:head><ns0:p>Mapping refined sequence reads to the reference genome is done by Snippy <ns0:ref type='bibr' target='#b40'>(Seemann, 2015)</ns0:ref> which is a variant calling and core genome alignment pipeline. Firstly, BWA MEM <ns0:ref type='bibr'>(Li and Durbin, 2009)</ns0:ref> maps reads to the provided reference genome and they are manipulated with Samtools <ns0:ref type='bibr'>(Li et al., 2009)</ns0:ref>. Variants are called from the resulting Binary Alignment Map (BAM) files using freebayes <ns0:ref type='bibr' target='#b12'>(Garrison and Marth, 2012)</ns0:ref> which takes short-read alignments in BAM formats and Phred+33 quality scores from each sample and determines the best combination of polymorphisms in each sample at each position in the reference genome with the aid of Bayesian variant detection model. The variants are annotated with SnpEff <ns0:ref type='bibr' target='#b7'>(Cingolani et al., 2012)</ns0:ref>, with reports generated in variant call format (VCF), TAB, CSV, HTML, BED, GFF and FASTA files. A unified VCF file that contains core and full genome alignments of individual SNP output files are generated with Snippy-core. Furthermore, BAGEP leverages on a function in Snippy-core that allows the user to filter out problematic or repetitive regions core genomes of the samples such as Proline-Glutamate (PE) and Proline-Proline-Glutamate (PPE) gene families in M. tuberculosis by providing a BED file to omit these regions.</ns0:p></ns0:div>
<ns0:div><ns0:head>Antimicrobial resistant gene screening</ns0:head><ns0:p>BAGEP uses abricate to screen the isolates for antimicrobial resistant (AMR) genes by using either one of the many AMR databases that comes with it. The great thing about abricate is that these databases can be updated from the help option and the user can select which databases suits their analysis. The results from this step is stored in a tab-separated file which can be viewed in a variety of software.</ns0:p></ns0:div>
<ns0:div><ns0:head>SNP visualisation</ns0:head><ns0:p>PeerJ reviewing PDF | (2020:04:48226:1:0:NEW 4 Aug 2020)</ns0:p><ns0:p>Manuscript to be reviewed One other key feature of BAGEP is the ability to generate an interactive SNP visualization that allows the end user to view substitution polymorphisms across core genomes in the population.</ns0:p><ns0:p>In the pipeline, a custom R script makes use of vcfR <ns0:ref type='bibr' target='#b19'>(Knaus and Grünwald, 2017)</ns0:ref> and heatmaply <ns0:ref type='bibr' target='#b10'>(Galili et al., 2017)</ns0:ref>, parses the VCF file output from Snippy-core to render a heatmap showing SNP positions in a HTML file (Figure <ns0:ref type='figure'>4</ns0:ref>).</ns0:p><ns0:p>The heatmap can be interacted with, with a number of tools such as zooming, hovering over the image to show SNPs and export a selected region as an image. This is very useful in a VCF file with high number of samples and reported SNPs as hovering around regions in the image can infer which position the polymorphism occur and in what sample.</ns0:p></ns0:div>
<ns0:div><ns0:head>Phylogeny</ns0:head><ns0:p>The final analysis done by BAGEP is the construction of a maximum-likelihood phylogenetic tree with IQ-TREE <ns0:ref type='bibr' target='#b34'>(Nguyen et al., 2015)</ns0:ref>, this step uses the core genome alignment generated by Snippy-core with ultra-fast bootstrap value of 1000. The output tree is deposited in the 'results' directory and can be viewed and annotated with any tree viewing software such as Figtree or Interactive Tree of Life (ITOL).</ns0:p></ns0:div>
<ns0:div><ns0:head>Performance</ns0:head><ns0:p>Running BAGEP on 20 M. tuberculosis and 20 S. enterica serovar Typhi genomes with an 8 GB RAM, 2.5 GHz quad core laptop took 122 and 61 minutes to complete the analysis, respectively;</ns0:p><ns0:p>M. tuberculosis has a 4.4 million base pair genome while S. enterica serovar Typhi has 4.81 million base pairs. BAGEP capitalizes on Snakemake's multi-threading feature which implies that it can be deployed on laptops with greater performance or a computing server to improve its speed. Comparing BAGEP's performance to Nullarbor pipeline on a benchmark computer using the same M. tuberculosis data sets, BAGEP took 122 minutes whilst Nullarbor spent 186 minutes PeerJ reviewing PDF | (2020:04:48226:1:0:NEW 4 Aug 2020)</ns0:p><ns0:p>Manuscript to be reviewed to complete the same analysis. In addition to speed, BAGEP outperforms Nullarbor in detecting more accurate SNPs. To assess the robustness of BAGEP, we ran the pipeline on 57 M. tuberculosis genomes using a 16 GB RAM computer and it completed the run in 225 minutes (Supplementary file 1).</ns0:p></ns0:div>
<ns0:div><ns0:head>Discussion</ns0:head><ns0:p>BAGEP was built around the core concepts of Snakemake which offers parallelization of several tasks in an orderly manner. It was designed to be user friendly, fast, customizable and reproducible. In addition, we wanted a pipeline that can be deployed on a personal computer and handle medium to large data and finally, an interactive tool for visualizing SNPs that were detected whilst running the pipeline. BAGEP requires minimal user input and runs from start to finish with a single command.</ns0:p><ns0:p>As the backbone of BAGEP is Snakemake, each rule is run in its own environment and allows the combination of other programming languages such as Python, R and Bash. This also implies that each rule can be customized to make it run faster depending on the configuration of the machine running it and basic understanding of tools used in the pipeline. For example, the rule that takes the longest time to run is the step where reads are mapped to a reference genome and variants are called. This is executed by Snippy and the number of threads can be increased to speed up the process. This pipeline has been tested with M. tuberculosis and S. enterica serovar Typhi and it is suitable for genetically monomorphic or monoclonal pathogens such as Yersinia pestis, Bacillus anthracis, Mycobacterium leprae and Treponema pallidum due to the limited amount of variations in their core genome as compared to highly recombinant pathogens <ns0:ref type='bibr' target='#b0'>(Achtman, 2008)</ns0:ref>. Manuscript to be reviewed For visualization, BAGEP generates a HTML file as output that contains observed SNPs in core genomes of the samples and the position where they can be found, in addition with dendrograms that highlight the relatedness of the genomes. The interactive image can be zoomed in and out, reveal SNP regions by hovering around the image with a mouse and also export the image as portable network graphics (PNG) image. This image can be easily interpreted by anyone with little or no knowledge in bioinformatics and gives a summary of the analysis.</ns0:p><ns0:p>It is worthy to note that BAGEP was compiled with speed in mind and ease-of-use and that is why the dependencies can be installed in a conda environment under the bioconda channel <ns0:ref type='bibr' target='#b8'>(Dale et al., 2018)</ns0:ref>. Performance comparison to Nullarbor shows that BAGEP identified 459 SNPs in a particular isolate which yielded 598 SNPs with Nullarbor (it is important to note that BAGEP and Nullarbor uses the same tool, Snippy for variant detection). In other to validate the accuracy of the SNPs, we ran the same sample as raw reads (that is without undergoing QC steps with fastp or Trimmomatic) and the end result was 598 SNPs, this trend was seen in other samples as well. This indicates that the QC step in Nullarbor is not as effective as the one embedded in BAGEP, which will give rise to false positives when using Nullarbor. Furthermore, BAGEP is quicker than Nullarbor when analyzing the same set of samples. Therefore, standard population genomics of low recombinant or genetically monomorphic bacteria can be conducted with BAGEP as it handles the fundamental analysis needed.</ns0:p></ns0:div>
<ns0:div><ns0:head>Conclusions</ns0:head><ns0:p>We present BAGEP, a fully automated and scalable pipeline that is built on Snakemake framework. This pipeline will be useful for researchers in low-to-middle income countries and people with little or no bioinformatics skills in analyzing raw genomics data.</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:04:48226:1:0:NEW 4 Aug 2020)</ns0:p><ns0:p>Manuscript to be reviewed It is effective for running medium to large data sets of paired-end raw reads of bacteria genomes, and has been tested with M. tuberculosis and S. enterica serovar Typhi. We have also shown that it is suitable for genetically monomorphic or monoclonal pathogens such as Yersinia pestis, Bacillus anthracis, Mycobacterium leprae and Treponema pallidum. Some of BAGEP's advantages is that it is quick, identifies SNPs with better accuracy, easy to run and can be deployed on a mid-range laptop computer. Our future plan is to improve this pipeline by adding new features and roll out updates as we collaborate with other scientists. We believe this will be useful for researchers in low-to-middle income countries and people with little or no bioinformatics skills in analyzing raw genomics data.</ns0:p><ns0:note type='other'>Figure 1</ns0:note><ns0:p>A schematic workflow of BAGEP.</ns0:p><ns0:p>Fastp is used for quality control on paired-end FASTQ reads. The processed reads are mapped against a reference genome provided by the user. Centrifuge is used to check the reads for contamination and generate a taxonomic visualized report with Krona. Variants are called using Freebayes and annotated with SnpEff (aided by Snippy). The resulting variant call format (VCF) files and genomes from each sample are collated with Snippy-core to produce a VCF file containing all samples, core and whole genome multiple sequence alignments. A maximum-likelihood phylogenetic tree is constructed with IQTREE and a HTML file containing an interactive SNP visualization with the VCF file. Finally, Abricate generates an AMR report from whole genome multiple sequence alignments.</ns0:p></ns0:div><ns0:figure xml:id='fig_0'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:04:48226:1:0:NEW 4 Aug 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_1'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:04:48226:1:0:NEW 4 Aug 2020)</ns0:figDesc></ns0:figure>
<ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='18,42.52,70.87,525.00,341.25' type='bitmap' /></ns0:figure>
<ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='20,42.52,178.87,525.00,273.75' type='bitmap' /></ns0:figure>
</ns0:body>
" | "Dear Editors,
We thank the reviewers for their generous comments on the manuscript and have edited the manuscript to address their concerns.
In particular, major improvement of the pipeline has been made to reflect the suitability of the manuscript.
We have also provided a point-by-point response to the reviewers’ queries and comments.
We strongly believe that this manuscript is now suitable for publication in PeerJ.
Reviewer 1 (Conor Meehan)
Basic reporting
Reviewer: The pipeline put forward by authors seems mostly suitable for the task at hand, which is assembly and visualisation of monomorphic bacterial WGS data. My main concern was that it did not take into account recombination for phylogenetics, which is very important for many bacteria. Authors state at the end of the discussion that the pipeline is only suitable for monomorphic bacteria: this should be clearly stated much earlier on, including in the abstract, so readers do not think this approach will work for bacteria that undergo any level of recombination.
Response: We thank the reviewer for this great suggestion, We have just indicated its suitability in the abstract as you suggested.
Reviewer: Similarly, the abstract and the introduction state 'However, there is the need to analyze multiple genomes within a short time, in order to provide critical information about a pathogen of interest such as drug resistance, mutations and patient-to-patient transmission in an outbreak setting'. I don't see how this pipeline addresses this need, as it does not do drug resistance prediction or transmission analyses, only SNP visualisation and phylogenetics. Thus, the need for another wrapper of existing software needs to be more clearly stated so the reader can better decide why to use this pipeline over a rival such as MTBseq or Nullarbor.
Response: This was a very great suggestion by the reviewer. We have now added a new feature of antimicrobial resistance gene detection to the pipeline (see Lines 210 to 215). Also, the statement, “patient-to-patient transmission” in line 51 has been changed to “genetic relationship of isolates”.
Experimental design
Reviewer: I was not able to install the pipeline on either an OSX laptop or a Linux sever. For both it hung at the 'solving environment' stage of the conda command. After 1 hour of waiting I cancelled the install.
Both the laptop and the server have regularly used conda installs and others were checked afterwards to ensure it was the package and not the conda or hardware. This should be fixed so the software can be evaluated.
Response: The issue with conda solving environment has now been resolved. We have also had it tested by independent users.
Reviewer: The authors state that the software has only been tested on a linux machine, yet say the target audience is LMIC researchers and those using a mid-range laptop. I would posit that most researchers that fall into this category are not using a Linux machine and thus the software should be tested on other platforms as well, or at least give instructions or pointers on their github to how to install virtual boxes or similar.
Response: We thank the reviewer for this comment and suggestion. Instruction for Microsoft Windows 10 users have been added to the github README file
Reviewer: Line 123-125 states that dependencies come from bioconda but a bash script is needed for the R libraries. R libraries can be installed through conda so I don't see why this is 2 steps, especially if the authors are putting this forward as a '1 install/pipeline only' solution.
Response: The R libraries have been included in the conda environment installation, so we have removed the bash script.
Reviewer: Similarly, I am unsure why the authors have a git install that then requires a conda environment for use, and ask that R already be installed. If using conda, they should use conda throughout the installation as a single command (e.g. conda install BAGEP), maybe from the bioconda channel, which would install the pipeline and the associated dependencies, including R and the R libraries. As it stands, it is not a single command install, which is what snakemake and conda were made to do.
Response:
The R installation has been removed from installation guideline as it has been embedded in the conda environment installation, as suggested by the reviewer.
Reviewer: Methods and github page do not indicate if parameters can be changed, such as model of evolution for IQ-TREE or cut-off parameters for snippy. This would be important as different bacteria require different parameters for phylogenetics. Without the ability to change IQ-TREE parameters, I see this as a massive drawback.
Response: We have included how users can modify the pipeline’s parameters to suit the type of isolates that are being analyzed in the github README page and methods (See Lines 156 to 161).
Validity of findings
Reviewer: No comparison to existing software was undertaken. Authors state that MTBseq installation as tried but failed. However, this is an M. tuberculosis-specific pipeline and not a general bacterial pipeline, which is what authors should be comparing to. The most obvious candidate for comparison would be nullarbor (https://github.com/tseemann/nullarbor) which undertakes most of what the authors are doing here, plush additional tasks and is created by the author of snippy. A comparison to this, to see the need for such a competing pipeline, should be done.
Response: A performance comparison to Nullarbor has been included in the results and concluding sections (See lines 238 to 243).
Reviewer: It has been shown previously that removal of contamination from read sets before mapping is required to reduce false SNPs for phylogenetics and other tasks (see Goig et al https://bmcbiol.biomedcentral.com/articles/10.1186/s12915-020-0748-z). This pipeline does not do this, and thus may give false results to the user. Authors should indicate why this is omitted or add a step that does this., However, contamination removal unfortunately tends to require large databases (this is why UVP is so large, as the authors state), so is incompatible with a laptop setup for such pipeline, which is one reason why such pipelines don't exist. Even still, authors need to justify why their pipeline doesn't do this, which likely will increase the rate of false SNP discovery.
Response: We have added a step that performs taxonomic classification by using Centrifuge which requires lesser memory (approximately less than 8 GB of RAM) to scan the data sets for contaminants (See lines 189 to 194).
Reviewer 2
Basic reporting
Introduction
Reviewer: Line 69. “as next generation sequencing gradually replaced Sanger sequencing”. NGS do not replaced Sanger sequencing. WGS is very useful for large scale works as epidemiology, infection control, contact tracing, surveillance etc, however most of research does not require WGS and Sanger sequencing is still an important tool for research. One of the problems of WGS are the ambiguous results due to sequences of low quality that can result from several incorrect procedures that frequently are solved/verified by Sanger sequencing. I suggest rephrase this statement as it can induce the readers in error.
Response: We thank the reviewer for this observation. The opening sentence in the introduction has been rephrased, and now reads “as next generation sequencing has rapidly become popular”.
Materials & Methods
In the MM should be included information about the genome sequences used to evaluate the pipeline. Their origin (published sequences or not – it is in the beginning of results but that information could be added here for contextualization), method used for DNA extraction (no need to be detailed), if the DNA was obtained from cultures or direct from clinical samples, were the sequences are deposited (database) and their accession numbers and study accession. Patient information should not be included.
Response: We agree with the reviewer. Information about genome sequenced used in this study has been moved from results section to materials & methods. We have included the nature of samples, sequencing platforms and accession numbers to sequences (See lines 132 to 139).
Other remarks
Reviewer: Designation of the species: some inconsistency was observed. The species need to be mentioned in full at the first time in the main text (excluding abstract). The same for Salmonella. In the middle of the results appears MTB; should be maintained the terminology used from the beginning. If want to designate the species name abbreviated, please change at the very beginning.
Response: Representation of species have been harmonized throughout the manuscript.
Reviewer: MTBseq and the remaining pipelines/software’s – without italic please; there is no need. Italic should be reserved for the name of the bacterial species.
Response: MTBseq and other pipelines/software mentioned in the manuscript have been reformatted to appear as regular text.
Reviewer 3 (Oren Tzfadia)
Basic reporting
1. Reviewer: Authors should acknowledge, more recent work in the field. For example, when referring to GWAS (lines 92-93), Farhat et al, 2019 (Nature Communications).
Response: We agree with the reviewer, and we have replaced older literatures with more recent research. For example, characterization of bacteria species (line 90) and GWAS reference (line 92) now have new citations.
2. Reviewer: The figures are in low resolution (for example the text in y axis is overriding itself in figure 2), and I would use better representation for SNPs along a chromosome (see example in Liu et al 2020, Scientific Advance).
Response: We thank the reviewer for this comment. However, the figures we provided can be zoomed as it is an interactive image. When zoomed, the y axis becomes more visible. We have also now added the HTML output as supplementary file.
3. Reviewer: Results does not demonstrate improvement in performance when compared to existing tools.
Response: A comparison in performance to a similar pipeline (Nullarbor) has been shown in results section. We see improvement in SNP accuracy and shorter run time using BAGEP (see lines 241 to 244)
Experimental design
Reviewer: When developing a new computational research methods / tools which aim to out-perform other existing methods, it is essential to compare performance in terms of accuracy, memory usage, running time, etc. This was not done here.
Response: We have demonstrated how BAGEP out-performs an existing tool in terms of accuracy and running time. It is well stated in lines 270 to 277.
Reviewer: Bacterial clinical samples tend to contain non-target contaminations, which generates noise and hampers downstream analyses. In order to address such contaminations, there are quality control measures that are used (example can be Centrifuge by Kim et al, 2016 Genome Research). Such measures are not included in the BAGEP pipeline.
Response: We agree with the reviewer and a new step has been added to the pipeline which detects non-target contaminations in sequencing reads being processed using Centrifuge. This step can be found in lines 189 to 194 with the heading “Taxonomic classification”.
Reviewer: Details and instructions for installation of BAGEP are only referred to a git page, which should also be included in the MS.
Response: We have now included instructions for installation of BAGEP in the manuscript. In lines 141 to 145 which reads “Installation of BAGEP requires the pipeline to be downloaded or cloned on a personal computer and installed by creating a conda environment...”
Validity of findings
Reviewer: The data sets used for demonstration of the applicability of the developed tool is too small to make a good assessment (n=20).
Response: We have tested BAGEP on larger datasets of over 50 genomes and it worked as it leverages on snakemake’s ability to scale up. The only caveat is that the more isolates in the dataset increases the total runtime (see lines 228 to 230).
" | Here is a paper. Please give your review comments after reading it. |
9,760 | "<ns0:abstract xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:p>Background: Studies have reported an increased risk of adverse pregnancy outcome associated with pre-pregnancy body mass index (BMI). However, the data on such associations in urban areas of southern Chinese women is limited, which drive us to clarify the associations of pre-pregnancy BMI and the risks of adverse pregnancy outcomes (preterm birth (PTB) and low birth weight (LBW)) and maternal health outcomes (gestational hypertension and cesarean delivery). Methods: We performed a hospitalbased case-control study including 3,864 Southern Chinese women who gave first birth to a live singleton infant from January 2015 to December 2015. PTB was stratified into three subgroups according to gestational age (extremely PTB, very PTB and moderate PTB).</ns0:p><ns0:p>Weight for gestational age were categorized into two group, small for gestational age (SGA, less than the tenth percentile of weight for gestation) and non-small for gestational age (NSGA, large than the tenth percentile of weight for gestation). We divided newborns into six groups: (1) term and NSGA; (2) 34-36 week gestation and NSGA; (3) below 34 week gestation and NSGA; (4) term and SAG; (5) 34-36 week gestation and SAG; (6) below 34 week gestation and SAG. Adjusted logistic regression models was used to estimate the odds ratios of adverse outcomes. Results: Underweight women were more likely to give LBW (AOR=1.44, 95% CI 1.11 to 1.89), the similar result was seen in term and SAG as compare with term and NSAG (AOR=1.78, 95% CI 1.45-2.17). Whereas underweight was significantly associated with a lower risk of gestational hypertension (AOR=0.45, 95% CI 0.25-0.82) and caesarean delivery (AOR=0.74, 95% CI 0.62-0.90). The risk of extremely PTB is relatively higher among overweight and obese mothers in a subgroup analysis of PTB (AOR=8.12, 95% CI=1.11 to 59.44; AOR=15.06, 95% CI=1.32 to 172.13, respectively).</ns0:p></ns0:div>
</ns0:abstract>
<ns0:body xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:head>Introduction</ns0:head><ns0:p>Preterm birth (PTB) is an important adverse pregnancy outcome with a significant impact on infant mortality and morbidity <ns0:ref type='bibr' target='#b4'>(Boghossian et al. 2016)</ns0:ref>. The incidence of PTB worldwide is expected to be 11.1%, and China has more than 1.1 million PTB each year, which ranks second in the world <ns0:ref type='bibr' target='#b3'>(Blencowe et al. 2012)</ns0:ref>. In spite of high-level advancement in healthcare services, the rate of PTB still seems to be climbing. According to statistics, perinatal mortality is as high as 70% in low birth weight (LBW) with neonatal weight below 2500 grams <ns0:ref type='bibr' target='#b12'>(Hack et al. 1994;</ns0:ref><ns0:ref type='bibr' target='#b48'>Rutter 1995</ns0:ref>), most of them are born preterm. Growing evidence has shown that the incidence and development of PTB is a complex process that is influenced by a variety of environmental and genetic factors <ns0:ref type='bibr' target='#b15'>(He et al. 2017;</ns0:ref><ns0:ref type='bibr' target='#b25'>Liu et al. 2017;</ns0:ref><ns0:ref type='bibr' target='#b42'>Qiu et al. 2017;</ns0:ref><ns0:ref type='bibr' target='#b57'>Xiao et al. 2016)</ns0:ref>. Therefore, it is helpful to develop approaches of effective prevention and treatment of neonatal morbidity and mortality by elucidating etiological factors contributing to PTB or LBW.</ns0:p><ns0:p>Recent results provide support to pre-pregnancy maternal body mass index (BMI) is one of the potential risk factors for PTB <ns0:ref type='bibr' target='#b17'>(Hendler et al. 2005;</ns0:ref><ns0:ref type='bibr' target='#b29'>Lynch et al. 2014;</ns0:ref><ns0:ref type='bibr' target='#b52'>Shaw et al. 2014)</ns0:ref> and LBW <ns0:ref type='bibr' target='#b13'>(Han et al. 2011;</ns0:ref><ns0:ref type='bibr' target='#b27'>Liu et al. 2016)</ns0:ref>. In recent decades, the prevalence of obesity and overweight among women in many countries has increased at an alarming rate, especially in developing countries. Different countries, regions and incomes have different patterns of overweight and obesity that is more common among women in developing countries and men in developed countries <ns0:ref type='bibr'>(Ng et al. 2014)</ns0:ref>. Moreover, epidemiological studies have been suggested that maternal overweight and obesity have been shown to be association with PTB <ns0:ref type='bibr' target='#b29'>(Lynch et al. 2014;</ns0:ref><ns0:ref type='bibr' target='#b52'>Shaw et al. 2014;</ns0:ref><ns0:ref type='bibr' target='#b53'>Su et al. 2020)</ns0:ref>, LBW <ns0:ref type='bibr' target='#b44'>(Rahman et al. 2015)</ns0:ref> and adverse maternal health time were included between January 2015 and December 2015. A total of 322 women with PTB and 3,362 women with term delivery controls were enrolled. These people also could be divided into 317 cases of LBW and 3,367 controls of normal birth weight. This study was approved by the institutional review board of The Third Affiliated Hospital of Guangzhou Medical University, <ns0:ref type='bibr'>Guangdong, China (Medical Ethics Hearing [2020]</ns0:ref>No.036). And Institutional Review Board waived the need for consent.</ns0:p><ns0:p>The preterm group was defined to deliver within 37 weeks of conception without congenital abnormalities or neurological damage. The control group between 37 and 42 weeks of gestation without congenital abnormalities or neurological damage was matched with the case in the residential area for one week at the same hospital. For both groups, we excluded women with a multiple pregnancy, stillbirth, prior deliveries, embryo transfer and in vitro fertilization.</ns0:p><ns0:p>All data for this study was collected from women who gave their first birth during hospitalization, including maternal age, education, gravidity, occupation, health insurance, height and weight before of pregnancy. Gestational hypertension, caesarean delivery, birth weight and gestational week at delivery were obtained from the medical records. Pre-pregnancy BMI defined as the body weight in kilograms divided by the square of the height in metres (kg/m 2 ). BMI is classified as underweight (<18.5 kg/m 2 ), normal weight (18.5-23.9 kg/m 2 ), overweight PeerJ reviewing PDF | (2020:04:47383:1:2:NEW 7 Aug 2020)</ns0:p><ns0:p>Manuscript to be reviewed (24.0-27.9 kg/m 2 ), and obesity (≥28.0 kg/m 2 ) according to the weight standard of Chinese adults (Zhou & Cooperative Meta-Analysis Group of the Working Group on Obesity in 2002). In this study, child birth weight below 2500 grams was considered LBW. Once the two occasions systolic blood pressure (BP) or diastolic BP values of pregnant women measured at intervals of 24 hours exceed 140 mmHg or 90 mmHg respectively, they are diagnosed as gestational hypertension.</ns0:p></ns0:div>
<ns0:div><ns0:head>Statistical analysis</ns0:head><ns0:p>t test (for continuous variables) or χ2 test (for categorical variables) was used to assess the difference in demographic characteristics between two groups. Adjusted odds ratios (OR) with 95% confidence intervals obtained from logistic regression model were used to quantify associations between pre-pregnancy BMI and adverse outcomes, adjusting for age, occupation, health insurance and education. We also further divided gestational age into four subtypes: extremely PTB (<28 gestational week), very PTB (28-31 gestational week), moderate PTB (32-36 gestational week) and normal (≧37 gestational week). Besides, we referred to a method application to weight for gestational age by Tanya Marchant et al <ns0:ref type='bibr' target='#b32'>(Marchant et al. 2012)</ns0:ref>. Which categorized weight for gestational age into two group, small for gestational age (SGA, less than the tenth percentile of weight for gestation) <ns0:ref type='bibr' target='#b38'>(Oken et al. 2003</ns0:ref>) and non-small for gestational age (NSGA, large than the tenth percentile of weight for gestation). We divided newborns into six groups: ( <ns0:ref type='formula'>1</ns0:ref> <ns0:ref type='table' target='#tab_9'>2020:04:47383:1:2:NEW 7 Aug 2020)</ns0:ref> Manuscript to be reviewed and SAG. A two-sided p-value of 0.05 or less was accepted to be statistically significant. Data were analyzed using Statistical Analysis Software 9.4. (SAS Institute, Cary, NC).</ns0:p><ns0:p>PeerJ reviewing <ns0:ref type='table' target='#tab_9'>PDF | (2020:04:47383:1:2:NEW 7 Aug 2020)</ns0:ref> Manuscript to be reviewed</ns0:p></ns0:div>
<ns0:div><ns0:head>Results</ns0:head></ns0:div>
<ns0:div><ns0:head>Characteristics of the study subjects</ns0:head><ns0:p>The baseline maternal characteristics are shown in table 1. Of the 3,684 live births, 8.7% (n=322) were PTBs, and the remaining 91.3% (n=3362) were term births. The proportions of underweight, normal weight, overweight, and obesity women were 23.29%, 63.36%, 10.72%, and 2.63%, respectively. There were significant differences between PTB group and terms group with respect to age, health insurance, occupation and education (all P<0.01). There were no significant differences between the two groups with regard to gravidity (P>0.05). The variables in the normal and the LBW group were basically the same as those in the term and preterm groups.</ns0:p></ns0:div>
<ns0:div><ns0:head>Association analysis between pre-pregnancy BMI and adverse outcomes</ns0:head><ns0:p>The association between pre-pregnancy BMI and the risk of adverse outcomes are considered in table 2 and table 3. Underweight, overweight and obesity did not increase the risk of PTB as compare with normal weight (AOR=1.01, 95% CI=0.76 to 1.34; AOR=1.25, 95% CI=0.87 to 1.80; and AOR=1.27, 95% CI=0.65 to 2.51, respectively). However, When PTB was stratified into three subgroups, both maternal overweight and obesity increased the risks of extremely PTB (AOR=8.12, 95% CI=1.11 to 59.44; AOR=15.06, 95% CI=1.32 to 172.13, respectively).</ns0:p><ns0:p>Pre-pregnancy underweight was significantly associated with the increased risk for LBW. In comparison with women who had normal pre-pregnancy BMI, women with low BMI category was more likely to deliver a LBW infant (crude OR=1.48, 95% CI=1.14 to 1.93). After the PeerJ reviewing PDF | (2020:04:47383:1:2:NEW 7 Aug 2020)</ns0:p><ns0:p>Manuscript to be reviewed adjustment for potential confounding factors, the AOR associated with the risk for giving birth to a LBW infant were 1.44 (95% CI=1.11 to 1.89), the similar result was seen in table 4 as compare with term and NSAG (AOR=1.78, 95% CI=1.45 to 2.17). Underweight was also significantly associated with a lower risk of gestational hypertension (AOR=0.45, 95% CI=0.25 to 0.82) and caesarean delivery (AOR=0.74, 95% CI=0.62 to 0.90). Both maternal overweight and obesity were found to be a risk factor for gestational hypertension (AOR=1.71, 95% CI=1.06 to 2.77; AOR=5.54, 95% CI=3.02 to 10.17, respectively) and caesarean delivery (AOR=1.91, 95% CI=1.53 to 2.38; AOR=1.85, 95% CI=1.21 to 2.82, respectively).</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:04:47383:1:2:NEW 7 Aug 2020)</ns0:p><ns0:p>Manuscript to be reviewed</ns0:p></ns0:div>
<ns0:div><ns0:head>Discussion</ns0:head><ns0:p>Our study demonstrated that maternal underweight prior to pregnancy, as compared with normal weight women, significantly elevated the risk for LBW for the first-time mothers among Southern Chinese. Maternal underweight were also found to be at lower risk of gestational hypertension and caesarean delivery. In women who had a high pre-pregnancy BMI, our study showed a significantly higher risk of gestational hypertension, caesarean delivery and extremely PTB.</ns0:p><ns0:p>The proportions of overweight and obesity are lower than underweight in our study, which are similar to previously reported data from other Chinese studies <ns0:ref type='bibr' target='#b40'>(Pan et al. 2016)</ns0:ref>. Our finding showed that underweight women increase the risk of LBW in subsequent pregnancy, which is consistent with the result of a review study by <ns0:ref type='bibr' target='#b27'>Liu et al. (Liu et al. 2016</ns0:ref>). The study, including 60 studies covering 1,392,799 women, showed that infants had a higher risk of having a LBW when their mothers were underweight (OR, 1.67, 95%CI, 1.39-2.02) as compare to women with normal weight. Previous studies have demonstrated that maternal nutrition during pregnancy has great influence on providing the essential nutrients for fetal growth <ns0:ref type='bibr' target='#b37'>(Nnam 2015)</ns0:ref> and pregnant women who are undernourished tend to have LBW infant <ns0:ref type='bibr' target='#b0'>(Allen 2001)</ns0:ref>. Maternal nutritional imbalance may be a key factor in the reduction of surface area and placental weight. In the lower surface area and placental weight, nutrient and waste transfer between the maternal and fetal circulation is reduced and other normal processes, such as fetal development and growth, are also restricted <ns0:ref type='bibr' target='#b21'>(Lechtig et al. 1975)</ns0:ref>. Thus, maternal malnutrition may lead to LBW infant. Moreover, the finding in our study on association between underweight and LBW was also in line with Manuscript to be reviewed previous literature <ns0:ref type='bibr' target='#b23'>(Li et al. 2013;</ns0:ref><ns0:ref type='bibr' target='#b28'>Liu et al. 2012)</ns0:ref>.</ns0:p><ns0:p>Previous literature has revealed that maternal overweight and obesity were associated with the increased risk for gestational hypertension and caesarian delivery <ns0:ref type='bibr' target='#b22'>(Lewandowska et al. 2020;</ns0:ref><ns0:ref type='bibr'>Machado et al. 2020;</ns0:ref><ns0:ref type='bibr' target='#b54'>Vince et al. 2020)</ns0:ref>. Our study also confirmed these findings. However, the risk of gestational hypertension and caesarean delivery were reduced among underweight mothers. Although mechanism by which obesity responsible for the increased risk of gestational hypertension or caesarean delivery is unclear, maternal obesity lead to an increase in the number and size of adipocytes or pelvic malacia. A significant amount of adipocytes has been proposed as a cause of excessive inflammatory reaction, pregnant female possibly experienced obstructive dystocia due to pelvic malacia lead to a relatively narrow birth canal. Therefore, both gestational hypertension and caesarean delivery were affected by obesity in obese mothers <ns0:ref type='bibr' target='#b19'>(Kriketos et al. 2004;</ns0:ref><ns0:ref type='bibr' target='#b58'>Young & Woodmansee 2002)</ns0:ref>.</ns0:p><ns0:p>In our study, the risk of PTB is relatively higher among overweight and obese mothers but the association was not statistically significant (table <ns0:ref type='table' target='#tab_1'>2</ns0:ref>). However, this association has dramatically changed in a subgroup analysis of gestational age (table <ns0:ref type='table' target='#tab_4'>3</ns0:ref>). Our result regarding extremely PTB is consistent with a previous study <ns0:ref type='bibr' target='#b53'>(Su et al. 2020)</ns0:ref>. Overweight and obesity are generally considered to be the risk factors for PTB due to the effects of placental insufficiency <ns0:ref type='bibr' target='#b20'>(Lassance et al. 2015;</ns0:ref><ns0:ref type='bibr' target='#b41'>Pereira et al. 2015)</ns0:ref>, inflammatory state <ns0:ref type='bibr' target='#b11'>(Gaillard et al. 2016)</ns0:ref>, insulin sensitivity <ns0:ref type='bibr' target='#b5'>(Catalano et al. 1999</ns0:ref>) and cellular oxidative stress <ns0:ref type='bibr' target='#b2'>(Ballesteros-Guzman et al. 2019</ns0:ref>).</ns0:p><ns0:p>However, conclusions on the relationship between pre-pregnancy BMI and PTB seem to be paradoxical among different studies. The differences emerged between studies can be attributed Manuscript to be reviewed to study design or power, recall bias, multiple comparisons, eating habits and different ethnicities.</ns0:p><ns0:p>Additionally, the category of BMI was different among the studies.</ns0:p><ns0:p>Although some confounding factors had been controlled, alcohol consumption and maternal smoking were not adjusted as only one woman claimed to have a history of alcohol consumption and smoking in our data. So we did not adjust for these two variables.</ns0:p><ns0:p>However, there are a number of potential limitations of this study that merit consideration.</ns0:p><ns0:p>One limitation of the current study is that it is difficult to distinguish between spontaneous and iatrogenic PTB, which we could not assess the association between special types of PTB and prepregnancy BMI. Additionally, the BMI we obtained in our study derived from weight and height information by women self-reported, which could lead to bias risk estimates of PTB.</ns0:p><ns0:p>Additionally, the BMI we obtained in our study derived from weight and height information by women self-reported, which could lead to bias risk estimates of PTB <ns0:ref type='bibr' target='#b34'>(Michels et al. 1998)</ns0:ref>.</ns0:p><ns0:p>In conclusion, our study suggested that maternal overweight and obesity were associated with a significantly higher risk of gestational hypertension, caesarean delivery and extremely PTB. Underweight was correlated with an increased risk of LBW and conferred a protective effect regarding the risk for gestational hypertension and caesarean delivery for the first-time mothers among Southern Chinese.</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:04:47383:1:2:NEW 7 Aug 2020)</ns0:p><ns0:p>Manuscript to be reviewed Manuscript to be reviewed Crude and AOR for the association between pre-pregnancy BMI and adverse outcomes Manuscript to be reviewed Manuscript to be reviewed Adjusted* a relative risk for the associations between pre-pregnancy BMI and weight for gestational age Manuscript to be reviewed </ns0:p></ns0:div><ns0:figure xml:id='fig_0'><ns0:head /><ns0:label /><ns0:figDesc>) term and NSGA; (2) 34-36 week gestation and NSGA; (3) below 34 week gestation and NSGA; (4) term and SAG; (5) 34-36 week gestation and SAG; (6) below 34 week gestation PeerJ reviewing PDF | (</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_1'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:04:47383:1:2:NEW 7 Aug 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_2'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:04:47383:1:2:NEW 7 Aug 2020)</ns0:figDesc></ns0:figure>
<ns0:figure type='table' xml:id='tab_0'><ns0:head>Table 1 :</ns0:head><ns0:label>1</ns0:label><ns0:figDesc>2 Baseline maternal characteristics of the first-time mothers between two groups 3 Abbreviation: PTB, preterm birth; LBW, low birth weight.</ns0:figDesc><ns0:table><ns0:row><ns0:cell>variable</ns0:cell><ns0:cell>Terms (N=3362)</ns0:cell><ns0:cell>PTB (N=322)</ns0:cell><ns0:cell>P</ns0:cell><ns0:cell>Normal (N=3367)</ns0:cell><ns0:cell>LBW (N=317)</ns0:cell><ns0:cell>P</ns0:cell></ns0:row><ns0:row><ns0:cell>Age</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell><0.001</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell><0.001</ns0:cell></ns0:row><ns0:row><ns0:cell>< 25</ns0:cell><ns0:cell>421 (12.52)</ns0:cell><ns0:cell>55 (17.08)</ns0:cell><ns0:cell /><ns0:cell>418 (12.41)</ns0:cell><ns0:cell>58 (18.30)</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>25-34</ns0:cell><ns0:cell>2628 (78.17)</ns0:cell><ns0:cell>220 (68.32)</ns0:cell><ns0:cell /><ns0:cell>2630 (78.11)</ns0:cell><ns0:cell>218 (68.77)</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>≧35</ns0:cell><ns0:cell>313 (9.31)</ns0:cell><ns0:cell>47 (14.60)</ns0:cell><ns0:cell /><ns0:cell>319 (9.48)</ns0:cell><ns0:cell>41 (12.93)</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>Health insurance</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell><0.001</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell><0.001</ns0:cell></ns0:row><ns0:row><ns0:cell>Care for urban employees</ns0:cell><ns0:cell>2269 (67.49)</ns0:cell><ns0:cell>161 (50.00)</ns0:cell><ns0:cell /><ns0:cell>2261 (67.15)</ns0:cell><ns0:cell>169 (53.31)</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>Free medical service</ns0:cell><ns0:cell>87 (2.59)</ns0:cell><ns0:cell>6 (1.86)</ns0:cell><ns0:cell /><ns0:cell>90 (2.67)</ns0:cell><ns0:cell>3 (0.95)</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>Full-cost</ns0:cell><ns0:cell>1006 (29.92)</ns0:cell><ns0:cell>155 (47.14)</ns0:cell><ns0:cell /><ns0:cell>1016 (30.18)</ns0:cell><ns0:cell>145 (45.74)</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>Occupation</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell>0.004</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell>0.007</ns0:cell></ns0:row><ns0:row><ns0:cell>Professional</ns0:cell><ns0:cell>1534 (45.63)</ns0:cell><ns0:cell>119 (36.96)</ns0:cell><ns0:cell /><ns0:cell>1532 (45.50)</ns0:cell><ns0:cell>121 (38.17)</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>Business</ns0:cell><ns0:cell>355 (10.56)</ns0:cell><ns0:cell>28 (8.70)</ns0:cell><ns0:cell /><ns0:cell>358 (10.63)</ns0:cell><ns0:cell>25 (7.89)</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>Housewife</ns0:cell><ns0:cell>389 (11.57)</ns0:cell><ns0:cell>46 (14.29)</ns0:cell><ns0:cell /><ns0:cell>388 (11.52)</ns0:cell><ns0:cell>47 (14.83)</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>Others</ns0:cell><ns0:cell>1084 (32.24)</ns0:cell><ns0:cell>129 (40.06)</ns0:cell><ns0:cell /><ns0:cell>1089 (32.35)</ns0:cell><ns0:cell>124 (39.12)</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>Gravidity</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell>0.214</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell>0.875</ns0:cell></ns0:row><ns0:row><ns0:cell>< 2</ns0:cell><ns0:cell>2245 (66.78)</ns0:cell><ns0:cell>204 (63.35)</ns0:cell><ns0:cell /><ns0:cell>2237 (66.44)</ns0:cell><ns0:cell>212 (66.88)</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>≧2</ns0:cell><ns0:cell>1117 (33.22)</ns0:cell><ns0:cell>118 (36.65)</ns0:cell><ns0:cell /><ns0:cell>1130 (33.56)</ns0:cell><ns0:cell>105 (33.12)</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>Education</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell><0.001</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell><0.001</ns0:cell></ns0:row><ns0:row><ns0:cell>Less than high school</ns0:cell><ns0:cell>439 (13.06)</ns0:cell><ns0:cell>86 (26.71)</ns0:cell><ns0:cell /><ns0:cell>444 (13.19)</ns0:cell><ns0:cell>81 (25.55)</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>High school</ns0:cell><ns0:cell>444 (13.21)</ns0:cell><ns0:cell>38 (11.80)</ns0:cell><ns0:cell /><ns0:cell>443 (13.16)</ns0:cell><ns0:cell>39 (12.30)</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>College</ns0:cell><ns0:cell>2479 (73.73)</ns0:cell><ns0:cell>198 (61.49)</ns0:cell><ns0:cell /><ns0:cell>2480 (73.65)</ns0:cell><ns0:cell>197 (62.15)</ns0:cell><ns0:cell /></ns0:row></ns0:table><ns0:note>1 PeerJ reviewing PDF | (2020:04:47383:1:2:NEW 7 Aug 2020)</ns0:note></ns0:figure>
<ns0:figure type='table' xml:id='tab_1'><ns0:head>Table 2 (on next page)</ns0:head><ns0:label>2</ns0:label><ns0:figDesc /><ns0:table /></ns0:figure>
<ns0:figure type='table' xml:id='tab_2'><ns0:head>Table 2</ns0:head><ns0:label>2</ns0:label><ns0:figDesc>PeerJ reviewing PDF | (2020:04:47383:1:2:NEW 7 Aug 2020)</ns0:figDesc><ns0:table><ns0:row><ns0:cell>Manuscript to be reviewed</ns0:cell></ns0:row></ns0:table><ns0:note>1</ns0:note></ns0:figure>
<ns0:figure type='table' xml:id='tab_3'><ns0:head>Table 2 :</ns0:head><ns0:label>2</ns0:label><ns0:figDesc>2 Crude and AOR for the association between pre-pregnancy BMI and adverse outcomes</ns0:figDesc><ns0:table><ns0:row><ns0:cell>Outcomes</ns0:cell><ns0:cell>BMI status</ns0:cell><ns0:cell>Case</ns0:cell><ns0:cell>Control</ns0:cell><ns0:cell>OR (95% CI)</ns0:cell><ns0:cell>P</ns0:cell><ns0:cell>AOR* (95% CI)</ns0:cell><ns0:cell>P</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell>783</ns0:cell><ns0:cell>1.04 (0.79-</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>PTB</ns0:cell><ns0:cell>Underweight</ns0:cell><ns0:cell>75 (23.29)</ns0:cell><ns0:cell>(23.29)</ns0:cell><ns0:cell>1.37)</ns0:cell><ns0:cell>0.787</ns0:cell><ns0:cell>1.01 (0.76-1.34)</ns0:cell><ns0:cell>0.943</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell>197</ns0:cell><ns0:cell>2137</ns0:cell><ns0:cell>1.00</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell /><ns0:cell>Normal weight</ns0:cell><ns0:cell>(61.18)</ns0:cell><ns0:cell>(63.56)</ns0:cell><ns0:cell>(reference)</ns0:cell><ns0:cell /><ns0:cell>1.00 (reference)</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell>1.22 (0.86-</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell /><ns0:cell>Overweight</ns0:cell><ns0:cell>40 (12.42)</ns0:cell><ns0:cell>355 (10.56)</ns0:cell><ns0:cell>1.75)</ns0:cell><ns0:cell>0.272</ns0:cell><ns0:cell>1.25 (0.87-1.80)</ns0:cell><ns0:cell>0.224</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell>1.25 (0.64-</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell /><ns0:cell>Obesity</ns0:cell><ns0:cell>10 (3.11)</ns0:cell><ns0:cell>87 (2.59)</ns0:cell><ns0:cell>2.44)</ns0:cell><ns0:cell>0.519</ns0:cell><ns0:cell>1.27 (0.65-2.51)</ns0:cell><ns0:cell>0.486</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell>1.48 (1.14-</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>LBW</ns0:cell><ns0:cell>Underweight</ns0:cell><ns0:cell>94 (29.65)</ns0:cell><ns0:cell>764 (22.69)</ns0:cell><ns0:cell>1.93)</ns0:cell><ns0:cell>0.003</ns0:cell><ns0:cell>1.44 (1.11-1.89)</ns0:cell><ns0:cell>0.007</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell>179</ns0:cell><ns0:cell>2155</ns0:cell><ns0:cell>1.00</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell /><ns0:cell>Normal weight</ns0:cell><ns0:cell>(56.47)</ns0:cell><ns0:cell>(64.00)</ns0:cell><ns0:cell>(reference)</ns0:cell><ns0:cell /><ns0:cell>1.00 (reference)</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell>1.13 (0.77-</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell /><ns0:cell>Overweight</ns0:cell><ns0:cell>34 (10.73)</ns0:cell><ns0:cell>361 (10.73)</ns0:cell><ns0:cell>1.66)</ns0:cell><ns0:cell>0.521</ns0:cell><ns0:cell>1.17 (0.80-1.73)</ns0:cell><ns0:cell>0.423</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell>1.38 (0.71-</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell /><ns0:cell>Obesity</ns0:cell><ns0:cell>10 (3.15)</ns0:cell><ns0:cell>87 (2.58)</ns0:cell><ns0:cell>2.71)</ns0:cell><ns0:cell>0.343</ns0:cell><ns0:cell>1.41 (0.72-2.78)</ns0:cell><ns0:cell>0.322</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell>0.43 (0.24-</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell cols='2'>Gestational hypertension Underweight</ns0:cell><ns0:cell>13 (9.92)</ns0:cell><ns0:cell>845 (23.78)</ns0:cell><ns0:cell>0.78)</ns0:cell><ns0:cell>0.006</ns0:cell><ns0:cell>0.45 (0.25-0.82)</ns0:cell><ns0:cell>0.009</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell>2254</ns0:cell><ns0:cell>1.00</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell /><ns0:cell>Normal weight</ns0:cell><ns0:cell>80 (61.07)</ns0:cell><ns0:cell>(63.44)</ns0:cell><ns0:cell>(reference)</ns0:cell><ns0:cell /><ns0:cell>1.00 (reference)</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell>1.74 (1.08-</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell /><ns0:cell>Overweight</ns0:cell><ns0:cell>23 (17.56)</ns0:cell><ns0:cell>372 (10.47)</ns0:cell><ns0:cell>2.81)</ns0:cell><ns0:cell>0.022</ns0:cell><ns0:cell>1.71 (1.06-2.77)</ns0:cell><ns0:cell>0.03</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell>5.15 (2.85-</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell /><ns0:cell>Obesity</ns0:cell><ns0:cell>15 (11.45)</ns0:cell><ns0:cell>82 (2.31)</ns0:cell><ns0:cell>9.33)</ns0:cell><ns0:cell><0.001</ns0:cell><ns0:cell>5.54 (3.02-10.17)</ns0:cell><ns0:cell><0.001</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell>194</ns0:cell><ns0:cell /><ns0:cell>0.70 (0.59-</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>Cesarean delivery</ns0:cell><ns0:cell>Underweight</ns0:cell><ns0:cell>(17.60)</ns0:cell><ns0:cell>664 (25.72)</ns0:cell><ns0:cell>0.84)</ns0:cell><ns0:cell><0.001</ns0:cell><ns0:cell>0.74 (0.62-0.90)</ns0:cell><ns0:cell>0.002</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell>686</ns0:cell><ns0:cell>1684</ns0:cell><ns0:cell>1.00</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell /><ns0:cell>Normal weight</ns0:cell><ns0:cell>(62.25)</ns0:cell><ns0:cell>(63.83)</ns0:cell><ns0:cell>(reference)</ns0:cell><ns0:cell /><ns0:cell>1.00 (reference)</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell>180</ns0:cell><ns0:cell /><ns0:cell>2.01 (1.62-</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell /><ns0:cell>Overweight</ns0:cell><ns0:cell>(16.33)</ns0:cell><ns0:cell>215 (8.33)</ns0:cell><ns0:cell>2.50)</ns0:cell><ns0:cell><0.001</ns0:cell><ns0:cell>1.91 (1.53-2.38)</ns0:cell><ns0:cell><0.001</ns0:cell></ns0:row></ns0:table><ns0:note>PeerJ reviewing PDF | (2020:04:47383:1:2:NEW 7 Aug 2020)Manuscript to be reviewed</ns0:note></ns0:figure>
<ns0:figure type='table' xml:id='tab_4'><ns0:head>Table 3 (on next page) Adjusted</ns0:head><ns0:label>3</ns0:label><ns0:figDesc /><ns0:table /><ns0:note>* a relative risk for the associations between pre-pregnancy BMI and PTB by gestational age</ns0:note></ns0:figure>
<ns0:figure type='table' xml:id='tab_5'><ns0:head>Table 3</ns0:head><ns0:label>3</ns0:label><ns0:figDesc>PeerJ</ns0:figDesc><ns0:table /><ns0:note>reviewing PDF | (2020:04:47383:1:2:NEW 7 Aug 2020)</ns0:note></ns0:figure>
<ns0:figure type='table' xml:id='tab_6'><ns0:head>Table 3 :</ns0:head><ns0:label>3</ns0:label><ns0:figDesc>Adjusted * a relative risk for the associations between pre-pregnancy BMI and PTB by gestational age Adjusted OR and 95% CI were calculated by the logistic regression model after adjusting for age, health insurance, occupation and education.</ns0:figDesc><ns0:table><ns0:row><ns0:cell>Gestational age</ns0:cell><ns0:cell>Underweight</ns0:cell><ns0:cell>Normal weight</ns0:cell><ns0:cell>Overweight</ns0:cell><ns0:cell>Obesity</ns0:cell></ns0:row><ns0:row><ns0:cell>Term</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>1.00 (reference)</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell></ns0:row><ns0:row><ns0:cell>Moderately PTB</ns0:cell><ns0:cell>0.97 (0.71-1.33)</ns0:cell><ns0:cell>1.00 (reference)</ns0:cell><ns0:cell>1.18 (0.79-1.77)</ns0:cell><ns0:cell>1.23 (0.58-2.59)</ns0:cell></ns0:row><ns0:row><ns0:cell>Very PTB</ns0:cell><ns0:cell>1.05 (0.54-2.02)</ns0:cell><ns0:cell>1.00 (reference)</ns0:cell><ns0:cell>1.28 (0.55-2.96)</ns0:cell><ns0:cell>0.77 (0.10-5.77)</ns0:cell></ns0:row><ns0:row><ns0:cell>Extremely PTB</ns0:cell><ns0:cell>3.22 (0.53-19.59)</ns0:cell><ns0:cell>1.00 (reference)</ns0:cell><ns0:cell>8.12 (1.11-59.44)</ns0:cell><ns0:cell>15.06 (1.32-172.13)</ns0:cell></ns0:row></ns0:table><ns0:note>Abbreviations: PTB, preterm birth. * PeerJ reviewing PDF | (2020:04:47383:1:2:NEW 7 Aug 2020)</ns0:note></ns0:figure>
<ns0:figure type='table' xml:id='tab_7'><ns0:head>Table 4 (on next page)</ns0:head><ns0:label>4</ns0:label><ns0:figDesc /><ns0:table /></ns0:figure>
<ns0:figure type='table' xml:id='tab_8'><ns0:head>Table 4</ns0:head><ns0:label>4</ns0:label><ns0:figDesc>PeerJ</ns0:figDesc><ns0:table /><ns0:note>reviewing PDF | (2020:04:47383:1:2:NEW 7 Aug 2020)</ns0:note></ns0:figure>
<ns0:figure type='table' xml:id='tab_9'><ns0:head>Table 4 :</ns0:head><ns0:label>4</ns0:label><ns0:figDesc>Adjusted* a relative risk for the associations between pre-pregnancy BMI and weight for gestational age Adjusted OR and 95% CI were calculated by the logistic regression model after adjusting for age, health insurance, occupation and education.</ns0:figDesc><ns0:table><ns0:row><ns0:cell>Weight for gestational age</ns0:cell><ns0:cell>Underweight</ns0:cell><ns0:cell>Normal weight</ns0:cell><ns0:cell>Overweight</ns0:cell><ns0:cell>Obesity</ns0:cell></ns0:row><ns0:row><ns0:cell>NSAG</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>Term</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>1.00 (reference)</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell></ns0:row><ns0:row><ns0:cell>34-36 week</ns0:cell><ns0:cell>0.88 (0.58-1.34)</ns0:cell><ns0:cell>1.00 (reference)</ns0:cell><ns0:cell>1.29 (0.80-2.06)</ns0:cell><ns0:cell>1.40 (0.59-3.29)</ns0:cell></ns0:row><ns0:row><ns0:cell><34 week</ns0:cell><ns0:cell>1.52 (0.94-2.47)</ns0:cell><ns0:cell>1.00 (reference)</ns0:cell><ns0:cell>1.48 (0.77-2.84)</ns0:cell><ns0:cell>0.97 (0.23-4.11)</ns0:cell></ns0:row><ns0:row><ns0:cell>SAG</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>Term</ns0:cell><ns0:cell>1.78 (1.45-2.17)</ns0:cell><ns0:cell>1.00 (reference)</ns0:cell><ns0:cell>0.88 (0.63-1.21)</ns0:cell><ns0:cell>1.89 (0.48-1.66)</ns0:cell></ns0:row><ns0:row><ns0:cell>34-36 week</ns0:cell><ns0:cell>1.49 (0.79-2.81)</ns0:cell><ns0:cell>1.00 (reference)</ns0:cell><ns0:cell>0.43 (0.10-1.82)</ns0:cell><ns0:cell>0.86 (0.12-6.41)</ns0:cell></ns0:row><ns0:row><ns0:cell><34 week</ns0:cell><ns0:cell>0.73 (0.20-2.63)</ns0:cell><ns0:cell>1.00 (reference)</ns0:cell><ns0:cell>1.62 (0.44-5.89)</ns0:cell><ns0:cell>2.23 (0.28-17.91)</ns0:cell></ns0:row></ns0:table><ns0:note>Abbreviations: NSAG, non-small for gestational age; SAG, small for gestational age.* PeerJ reviewing PDF | (2020:04:47383:1:2:NEW 7 Aug 2020)</ns0:note></ns0:figure>
<ns0:note place='foot' n='3'>Abbreviations: OR, odds ratio; CI, confidence interval; AOR, adjusted odds ratio; PTB, preterm birth; LBW, low birth weight. 4 * Adjusted OR and 95% CI were calculated by the logistic regression model after adjusting for age, health insurance, occupation and education.PeerJ reviewing PDF | (2020:04:47383:1:2:NEW 7 Aug 2020)</ns0:note>
</ns0:body>
" | "Guangzhou, 7th, August 2020
Dear Professor Qing-Yuan Sun,
Thank you very much for your letter and your efforts in review of our manuscript. We would like to resubmit the manuscript entitled “Association of Pre-Pregnancy Body Mass Index with adverse pregnancy outcome among first-time mothers” (ID: 47383).
We appreciate the reviewer for the valuable comments. We have addressed, point-by-point, the issues raised by the reviewer. The amendments made in the original version of the manuscript are highlighted in red in the revised version. We hope that these changes will make the manuscript acceptable for publication in PeerJ. If there are additional changes that we should make, please let us know. Thank you very much for your consideration.
Yours sincerely
Dr. Ke Li
Key Laboratory for Major Obstetric Diseases of Guangdong Province, the Third Affiliated Hospital of Guangzhou Medical University, Guangzhou, China
Phone: 86-20-81292138
FAX: 86-20-81292908
E-mail: keli1221@126.com
Reviewer 1
Basic reporting
In this manuscript, Li Li et al conducted a study on a total of 3,684 women who gave birth for the first time in south of China to investigate the associations of preterm birth (PTB), low birth weight (LBW) with pre-pregnancy Body Mass Index (BMI). Their study suggests that underweight women increase the risk of LBW.
Although their result is consistent with many others, there are several major concerns as outlined below:
1. No big differences/advantages of this study are from the literatures (J Nutr. 2003 Nov;133(11):3449-55; BMC Pregnancy Childbirth. 2019 Mar 29;19(1):105....).
Response: Thanks for your positive comments and constructive suggestions. Notably, the first study mentioned above was conducted in a eastern Chinese women rather than southern Chinese women. Secondly, the sample size of the study is limited. However, The advantages of our study are the relatively large sample size as compare with the previous study.
As for the second article mentioned above, the study is a meta-analysis, only three Chinese articles are included, However, these three previous papers were mainly limited to earlier data or pregnant women in rural areas. In contrast, our study was conducted in an urban centre.
We have added these two references into the list of references. In summary, our study had some advantages that prior research mentioned above has not offered. These have been added into the introduction part.
2. This manuscript should be clearer, more succinct, and professional. For example: in Abstract, 'Risks of PTB and LBW in the first subsequent pregnancy in women who had a abnomal BMI were compared with risks in women who had a ideal BMI.”
Response: We have tried to go over the manuscript carefully and make it clearer and more accessible to readers. The abstract has been rewritten and this text removed.
3. It is better to show some data in Figure although the tables are accepted.
Response: This is a very good advice. We will use graphics to show the results as much as possible in future research.
We highly appreciate your careful review and critical comments, which are very helpful for improving our manuscript. And we have carefully revised the manuscript as requested. Once again, thank you very much for your comments and suggestions.
Reviewer 2
Basic reporting
This study was designed to elucidate the association between maternal BMI and adverse pregnancy outcomes for first-time mothers in Southern China. The topic is valuable taking into account that obesity is increasing rapidly worldwide, particularly in rapid developing countries like China. However, I have some comments/concerns:
There have been many studies regarding the association of maternal BMI and offspring birth outcomes, in and out of China, such as Rahman MM, Abe SK, et al. 10.1111/obr.12293, Pan Y, Zhang S et al, 10.1136/bmjopen-2016-011227. The authors did not provide a comprehensive background in Introduction. Several latest published papers were not incorporated.
Response: Thanks for your positive comments. We have added these two references into the list of references. Besides, we have incorporated several latest published papers in the introduction section. And the introduction section have been partly rewritten.
Moreover, the language is hard to follow. There are flaws and mistakes in written English. For ex, Introduction, line 58 “both” should be not capitalized; line 62, “is” should be “was”; Materials and methods, line 81 and 83, repeated “the control group”; Discussion Line 157, “showed” is in fact “showing”, line 158, “ruduced” should be “reduced”. It would be useful to have a native English speaker edit the manuscript for clarity and concise language.
Response: Thank you for your careful review of the manuscript. These have been corrected now. We have tried to go over the manuscript carefully and make it clearer and more accessible to readers.
Experimental design
1) First, the meaning of this study is not well explained in Introduction. Given the many reports on this topic, the authors did not address clearly the importance and novelty for this study. Is it for Southern Chinese, or first-time mothers?
Response: The introduction section has been partly rewritten. Although several researcher have conducted relevant research in domestic. However, these previous papers were mainly limited to earlier data or pregnant women in rural areas. Study is still scarce among pregnant women in urban areas for first-time mothers among Southern Chinese.
2) As the authors mentioned, preterm birth is related to a variety of factors. Gestational hypertension may be a mediator between maternal BMI and PTB or low birth weight. To simply adjust for gestational hypertension in the regression is not suitable.
Response: Good advice. Gestational hypertension and cesarean delivery have been used as a maternal health outcome. To explore the effect of pre-pregnancy BMI on gestational hypertension and cesarean delivery.
3) Gestational diabetes, twin pregnancy, and gestational weight gain are also important factors influencing the association. Given your dataset, are you able to analyze the association taking into account these factors?
Response: We apologize for this omission, these data are not included in our dataset, but this suggestion is instructive for our future research, we will pay more attention to the early research design.
4) Birth weight is a key indicator for intrauterine environment. However, birth weight adjusted for gestational age would be more appropriate than birth weight itself.
Response: We completely agree. We have added the detail at method section“We referred to a method application to weight for gestational age by Tanya Marchant et al (Marchant et al. 2012). Which categorized weight for gestational age into two group, small for gestational age (SGA, less than the tenth percentile of weight for gestation) (Oken et al. 2003) and non-small for gestational age (NSGA, large than the tenth percentile of weight for gestation). We divided newborns into six groups: (1) term and NSGA; (2) 34–36 week gestation and NSGA; (3) below 34 week gestation and NSGA; (4) term and SAG; (5) 34–36 week gestation and SAG; (6) below 34 week gestation and SAG”. We found that maternal underweight significantly elevated the risk for LBW in the term and SAG group as compared with the term and NSAG group (AOR=1.78, 95% CI 1.45–2.17).
5) Also, information on early and late PTB is useful for health professionals. It would be useful to examine the risk of PTB according to gestational age with maternal BMI strata.
Response: Good advice. We divided gestational age into four subtypes: extremely PTB (<28 gestational week), very PTB (28–31 gestational week), moderate PTB (32–36 gestational week) and normal (≧37 gestational week). We found that the risk of extremely PTB is relatively higher among overweight and obese mothers in a subgroup analysis of PTB (AOR=8.12, 95% CI 1.11–59.44; AOR=15.06, 95% CI 1.32–172.13, respectively).
Validity of the findings
1) Table 2 and 3 are not well structured. In Results, the associations between overweight and obese mothers with offspring birth outcomes are not presented.
Response: Thank you for your comment. We have restructured the table according to the reviewer’s comment. We have integrated the original two tables (table2 and table3) into a single table, the detail was seen in new version of table 2.
2) Please explain why this study did not find an association between high BMI and PTB.
Response: In our study, the risk of PTB is relatively higher among overweight and obese mothers but the association was not statistically significant (table 2). However, When preterm birth was stratified into three subgroups, this association has dramatically changed in a subgroup analysis of gestational age (table 3). both maternal overweight and obesity increased the risks of extremely PTB (AOR=8.12, 95% CI 1.11–59.44; AOR=15.06, 95% CI 1.32–172.13, respectively). However, conclusions on the relationship between pre-pregnancy BMI and PTB seem to be paradoxical among different studies. We try to explain the reason for this result. The differences emerged between studies can be attributed to study design or power, recall bias, multiple comparisons, eating habits and different ethnicities. Additionally, the category of BMI was different among the studies.
3) The discussion part is not well organized and hard to follow
Response: Thank you for your comment. We have restructured the discussion part according to the reviewer’s comment. The discussion section is divided into seven paragraphs. The main content are listed below in order:
(1) Paragraph 1: Summary of results, maternal underweight significantly elevated the risk for LBW for the first-time mothers among Southern Chinese. Maternal underweight were also found to be at lower risk of gestational hypertension and caesarean delivery. In women who had a high pre-pregnancy BMI, our study showed a significantly higher risk of gestational hypertension, caesarean delivery and extremely PTB.
(2) Paragraph 2: We discuss the association between maternal underweight and LBW.
(3) Paragraph 3: We discuss the association between pre-pregnancy BMI and gestational hypertension and caesarean delivery.
(4) Paragraph 4: We discuss the association between high pre-pregnancy BMI and PTB.
(5) Paragraph 5: Reasons for not adjusting smoking and drinking.
(6) Paragraph 6: Limitations of our study.
(7) Paragraph 7: Draw conclusions.
We highly appreciate your careful review and critical comments, which are very helpful for improving our manuscript. And we have carefully revised the manuscript as requested. Once again, thank you very much for your comments and suggestions.
" | Here is a paper. Please give your review comments after reading it. |
9,761 | "<ns0:abstract xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:p>Background: Studies have reported an increased risk of adverse pregnancy outcome associated with pre-pregnancy body mass index (BMI). However, the data on such associations in urban areas of southern Chinese women is limited, which drive us to clarify the associations of pre-pregnancy BMI and the risks of adverse pregnancy outcomes (preterm birth (PTB) and low birth weight (LBW)) and maternal health outcomes (gestational hypertension and cesarean delivery). Methods: We performed a hospitalbased case-control study including 3,864 Southern Chinese women who gave first birth to a live singleton infant from January 2015 to December 2015. PTB was stratified into three subgroups according to gestational age (extremely PTB, very PTB and moderate PTB).</ns0:p><ns0:p>Besides, we combined birth weight and gestational age to dichotomise as being small for gestational age (SGA, less than the tenth percentile of weight for gestation) and non-small for gestational age (NSGA, large than the tenth percentile of weight for gestation), gestational week was also classified into categories of term, 34-36 week and below 34 week.. We then divided newborns into six groups: (1) term and NSGA; (2) 34-36 week gestation and NSGA; (3) below 34 week gestation and NSGA; (4) term and SAG; (5) 34-36 week gestation and SAG; (6) below 34 week gestation and SAG. Adjusted logistic regression models was used to estimate the odds ratios of adverse outcomes. Results:</ns0:p><ns0:p>Underweight women were more likely to give LBW (AOR=1.44, 95% CI 1.11 to 1.89), the similar result was seen in term and SAG as compare with term and NSAG (AOR=1.78, 95% CI 1.45-2.17). Whereas underweight was significantly associated with a lower risk of gestational hypertension (AOR=0.45, 95% CI 0.25-0.82) and caesarean delivery (AOR=0.74, 95% CI 0.62-0.90). The risk of extremely PTB is relatively higher among overweight and obese mothers in a subgroup analysis of PTB (AOR=8.12, 95% CI=1.11 to</ns0:p></ns0:div>
</ns0:abstract>
<ns0:body xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:head>Introduction</ns0:head><ns0:p>Preterm birth (PTB) is an important adverse pregnancy outcome with a significant impact on infant mortality and morbidity <ns0:ref type='bibr' target='#b4'>(Boghossian et al. 2016)</ns0:ref>. The incidence of PTB worldwide is expected to be 11.1%, and China has more than 1.1 million PTB each year, which ranks second in the world <ns0:ref type='bibr' target='#b3'>(Blencowe et al. 2012)</ns0:ref>. In spite of high-level advancement in healthcare services, the rate of PTB still seems to be climbing. According to statistics, perinatal mortality is as high as 70% in low birth weight (LBW) with neonatal weight below 2500 grams <ns0:ref type='bibr' target='#b12'>(Hack et al. 1994;</ns0:ref><ns0:ref type='bibr' target='#b48'>Rutter 1995</ns0:ref>), most of them are born preterm. Growing evidence has shown that the incidence and development of PTB is a complex process that is influenced by a variety of environmental and genetic factors <ns0:ref type='bibr' target='#b15'>(He et al. 2017;</ns0:ref><ns0:ref type='bibr' target='#b25'>Liu et al. 2017;</ns0:ref><ns0:ref type='bibr' target='#b42'>Qiu et al. 2017;</ns0:ref><ns0:ref type='bibr' target='#b57'>Xiao et al. 2016)</ns0:ref>. Therefore, it is helpful to develop approaches of effective prevention and treatment of neonatal morbidity and mortality by elucidating etiological factors contributing to PTB or LBW.</ns0:p><ns0:p>Recent results provide support to pre-pregnancy maternal body mass index (BMI) is one of the potential risk factors for PTB <ns0:ref type='bibr' target='#b17'>(Hendler et al. 2005;</ns0:ref><ns0:ref type='bibr' target='#b29'>Lynch et al. 2014;</ns0:ref><ns0:ref type='bibr' target='#b52'>Shaw et al. 2014)</ns0:ref> and LBW <ns0:ref type='bibr' target='#b13'>(Han et al. 2011;</ns0:ref><ns0:ref type='bibr' target='#b27'>Liu et al. 2016)</ns0:ref>. In recent decades, the prevalence of obesity and overweight among women in many countries has increased at an alarming rate, especially in developing countries. Different countries, regions and incomes have different patterns of overweight and obesity that is more common among women in developing countries and men in developed countries <ns0:ref type='bibr'>(Ng et al. 2014)</ns0:ref>. Moreover, epidemiological studies have been suggested that maternal overweight and obesity have been shown to be association with PTB <ns0:ref type='bibr' target='#b29'>(Lynch et al. 2014;</ns0:ref><ns0:ref type='bibr' target='#b52'>Shaw et al. 2014;</ns0:ref><ns0:ref type='bibr' target='#b53'>Su et al. 2020)</ns0:ref>, LBW <ns0:ref type='bibr' target='#b44'>(Rahman et al. 2015)</ns0:ref> and adverse maternal health outcomes, such as gestational hypertension <ns0:ref type='bibr' target='#b50'>(Santos et al. 2019</ns0:ref>) and cesarean delivery (Paidas <ns0:ref type='bibr' target='#b39'>Teefey et al. 2020;</ns0:ref><ns0:ref type='bibr' target='#b44'>Rahman et al. 2015)</ns0:ref>. Similarly, several observational studies show that underweight women in pre-pregnancy is the major risk factor for LBW and PTB <ns0:ref type='bibr' target='#b10'>(Ehrenberg et al. 2003;</ns0:ref><ns0:ref type='bibr' target='#b31'>Madzia et al. 2020;</ns0:ref><ns0:ref type='bibr' target='#b43'>Qu et al. 2019)</ns0:ref>. However, the conclusions of various studies on the correlation between pre-pregnancy BMI and PTB appear to be paradoxical. It was reported that the risk of PTB in women with pre-pregnancy high BMI was significantly increased <ns0:ref type='bibr' target='#b1'>(Baeten et al. 2001;</ns0:ref><ns0:ref type='bibr' target='#b6'>Chen et al. 2020;</ns0:ref><ns0:ref type='bibr' target='#b8'>Cnattingius et al. 1998;</ns0:ref><ns0:ref type='bibr' target='#b9'>Cnattingius et al. 2013;</ns0:ref><ns0:ref type='bibr' target='#b55'>Weiss et al. 2004;</ns0:ref><ns0:ref type='bibr'>Zhou et al. 2019)</ns0:ref>, whereas other studies contradicted this result, suggesting that women with a high pre-pregnancy BMI could have a protective impact on PTB <ns0:ref type='bibr' target='#b7'>(Chen et al. 2013;</ns0:ref><ns0:ref type='bibr' target='#b18'>Khashan & Kenny 2009;</ns0:ref><ns0:ref type='bibr' target='#b51'>Sebire et al. 2001)</ns0:ref>. Furthermore, similar results have also been reported on association between pre-pregnancy BMI and LBW <ns0:ref type='bibr' target='#b24'>(Li et al. 2020;</ns0:ref><ns0:ref type='bibr' target='#b46'>Ronnenberg et al. 2003a;</ns0:ref><ns0:ref type='bibr' target='#b56'>Wu et al. 2020)</ns0:ref>.</ns0:p><ns0:p>However, there are relatively few studies of the effects of pre-pregnancy BMI on subsequent pregnancies for first-time mothers among Southern Chinese women. Although several researchers <ns0:ref type='bibr' target='#b26'>(Liu et al. 2019;</ns0:ref><ns0:ref type='bibr' target='#b40'>Pan et al. 2016;</ns0:ref><ns0:ref type='bibr' target='#b47'>Ronnenberg et al. 2003b</ns0:ref>) have conducted relevant research in domestic, these previous papers are mainly limited to earlier data or pregnant women in rural areas. Thus, studies on the role of pre-pregnancy BMI in adverse pregnancy outcomes for first-time mothers in urban areas of Southern China remain scarce.</ns0:p><ns0:p>Given above controversial results, we conducted a hospital-based case-control study to determine whether there is a higher risk of adverse outcome in women with abnormal BMI as compared with normal BMI women among first-time mothers in urban areas of southern China. Manuscript to be reviewed</ns0:p></ns0:div>
<ns0:div><ns0:head>Methods</ns0:head></ns0:div>
<ns0:div><ns0:head>Study subjects</ns0:head><ns0:p>This was a hospital-based, case-control study conducted in The Third Affiliated Hospital of Guangzhou Medical University, Guangdong, China. Pregnant women who gave birth for the first time were included between January 2015 and December 2015. A total of 322 women with PTB and 3,362 women with term delivery controls were enrolled. These people also could be divided into 317 cases of LBW and 3,367 controls of normal birth weight. This study was approved by the institutional review board of The Third Affiliated Hospital of Guangzhou Medical University, <ns0:ref type='bibr'>Guangdong, China (Medical Ethics Hearing [2020]</ns0:ref>No.036). And Institutional Review Board waived the need for consent.</ns0:p><ns0:p>The preterm group was defined to deliver within 37 weeks of conception without congenital abnormalities or neurological damage. The control group between 37 and 42 weeks of gestation without congenital abnormalities or neurological damage was matched with the case in the residential area for one week at the same hospital. For both groups, we excluded women with a multiple pregnancy, stillbirth, prior deliveries, embryo transfer and in vitro fertilization.</ns0:p><ns0:p>All data for this study was collected from women who gave their first birth during hospitalization, including maternal age, education, gravidity, occupation, health insurance, height and weight before of pregnancy. Gestational hypertension, caesarean delivery, birth weight and gestational week at delivery were obtained from the medical records. Pre-pregnancy BMI defined as the body weight in kilograms divided by the square of the height in metres (kg/m 2 ). BMI is classified as underweight (<18.5 kg/m 2 ), normal weight (18.5-23.9 kg/m 2 ), overweight <ns0:ref type='table' target='#tab_9'>PeerJ reviewing PDF | (2020:04:47383:2:0:NEW 10 Sep 2020)</ns0:ref> Manuscript to be reviewed (24.0-27.9 kg/m 2 ), and obesity (≥28.0 kg/m 2 ) according to the weight standard of Chinese adults (Zhou & Cooperative Meta-Analysis Group of the Working Group on Obesity in 2002). In this study, child birth weight below 2500 grams was considered LBW. Once the two occasions systolic blood pressure (BP) or diastolic BP values of pregnant women measured at intervals of 24 hours exceed 140 mmHg or 90 mmHg respectively, they are diagnosed as gestational hypertension.</ns0:p></ns0:div>
<ns0:div><ns0:head>Statistical analysis</ns0:head><ns0:p>t test (for continuous variables) or χ2 test (for categorical variables) was used to assess the difference in demographic characteristics between two groups. Adjusted odds ratios (OR) with 95% confidence intervals obtained from logistic regression model were used to quantify associations between pre-pregnancy BMI and adverse outcomes, adjusting for age, occupation, health insurance and education. We also further divided gestational age into four subtypes: extremely PTB (<28 gestational week), very PTB (28-31 gestational week), moderate PTB (32-36 gestational week) and normal (≧37 gestational week). Besides, we referred to the method provided by Tanya Marchant et al <ns0:ref type='bibr' target='#b32'>(Marchant et al. 2012)</ns0:ref> apply to weight for gestational age, details were described as below: we combined birth weight and gestational age to dichotomise as being small for gestational age (SGA, less than the tenth percentile of weight for gestation) <ns0:ref type='bibr' target='#b38'>(Oken et al. 2003</ns0:ref>) and non-small for gestational age (NSGA, large than the tenth percentile of weight for gestation). As in previous studies, gestational week was classified into categories of term, 34-36 week and below 34 week. We then divided newborns into six groups: (1) term and Manuscript to be reviewed</ns0:p></ns0:div>
<ns0:div><ns0:head>Results</ns0:head></ns0:div>
<ns0:div><ns0:head>Characteristics of the study subjects</ns0:head><ns0:p>The baseline maternal characteristics are shown in table 1. Of the 3,684 live births, 8.7% (n=322) were PTBs, and the remaining 91.3% (n=3362) were term births. The proportions of underweight, normal weight, overweight, and obesity women were 23.29%, 63.36%, 10.72%, and 2.63%, respectively. There were significant differences between PTB group and terms group with respect to age, health insurance, occupation and education (all P<0.01). There were no significant differences between the two groups with regard to gravidity (P>0.05). The variables in the normal and the LBW group were basically the same as those in the term and preterm groups.</ns0:p></ns0:div>
<ns0:div><ns0:head>Association analysis between pre-pregnancy BMI and adverse outcomes</ns0:head><ns0:p>The association between pre-pregnancy BMI and the risk of adverse outcomes are considered in table 2 and table 3. Underweight, overweight and obesity did not increase the risk of PTB as compare with normal weight (AOR=1.01, 95% CI=0.76 to 1.34; AOR=1.25, 95% CI=0.87 to 1.80; and AOR=1.27, 95% CI=0.65 to 2.51, respectively). However, When PTB was stratified into three subgroups, both maternal overweight and obesity increased the risks of extremely PTB (AOR=8.12, 95% CI=1.11 to 59.44; AOR=15.06, 95% CI=1.32 to 172.13, respectively).</ns0:p><ns0:p>Pre-pregnancy underweight was significantly associated with the increased risk for LBW. In comparison with women who had normal pre-pregnancy BMI, women with low BMI category was more likely to deliver a LBW infant (crude OR=1.48, 95% CI=1.14 to 1.93). After the Manuscript to be reviewed adjustment for potential confounding factors, the AOR associated with the risk for giving birth to a LBW infant were 1.44 (95% CI=1.11 to 1.89), the similar result was seen in table 4 as compared with term and NSAG (AOR=1.78, 95% CI=1.45 to 2.17). Underweight was also significantly associated with a lower risk of gestational hypertension (AOR=0.45, 95% CI=0.25 to 0.82) and caesarean delivery (AOR=0.74, 95% CI=0.62 to 0.90). Both maternal overweight and obesity were found to be a risk factor for gestational hypertension (AOR=1.71, 95% CI=1.06 to 2.77; AOR=5.54, 95% CI=3.02 to 10.17, respectively) and caesarean delivery (AOR=1.91, 95% CI=1.53 to 2.38; AOR=1.85, 95% CI=1.21 to 2.82, respectively).</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:04:47383:2:0:NEW 10 Sep 2020)</ns0:p><ns0:p>Manuscript to be reviewed</ns0:p></ns0:div>
<ns0:div><ns0:head>Discussion</ns0:head><ns0:p>Our study demonstrated that maternal underweight prior to pregnancy, as compared with normal weight women, significantly elevated the risk for LBW for the first-time mothers among Southern Chinese. Maternal underweight were also found to be at lower risk of gestational hypertension and caesarean delivery. In women who had a high pre-pregnancy BMI, our study showed a significantly higher risk of gestational hypertension, caesarean delivery and extremely PTB.</ns0:p><ns0:p>The proportions of overweight and obesity were lower than underweight in our study, which were similar to previously reported data from other Chinese studies <ns0:ref type='bibr' target='#b40'>(Pan et al. 2016)</ns0:ref>. Our finding showed that underweight women increase the risk of LBW in subsequent pregnancy, which is consistent with the result of a review study by <ns0:ref type='bibr' target='#b27'>Liu et al. (Liu et al. 2016</ns0:ref>). The study, including 60 studies covering 1,392,799 women, showed that infants had a higher risk of having a LBW when their mothers were underweight (OR, 1.67, 95%CI, 1.39-2.02) as compare to women with normal weight. Previous studies have demonstrated that maternal nutrition during pregnancy has great influence on providing the essential nutrients for fetal growth <ns0:ref type='bibr' target='#b37'>(Nnam 2015)</ns0:ref> and pregnant women who are undernourished tend to have LBW infant <ns0:ref type='bibr' target='#b0'>(Allen 2001)</ns0:ref>. Maternal nutritional imbalance may be a key factor in the reduction of surface area and placental weight. In the lower surface area and placental weight, nutrient and waste transfer between the maternal and fetal circulation is reduced and other normal processes, such as fetal development and growth, are also restricted <ns0:ref type='bibr' target='#b21'>(Lechtig et al. 1975)</ns0:ref>. Thus, maternal malnutrition may lead to LBW infant. Moreover, the finding in our study on association between underweight and LBW was also in line with Manuscript to be reviewed previous literature <ns0:ref type='bibr' target='#b23'>(Li et al. 2013;</ns0:ref><ns0:ref type='bibr' target='#b28'>Liu et al. 2012)</ns0:ref>.</ns0:p><ns0:p>Previous literature has revealed that maternal overweight and obesity were associated with the increased risk for gestational hypertension and caesarian delivery <ns0:ref type='bibr' target='#b22'>(Lewandowska et al. 2020;</ns0:ref><ns0:ref type='bibr'>Machado et al. 2020;</ns0:ref><ns0:ref type='bibr' target='#b54'>Vince et al. 2020)</ns0:ref>. Our study also confirmed these findings. However, the risk of gestational hypertension and caesarean delivery were reduced among underweight mothers. Although mechanism by which obesity responsible for the increased risk of gestational hypertension or caesarean delivery is unclear, maternal obesity lead to an increase in the number and size of adipocytes or pelvic malacia. A significant amount of adipocytes has been proposed as a cause of excessive inflammatory reaction, pregnant female possibly experienced obstructive dystocia due to pelvic malacia lead to a relatively narrow birth canal. Therefore, both gestational hypertension and caesarean delivery were affected by obesity in obese mothers <ns0:ref type='bibr' target='#b19'>(Kriketos et al. 2004;</ns0:ref><ns0:ref type='bibr' target='#b58'>Young & Woodmansee 2002)</ns0:ref>.</ns0:p><ns0:p>In our study, the risk of PTB is relatively higher among overweight and obese mothers but the association was not statistically significant (table <ns0:ref type='table' target='#tab_1'>2</ns0:ref>). However, this association has dramatically changed in a subgroup analysis of gestational age (table <ns0:ref type='table' target='#tab_4'>3</ns0:ref>). Our result regarding extremely PTB is consistent with a previous study <ns0:ref type='bibr' target='#b53'>(Su et al. 2020)</ns0:ref>. Overweight and obesity are generally considered to be the risk factors for PTB due to the effects of placental insufficiency <ns0:ref type='bibr' target='#b20'>(Lassance et al. 2015;</ns0:ref><ns0:ref type='bibr' target='#b41'>Pereira et al. 2015)</ns0:ref>, inflammatory state <ns0:ref type='bibr' target='#b11'>(Gaillard et al. 2016)</ns0:ref>, insulin sensitivity <ns0:ref type='bibr' target='#b5'>(Catalano et al. 1999</ns0:ref>) and cellular oxidative stress <ns0:ref type='bibr' target='#b2'>(Ballesteros-Guzman et al. 2019</ns0:ref>).</ns0:p><ns0:p>However, conclusions on the relationship between pre-pregnancy BMI and PTB seem to be paradoxical among different studies. The differences emerged between studies could be PeerJ reviewing PDF | (2020:04:47383:2:0:NEW 10 Sep 2020)</ns0:p><ns0:p>Manuscript to be reviewed attributed to study design or power, recall bias, multiple comparisons, eating habits and different ethnicities. Additionally, the category of BMI was different among the studies.</ns0:p><ns0:p>Although some confounding factors had been controlled, alcohol consumption and maternal smoking were not adjusted as only one woman claimed to have a history of alcohol consumption and smoking in our data. So we did not adjust for these two variables.</ns0:p><ns0:p>However, there are a number of potential limitations of this study that merit consideration.</ns0:p><ns0:p>One limitation of the current study is that it is difficult to distinguish between spontaneous and iatrogenic PTB, which we could not assess the association between special types of PTB and prepregnancy BMI. Additionally, the BMI we obtained in our study derived from weight and height information by women self-reported, which could lead to bias risk estimates of PTB <ns0:ref type='bibr' target='#b34'>(Michels et al. 1998)</ns0:ref>.</ns0:p><ns0:p>In conclusion, our study suggested that maternal overweight and obesity were associated with a significantly higher risk of gestational hypertension, caesarean delivery and extremely PTB. Underweight was correlated with an increased risk of LBW and conferred a protective effect regarding the risk for gestational hypertension and caesarean delivery for the first-time mothers among Southern Chinese.</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:04:47383:2:0:NEW 10 Sep 2020)</ns0:p><ns0:p>Manuscript to be reviewed Manuscript to be reviewed Crude and AOR for the association between pre-pregnancy BMI and adverse outcomes Manuscript to be reviewed Manuscript to be reviewed Adjusted* a relative risk for the associations between pre-pregnancy BMI and weight for gestational age Manuscript to be reviewed </ns0:p></ns0:div><ns0:figure xml:id='fig_0'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:04:47383:2:0:NEW 10 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_1'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:04:47383:2:0:NEW 10 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_2'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:04:47383:2:0:NEW 10 Sep 2020) Manuscript to be reviewed NSGA; (2) 34-36 week gestation and NSGA; (3) below 34 week gestation and NSGA; (4) term and SAG; (5) 34-36 week gestation and SAG; (6) below 34 week gestation and SAG. A twosided p-value of 0.05 or less was accepted to be statistically significant. Data were analyzed using Statistical Analysis Software 9.4. (SAS Institute, Cary, NC). PeerJ reviewing PDF | (2020:04:47383:2:0:NEW 10 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_3'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:04:47383:2:0:NEW 10 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_4'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:04:47383:2:0:NEW 10 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure type='table' xml:id='tab_0'><ns0:head>Table 1 :</ns0:head><ns0:label>1</ns0:label><ns0:figDesc>2 Baseline maternal characteristics of the first-time mothers between two groups 3 Abbreviation: PTB, preterm birth; LBW, low birth weight.</ns0:figDesc><ns0:table><ns0:row><ns0:cell>variable</ns0:cell><ns0:cell>Terms (N=3362)</ns0:cell><ns0:cell>PTB (N=322)</ns0:cell><ns0:cell>P</ns0:cell><ns0:cell>Normal (N=3367)</ns0:cell><ns0:cell>LBW (N=317)</ns0:cell><ns0:cell>P</ns0:cell></ns0:row><ns0:row><ns0:cell>Age</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell><0.001</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell><0.001</ns0:cell></ns0:row><ns0:row><ns0:cell>< 25</ns0:cell><ns0:cell>421 (12.52)</ns0:cell><ns0:cell>55 (17.08)</ns0:cell><ns0:cell /><ns0:cell>418 (12.41)</ns0:cell><ns0:cell>58 (18.30)</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>25-34</ns0:cell><ns0:cell>2628 (78.17)</ns0:cell><ns0:cell>220 (68.32)</ns0:cell><ns0:cell /><ns0:cell>2630 (78.11)</ns0:cell><ns0:cell>218 (68.77)</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>≧35</ns0:cell><ns0:cell>313 (9.31)</ns0:cell><ns0:cell>47 (14.60)</ns0:cell><ns0:cell /><ns0:cell>319 (9.48)</ns0:cell><ns0:cell>41 (12.93)</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>Health insurance</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell><0.001</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell><0.001</ns0:cell></ns0:row><ns0:row><ns0:cell>Care for urban employees</ns0:cell><ns0:cell>2269 (67.49)</ns0:cell><ns0:cell>161 (50.00)</ns0:cell><ns0:cell /><ns0:cell>2261 (67.15)</ns0:cell><ns0:cell>169 (53.31)</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>Free medical service</ns0:cell><ns0:cell>87 (2.59)</ns0:cell><ns0:cell>6 (1.86)</ns0:cell><ns0:cell /><ns0:cell>90 (2.67)</ns0:cell><ns0:cell>3 (0.95)</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>Full-cost</ns0:cell><ns0:cell>1006 (29.92)</ns0:cell><ns0:cell>155 (47.14)</ns0:cell><ns0:cell /><ns0:cell>1016 (30.18)</ns0:cell><ns0:cell>145 (45.74)</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>Occupation</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell>0.004</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell>0.007</ns0:cell></ns0:row><ns0:row><ns0:cell>Professional</ns0:cell><ns0:cell>1534 (45.63)</ns0:cell><ns0:cell>119 (36.96)</ns0:cell><ns0:cell /><ns0:cell>1532 (45.50)</ns0:cell><ns0:cell>121 (38.17)</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>Business</ns0:cell><ns0:cell>355 (10.56)</ns0:cell><ns0:cell>28 (8.70)</ns0:cell><ns0:cell /><ns0:cell>358 (10.63)</ns0:cell><ns0:cell>25 (7.89)</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>Housewife</ns0:cell><ns0:cell>389 (11.57)</ns0:cell><ns0:cell>46 (14.29)</ns0:cell><ns0:cell /><ns0:cell>388 (11.52)</ns0:cell><ns0:cell>47 (14.83)</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>Others</ns0:cell><ns0:cell>1084 (32.24)</ns0:cell><ns0:cell>129 (40.06)</ns0:cell><ns0:cell /><ns0:cell>1089 (32.35)</ns0:cell><ns0:cell>124 (39.12)</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>Gravidity</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell>0.214</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell>0.875</ns0:cell></ns0:row><ns0:row><ns0:cell>< 2</ns0:cell><ns0:cell>2245 (66.78)</ns0:cell><ns0:cell>204 (63.35)</ns0:cell><ns0:cell /><ns0:cell>2237 (66.44)</ns0:cell><ns0:cell>212 (66.88)</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>≧2</ns0:cell><ns0:cell>1117 (33.22)</ns0:cell><ns0:cell>118 (36.65)</ns0:cell><ns0:cell /><ns0:cell>1130 (33.56)</ns0:cell><ns0:cell>105 (33.12)</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>Education</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell><0.001</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell><0.001</ns0:cell></ns0:row><ns0:row><ns0:cell>Less than high school</ns0:cell><ns0:cell>439 (13.06)</ns0:cell><ns0:cell>86 (26.71)</ns0:cell><ns0:cell /><ns0:cell>444 (13.19)</ns0:cell><ns0:cell>81 (25.55)</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>High school</ns0:cell><ns0:cell>444 (13.21)</ns0:cell><ns0:cell>38 (11.80)</ns0:cell><ns0:cell /><ns0:cell>443 (13.16)</ns0:cell><ns0:cell>39 (12.30)</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>College</ns0:cell><ns0:cell>2479 (73.73)</ns0:cell><ns0:cell>198 (61.49)</ns0:cell><ns0:cell /><ns0:cell>2480 (73.65)</ns0:cell><ns0:cell>197 (62.15)</ns0:cell><ns0:cell /></ns0:row></ns0:table><ns0:note>1 PeerJ reviewing PDF | (2020:04:47383:2:0:NEW 10 Sep 2020)</ns0:note></ns0:figure>
<ns0:figure type='table' xml:id='tab_1'><ns0:head>Table 2 (on next page)</ns0:head><ns0:label>2</ns0:label><ns0:figDesc /><ns0:table /></ns0:figure>
<ns0:figure type='table' xml:id='tab_2'><ns0:head>Table 2</ns0:head><ns0:label>2</ns0:label><ns0:figDesc>PeerJ reviewing PDF | (2020:04:47383:2:0:NEW 10 Sep 2020)</ns0:figDesc><ns0:table><ns0:row><ns0:cell>Manuscript to be reviewed</ns0:cell></ns0:row></ns0:table><ns0:note>1</ns0:note></ns0:figure>
<ns0:figure type='table' xml:id='tab_3'><ns0:head>Table 2 :</ns0:head><ns0:label>2</ns0:label><ns0:figDesc>2 Crude and AOR for the association between pre-pregnancy BMI and adverse outcomes</ns0:figDesc><ns0:table><ns0:row><ns0:cell>Outcomes</ns0:cell><ns0:cell>BMI status</ns0:cell><ns0:cell>Case</ns0:cell><ns0:cell>Control</ns0:cell><ns0:cell>OR (95% CI)</ns0:cell><ns0:cell>P</ns0:cell><ns0:cell>AOR* (95% CI)</ns0:cell><ns0:cell>P</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell>783</ns0:cell><ns0:cell>1.04 (0.79-</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>PTB</ns0:cell><ns0:cell>Underweight</ns0:cell><ns0:cell>75 (23.29)</ns0:cell><ns0:cell>(23.29)</ns0:cell><ns0:cell>1.37)</ns0:cell><ns0:cell>0.787</ns0:cell><ns0:cell>1.01 (0.76-1.34)</ns0:cell><ns0:cell>0.943</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell>197</ns0:cell><ns0:cell>2137</ns0:cell><ns0:cell>1.00</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell /><ns0:cell>Normal weight</ns0:cell><ns0:cell>(61.18)</ns0:cell><ns0:cell>(63.56)</ns0:cell><ns0:cell>(reference)</ns0:cell><ns0:cell /><ns0:cell>1.00 (reference)</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell>1.22 (0.86-</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell /><ns0:cell>Overweight</ns0:cell><ns0:cell>40 (12.42)</ns0:cell><ns0:cell>355 (10.56)</ns0:cell><ns0:cell>1.75)</ns0:cell><ns0:cell>0.272</ns0:cell><ns0:cell>1.25 (0.87-1.80)</ns0:cell><ns0:cell>0.224</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell>1.25 (0.64-</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell /><ns0:cell>Obesity</ns0:cell><ns0:cell>10 (3.11)</ns0:cell><ns0:cell>87 (2.59)</ns0:cell><ns0:cell>2.44)</ns0:cell><ns0:cell>0.519</ns0:cell><ns0:cell>1.27 (0.65-2.51)</ns0:cell><ns0:cell>0.486</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell>1.48 (1.14-</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>LBW</ns0:cell><ns0:cell>Underweight</ns0:cell><ns0:cell>94 (29.65)</ns0:cell><ns0:cell>764 (22.69)</ns0:cell><ns0:cell>1.93)</ns0:cell><ns0:cell>0.003</ns0:cell><ns0:cell>1.44 (1.11-1.89)</ns0:cell><ns0:cell>0.007</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell>179</ns0:cell><ns0:cell>2155</ns0:cell><ns0:cell>1.00</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell /><ns0:cell>Normal weight</ns0:cell><ns0:cell>(56.47)</ns0:cell><ns0:cell>(64.00)</ns0:cell><ns0:cell>(reference)</ns0:cell><ns0:cell /><ns0:cell>1.00 (reference)</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell>1.13 (0.77-</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell /><ns0:cell>Overweight</ns0:cell><ns0:cell>34 (10.73)</ns0:cell><ns0:cell>361 (10.73)</ns0:cell><ns0:cell>1.66)</ns0:cell><ns0:cell>0.521</ns0:cell><ns0:cell>1.17 (0.80-1.73)</ns0:cell><ns0:cell>0.423</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell>1.38 (0.71-</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell /><ns0:cell>Obesity</ns0:cell><ns0:cell>10 (3.15)</ns0:cell><ns0:cell>87 (2.58)</ns0:cell><ns0:cell>2.71)</ns0:cell><ns0:cell>0.343</ns0:cell><ns0:cell>1.41 (0.72-2.78)</ns0:cell><ns0:cell>0.322</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell>0.43 (0.24-</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell cols='2'>Gestational hypertension Underweight</ns0:cell><ns0:cell>13 (9.92)</ns0:cell><ns0:cell>845 (23.78)</ns0:cell><ns0:cell>0.78)</ns0:cell><ns0:cell>0.006</ns0:cell><ns0:cell>0.45 (0.25-0.82)</ns0:cell><ns0:cell>0.009</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell>2254</ns0:cell><ns0:cell>1.00</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell /><ns0:cell>Normal weight</ns0:cell><ns0:cell>80 (61.07)</ns0:cell><ns0:cell>(63.44)</ns0:cell><ns0:cell>(reference)</ns0:cell><ns0:cell /><ns0:cell>1.00 (reference)</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell>1.74 (1.08-</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell /><ns0:cell>Overweight</ns0:cell><ns0:cell>23 (17.56)</ns0:cell><ns0:cell>372 (10.47)</ns0:cell><ns0:cell>2.81)</ns0:cell><ns0:cell>0.022</ns0:cell><ns0:cell>1.71 (1.06-2.77)</ns0:cell><ns0:cell>0.03</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell>5.15 (2.85-</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell /><ns0:cell>Obesity</ns0:cell><ns0:cell>15 (11.45)</ns0:cell><ns0:cell>82 (2.31)</ns0:cell><ns0:cell>9.33)</ns0:cell><ns0:cell><0.001</ns0:cell><ns0:cell>5.54 (3.02-10.17)</ns0:cell><ns0:cell><0.001</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell>194</ns0:cell><ns0:cell /><ns0:cell>0.70 (0.59-</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>Cesarean delivery</ns0:cell><ns0:cell>Underweight</ns0:cell><ns0:cell>(17.60)</ns0:cell><ns0:cell>664 (25.72)</ns0:cell><ns0:cell>0.84)</ns0:cell><ns0:cell><0.001</ns0:cell><ns0:cell>0.74 (0.62-0.90)</ns0:cell><ns0:cell>0.002</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell>686</ns0:cell><ns0:cell>1684</ns0:cell><ns0:cell>1.00</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell /><ns0:cell>Normal weight</ns0:cell><ns0:cell>(62.25)</ns0:cell><ns0:cell>(63.83)</ns0:cell><ns0:cell>(reference)</ns0:cell><ns0:cell /><ns0:cell>1.00 (reference)</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell>180</ns0:cell><ns0:cell /><ns0:cell>2.01 (1.62-</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell /><ns0:cell>Overweight</ns0:cell><ns0:cell>(16.33)</ns0:cell><ns0:cell>215 (8.33)</ns0:cell><ns0:cell>2.50)</ns0:cell><ns0:cell><0.001</ns0:cell><ns0:cell>1.91 (1.53-2.38)</ns0:cell><ns0:cell><0.001</ns0:cell></ns0:row></ns0:table><ns0:note>PeerJ reviewing PDF | (2020:04:47383:2:0:NEW 10 Sep 2020)Manuscript to be reviewed</ns0:note></ns0:figure>
<ns0:figure type='table' xml:id='tab_4'><ns0:head>Table 3 (on next page) Adjusted</ns0:head><ns0:label>3</ns0:label><ns0:figDesc /><ns0:table /><ns0:note>* a relative risk for the associations between pre-pregnancy BMI and PTB by gestational age</ns0:note></ns0:figure>
<ns0:figure type='table' xml:id='tab_5'><ns0:head>Table 3</ns0:head><ns0:label>3</ns0:label><ns0:figDesc>PeerJ</ns0:figDesc><ns0:table /><ns0:note>reviewing PDF | (2020:04:47383:2:0:NEW 10 Sep 2020)</ns0:note></ns0:figure>
<ns0:figure type='table' xml:id='tab_6'><ns0:head>Table 3 :</ns0:head><ns0:label>3</ns0:label><ns0:figDesc>Adjusted * a relative risk for the associations between pre-pregnancy BMI and PTB by gestational age Adjusted OR and 95% CI were calculated by the logistic regression model after adjusting for age, health insurance, occupation and education.</ns0:figDesc><ns0:table><ns0:row><ns0:cell>Gestational age</ns0:cell><ns0:cell>Underweight</ns0:cell><ns0:cell>Normal weight</ns0:cell><ns0:cell>Overweight</ns0:cell><ns0:cell>Obesity</ns0:cell></ns0:row><ns0:row><ns0:cell>Term</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>1.00 (reference)</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell></ns0:row><ns0:row><ns0:cell>Moderately PTB</ns0:cell><ns0:cell>0.97 (0.71-1.33)</ns0:cell><ns0:cell>1.00 (reference)</ns0:cell><ns0:cell>1.18 (0.79-1.77)</ns0:cell><ns0:cell>1.23 (0.58-2.59)</ns0:cell></ns0:row><ns0:row><ns0:cell>Very PTB</ns0:cell><ns0:cell>1.05 (0.54-2.02)</ns0:cell><ns0:cell>1.00 (reference)</ns0:cell><ns0:cell>1.28 (0.55-2.96)</ns0:cell><ns0:cell>0.77 (0.10-5.77)</ns0:cell></ns0:row><ns0:row><ns0:cell>Extremely PTB</ns0:cell><ns0:cell>3.22 (0.53-19.59)</ns0:cell><ns0:cell>1.00 (reference)</ns0:cell><ns0:cell>8.12 (1.11-59.44)</ns0:cell><ns0:cell>15.06 (1.32-172.13)</ns0:cell></ns0:row></ns0:table><ns0:note>Abbreviations: PTB, preterm birth. * PeerJ reviewing PDF | (2020:04:47383:2:0:NEW 10 Sep 2020)</ns0:note></ns0:figure>
<ns0:figure type='table' xml:id='tab_7'><ns0:head>Table 4 (on next page)</ns0:head><ns0:label>4</ns0:label><ns0:figDesc /><ns0:table /></ns0:figure>
<ns0:figure type='table' xml:id='tab_8'><ns0:head>Table 4</ns0:head><ns0:label>4</ns0:label><ns0:figDesc>PeerJ</ns0:figDesc><ns0:table /><ns0:note>reviewing PDF | (2020:04:47383:2:0:NEW 10 Sep 2020)</ns0:note></ns0:figure>
<ns0:figure type='table' xml:id='tab_9'><ns0:head>Table 4 :</ns0:head><ns0:label>4</ns0:label><ns0:figDesc>Adjusted* a relative risk for the associations between pre-pregnancy BMI and weight for gestational age Adjusted OR and 95% CI were calculated by the logistic regression model after adjusting for age, health insurance, occupation and education.</ns0:figDesc><ns0:table><ns0:row><ns0:cell>Weight for gestational age</ns0:cell><ns0:cell>Underweight</ns0:cell><ns0:cell>Normal weight</ns0:cell><ns0:cell>Overweight</ns0:cell><ns0:cell>Obesity</ns0:cell></ns0:row><ns0:row><ns0:cell>NSAG</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>Term</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>1.00 (reference)</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell></ns0:row><ns0:row><ns0:cell>34-36 week</ns0:cell><ns0:cell>0.88 (0.58-1.34)</ns0:cell><ns0:cell>1.00 (reference)</ns0:cell><ns0:cell>1.29 (0.80-2.06)</ns0:cell><ns0:cell>1.40 (0.59-3.29)</ns0:cell></ns0:row><ns0:row><ns0:cell><34 week</ns0:cell><ns0:cell>1.52 (0.94-2.47)</ns0:cell><ns0:cell>1.00 (reference)</ns0:cell><ns0:cell>1.48 (0.77-2.84)</ns0:cell><ns0:cell>0.97 (0.23-4.11)</ns0:cell></ns0:row><ns0:row><ns0:cell>SAG</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>Term</ns0:cell><ns0:cell>1.78 (1.45-2.17)</ns0:cell><ns0:cell>1.00 (reference)</ns0:cell><ns0:cell>0.88 (0.63-1.21)</ns0:cell><ns0:cell>1.89 (0.48-1.66)</ns0:cell></ns0:row><ns0:row><ns0:cell>34-36 week</ns0:cell><ns0:cell>1.49 (0.79-2.81)</ns0:cell><ns0:cell>1.00 (reference)</ns0:cell><ns0:cell>0.43 (0.10-1.82)</ns0:cell><ns0:cell>0.86 (0.12-6.41)</ns0:cell></ns0:row><ns0:row><ns0:cell><34 week</ns0:cell><ns0:cell>0.73 (0.20-2.63)</ns0:cell><ns0:cell>1.00 (reference)</ns0:cell><ns0:cell>1.62 (0.44-5.89)</ns0:cell><ns0:cell>2.23 (0.28-17.91)</ns0:cell></ns0:row></ns0:table><ns0:note>Abbreviations: NSAG, non-small for gestational age; SAG, small for gestational age.* PeerJ reviewing PDF | (2020:04:47383:2:0:NEW 10 Sep 2020)</ns0:note></ns0:figure>
<ns0:note place='foot' n='3'>Abbreviations: OR, odds ratio; CI, confidence interval; AOR, adjusted odds ratio; PTB, preterm birth; LBW, low birth weight. 4 * Adjusted OR and 95% CI were calculated by the logistic regression model after adjusting for age, health insurance, occupation and education.PeerJ reviewing PDF | (2020:04:47383:2:0:NEW 10 Sep 2020)</ns0:note>
</ns0:body>
" | "Guangzhou, 10th, September 2020
Dear Professor Qing-Yuan Sun,
Thank you very much for your letter and your efforts in review of our manuscript. We would like to resubmit the manuscript entitled “Association of Pre-Pregnancy Body Mass Index with adverse pregnancy outcome among first-time mothers” (ID: 47383).
We appreciate the reviewer for the valuable comments. We have addressed, point-by-point, the issues raised by the reviewer. The amendments made in the original version of the manuscript are tracked in the revised version. We hope that these changes will make the manuscript acceptable for publication in PeerJ. Thank you very much for your consideration.
Yours sincerely
Li Ke
Key Laboratory for Major Obstetric Diseases of Guangdong Province, the Third Affiliated Hospital of Guangzhou Medical University, Guangzhou, China
Phone: 86-20-81292138
FAX: 86-20-81292908
E-mail: keli1221@126.com
Reviewer 1 (Anonymous)
Basic reporting
The revision improved the paper as written. There are several issues although the authors addressed most of concerns and suggestions in the revision.
1. Word document of the revised manuscript is not the same version of the PDF
Response: Thank you for your careful review of the manuscript. We have tried to go over the manuscript carefully and find the differences between version of the word document and version of the PDF (for example line 81-82, line 112, line 159-160, line 174, line 184). We apologize for our mistake, and have now corrected.
2. One sentence is duplicated with the next one (line 221-222)
Response: Thank you for your careful review of the manuscript. Duplicated sentence have been removed.
3. Grammatical issues will need to be addressed. For example, line85-90; line131, line 132, line 162, line 178-179, line 212.
Response: Thanks for your positive comments and constructive suggestions. The corrections are listed below in order:
(1) We have revised line 85-90 as “However, there are relatively few studies of the effects of pre-pregnancy BMI on subsequent pregnancies for first-time mothers among Southern Chinese women. Although several researchers (Liu et al. 2019; Pan et al. 2016; Ronnenberg et al. 2003b) have conducted relevant research in domestic, these previous papers are mainly limited to earlier data or pregnant women in rural areas. Thus, studies on the role of pre-pregnancy BMI in adverse pregnancy outcomes for first-time mothers in urban areas of Southern China remain scarce.”
(2) We have revised line 131-132 as “Besides, we referred to the method provided by Tanya Marchant et al (Marchant et al. 2012) apply to weight for gestational age, details were described as below: we combined birth weight and gestational age to dichotomise as being small for gestational age (SGA, less than the tenth percentile of weight for gestation) (Oken et al. 2003) and non-small for gestational age (NSGA, large than the tenth percentile of weight for gestation). ”
(3) We have revised line 162 as “the similar result was seen in table 4 as compared with term and NSAG (AOR=1.78, 95% CI=1.45 to 2.17).”
(4) We have revised line 178-179 as “The proportions of overweight and obesity were lower than underweight in our study, which were similar to previously reported data from other Chinese studies (Pan et al. 2016).”
(5) We have revised line 212 as “The differences emerged between studies could be attributed to study design or power, recall bias, multiple comparisons, eating habits and different ethnicities.”
Experimental design
What is the logical explanation for dividing newborns into six groups, not 8 groups? The authors divided gestational ages into four subtypes: extremely PTB (<28 gestational week), very PTB (28–31 gestational week), moderate PTB (32–36 gestational week) and normal (≧37 gestational week), and categorized newborn weight for gestational age into two groups: SGA and NSGA.
Response: Thank you for your comment. We divided gestational age into four subtypes: extremely PTB (<28 gestational week), very PTB (28–31 gestational week), moderate PTB (32–36 gestational week) and normal (≧37 gestational week). Besides, we referred to the method provided by Tanya Marchant et al (Marchant et al. 2012) apply to weight for gestational age, details were described as below: we combined birth weight and gestational age to dichotomise as being small for gestational age (SGA, less than the tenth percentile of weight for gestation) (Oken et al. 2003) and non-small for gestational age (NSGA, large than the tenth percentile of weight for gestation). As in previous studies, gestational week was classified into categories of term, 34-36 week and below 34 week. We then divided newborns into six groups: (1) term and NSGA; (2) 34–36 week gestation and NSGA; (3) below 34 week gestation and NSGA; (4) term and SAG; (5) 34–36 week gestation and SAG; (6) below 34 week gestation and SAG. To eliminate the misunderstanding, we have added the text in abstract and method section.
We highly appreciate your careful review and critical comments, which are very helpful for improving our manuscript. And we have carefully revised the manuscript as requested. Once again, thank you very much for your comments and suggestions.
" | Here is a paper. Please give your review comments after reading it. |
9,763 | "<ns0:abstract xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:p>Background. Excessive fluoride can lead to chronic neurodegeneration characterized by neuron and myelin loss and memory dysfunction. The gut-brain axis hypothesis suggests that gut microbiota plays a crucial role in regulating brain function. Thus, using probiotics to adjust the gut microenvironment may be a potential therapy for mental diseases.</ns0:p><ns0:p>Methods. Mice in the prob group were administrated with Lactobacillus johnsonii BS15 for 28 days prior to and throughout a 70-day exposure to sodium fluoride. The drinking water of all groups (F and prob groups) except the control group were replaced by high-fluoride water (100 mg NaF/L) on day 28. Animals in each group were divided into two subsets: one underwent behavioral test, and the other was sacrificed for sampling. The mRNA expression level and protein content related to inflammatory reaction in the ileum and hippocampus were respectively detected by reverse transcription quantitative polymerase chain reaction (RT-qPCR) and enzyme-linked immunosorbent assay (ELISA). The mRNA expression levels of proteins related to myelin structure, apoptosis, and memory in the hippocampus and tight junction proteins in the ileum were determined by RT-qPCR and/or immunohistochemistry. Gut permeability markers (D-lactate and diamine oxidase [DAO]) in the serum were also examined by ELISA.</ns0:p><ns0:p>Results. BS15 treatment exerted significant preventive effects on reversing the gut inflammation induced by excessive fluoride intake by reducing (P<0.05) the levels of pro-inflammatory cytokines (tumor necrosis factor-alpha [TNF-α] and interferon-gamma [IFN-γ]) and remarkably increasing (P<0.05) the level of anti-inflammatory cytokines (IL-10). Moreover, the serum DAO activity and D-lactate concentration significantly increased by fluoride were also reduced (P<0.05) by BS15. This result indicated the profitable effect of BS15 on gut permeability. Memory ability was significantly decreased (P<0.05) by fluoride as observed by T-maze test. BS15 reversed memory impairment by increasing (P<0.05) the downregulated expression levels of myelin structural protein (proteolipid protein) and neurogenesis-associated proteins (brain-derived neurotrophic factor and cAMP/Ca 2+ responsive elementbinding protein), balancing the hippocampal inflammatory cytokines (TNF-α, IFN-γ, and IL-6; P<0.05), reducing pro-apoptotic genes (caspase-3; P<0.05), and increasing anti-apoptotic genes (Bcl-2; P<0.05) in the hippocampus of fluoride-infected mice.</ns0:p></ns0:div>
</ns0:abstract>
<ns0:body xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:head>Introduction</ns0:head><ns0:p>Excessive fluoride intake has attracted increasing attention because of its widely introduced sources and adverse effects on human health. Fluoride concentration in the groundwater could range from under 1 mg/L to more than 35 mg/L <ns0:ref type='bibr' target='#b50'>(Petrone et al., 2013)</ns0:ref>. In addition, fluoride concentration in drinking tea (especially Chinese brick tea) ranges from 600-2800 mg/kg <ns0:ref type='bibr' target='#b23'>(Fung et al.,1999)</ns0:ref>. The risk of high fluoride intake from food is also increasing because of the increased fluorine-containing crop protection compounds <ns0:ref type='bibr' target='#b45'>(Maienfisch & Hall, 2004)</ns0:ref>. The most well-known fluoride-induced negative influences are on teeth (dental fluorosis) <ns0:ref type='bibr' target='#b55'>(Sabokseir , Golkari & Sheiham, 2016)</ns0:ref> and skeleton (skeletal fluorosis) <ns0:ref type='bibr' target='#b20'>(Littleton et al., 1999;</ns0:ref><ns0:ref type='bibr' target='#b64'>Wang et al., 2019)</ns0:ref>. Workers exposed to high fluoride suffer from drowsiness and impaired learning and memory <ns0:ref type='bibr' target='#b11'>(Czerwiński & Lankosz, 1978)</ns0:ref>. Furthermore, epidemiological investigation from China <ns0:ref type='bibr' target='#b63'>(Wang et al., 2007)</ns0:ref>, India <ns0:ref type='bibr' target='#b57'>(Sebastian & Sunitha, 2015)</ns0:ref>, Mexico <ns0:ref type='bibr' target='#b2'>(Bashash et al., 2017)</ns0:ref>, and Iran <ns0:ref type='bibr' target='#b52'>(Razdan et al., 2017)</ns0:ref> reported that children residing in endemic areas show deficits in learning and memory abilities. Rodents exposed to excessive fluoride also showed poor performances in memory-related behavioral tests in animal experiments <ns0:ref type='bibr' target='#b6'>(Chen et al., 2018;</ns0:ref><ns0:ref type='bibr' target='#b39'>Liu et al., 2010)</ns0:ref>. Animal experiments concerning fluoride neurotoxicity demonstrated that high fluoride exposure causes the pathological alteration of the synaptic ultrastructure of the hippocampus <ns0:ref type='bibr' target='#b68'>(Qian et al., 2013)</ns0:ref>. <ns0:ref type='bibr' target='#b49'>Niu et al. (2018)</ns0:ref> suggested that fluoride could reduce neurotrophy and neuron adhesion and consequently damage the myelin in the hippocampus of mice.</ns0:p><ns0:p>Few safe and effective methods can protect the brain from fluoride neurotoxicity. The gut-brain axis, a bi-directional communication system linking the gut and brain, has provided a new insight into the treatment of brain-derived diseases <ns0:ref type='bibr' target='#b21'>(Forsythe &Kunze, 2013)</ns0:ref>. Recent advances in metagenomics confirmed that the relationship between diet and gut microbiota is a critical modulator underlying neurodevelopmental and psychiatric disorders in adults <ns0:ref type='bibr' target='#b47'>(Mayer, 2011)</ns0:ref>. Germ-free mice and antibiotic-treated mice have exhibited lower performance in a series of memory behavioral test than normal animals. This finding suggested that a disturbed gut microbiota is associated with memory dysfunction and brain-derived neurotrophic factor (BDNF) reduction <ns0:ref type='bibr' target='#b1'>(Arentsen et al, 2015;</ns0:ref><ns0:ref type='bibr' target='#b3'>Bercik et al, 2011;</ns0:ref><ns0:ref type='bibr' target='#b13'>Desbonnet et al, 2014)</ns0:ref>. Similarly, exposure to fluoride can induce the imbalance of microbiological composition. <ns0:ref type='bibr' target='#b71'>Yasuda et al. (2017)</ns0:ref> found that fluoride exposure causes a depletion of acidogenic bacterial genera in oral community. <ns0:ref type='bibr' target='#b43'>Luo et al. (2016)</ns0:ref> reported that excessive fluoride intake induces a reduction of Lactobacillus spp. in the gut of broiler chickens. Moreover, previous studies demonstrated that mental diseases and cognitive functions can be effectively modulated by supplying probiotics or prebiotics to enhance the intestinal environment <ns0:ref type='bibr' target='#b60'>(Sgritta et al, 2019;</ns0:ref><ns0:ref type='bibr' target='#b26'>Gareau et al, 2011)</ns0:ref>. Therefore, we speculated that a potential link between disordered gut physiology and fluoride neurotoxicity could be utilized in preventing fluoride-induced memory dysfunction. However, little evidence has clearly demonstrated this relationship, and no probiotic has been proven effective on preventing fluoride-induced dysfunctions in the brain.</ns0:p><ns0:p>Therefore, the present study aimed to assess whether fluoride could alter gut physiology. Lactobacillus johnsonii BS15 was used to revert the altered intestinal physiology to further reveal whether fluoride neurotoxicity is associated with intestinal physiology and assess whether BS15 could be an effective strategy to control fluoride-induced memory dysfunction and hippocampal injury. The hippocampus is critical for bacteria-cognition link, as well as learning and memory <ns0:ref type='bibr' target='#b62'>(Stachenfeld, Botvinick & Gershman, 2017)</ns0:ref>, because of its lifetime synaptic plasticity and neurogenesis <ns0:ref type='bibr' target='#b28'>(Greenberg, Ziff & Greene, 1986;</ns0:ref><ns0:ref type='bibr' target='#b15'>Deisseroth & Tsien 2002;</ns0:ref><ns0:ref type='bibr' target='#b30'>Hong et al., 2008)</ns0:ref>, thus, changes in hippocampal chemistry were given more attention in this study. The expression of neuronal activity-regulated genes, such as the immediate-early gene c-fos, BDNF, and neuronal cell adhesion molecule (NCAM), play a critical role in synaptic plasticity <ns0:ref type='bibr' target='#b30'>(Hong et al., 2008;</ns0:ref><ns0:ref type='bibr' target='#b59'>Simpson & Morris 2000)</ns0:ref>. Neurotrophin-and activity-dependent gene expression is mediated by cAMP/Ca 2+ responsive element-binding protein (CREB) <ns0:ref type='bibr' target='#b46'>(Mantamadiotis et al., 2002)</ns0:ref>. Moreover, BDNF, CREB, NCAM, and stem cell factor (SCF) are essential for neurogenesis.</ns0:p></ns0:div>
<ns0:div><ns0:head>Materials & Methods</ns0:head></ns0:div>
<ns0:div><ns0:head>Materials and methods</ns0:head></ns0:div>
<ns0:div><ns0:head>Culture and treatment with BS15</ns0:head><ns0:p>L. johnsonii BS15 (CCTTCC M2013663) was isolated from homemade yogurt collected from Hongyuan Prairie, Aba Autonomous Prefecture, China. Our previous study demonstrated that BS15 can effectively prevent non-alcoholic fatty liver disease by attenuating mitochondrial lesion and inflammation in the liver, lowering intestinal permeability, and adjusting gut microbiota <ns0:ref type='bibr' target='#b69'>(Xin et al., 2014)</ns0:ref>. Thus, BS15 was selected as the potential treatment strategy to improve gut flora in the present study. L. johnsonii BS15 was cultured anaerobically in de Man-Rogosa-Sharpe broth (Qingdao Rishui Bio-technologies Co., Ltd., Qingdao, China) at 37 °C. The amounts of bacterial cells were evaluated by heterotrophic plate count. Briefly, the cultures were centrifuged, washed, and resuspended in phosphate buffered saline (PBS; pH 7.0) for experimental use. The concentration of BS15 suspension was 1×10 9 cfu/mL (daily consumption dose: 0.2 mL/mice).</ns0:p></ns0:div>
<ns0:div><ns0:head>Behavioral tests</ns0:head></ns0:div>
<ns0:div><ns0:head>Novel object recognition (NOR) test</ns0:head><ns0:p>The NOR test is a widely used method for the investigation of working memory alteration. The results of NOR test reflects the function of the hippocampus based on the nature propensity of mice to a novel object rather than a familiar one. The task procedure consists three phases <ns0:ref type='bibr' target='#b0'>(Antunes & Biala, 2012)</ns0:ref>: habituation, familiarization, and test phase. Briefly, in the habituation phase, each mouse was allowed to freely explore the open-field arena (40×40×45 cm, l×b×h) for 1 h in the absence of objects. The mouse was then removed from the arena and placed in its home cage. During the familiarization phase, each mouse was placed in the arena to freely explore two different objects (#A+#B) for 5 min. The two objects were placed in the opposite corners of the cage. The mouse was given an intermediate retention interval of 20 min and then returned to the arena and re-exposed to object B along with a completely new object (object #C, distinguishable from object #A). Exploration ratio (F#C / (F#C + F#B) × 100, where F#C = frequency of exploring object #C, and F#B = frequency of exploring object #B) was calculated to assess memory. The objects used included a green bottle cap (#A), an orange bottle cap (#B), and a small smooth stone (#C).</ns0:p></ns0:div>
<ns0:div><ns0:head>T-maze test</ns0:head><ns0:p>An enclosed T-maze, which is an equipment with 10 cm-wide floor and 20 cm-high walls in the form of a 'T', was placed horizontally. The stems of two goal arms and a start arm were 30 cm long. A central partition in the middle of the two goal arms extended into the start arm (7 cm). Every arm had a guillotine door. The equipment and the operating steps were consistent with those of <ns0:ref type='bibr' target='#b14'>Deacon and Rawlins (2006)</ns0:ref>. First, the central partition was put in the T-maze with all doors open. Then, each mouse was placed in the start area directly from its home cage and allowed to choose the left or right arm. The mouse was kept in the chosen arm by quietly sliding the door down. After 30 s, the mouse and central partition were removed, and the mouse was placed back into its holding cage. After a retention interval of 1 min, the mouse was placed back into the start area for a second trial with all doors open. Each mouse was given ten trials over five days and allowed to explore the maze before sated. The trial was marked as 'correct' if the mouse chooses the other goal arm in consecutive trials. Each exploration should take no more than 2 min.</ns0:p></ns0:div>
<ns0:div><ns0:head>Establishment of animal model and study design</ns0:head><ns0:p>Forty-eight male ICR mice (3 week-old, Dashuo Biological Institute, Chengdu, Sichuan, China) were fed with normal chow diet (Dashuo Biological Institute, Chengdu, Sichuan, China) for 1 week to acclimatize to the new environment. After the adaptation period, the mice were equally and randomly divided into three groups and administered with either 0.2 mL of PBS (control group, F group) or BS15 (prob group) every day by gavage throughout a 98-day experimental period. Animals in the F and prob groups were exposed to 100 mg/L fluoride in drinking water from the 28th day to the 98th day. The mice were housed in an animal facility with a humidity of 40%-60%, a temperature of 20-22 °C and a 12-h light/dark cycle (lights off at 6:00 a.m. and on at 6:00 p.m.). We housed five or six mice per cage bedded by wood shavings and with food and water available ad libitum. The wood shavings were replaced every 2 days. Drinking water was replaced and water bottles were washed every 5 days. All animal experiments were performed according to the guidelines for the care and use of laboratory animals approved by the Institutional Animal Care and Use Committee of Sichuan Agricultural University (Approval number: SYXKchuan2019-187). Ten mice from each group were selected for behavioral test on the 98th day of the experiment. The other six mice of each group were sacrificed by cervical dislocation to collect tissues. The behavior test and sampling were carried out from 7:00 a.m. to 11:30 a.m.</ns0:p><ns0:p>Blood was sampled from the mice orbit, and the serum was separated by incubation at 4 °C for 30 min followed by centrifugation at 2,000×g for 20 min and stored at −30 °C. Tissues from the left hippocampus and partial ileum were removed and washed with ice-cold sterilized saline and then immediately frozen in liquid nitrogen for gene expression analysis. Tissues from the right hippocampus and partial ileum were ground (pH 7.4) into 5% and 10% homogenates, respectively, with PBS and then centrifuged at 12,000×g for 5 min at 4 °C. The obtained supernatant was stored at −80 °C for further detection.</ns0:p></ns0:div>
<ns0:div><ns0:head>Biochemical evaluation</ns0:head><ns0:p>The contents of corticosterone, D-lactate, and diamine oxidase (DAO) in the serum; inflammatory cytokines in the supernatant of hippocampal and ileal homogenates; and the apoptosis-regulated proteins in the supernatant of the hippocampal homogenate were measured by commercial enzyme-linked immunosorbent assay (ELISA) kit (Enzyme-linked Biotechnology Co., Ltd., Shanghai, China) specific for mice. The inflammatory cytokines included tumor necrosis factor-alpha (TNF-α), interferon-gamma (IFN-γ), interleukin-1β (IL-1β), IL-6, and IL-10 (only detected in ileum tissue). The operation was performed strictly according to the manufacturer's instructions.</ns0:p></ns0:div>
<ns0:div><ns0:head>Real-time quantitative polymerase chain reaction (RT-qPCR) analysis of gene expression</ns0:head><ns0:p>Total hippocampal RNA and ileal RNA were isolated using E.Z.N.A.® Total RNA Kit (OMEGA Bio-Tek, Doraville, GA, USA) according to the manufacturer's instructions. The isolated RNA was assessed for the ratio of absorbances at 260 and 280 nm and by agarose gel electrophoresis for quantitative and qualitative analyses. The isolated RNA was transcribed into first-strand complementary DNA (cDNA) with PrimeScript RT reagent kit with gDNA Eraser (Thermo Scientific, Waltham, Massachusetts, USA) according to the manufacturer's instructions. The cDNA products were stored in −80 °C for further use. qPCR was performed using CFX96 RT PCR Detection System (Bio-Rad, Hercules, CA, USA) and SYBR Premix Ex TaqTM PCR Kit (Bio-Rad, Hercules CA, USA) to quantify the relative expression levels of neuroplasticity-related factors (BDNF, CREB, SCF, and NCAM), early gene (c-fos), molecular proteins related to myelin structure (myelin oligodendrocyte glycoprotein Bcl-2-associated X protein [Bax], Bcl-xl/Bcl-2-asociated death promoter [Bad], caspase-9, and caspase-3) and cytokines (IFN-γ, TNF-α, in the hippocampus tissue and cytokines and tight junction (TJ) proteins (zonula occludens protein 1 [ZO-1], claudin-1, and occludin) in ileum tissue with 10 µL total reaction volume. The thermocycle protocol was performed as follows: 5 min at 95 °C, followed by 40 cycles of 10 s denaturation at 95 °C, and 30 s annealing/extension at optimum temperature (Table <ns0:ref type='table'>1</ns0:ref>). A final melting curve analysis was performed to monitor the purity of the PCR product. β-actin was used as reference gene to normalize the relative mRNA expression levels of target genes with values presented as 2 −ΔΔCq . The primer sequences and optimum annealing temperatures are shown in Table <ns0:ref type='table'>1</ns0:ref>.</ns0:p></ns0:div>
<ns0:div><ns0:head>Immunohistochemistry</ns0:head><ns0:p>A subset of mice in each group was sacrificed, and their brain was removed, fixed in 4% paraformaldehyde solution, and stored in 4 °C for immunohistochemical assay. The brain tissues were embedded by paraffin and cut by microtome. Slices were submerged in citrate antigen retrieval solution and heated on medium until boiling using a microwave (model: P70D20TL-P4; Galanz, Guangdong, China). The fire was ceased, and the tissues were kept warm for 8 min. Then, the tissues were heated at medium-low heat for 7 min. After free cooling, the slices were placed into PBS (pH 7.4) and shaken for 5 min for decoloration, which was repeated three times. Then, the sections were incubated in 3% oxydol for 25 min at room temperature and away from light for blocking endogenous peroxidase. The slices were washed three times in PBS by shaking for 5 min, then sealed for 30 min by 3% bull serum albumin, and incubated with monoclonal rabbit anti-BDNF (1:400) or polyclonal rabbit anti-CREB (1:500) at 4 °C overnight. Speciesspecific biotinylated anti-rabbit immunoglobulin (horseradish peroxidase-labeled) was used for immuno-detection. Following the second antibody incubation, 3,3′-diaminobenzidine staining kit was used to complete the reaction according to the manufacturer's instructions. Hematoxylin stain was performed to re-stain the nucleus. BDNF and CREB were quantified by calculating their integral optical density (IOD) in the object region (ImageJ, National Institutes of Health, USA). The average optical density (IOD/object region areas) was calculated, and the results were presented as levels of expression.</ns0:p></ns0:div>
<ns0:div><ns0:head>Statistics</ns0:head><ns0:p>Results were expressed as mean ± standard deviation. One-way ANOVA was performed between different groups with IBM SPSS Statistics 25 (IBM Corporation). Differences of P<0.05 were considered statistically significant. The figures were plotted using GraphPad Prism version 7.04 (San Diego, CA, USA).</ns0:p></ns0:div>
<ns0:div><ns0:head>Results</ns0:head></ns0:div>
<ns0:div><ns0:head>Behavioral results</ns0:head><ns0:p>Figure <ns0:ref type='figure' target='#fig_2'>1A</ns0:ref> shows that the F group showed a lower spontaneous exploration (control vs. F, P<0.01; F vs. prob, P=0.019) than the other two groups, but no difference (P>0.05) was found between the control and prob groups (P=0.104). The exploration ratio of the three groups were not significantly different (control vs. F, P=0.125; control vs. prob, P=0.285; F vs. prob, P=0.626; Fig. <ns0:ref type='figure' target='#fig_2'>1B</ns0:ref>).</ns0:p></ns0:div>
<ns0:div><ns0:head>mRNA and protein expressions of BDNF in the hippocampus</ns0:head><ns0:p>Figure <ns0:ref type='figure'>2A</ns0:ref> shows that the F group presented a significant decrease in BDNF mRNA level than the other two groups (control vs. F, P<0.01; F vs. prob, P=0.014), whereas the control and prob groups had no difference (P=0.414). As shown in Fig. <ns0:ref type='figure'>2B-E</ns0:ref>, the BDNF protein level was decreased in the F group compared with the control (P<0.01) and prob groups (P=0.018). No difference was observed between the control and prob groups (P=0.513).</ns0:p></ns0:div>
<ns0:div><ns0:head>mRNA and protein expressions of CREB in the hippocampus</ns0:head><ns0:p>As shown in Fig. <ns0:ref type='figure'>3A</ns0:ref>, CREB mRNA level was slightly reduced in the F group (P=0.484) and increased in the prob group (P=0.065) compared with the control group. The CREB mRNA level in the prob group was significantly (P=0.026) higher than that in the F group. Figure <ns0:ref type='figure'>3B-E</ns0:ref> shows that the F group presented a significant decrease in CREB protein level than the other two groups (control vs. F, P=0.048; F vs. prob, P<0.01). CREB protein level in the prob group was remarkably higher than that in the control group (P=0.032).</ns0:p></ns0:div>
<ns0:div><ns0:head>mRNA expressions of NCAM, SCF, and c-fos in the hippocampus</ns0:head><ns0:p>As shown in Fig. <ns0:ref type='figure' target='#fig_3'>4A</ns0:ref>, the control group had a higher NCAM mRNA level than the other two groups (control vs. F, P=0.001; control vs. prob, P=0.001), but no difference was found between the F and prob groups (P=0.783). Figure <ns0:ref type='figure' target='#fig_3'>4B</ns0:ref> shows that the SCF mRNA level of the three groups were not significantly different (control vs. F, P=0.105; control vs. prob, P=0.842; F vs. prob, P=0.181). As shown in Fig. <ns0:ref type='figure' target='#fig_3'>4C</ns0:ref>, the control group presented a higher mRNA expression of c-fos than the other two groups (control vs. F, P<0.01; control vs. prob, P<0.01), but the expression of c-fos between the other two groups were not significantly different (F vs. prob, P=0.449).</ns0:p></ns0:div>
<ns0:div><ns0:head>Hippocampal inflammation</ns0:head><ns0:p>Figures 5A and 5D show that the TNF-α (mRNA level: control vs. F, P=0.013; F vs. prob, P=0.252; protein level: control vs. F, P<0.001; F vs. prob, P=0.001) and IFN-γ (mRNA level: control vs. F, P=0.007; F vs. prob, P=0.037; protein level: control vs. F, P<0.001; F vs. prob, P<0.001) in the F group were significantly or slightly higher than those in the other two groups in mRNA and protein levels. The TNF-α of the prob group was significantly increased in the protein level (P<0.001) but not in the mRNA level (P=0.115) compared with the control group (Fig. <ns0:ref type='figure' target='#fig_4'>5A</ns0:ref>). No difference in IFN-γ (Fig. <ns0:ref type='figure' target='#fig_4'>5D</ns0:ref>) was observed between the control and prob groups in the mRNA (P=0.381) and protein levels (P=0.931). Figure <ns0:ref type='figure' target='#fig_4'>5C</ns0:ref> reveals that the F group had a remarkably lesser IL-6 than the other two groups in mRNA (control vs. F, P<0.001; F vs. prob, P=0.011) and protein levels (control vs. F, P<0.001; F vs. prob, P<0.001). As shown in Fig. <ns0:ref type='figure' target='#fig_4'>5C</ns0:ref>, the prob group presented a significantly lesser IL-6 than the control group in mRNA (P<0.001) and protein levels (P<0.001). The three groups had no remarkable change in IL-1β in the mRNA and protein levels (mRNA level: control vs. F, P=0.089; control vs. prob, P=0.097; F vs. prob, P=0.962; protein level: control vs. F, P=0.420; control vs. prob, P=0.733; F vs. prob, P=0.636). IL-10 did not reach the detection threshold (Fig. <ns0:ref type='figure' target='#fig_4'>5B</ns0:ref>).</ns0:p></ns0:div>
<ns0:div><ns0:head>Hippocampal myelin and apoptosis-related proteins</ns0:head><ns0:p>Fig. <ns0:ref type='figure' target='#fig_5'>6A</ns0:ref> shows the clear decreasing trend in the mRNA level of PLP in the F group compared with the other two groups (control vs. F, P<0.013; F vs. prob, P=0.0039). A difference (control vs.prob, P=0.568) in the mRNA level of PLP between the control and prob groups was not detected. Figure <ns0:ref type='figure' target='#fig_5'>6B</ns0:ref> shows that the control group had a higher MOG mRNA level than the F and prob group (control vs. F, P=0.005; control vs. prob, P=0.002), whereas the MOG mRNA level of the F group was not different (P=0.778) from that of the prob group. Differences in the mRNA levels of MBP (control vs. F, P=0.277; control vs.prob, P=0.706; F vs. prob, P=0.415; Fig. <ns0:ref type='figure' target='#fig_5'>6C</ns0:ref>) and MAG (control vs. F, P=0.904; control vs. prob, P=0.919; F vs. prob, P=0.970; Fig. <ns0:ref type='figure' target='#fig_5'>6D</ns0:ref>) were not observed among the three groups. Figures <ns0:ref type='figure' target='#fig_6'>7A and 7E</ns0:ref> present a remarkable (control vs. F, P<0.026; F vs. prob, P=0.038) decrease in Bcl-2 mRNA level and a significant (control vs. F, P<0.033; F vs. prob, P=0.001) increase in caspase-3 mRNA level in F group compared with the other two groups, and the Bcl-2 (P=0.948) and caspase-3 mRNA levels (P=0.127) of the control and prob groups were not significantly different. Figures <ns0:ref type='figure' target='#fig_6'>7B, 7C</ns0:ref>, 7D, and 7F reveal no significant differences (P>0.05) among the three group in the mRNA levels of Bcl-xl (control vs. F, P=0.290; control vs. prob, P=0.463; F vs. prob, P=0.735), Bax (control vs. F, P=0.784; control vs. prob, P=0.681; F vs. prob, P=0.891), Bad (control vs. F, P=0.954; control vs. prob, P=0.137; F vs. prob, P=0.124), and caspase-9 (control vs. F, P=0.128; control vs. prob, P=0.684; F vs. prob, P=0.197).</ns0:p></ns0:div>
<ns0:div><ns0:head>Inflammatory factors in the ileum</ns0:head><ns0:p>As shown in Fig. <ns0:ref type='figure' target='#fig_7'>8A</ns0:ref>, the F group exhibited a slight increase in TNF-α mRNA level (P=0.079) and a significantly enhanced TNF-α protein level (P<0.001) compared with the control group. TNF-α in the mRNA (P=0.008) and protein levels (P<0.001) were significantly reduced in the prob group compared with the F group. Differences in the mRNA (P=0.197) and protein levels (P=0.348) of TNF-α were not observed between the control and prob groups. The mRNA (P=0.888) and protein levels (P=0.781) of IL-1β (Fig. <ns0:ref type='figure' target='#fig_7'>8B</ns0:ref>) between the F and prob groups were not significantly different. IL-1β (Fig. <ns0:ref type='figure' target='#fig_7'>8B</ns0:ref>) in the control group showed a slight decrease in mRNA level (control vs. F, P=0.132; control vs. prob, P=0.103) and a remarkable (control vs. F, P=0.003; control vs. prob, P=0.002) decline in protein level compared with those in the other two groups (Fig. <ns0:ref type='figure' target='#fig_7'>8B</ns0:ref>). No difference was observed in IL-6 (Fig. <ns0:ref type='figure' target='#fig_7'>8C</ns0:ref>) in mRNA (control vs. F, P=0.635; control vs. prob, P=0.615; F vs. prob, P=0.975) and protein levels (control vs. F, P=0.407; control vs. prob, P=0.908; F vs. prob, P=0.347) among the three groups. The F group exhibited a significantly increased IFN-γ (mRNA level: control vs. F, P<0.001; F vs. prob, P<0.001; protein level: control vs. F, P<0.001; F vs. prob, P<0.001; Fig. <ns0:ref type='figure' target='#fig_7'>8D</ns0:ref>) and a sharply decreased IL-10 (mRNA level: control vs. F, P=0.01; F vs. prob, P=0.02; protein level: control vs. F, P<0.001; F vs. prob, P<0.001; Fig. <ns0:ref type='figure' target='#fig_7'>8E</ns0:ref>) compared with the other two groups in mRNA and protein levels. These differences (IFN-γ mRNA level: control vs. prob, P=0.062; IL-10 mRNA level: control vs. prob, P=0.838; IL-10 protein level: control vs. prob, P=0.535) were not observed in the other two groups except for IFN-γ protein level (control vs. prob, P=0.005).</ns0:p></ns0:div>
<ns0:div><ns0:head>Intestinal permeability</ns0:head><ns0:p>As shown in Fig. <ns0:ref type='figure'>9A</ns0:ref>, the control group reported a slightly higher claudin-1 than the other two groups <ns0:ref type='bibr'>(control vs. F, P=0.111; control vs. prob, P=0.129)</ns0:ref>. The mRNA expression of claudin-1 between the F and prob groups had no difference <ns0:ref type='bibr'>(F vs. prob, P=0.855)</ns0:ref>. Figure 9A also shows that ZO-1 and occludin in the control group were remarkably higher than those in the F group (P=0.008, P=0.004) and slightly (P=0.121, P=0.061) higher than the prob group. Figure <ns0:ref type='figure'>9A</ns0:ref> shows that the prob group presented a slightly more ZO-1 (P=0.125) and occludin (P=0.114) than the F group. Figure <ns0:ref type='figure'>9B</ns0:ref> and9 C show that the F group exhibited a significant increase in serum DAO activity (control vs. F, P=0.002; F vs. prob, P=0.048) and D-lactate concentration (control vs. F, P<0.001; F vs. prob, P=0.011) compared with the other two groups. No significant (P=0.128) difference in the DAO activity between the control and prob groups was observed. The D-lactate concentration in the prob group was significantly (P=0.001) higher than that in the control group.</ns0:p></ns0:div>
<ns0:div><ns0:head>Discussion</ns0:head><ns0:p>Although a large number of studies have found fluoride-induced brain lesions, few studies focused on the link between the gut changes and neurotoxicity induced by fluoride. In view of the gut-brain axis, changes in the intestinal microenvironment, including gut microbiota, inflammatory cytokines, and hormones, can influence brain chemistry and behaviors. The current study further explored fluoride-induced brain lesion and its link with the gut. We found that the mice exposed to 100 mg/L sodium fluoride for 10 weeks had hippocampal lesions, which caused memory impairment as indicated by their lower performance in the T-maze test and the reduced PLP mRNA level. Moreover, high fluoride exposure caused intestinal inflammation and increased the intestinal permeability as indicated by the increased inflammatory cytokines, serum DAO activity, and D-lactate content and reduced mRNA levels of TJ protein. BS15, a probiotic capable of improving the gut microbiome, reversed the gut changes and alleviated the brain lesion and memory impairment induced by fluoride. The current study confirmed our hypothesis that gut changes may play a key role in memory dysfunction during high fluoride exposure, and BS15 administration is a potential method of preventing these fluoride-induced damages on memory ability.</ns0:p><ns0:p>Although the memory abilities of the three groups were not remarkably different as shown in the NOR test, the fluoride-infected mice in the F group showed memory impairment as indicated by their lower spontaneous exploration in the T-maze test compared with the other groups. The poor performances of the fluoride-infected mice in various memory-related behavioral tests were also reported in previous studies <ns0:ref type='bibr' target='#b39'>(Liu et al., 2010;</ns0:ref><ns0:ref type='bibr' target='#b6'>Chen et al., 2018)</ns0:ref>. In our study, BS15 substantially reversed the performance of mice in the T-maze and thus has a beneficial effect on the memory ability of fluoride-infected mice. Neuronal activation was also decreased as indicated by the lower mRNA level of c-fos in the hippocampus of fluoride-infected mice. <ns0:ref type='bibr' target='#b19'>Fleischmann et al. (2003)</ns0:ref> demonstrated that mice lacking c-fos exhibit remarkable memory deficits; hence, c-fos gene has a critical role in memory. The hippocampus is a unique area of the brain that is able of neuroplasticity <ns0:ref type='bibr' target='#b25'>(Galea et al., 2013)</ns0:ref>. Neuroplasticity usually increases under the conditions that increase memory abilities, and its ablation often induces lesions that affect memory <ns0:ref type='bibr' target='#b16'>(Deng, Aimone & Gage, 2010;</ns0:ref><ns0:ref type='bibr' target='#b38'>Leuner & Gould, 2010)</ns0:ref>. These findings suggested that the hippocampus play a critical role in memory and synaptic plasticity <ns0:ref type='bibr' target='#b73'>(Yirmiya & Goshen, 2011)</ns0:ref>. Neurotrophic factors, especially the BDNF, are important in neurogenesis, and their production are involved in almost every aspect of neural and behavioral plasticity <ns0:ref type='bibr' target='#b41'>(Lu, Christian & Lu, 2008;</ns0:ref><ns0:ref type='bibr' target='#b40'>Li et al., 2008)</ns0:ref>, especially for hippocampal-dependent memory <ns0:ref type='bibr' target='#b29'>(Heldt et al., 2007)</ns0:ref>. An increasing body of data indicated that probiotic consumption can regulate anxiety and memory functions via changes in the expression of key components, such as BDNF, CREB, and N-methyl-d-aspartate receptors <ns0:ref type='bibr' target='#b10'>(Clarke et al., 2013;</ns0:ref><ns0:ref type='bibr' target='#b65'>Wall et al., 2010)</ns0:ref>. The underlying mechanism is that probiotics can synthesize and recognize an array of neurochemicals, including neurotransmitters, secondary bile acids, neuroactive short chain fatty acids (SCFAs), and other biologically active small molecules. For example, <ns0:ref type='bibr' target='#b37'>Kumar et al. (2017)</ns0:ref> reported that L. johnsonii could increase the concentrations of acetate and butyrate in feces. Butyrate, an SCFA, can decrease BDNF methylation and consequently cause an overexpression of BDNF by decreasing ten-eleven translocation methylcytosine dioxygenase 1, which is the enzyme responsible for catalyzing the conversion of DNA methylation to hydroxymethylation <ns0:ref type='bibr' target='#b66'>(Wei et al., 2015)</ns0:ref>. In the present study, the expression of BDNF in the hippocampus was reduced by fluoride. This finding is consistent with the result found by Niu et al. <ns0:ref type='bibr' target='#b9'>(2018)</ns0:ref> CREB is the transcriptional regulator of BDNF, and similar genomic network analysis reported that CREB is the center of Alzheimer's disease's pathology <ns0:ref type='bibr' target='#b34'>(Jeong et al., 2001)</ns0:ref>. The mRNA and protein levels of BDNF and CREB in the BS15treated mice were substantially increased compared with the fluoride-infected mice, and the level of CREB was even slightly higher than the control group. Similarly, Kadry and Megeed <ns0:ref type='bibr' target='#b9'>(2018)</ns0:ref> found that Lactobacillus could effectively inhibit the reduction of BDNF induced by cadmium chloride in the hippocampus of mice. Moreover, Corpuz et al. <ns0:ref type='bibr' target='#b9'>(2018)</ns0:ref> reported that Lactobacillus paracasei K71 prevents age-related cognitive decline in senescence-accelerated mouse prone 8 by increasing the protein expression of BDNF and the phosphorylation of CREB in the hippocampus. Neuroplasticity also needs the regulation of NCAM <ns0:ref type='bibr' target='#b58'>(Seidenfaden, Krauter & Hildebrandt, 2006)</ns0:ref> and the stimulation of SCF <ns0:ref type='bibr' target='#b35'>(Jin et al., 2002)</ns0:ref>. In line with the results of Niu et al. <ns0:ref type='bibr' target='#b9'>(2018)</ns0:ref>, fluoride significantly reduced the mRNA level of NCAM.</ns0:p><ns0:p>Many evidence indicated that neuroinflammation may be one of the pathogenesis underlying cognitive changes and the development of many neurodegenerative diseases <ns0:ref type='bibr' target='#b22'>(Frank-Cannon et al., 2009)</ns0:ref>. The hippocampus is vulnerable to the insults by inflammatory cytokines because of its high expression of cytokine receptors <ns0:ref type='bibr' target='#b12'>(Das & Research, 2008)</ns0:ref>. Few in vivo studies focused on the neuroinflammation caused by fluorosis. Excessive exposure to fluoride triggered neuroinflammation as indicated by the increase in the TNF-α and IFN-γ in the hippocampus of the F group. The result on TNF-α was in agreement with the finding of <ns0:ref type='bibr' target='#b72'>Yan et al. (2016)</ns0:ref>, who found that TNF-α immunoreactivity is increased in the hippocampus of rat exposed to 120 ppm fluoride for 10 weeks. However, by contrast to the results shown by <ns0:ref type='bibr' target='#b72'>Yan et al. (2016)</ns0:ref>, the present study reported a reduced IL-6 in the hippocampus of mice exposed to 100 mg/L sodium fluoride for 10 weeks. Differences in drug doses, delivery cycles, and breeds of rodents may be associated with the contradictory results. A previous research confirmed that IL-6-deficient mice present a weakened neuroprotection of the hippocampus <ns0:ref type='bibr' target='#b33'>(Jean Harry, Bruccoleri & Lefebvre d'Hellencourt, 2003)</ns0:ref>. <ns0:ref type='bibr' target='#b24'>Funk et al. (2011)</ns0:ref> indicated the potential role of IL-6 in modulating TNFα-mediated neurotoxicity. Intestinal microbiome and some probiotics can influence health status and disease risk by activating immune response against dangerous stimuli and activating regulatory mechanisms to avoid uncontrolled inflammation. Intestinal permeability and potentially beneficial metabolites may be the underlying mechanisms of the anti-inflammation. Intestinal microbiome can ferment dietary fiber and starch in the large intestines and produce SCFAs <ns0:ref type='bibr' target='#b7'>(Chen, Faller & Spanjaard , 2003)</ns0:ref>. The effect of butyrate and other SCFAs on preventing inflammation in colon diseases and different neural inflammation models in cell cultures have been demonstrated <ns0:ref type='bibr' target='#b31'>(Huuskonen, 2004)</ns0:ref>. In the present study, BS15 administration had profitable effect on balancing the inflammatory cytokines in fluoride-infected mice.</ns0:p><ns0:p>The mRNA expression levels of myelin-and apoptosis-related proteins were also detected to further investigate the effect of BS15 on the hippocampal impairment. Myelin sheaths enwrap the nerve fiber to guarantee interneuronal transmission efficiency <ns0:ref type='bibr' target='#b48'>(Nguyen et al., 2009)</ns0:ref>. Myelin is consisted of PLP (a transmembrane protein), MBP (a peripheral membrane protein), the outermost MOG, and the innermost MAG <ns0:ref type='bibr' target='#b49'>(Niu et al., 2018)</ns0:ref>. The remarkably reduced mRNA expression levels of PLP and MOG in the F group suggested that myelin lesion occurred in the hippocampus. The changes in PLP induced by fluoride were consistent with the findings of Niu et al. <ns0:ref type='bibr' target='#b9'>(2018)</ns0:ref>. The reduced tendency of PLP was inhibited by BS15; hence, BS15 may have a protective effect on the myelin. A previous study demonstrated that mice present increased positive apoptotic neurons following 10 weeks of exposure to 120 ppm fluoride in drinking water <ns0:ref type='bibr' target='#b72'>(Yan et al., 2016)</ns0:ref>. In the present study, the reduced Bcl-2 (anti-apoptosis protein) and increased caspase-3 (pro-apoptosis protein) in fluoride-infected mice created conditions for apoptotic neurons, and these changes were remarkably reversed by BS15 treatment.</ns0:p><ns0:p>Intestinal leakage can facilitate the translocation of bacterial composition, such as microorganisms and their products <ns0:ref type='bibr' target='#b5'>(Carvalho AF, Berk M & Maes M 2016)</ns0:ref>, and is considered a key factor in mental disease <ns0:ref type='bibr' target='#b4'>(Braniste et al., 2014;</ns0:ref><ns0:ref type='bibr'>Zhan et al., 2016;</ns0:ref><ns0:ref type='bibr' target='#b17'>Emery et al., 2017)</ns0:ref>. Inflammation can enhance epithelial permeability <ns0:ref type='bibr' target='#b70'>(Xue et al., 2014;</ns0:ref><ns0:ref type='bibr' target='#b56'>Schulzke et al., 2009)</ns0:ref>. Inflammatory cytokines, such as IL-1β, TNF-α, and IFN-γ, can increase gut permeability <ns0:ref type='bibr' target='#b44'>(Ma et al., 2004;</ns0:ref><ns0:ref type='bibr' target='#b56'>Schulzke et al., 2009;</ns0:ref><ns0:ref type='bibr' target='#b67'>Weber et al., 2010)</ns0:ref>. IL-10, an anti-inflammatory cytokine, plays a critical role in the homeostasis of the gut, which was illustrated by the finding that spontaneous colitis occurs in IL-10 −/− mice <ns0:ref type='bibr' target='#b27'>(Gomes-Santos et al., 2012)</ns0:ref>. The current study found that excessive fluoride intake resulted in intestinal inflammation by increasing pro-inflammatory cytokines (TNF-α, IL-1β, and IFN-γ) and reducing anti-inflammatory cytokine (IL-10). Treatment with BS15 efficiently lowered the inflammatory reaction caused by fluoride. TJ proteins act as a barrier that mediates the cell-to-cell adhesion and prevents molecules from crossing through the epithelial sheet between adjacent cells into systemic circulation <ns0:ref type='bibr' target='#b51'>(Piche, 2014)</ns0:ref>. The mRNA level of two TJ proteins, namely, ZO-1 and occludin, in the ileum of the fluoride-infected mice were also remarkably reduced with gut inflammation enhancement and therefore led to higher levels of DAO activity and D-lactate content in the serum. The tissue of the small intestine contains the highest DAO activity, and serum DAO is derived primarily from the small intestines in many mammalian species <ns0:ref type='bibr' target='#b42'>(Luk, Bayless & Baylin, 1980)</ns0:ref>. Moreover, mammalian species cannot produce D-lactate, and the main source of D-lactate is from the commensal bacteria in the gastrointestinal tract <ns0:ref type='bibr' target='#b61'>(Sun et al, 2001)</ns0:ref>. The metabolism of serum Dlactate is very slow. The increases in serum DAO activity and D-lactate content occurred when the intestinal mucosal integrity was damaged and served as useful plasma markers of mucosal integrity <ns0:ref type='bibr' target='#b18'>(Ewaschuk, Naylor & Zello, 2005;</ns0:ref><ns0:ref type='bibr' target='#b42'>Luk, Bayless & Baylin, 1980)</ns0:ref>. In this study, we found that BS15 effectively improved intestinal permeability as shown by the remarkably lower serum DAO activity and D-lactate concentration in the prob group compared with the F group. The result may be explained in part by the slightly increased TJ proteins in the prob group. Apoptosis is another possible reason that may have caused barrier dysfunction <ns0:ref type='bibr' target='#b56'>(Schulzke et al., 2009)</ns0:ref>. These results suggested that fluoride could cause intestinal inflammation and damage mucosal integrity, which results in enhanced intestinal permeability, and BS15 administration could alleviate these pathological changes.</ns0:p></ns0:div>
<ns0:div><ns0:head>Conclusions</ns0:head><ns0:p>The study deepened our understanding of the link between fluoride neurotoxicity on memory function and gut microenvironment. BS15 exerted beneficial effects against excessive fluoride intake-induced memory impairment, related neural inflammation, and demyelination by improving intestinal inflammation and integrity and increasing apoptosis markers in the hippocampus of mice. Manuscript to be reviewed Manuscript to be reviewed </ns0:p><ns0:note type='other'>Figure 3</ns0:note></ns0:div><ns0:figure xml:id='fig_0'><ns0:head /><ns0:label /><ns0:figDesc>[MOG], proteolipid protein [PLP], myelin basic protein [MBP], and myelin-associated glycoprotein [MAG]), and apoptosis-related proteins (B-cell lymphoma-2 [Bcl-2], B-cell lymphoma-extra large [Bcl-xl],</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_1'><ns0:head /><ns0:label /><ns0:figDesc>Zhan X, Stamova B, Jin LW, Decarli C, Phinney B, Sharp FR. 2016. Gram-negative bacterial molecules associate with Alzheimer disease pathology. Neurology 87(22):2324-2332 DOI: 10.1212/WNL.0000000000003391. Zhong W, Zhao Y, McClain CJ, Kang YJ, Zhou Z. 2010. Inactivation of hepatocyte nuclear factor-42 mediates alcohol-induced downregulation of intestinal tight junction proteins. Am J Physiol Gastrointest Liver Physiol 299(3):1251-1254 DOI: 10.1152/ajpgi.00515.2009.</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_2'><ns0:head>Figure 1</ns0:head><ns0:label>1</ns0:label><ns0:figDesc>Figure 1</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_3'><ns0:head>Figure 4 mRNA</ns0:head><ns0:label>4</ns0:label><ns0:figDesc>Figure 4</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_4'><ns0:head>Figure 5 mRNA</ns0:head><ns0:label>5</ns0:label><ns0:figDesc>Figure 5</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_5'><ns0:head>Figure 6 mRNA</ns0:head><ns0:label>6</ns0:label><ns0:figDesc>Figure 6</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_6'><ns0:head>Figure 7 mRNA</ns0:head><ns0:label>7</ns0:label><ns0:figDesc>Figure 7</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_7'><ns0:head>Figure 8 mRNA</ns0:head><ns0:label>8</ns0:label><ns0:figDesc>Figure 8</ns0:figDesc></ns0:figure>
<ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='28,42.52,303.41,525.00,384.00' type='bitmap' /></ns0:figure>
<ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='29,42.52,288.81,525.00,376.50' type='bitmap' /></ns0:figure>
<ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='36,42.52,280.87,525.00,142.50' type='bitmap' /></ns0:figure>
<ns0:note place='foot' n='2'>a The primer sequences of c-fos is designed by National Center for Biotechnology Information 3 (NCBI) and the referenced gene ID is 14281. PeerJ reviewing PDF | (2020:02:46283:2:0:CHECK 22 Jul 2020)</ns0:note>
</ns0:body>
" | "Dear Editors and Reviewers:
Thank you very much for sending us the Reviewers’ reports of our manuscript.
We would like to thank the Reviewers for their valuable comments. We have revised carefully our manuscript according to the comments. The following is a detailed list of responses to all comments and changes we have made. We hope that the revised manuscript could be satisfactory enough for publication in PeerJ. If any question arises, please let us know.
Thank you very much again for your consideration and precious time.
Sincerely yours,
Xueqin Ni
Tel.: +86 2886291162
E- mail：xueqinni@foxmail.com
Responses to reviewer 1：
1. The authors should present the research results following certain logics such as protection, then the underlying mechanisms.
We have modified the order of research results following the logic as protective effects on memory ability first, then the mechanism underlying hippocampus and intestinal health. Accordingly, we changed all the orders of expression in the revised manuscript, including the sections of Abstract, Results and Discussion. We also renumbered the figures.
2. Figure legends didn’t clearly show the meaning of figure. For example, what’s the meaning of a and b in figures?
Bars with different letters (a, b, and c) are significantly different on the basis of Duncan’s multiple range test (P<0.05). We have added a bracket with “a, b and c” inside to explain the meaning of different letters and make it clear.
3. Figure 3C is wrong.
We have corrected it. We are sorry for our careless mistake.
4. Figure 5 is not clear. Higher magnification is needed.
Figure 5 has been edited according to the comment. However, the figure may be still not very clear when the PDF is systematically generated. If the new figure is still unclear, please check the orginally uploaded figure.
5. Data of a molecule of mRNA and protein should put in one figure. And all protein level detection should provide western blot band and corresponding statistical chart.
According to the comment, we re-organized the figures to make data of mRNA and protein levels of BDNF and CREB in one figure. We also removed the result of c-fos from Fig.1 to Fig.4 for better logic of this paper. However, no western blot test was invovled in our present study. All the protein levels of inflammatory factors in ileum and hippocampus were detected by ELISA. Please check it in case of any further misunderstanding.
" | Here is a paper. Please give your review comments after reading it. |
9,764 | "<ns0:abstract xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:p>Background. Excessive fluoride can lead to chronic neurodegeneration characterized by neuron and myelin loss and memory dysfunction. The gut-brain axis hypothesis suggests that gut microbiota plays a crucial role in regulating brain function . Thus, using probiotics to adjust the gut microenvironment may be a potential therapy for mental diseases.</ns0:p><ns0:p>Methods. Mice in the prob group were administrated with Lactobacillus johnsonii BS15 for 28 days prior to and throughout a 70-day exposure to sodium fluoride. The drinking water of all groups (F and prob groups) except the control group were replaced by high-fluoride water (100 mg NaF/L) on day 28. Animals in each group were divided into two subsets: one underwent behavioral test, and the other was sacrificed for sampling. The mRNA expression level and protein content related to inflammatory reaction in the ileum and hippocampus were respectively detected by reverse transcription quantitative polymerase chain reaction (RT-qPCR) and enzyme-linked immunosorbent assay (ELISA). The mRNA expression levels of proteins related to myelin structure, apoptosis, and memory in the hippocampus and tight junction proteins in the ileum were determined by RT-qPCR and/or immunohistochemistry. Gut permeability markers (D-lactate and diamine oxidase [DAO]) in the serum were also examined by ELISA.</ns0:p><ns0:p>Results. The results showed that fluoride exposure induced a lower spontaneous exploration (P<0.05) in T-maze test, which indicated an impairment of memory. Spontaneous exploration of BS15-treated mice was significantly higher (P<0.05) than that in F group. Fluoride reduced (P<0.05) levels of myelin structural protein (proteolipid protein) and neurogenesis-associated proteins (brain-derived neurotrophic factor and cAMP/Ca 2+ responsive element-binding protein), induced disordered inflammatory cytokines (TNF-α, IFN-γ, and IL-6; P<0.05), increased pro-apoptotic genes (caspase-3; P<0.05), and decreased antiapoptotic genes (Bcl-2; P<0.05) in the hippocampus, of which the influences were reversed by BS15. BS15 treatment exerted significant preventive effects on reversing the gut inflammation induced by excessive fluoride intake by reducing (P<0.05) the levels of pro-inflammatory cytokines (tumor necrosis factor-alpha [TNF-α] and interferon-gamma [IFN-γ ]) and remarkably increasing (P<0.05) the level of antiinflammatory cytokines (IL-10). Moreover, the serum DAO activity and D-lactate concentration significantly increased by fluoride were also reduced (P<0.05) by BS15. This result indicated the profitable effect of BS15 on gut permeability.</ns0:p></ns0:div>
<ns0:div><ns0:head>Conclusion:</ns0:head><ns0:p>L. johnsonii BS15 intake could benefit the neuroinflammation and demyelination in the hippocampus by improving the gut environment and ameliorating fluorine-induced memory dysfunction.</ns0:p></ns0:div>
</ns0:abstract>
<ns0:body xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:head>Introduction</ns0:head><ns0:p>Excessive fluoride intake has attracted increasing attention because of its widely introduced sources and adverse effects on human health. Fluoride concentration in the groundwater could range from under 1 mg/L to more than 35 mg/L <ns0:ref type='bibr' target='#b50'>(Petrone et al., 2013)</ns0:ref>. In addition, fluoride concentration in drinking tea (especially Chinese brick tea) ranges from 600-2800 mg/kg <ns0:ref type='bibr' target='#b23'>(Fung et al.,1999)</ns0:ref>. The risk of high fluoride intake from food is also increasing because of the increased fluorine-containing crop protection compounds <ns0:ref type='bibr' target='#b45'>(Maienfisch & Hall, 2004)</ns0:ref>. The most well-known fluoride-induced negative influences are on teeth (dental fluorosis) <ns0:ref type='bibr' target='#b54'>(Sabokseir , Golkari & Sheiham, 2016)</ns0:ref> and skeleton (skeletal fluorosis) <ns0:ref type='bibr' target='#b20'>(Littleton et al., 1999;</ns0:ref><ns0:ref type='bibr' target='#b63'>Wang et al., 2019)</ns0:ref>. Workers exposed to high fluoride suffer from drowsiness and impaired learning and memory <ns0:ref type='bibr' target='#b11'>(Czerwiński & Lankosz, 1978)</ns0:ref>. Furthermore, epidemiological investigation from China <ns0:ref type='bibr' target='#b62'>(Wang et al., 2007)</ns0:ref>, India <ns0:ref type='bibr' target='#b56'>(Sebastian & Sunitha, 2015)</ns0:ref>, Mexico <ns0:ref type='bibr' target='#b2'>(Bashash et al., 2017)</ns0:ref>, and Iran <ns0:ref type='bibr' target='#b52'>(Razdan et al., 2017)</ns0:ref> reported that children residing in endemic areas show deficits in learning and memory abilities. Rodents exposed to excessive fluoride also showed poor performances in memory-related behavioral tests in animal experiments <ns0:ref type='bibr' target='#b6'>(Chen et al., 2018;</ns0:ref><ns0:ref type='bibr' target='#b39'>Liu et al., 2010)</ns0:ref>. Animal experiments concerning fluoride neurotoxicity demonstrated that high fluoride exposure causes the pathological alteration of the synaptic ultrastructure of the hippocampus <ns0:ref type='bibr' target='#b67'>(Qian et al., 2013)</ns0:ref>. <ns0:ref type='bibr' target='#b49'>Niu et al. (2018)</ns0:ref> suggested that fluoride could reduce neurotrophy and neuron adhesion and consequently damage the myelin in the hippocampus of mice.</ns0:p><ns0:p>Few safe and effective methods can protect the brain from fluoride neurotoxicity. The gut-brain axis, a bi-directional communication system linking the gut and brain, has provided a new insight into the treatment of brain-derived diseases <ns0:ref type='bibr' target='#b21'>(Forsythe &Kunze, 2013)</ns0:ref>. Recent advances in metagenomics confirmed that the relationship between diet and gut microbiota is a critical modulator underlying neurodevelopmental and psychiatric disorders in adults <ns0:ref type='bibr' target='#b47'>(Mayer, 2011)</ns0:ref>. Germ-free mice and antibiotic-treated mice have exhibited lower performance in a series of memory behavioral test than normal animals. This finding suggested that a disturbed gut microbiota is associated with memory dysfunction and brain-derived neurotrophic factor (BDNF) reduction <ns0:ref type='bibr' target='#b1'>(Arentsen et al, 2015;</ns0:ref><ns0:ref type='bibr' target='#b3'>Bercik et al, 2011;</ns0:ref><ns0:ref type='bibr' target='#b13'>Desbonnet et al, 2014)</ns0:ref>. Similarly, exposure to fluoride can induce the imbalance of microbiological composition. <ns0:ref type='bibr' target='#b70'>Yasuda et al. (2017)</ns0:ref> found that fluoride exposure causes a depletion of acidogenic bacterial genera in oral community. <ns0:ref type='bibr' target='#b43'>Luo et al. (2016)</ns0:ref> reported that excessive fluoride intake induces a reduction of Lactobacillus spp. in the gut of broiler chickens. Moreover, previous studies demonstrated that mental diseases and cognitive functions can be effectively modulated by supplying probiotics or prebiotics to enhance the intestinal environment <ns0:ref type='bibr' target='#b59'>(Sgritta et al, 2019;</ns0:ref><ns0:ref type='bibr' target='#b26'>Gareau et al, 2011)</ns0:ref>. Therefore, we speculated that a potential link between disordered gut physiology and fluoride neurotoxicity could be utilized in preventing fluoride-induced memory dysfunction. However, little evidence has clearly demonstrated this relationship, and no probiotic has been proven effective on preventing fluoride-induced dysfunctions in the brain.</ns0:p><ns0:p>Therefore, the present study aimed to assess whether fluoride could alter gut physiology. Lactobacillus johnsonii BS15 was used to revert the altered intestinal physiology to further reveal whether fluoride neurotoxicity is associated with intestinal physiology and assess whether BS15 could be an effective strategy to control fluoride-induced memory dysfunction and hippocampal injury. The hippocampus is critical for bacteria-cognition link, as well as learning and memory <ns0:ref type='bibr' target='#b61'>(Stachenfeld, Botvinick & Gershman, 2017)</ns0:ref>, because of its lifetime synaptic plasticity and neurogenesis <ns0:ref type='bibr' target='#b28'>(Greenberg, Ziff & Greene, 1986;</ns0:ref><ns0:ref type='bibr' target='#b15'>Deisseroth & Tsien 2002;</ns0:ref><ns0:ref type='bibr' target='#b30'>Hong et al., 2008)</ns0:ref>, thus, changes in hippocampal chemistry were given more attention in this study. The expression of neuronal activity-regulated genes, such as the immediate-early gene c-fos, BDNF, and neuronal cell adhesion molecule (NCAM), play a critical role in synaptic plasticity <ns0:ref type='bibr' target='#b30'>(Hong et al., 2008;</ns0:ref><ns0:ref type='bibr' target='#b58'>Simpson & Morris 2000)</ns0:ref>. Neurotrophin-and activity-dependent gene expression is mediated by cAMP/Ca 2+ responsive element-binding protein (CREB) <ns0:ref type='bibr' target='#b46'>(Mantamadiotis et al., 2002)</ns0:ref>. Moreover, BDNF, CREB, NCAM, and stem cell factor (SCF) are essential for neurogenesis.</ns0:p></ns0:div>
<ns0:div><ns0:head>Materials & Methods</ns0:head></ns0:div>
<ns0:div><ns0:head>Materials and methods</ns0:head></ns0:div>
<ns0:div><ns0:head>Culture and treatment with BS15</ns0:head><ns0:p>L. johnsonii BS15 (CCTTCC M2013663) was isolated from homemade yogurt collected from Hongyuan Prairie, Aba Autonomous Prefecture, China. Our previous study demonstrated that BS15 can effectively prevent non-alcoholic fatty liver disease by attenuating mitochondrial lesion and inflammation in the liver, lowering intestinal permeability, and adjusting gut microbiota <ns0:ref type='bibr' target='#b68'>(Xin et al., 2014)</ns0:ref>. Thus, BS15 was selected as the potential treatment strategy to improve gut flora in the present study. L. johnsonii BS15 was cultured anaerobically in de Man-Rogosa-Sharpe broth (Qingdao Rishui Bio-technologies Co., Ltd., Qingdao, China) at 37 °C. The amounts of bacterial cells were evaluated by heterotrophic plate count. Briefly, the cultures were centrifuged, washed, and resuspended in phosphate buffered saline (PBS; pH 7.0) for experimental use. The concentration of BS15 suspension was 1×10 9 cfu/mL (daily consumption dose: 0.2 mL/mice).</ns0:p></ns0:div>
<ns0:div><ns0:head>Behavioral tests</ns0:head></ns0:div>
<ns0:div><ns0:head>Novel object recognition (NOR) test</ns0:head><ns0:p>The NOR test is a widely used method for the investigation of working memory alteration. The results of NOR test reflects the function of the hippocampus based on the nature propensity of mice to a novel object rather than a familiar one. The task procedure consists three phases <ns0:ref type='bibr' target='#b0'>(Antunes & Biala, 2012)</ns0:ref>: habituation, familiarization, and test phase. Briefly, in the habituation phase, each mouse was allowed to freely explore the open-field arena (40×40×45 cm, l×b×h) for 1 h in the absence of objects. The mouse was then removed from the arena and placed in its home cage. During the familiarization phase, each mouse was placed in the arena to freely explore two different objects (#A+#B) for 5 min. The two objects were placed in the opposite corners of the cage. The mouse was given an intermediate retention interval of 20 min and then returned to the arena and re-exposed to object B along with a completely new object (object #C, distinguishable from object #A). Exploration ratio (F#C / (F#C + F#B) × 100, where F#C = frequency of exploring object #C, and F#B = frequency of exploring object #B) was calculated to assess memory. The objects used included a green bottle cap (#A), an orange bottle cap (#B), and a small smooth stone (#C).</ns0:p></ns0:div>
<ns0:div><ns0:head>T-maze test</ns0:head><ns0:p>An enclosed T-maze, which is an equipment with 10 cm-wide floor and 20 cm-high walls in the form of a 'T', was placed horizontally. The stems of two goal arms and a start arm were 30 cm long. A central partition in the middle of the two goal arms extended into the start arm (7 cm). Every arm had a guillotine door. The equipment and the operating steps were consistent with those of <ns0:ref type='bibr' target='#b14'>Deacon and Rawlins (2006)</ns0:ref>. First, the central partition was put in the T-maze with all doors open. Then, each mouse was placed in the start area directly from its home cage and allowed to choose the left or right arm. The mouse was kept in the chosen arm by quietly sliding the door down. After 30 s, the mouse and central partition were removed, and the mouse was placed back into its holding cage. After a retention interval of 1 min, the mouse was placed back into the start area for a second trial with all doors open. Each mouse was given ten trials over five days and allowed to explore the maze before sated. The trial was marked as 'correct' if the mouse chooses the other goal arm in consecutive trials. Each exploration should take no more than 2 min.</ns0:p></ns0:div>
<ns0:div><ns0:head>Establishment of animal model and study design</ns0:head><ns0:p>Forty-eight male ICR mice (3 week-old, Dashuo Biological Institute, Chengdu, Sichuan, China) were fed with normal chow diet (Dashuo Biological Institute, Chengdu, Sichuan, China) for 1 week to acclimatize to the new environment. After the adaptation period, the mice were equally and randomly divided into three groups and administered with either 0.2 mL of PBS (control group, F group) or BS15 (prob group) every day by gavage throughout a 98-day experimental period. Animals in the F and prob groups were exposed to 100 mg/L fluoride in drinking water from the 28th day to the 98th day. The mice were housed in an animal facility with a humidity of 40%-60%, a temperature of 20-22 °C and a 12-h light/dark cycle (lights off at 6:00 a.m. and on at 6:00 p.m.). We housed five or six mice per cage bedded by wood shavings and with food and water available ad libitum. The wood shavings were replaced every 2 days. Drinking water was replaced and water bottles were washed every 5 days. All animal experiments were performed according to the guidelines for the care and use of laboratory animals approved by the Institutional Animal Care and Use Committee of Sichuan Agricultural University (Approval number: SYXKchuan2019-187). Ten mice from each group were selected for behavioral test on the 98th day of the experiment. The other six mice of each group were sacrificed by cervical dislocation to collect tissues. The behavior test and sampling were carried out from 7:00 a.m. to 11:30 a.m.</ns0:p><ns0:p>Blood was sampled from the mice orbit, and the serum was separated by incubation at 4 °C for 30 min followed by centrifugation at 2,000×g for 20 min and stored at −30 °C. Tissues from the left hippocampus and partial ileum were removed and washed with ice-cold sterilized saline and then immediately frozen in liquid nitrogen for gene expression analysis. Tissues from the right hippocampus and partial ileum were ground (pH 7.4) into 5% and 10% homogenates, respectively, with PBS and then centrifuged at 12,000×g for 5 min at 4 °C. The obtained supernatant was stored at −80 °C for further detection.</ns0:p></ns0:div>
<ns0:div><ns0:head>Biochemical evaluation</ns0:head><ns0:p>The contents of corticosterone, D-lactate, and diamine oxidase (DAO) in the serum; inflammatory cytokines in the supernatant of hippocampal and ileal homogenates; and the apoptosis-regulated proteins in the supernatant of the hippocampal homogenate were measured by commercial enzyme-linked immunosorbent assay (ELISA) kit (Enzyme-linked Biotechnology Co., Ltd., Shanghai, China) specific for mice. The inflammatory cytokines included tumor necrosis factor-alpha (TNF-α), interferon-gamma (IFN-γ), interleukin-1β (IL-1β), IL-6, and IL-10 (only detected in ileum tissue). The operation was performed strictly according to the manufacturer's instructions.</ns0:p></ns0:div>
<ns0:div><ns0:head>Real-time quantitative polymerase chain reaction (RT-qPCR) analysis of gene expression</ns0:head><ns0:p>Total hippocampal RNA and ileal RNA were isolated using E.Z.N.A.® Total RNA Kit (OMEGA Bio-Tek, Doraville, GA, USA) according to the manufacturer's instructions. The isolated RNA was assessed for the ratio of absorbances at 260 and 280 nm and by agarose gel electrophoresis for quantitative and qualitative analyses. The isolated RNA was transcribed into first-strand complementary DNA (cDNA) with PrimeScript RT reagent kit with gDNA Eraser (Thermo Scientific, Waltham, Massachusetts, USA) according to the manufacturer's instructions. The cDNA products were stored in −80 °C for further use. qPCR was performed using CFX96 RT PCR Detection System (Bio-Rad, Hercules, CA, USA) and SYBR Premix Ex TaqTM PCR Kit (Bio-Rad, Hercules CA, USA) to quantify the relative expression levels of neuroplasticity-related factors (BDNF, CREB, SCF, and NCAM), early gene (c-fos), molecular proteins related to myelin structure (myelin oligodendrocyte glycoprotein <ns0:ref type='bibr'>[MOG]</ns0:ref></ns0:p></ns0:div>
<ns0:div><ns0:head>, proteolipid protein [PLP], myelin basic protein [MBP], and myelin-associated glycoprotein [MAG]), and apoptosis-related proteins (B-cell lymphoma-2 [Bcl-2], B-cell lymphoma-extra large [Bcl-xl],</ns0:head><ns0:p>Bcl-2-associated X protein [Bax], Bcl-xl/Bcl-2-asociated death promoter [Bad], caspase-9, and caspase-3) and cytokines (IFN-γ, TNF-α, in the hippocampus tissue and cytokines and tight junction (TJ) proteins (zonula occludens protein 1 [ZO-1], claudin-1, and occludin) in ileum tissue with 10 µL total reaction volume. The thermocycle protocol was performed as follows: 5 min at 95 °C, followed by 40 cycles of 10 s denaturation at 95 °C, and 30 s annealing/extension at optimum temperature (Table <ns0:ref type='table'>1</ns0:ref>). A final melting curve analysis was performed to monitor the purity of the PCR product. β-actin was used as reference gene to normalize the relative mRNA expression levels of target genes with values presented as 2 −ΔΔCq . The primer sequences and optimum annealing temperatures are shown in Table <ns0:ref type='table'>1</ns0:ref>.</ns0:p></ns0:div>
<ns0:div><ns0:head>Immunohistochemistry</ns0:head><ns0:p>A subset of mice in each group was sacrificed, and their brain was removed, fixed in 4% paraformaldehyde solution, and stored in 4 °C for immunohistochemical assay. The brain tissues were embedded by paraffin and cut by microtome. Slices were submerged in citrate antigen retrieval solution and heated on medium until boiling using a microwave (model: P70D20TL-P4; Galanz, Guangdong, China). The fire was ceased, and the tissues were kept warm for 8 min. Then, the tissues were heated at medium-low heat for 7 min. After free cooling, the slices were placed into PBS (pH 7.4) and shaken for 5 min for decoloration, which was repeated three times. Then, the sections were incubated in 3% oxydol for 25 min at room temperature and away from light for blocking endogenous peroxidase. The slices were washed three times in PBS by shaking for 5 min, then sealed for 30 min by 3% bull serum albumin, and incubated with monoclonal rabbit anti-BDNF (1:400) or polyclonal rabbit anti-CREB (1:500) at 4 °C overnight. Speciesspecific biotinylated anti-rabbit immunoglobulin (horseradish peroxidase-labeled) was used for immuno-detection. Following the second antibody incubation, 3,3′-diaminobenzidine staining kit was used to complete the reaction according to the manufacturer's instructions. Hematoxylin stain was performed to re-stain the nucleus. BDNF and CREB were quantified by calculating their integral optical density (IOD) in the object region (ImageJ, National Institutes of Health, USA). The average optical density (IOD/object region areas) was calculated, and the results were presented as levels of expression.</ns0:p></ns0:div>
<ns0:div><ns0:head>Statistics</ns0:head><ns0:p>Results were expressed as mean ± standard deviation. One-way ANOVA was performed between different groups with IBM SPSS Statistics 25 (IBM Corporation). Differences of P<0.05 were considered statistically significant. The figures were plotted using GraphPad Prism version 7.04 (San Diego, CA, USA).</ns0:p></ns0:div>
<ns0:div><ns0:head>Results</ns0:head></ns0:div>
<ns0:div><ns0:head>Behavioral results</ns0:head><ns0:p>Figure <ns0:ref type='figure' target='#fig_0'>1A</ns0:ref> shows that the F group showed a lower spontaneous exploration (control vs. F, P<0.01; F vs. prob, P=0.019) than the other two groups, but no difference (P>0.05) was found between the control and prob groups (P=0.104). The exploration ratio of the three groups were not significantly different (control vs. F, P=0.125; control vs. prob, P=0.285; F vs. prob, P=0.626; Fig. <ns0:ref type='figure' target='#fig_0'>1B</ns0:ref>).</ns0:p></ns0:div>
<ns0:div><ns0:head>mRNA and protein expressions of BDNF in the hippocampus</ns0:head><ns0:p>Figure <ns0:ref type='figure'>2A</ns0:ref> shows that the F group presented a significant decrease in BDNF mRNA level than the other two groups (control vs. F, P<0.01; F vs. prob, P=0.014), whereas the control and prob groups had no difference (P=0.414). As shown in Fig. <ns0:ref type='figure'>2B-E</ns0:ref>, the BDNF protein level was decreased in the F group compared with the control (P<0.01) and prob groups (P=0.018). No difference was observed between the control and prob groups (P=0.513).</ns0:p></ns0:div>
<ns0:div><ns0:head>mRNA and protein expressions of CREB in the hippocampus</ns0:head><ns0:p>As shown in Fig. <ns0:ref type='figure'>3A</ns0:ref>, CREB mRNA level was slightly reduced in the F group (P=0.484) and increased in the prob group (P=0.065) compared with the control group. The CREB mRNA level in the prob group was significantly (P=0.026) higher than that in the F group. Figure <ns0:ref type='figure'>3B-E</ns0:ref> shows that the F group presented a significant decrease in CREB protein level than the other two groups (control vs. F, P=0.048; F vs. prob, P<0.01). CREB protein level in the prob group was remarkably higher than that in the control group (P=0.032).</ns0:p></ns0:div>
<ns0:div><ns0:head>mRNA expressions of NCAM, SCF, and c-fos in the hippocampus</ns0:head><ns0:p>As shown in Fig. <ns0:ref type='figure' target='#fig_1'>4A</ns0:ref>, the control group had a higher NCAM mRNA level than the other two groups (control vs. F, P=0.001; control vs. prob, P=0.001), but no difference was found between the F and prob groups (P=0.783). Figure <ns0:ref type='figure' target='#fig_1'>4B</ns0:ref> shows that the SCF mRNA level of the three groups were not significantly different (control vs. F, P=0.105; control vs. prob, P=0.842; F vs. prob, P=0.181). As shown in Fig. <ns0:ref type='figure' target='#fig_1'>4C</ns0:ref>, the control group presented a higher mRNA expression of c-fos than the other two groups (control vs. F, P<0.01; control vs. prob, P<0.01), but the expression of c-fos between the other two groups were not significantly different (F vs. prob, P=0.449).</ns0:p></ns0:div>
<ns0:div><ns0:head>Hippocampal inflammation</ns0:head><ns0:p>Figures 5A and 5D show that the TNF-α (mRNA level: control vs. F, P=0.013; F vs. prob, P=0.252; protein level: control vs. F, P<0.001; F vs. prob, P=0.001) and IFN-γ (mRNA level: control vs. F, P=0.007; F vs. prob, P=0.037; protein level: control vs. F, P<0.001; F vs. prob, P<0.001) in the F group were significantly or slightly higher than those in the other two groups in mRNA and protein levels. The TNF-α of the prob group was significantly increased in the protein level (P<0.001) but not in the mRNA level (P=0.115) compared with the control group (Fig. <ns0:ref type='figure' target='#fig_2'>5A</ns0:ref>). No difference in IFN-γ (Fig. <ns0:ref type='figure' target='#fig_2'>5D</ns0:ref>) was observed between the control and prob groups in the mRNA (P=0.381) and protein levels (P=0.931). Figure <ns0:ref type='figure' target='#fig_2'>5C</ns0:ref> reveals that the F group had a remarkably lesser IL-6 than the other two groups in mRNA (control vs. F, P<0.001; F vs. prob, P=0.011) and protein levels (control vs. F, P<0.001; F vs. prob, P<0.001). As shown in Fig. <ns0:ref type='figure' target='#fig_2'>5C</ns0:ref>, the prob group presented a significantly lesser IL-6 than the control group in mRNA (P<0.001) and protein levels (P<0.001). The three groups had no remarkable change in IL-1β in the mRNA and protein levels (mRNA level: control vs. F, P=0.089; control vs. prob, P=0.097; F vs. prob, P=0.962; protein level: control vs. F, P=0.420; control vs. prob, P=0.733; F vs. prob, P=0.636). IL-10 did not reach the detection threshold (Fig. <ns0:ref type='figure' target='#fig_2'>5B</ns0:ref>).</ns0:p></ns0:div>
<ns0:div><ns0:head>Hippocampal myelin and apoptosis-related proteins</ns0:head><ns0:p>Fig. <ns0:ref type='figure' target='#fig_3'>6A</ns0:ref> shows the clear decreasing trend in the mRNA level of PLP in the F group compared with the other two groups (control vs. F, P<0.013; F vs. prob, P=0.0039). A difference (control vs.prob, P=0.568) in the mRNA level of PLP between the control and prob groups was not detected. Figure <ns0:ref type='figure' target='#fig_3'>6B</ns0:ref> shows that the control group had a higher MOG mRNA level than the F and prob group (control vs. F, P=0.005; control vs. prob, P=0.002), whereas the MOG mRNA level of the F group was not different (P=0.778) from that of the prob group. Differences in the mRNA levels of MBP (control vs. F, P=0.277; control vs.prob, P=0.706; F vs. prob, P=0.415; Fig. <ns0:ref type='figure' target='#fig_3'>6C</ns0:ref>) and MAG (control vs. F, P=0.904; control vs. prob, P=0.919; F vs. prob, P=0.970; Fig. <ns0:ref type='figure' target='#fig_3'>6D</ns0:ref>) were not observed among the three groups. Figures <ns0:ref type='figure' target='#fig_4'>7A and 7E</ns0:ref> present a remarkable (control vs. F, P<0.026; F vs. prob, P=0.038) decrease in Bcl-2 mRNA level and a significant (control vs. F, P<0.033; F vs. prob, P=0.001) increase in caspase-3 mRNA level in F group compared with the other two groups, and the Bcl-2 (P=0.948) and caspase-3 mRNA levels (P=0.127) of the control and prob groups were not significantly different. Figures <ns0:ref type='figure' target='#fig_4'>7B, 7C</ns0:ref>, 7D, and 7F reveal no significant differences (P>0.05) among the three group in the mRNA levels of Bcl-xl (control vs. F, P=0.290; control vs. prob, P=0.463; F vs. prob, P=0.735), Bax (control vs. F, P=0.784; control vs. prob, P=0.681; F vs. prob, P=0.891), Bad (control vs. F, P=0.954; control vs. prob, P=0.137; F vs. prob, P=0.124), and caspase-9 (control vs. F, P=0.128; control vs. prob, P=0.684; F vs. prob, P=0.197).</ns0:p></ns0:div>
<ns0:div><ns0:head>Inflammatory factors in the ileum</ns0:head><ns0:p>As shown in Fig. <ns0:ref type='figure' target='#fig_5'>8A</ns0:ref>, the F group exhibited a slight increase in TNF-α mRNA level (P=0.079) and a significantly enhanced TNF-α protein level (P<0.001) compared with the control group. TNF-α in the mRNA (P=0.008) and protein levels (P<0.001) were significantly reduced in the prob group compared with the F group. Differences in the mRNA (P=0.197) and protein levels (P=0.348) of TNF-α were not observed between the control and prob groups. The mRNA (P=0.888) and protein levels (P=0.781) of IL-1β (Fig. <ns0:ref type='figure' target='#fig_5'>8B</ns0:ref>) between the F and prob groups were not significantly different. IL-1β (Fig. <ns0:ref type='figure' target='#fig_5'>8B</ns0:ref>) in the control group showed a slight decrease in mRNA level (control vs. F, P=0.132; control vs. prob, P=0.103) and a remarkable (control vs. F, P=0.003; control vs. prob, P=0.002) decline in protein level compared with those in the other two groups (Fig. <ns0:ref type='figure' target='#fig_5'>8B</ns0:ref>). No difference was observed in IL-6 (Fig. <ns0:ref type='figure' target='#fig_5'>8C</ns0:ref>) in mRNA (control vs. F, P=0.635; control vs. prob, P=0.615; F vs. prob, P=0.975) and protein levels (control vs. F, P=0.407; control vs. prob, P=0.908; F vs. prob, P=0.347) among the three groups. The F group exhibited a significantly increased IFN-γ (mRNA level: control vs. F, P<0.001; F vs. prob, P<0.001; protein level: control vs. F, P<0.001; F vs. prob, P<0.001; Fig. <ns0:ref type='figure' target='#fig_5'>8D</ns0:ref>) and a sharply decreased IL-10 (mRNA level: control vs. F, P=0.01; F vs. prob, P=0.02; protein level: control vs. F, P<0.001; F vs. prob, P<0.001; Fig. <ns0:ref type='figure' target='#fig_5'>8E</ns0:ref>) compared with the other two groups in mRNA and protein levels. These differences (IFN-γ mRNA level: control vs. prob, P=0.062; IL-10 mRNA level: control vs. prob, P=0.838; IL-10 protein level: control vs. prob, P=0.535) were not observed in the other two groups except for IFN-γ protein level (control vs. prob, P=0.005).</ns0:p></ns0:div>
<ns0:div><ns0:head>Intestinal permeability</ns0:head><ns0:p>As shown in Fig. <ns0:ref type='figure'>9A</ns0:ref>, the control group reported a slightly higher claudin-1 than the other two groups <ns0:ref type='bibr'>(control vs. F, P=0.111; control vs. prob, P=0.129)</ns0:ref>. The mRNA expression of claudin-1 between the F and prob groups had no difference <ns0:ref type='bibr'>(F vs. prob, P=0.855)</ns0:ref>. Figure 9A also shows that ZO-1 and occludin in the control group were remarkably higher than those in the F group (P=0.008, P=0.004) and slightly (P=0.121, P=0.061) higher than the prob group. Figure <ns0:ref type='figure'>9A</ns0:ref> shows that the prob group presented a slightly more ZO-1 (P=0.125) and occludin (P=0.114) than the F group. Figure <ns0:ref type='figure'>9B</ns0:ref> and9 C show that the F group exhibited a significant increase in serum DAO activity (control vs. F, P=0.002; F vs. prob, P=0.048) and D-lactate concentration (control vs. F, P<0.001; F vs. prob, P=0.011) compared with the other two groups. No significant (P=0.128) difference in the DAO activity between the control and prob groups was observed. The D-lactate concentration in the prob group was significantly (P=0.001) higher than that in the control group.</ns0:p></ns0:div>
<ns0:div><ns0:head>Discussion</ns0:head><ns0:p>Although a large number of studies have found fluoride-induced brain lesions, few studies focused on the link between the gut changes and neurotoxicity induced by fluoride. In view of the gut-brain axis, changes in the intestinal microenvironment, including gut microbiota, inflammatory cytokines, and hormones, can influence brain chemistry and behaviors. The current study further explored fluoride-induced brain lesion and its link with the gut. We found that the mice exposed to 100 mg/L sodium fluoride for 10 weeks had hippocampal lesions, which caused memory impairment as indicated by their lower performance in the T-maze test and the reduced PLP mRNA level. Moreover, high fluoride exposure caused intestinal inflammation and increased the intestinal permeability as indicated by the increased inflammatory cytokines, serum DAO activity, and D-lactate content and reduced mRNA levels of TJ protein. BS15, a probiotic capable of improving the gut microbiome, reversed the gut changes and alleviated the brain lesion and memory impairment induced by fluoride. The current study confirmed our hypothesis that gut changes may play a key role in memory dysfunction during high fluoride exposure, and BS15 administration is a potential method of preventing these fluoride-induced damages on memory ability.</ns0:p><ns0:p>Although the memory abilities of the three groups were not remarkably different as shown in the NOR test, the fluoride-infected mice in the F group showed memory impairment as indicated by their lower spontaneous exploration in the T-maze test compared with the other groups. The poor performances of the fluoride-infected mice in various memory-related behavioral tests were also reported in previous studies <ns0:ref type='bibr' target='#b39'>(Liu et al., 2010;</ns0:ref><ns0:ref type='bibr' target='#b6'>Chen et al., 2018)</ns0:ref>. In our study, BS15 substantially reversed the performance of mice in the T-maze and thus has a beneficial effect on the memory ability of fluoride-infected mice. Neuronal activation was also decreased as indicated by the lower mRNA level of c-fos in the hippocampus of fluoride-infected mice. <ns0:ref type='bibr' target='#b19'>Fleischmann et al. (2003)</ns0:ref> demonstrated that mice lacking c-fos exhibit remarkable memory deficits; hence, c-fos gene has a critical role in memory. The hippocampus is a unique area of the brain that is able of neuroplasticity <ns0:ref type='bibr' target='#b25'>(Galea et al., 2013)</ns0:ref>. Neuroplasticity usually increases under the conditions that increase memory abilities, and its ablation often induces lesions that affect memory <ns0:ref type='bibr' target='#b16'>(Deng, Aimone & Gage, 2010;</ns0:ref><ns0:ref type='bibr' target='#b38'>Leuner & Gould, 2010)</ns0:ref>. These findings suggested that the hippocampus play a critical role in memory and synaptic plasticity <ns0:ref type='bibr' target='#b73'>(Yirmiya & Goshen, 2011)</ns0:ref>. Neurotrophic factors, especially the BDNF, are important in neurogenesis, and their production are involved in almost every aspect of neural and behavioral plasticity <ns0:ref type='bibr' target='#b41'>(Lu, Christian & Lu, 2008;</ns0:ref><ns0:ref type='bibr' target='#b40'>Li et al., 2008)</ns0:ref>, especially for hippocampal-dependent memory <ns0:ref type='bibr' target='#b29'>(Heldt et al., 2007)</ns0:ref>. An increasing body of data indicated that probiotic consumption can regulate anxiety and memory functions via changes in the expression of key components, such as BDNF, CREB, and N-methyl-d-aspartate receptors <ns0:ref type='bibr' target='#b10'>(Clarke et al., 2013;</ns0:ref><ns0:ref type='bibr' target='#b64'>Wall et al., 2010)</ns0:ref>. The underlying mechanism is that probiotics can synthesize and recognize an array of neurochemicals, including neurotransmitters, secondary bile acids, neuroactive short chain fatty acids (SCFAs), and other biologically active small molecules. For example, <ns0:ref type='bibr' target='#b37'>Kumar et al. (2017)</ns0:ref> reported that L. johnsonii could increase the concentrations of acetate and butyrate in feces. Butyrate, an SCFA, can decrease BDNF methylation and consequently cause an overexpression of BDNF by decreasing ten-eleven translocation methylcytosine dioxygenase 1, which is the enzyme responsible for catalyzing the conversion of DNA methylation to hydroxymethylation <ns0:ref type='bibr' target='#b65'>(Wei et al., 2015)</ns0:ref>. In the present study, the expression of BDNF in the hippocampus was reduced by fluoride. This finding is consistent with the result found by <ns0:ref type='bibr' target='#b49'>Niu et al. (2018)</ns0:ref> CREB is the transcriptional regulator of BDNF, and similar genomic network analysis reported that CREB is the center of Alzheimer's disease's pathology <ns0:ref type='bibr' target='#b34'>(Jeong et al., 2001)</ns0:ref>. The mRNA and protein levels of BDNF and CREB in the BS15treated mice were substantially increased compared with the fluoride-infected mice, and the level of CREB was even slightly higher than the control group. Similarly, Kadry and Megeed <ns0:ref type='bibr' target='#b9'>(2018)</ns0:ref> found that Lactobacillus could effectively inhibit the reduction of BDNF induced by cadmium chloride in the hippocampus of mice. Moreover, Corpuz et al. <ns0:ref type='bibr' target='#b9'>(2018)</ns0:ref> reported that Lactobacillus paracasei K71 prevents age-related cognitive decline in senescence-accelerated mouse prone 8 by increasing the protein expression of BDNF and the phosphorylation of CREB in the hippocampus. Neuroplasticity also needs the regulation of NCAM <ns0:ref type='bibr' target='#b57'>(Seidenfaden, Krauter & Hildebrandt, 2006)</ns0:ref> and the stimulation of SCF <ns0:ref type='bibr' target='#b35'>(Jin et al., 2002)</ns0:ref>. In line with the results of Niu et al. <ns0:ref type='bibr' target='#b9'>(2018)</ns0:ref>, fluoride significantly reduced the mRNA level of NCAM.</ns0:p><ns0:p>Many evidence indicated that neuroinflammation may be one of the pathogenesis underlying cognitive changes and the development of many neurodegenerative diseases <ns0:ref type='bibr' target='#b22'>(Frank-Cannon et al., 2009)</ns0:ref>. The hippocampus is vulnerable to the insults by inflammatory cytokines because of its high expression of cytokine receptors <ns0:ref type='bibr' target='#b12'>(Das & Research, 2008)</ns0:ref>. Few in vivo studies focused on the neuroinflammation caused by fluorosis. Excessive exposure to fluoride triggered neuroinflammation as indicated by the increase in the TNF-α and IFN-γ in the hippocampus of the F group. The result on TNF-α was in agreement with the finding of <ns0:ref type='bibr' target='#b71'>Yan et al. (2016)</ns0:ref>, who found that TNF-α immunoreactivity is increased in the hippocampus of rat exposed to 120 ppm fluoride for 10 weeks. However, by contrast to the results shown by <ns0:ref type='bibr' target='#b71'>Yan et al. (2016)</ns0:ref>, the present study reported a reduced IL-6 in the hippocampus of mice exposed to 100 mg/L sodium fluoride for 10 weeks. Differences in drug doses, delivery cycles, and breeds of rodents may be associated with the contradictory results. A previous research confirmed that IL-6-deficient mice present a weakened neuroprotection of the hippocampus <ns0:ref type='bibr' target='#b33'>(Jean Harry, Bruccoleri & Lefebvre d'Hellencourt, 2003)</ns0:ref>. <ns0:ref type='bibr' target='#b24'>Funk et al. (2011)</ns0:ref> indicated the potential role of IL-6 in modulating TNFα-mediated neurotoxicity. Intestinal microbiome and some probiotics can influence health status and disease risk by activating immune response against dangerous stimuli and activating regulatory mechanisms to avoid uncontrolled inflammation. Intestinal permeability and potentially beneficial metabolites may be the underlying mechanisms of the anti-inflammation. Intestinal microbiome can ferment dietary fiber and starch in the large intestines and produce SCFAs <ns0:ref type='bibr' target='#b7'>(Chen, Faller & Spanjaard , 2003)</ns0:ref>. The effect of butyrate and other SCFAs on preventing inflammation in colon diseases and different neural inflammation models in cell cultures have been demonstrated <ns0:ref type='bibr' target='#b31'>(Huuskonen, 2004)</ns0:ref>. In the present study, BS15 administration had profitable effect on balancing the inflammatory cytokines in fluoride-infected mice.</ns0:p><ns0:p>The mRNA expression levels of myelin-and apoptosis-related proteins were also detected to further investigate the effect of BS15 on the hippocampal impairment. Myelin sheaths enwrap the nerve fiber to guarantee interneuronal transmission efficiency <ns0:ref type='bibr' target='#b48'>(Nguyen et al., 2009)</ns0:ref>. Myelin is consisted of PLP (a transmembrane protein), MBP (a peripheral membrane protein), the outermost MOG, and the innermost MAG <ns0:ref type='bibr' target='#b49'>(Niu et al., 2018)</ns0:ref>. The remarkably reduced mRNA expression levels of PLP and MOG in the F group suggested that myelin lesion occurred in the hippocampus. The changes in PLP induced by fluoride were consistent with the findings of Niu et al. <ns0:ref type='bibr' target='#b9'>(2018)</ns0:ref>. The reduced tendency of PLP was inhibited by BS15; hence, BS15 may have a protective effect on the myelin. A previous study demonstrated that mice present increased positive apoptotic neurons following 10 weeks of exposure to 120 ppm fluoride in drinking water <ns0:ref type='bibr' target='#b71'>(Yan et al., 2016)</ns0:ref>. In the present study, the reduced Bcl-2 (anti-apoptosis protein) and increased caspase-3 (pro-apoptosis protein) in fluoride-infected mice created conditions for apoptotic neurons, and these changes were remarkably reversed by BS15 treatment.</ns0:p><ns0:p>Intestinal leakage can facilitate the translocation of bacterial composition, such as microorganisms and their products <ns0:ref type='bibr' target='#b5'>(Carvalho AF, Berk M & Maes M 2016)</ns0:ref>, and is considered a key factor in mental disease <ns0:ref type='bibr' target='#b4'>(Braniste et al., 2014;</ns0:ref><ns0:ref type='bibr' target='#b74'>Zhan et al., 2016;</ns0:ref><ns0:ref type='bibr' target='#b17'>Emery et al., 2017)</ns0:ref>. Inflammation can enhance epithelial permeability <ns0:ref type='bibr' target='#b69'>(Xue et al., 2014;</ns0:ref><ns0:ref type='bibr' target='#b55'>Schulzke et al., 2009)</ns0:ref>. Inflammatory cytokines, such as IL-1β, TNF-α, and IFN-γ, can increase gut permeability <ns0:ref type='bibr' target='#b44'>(Ma et al., 2004;</ns0:ref><ns0:ref type='bibr' target='#b55'>Schulzke et al., 2009;</ns0:ref><ns0:ref type='bibr' target='#b66'>Weber et al., 2010)</ns0:ref>. IL-10, an anti-inflammatory cytokine, plays a critical role in the homeostasis of the gut, which was illustrated by the finding that spontaneous colitis occurs in IL-10 −/− mice <ns0:ref type='bibr' target='#b27'>(Gomes-Santos et al., 2012)</ns0:ref>. The current study found that excessive fluoride intake resulted in intestinal inflammation by increasing pro-inflammatory cytokines (TNF-α, IL-1β, and IFN-γ) and reducing anti-inflammatory cytokine (IL-10). Treatment with BS15 efficiently lowered the inflammatory reaction caused by fluoride. TJ proteins act as a barrier that mediates the cell-to-cell adhesion and prevents molecules from crossing through the epithelial sheet between adjacent cells into systemic circulation <ns0:ref type='bibr' target='#b51'>(Piche, 2014)</ns0:ref>. The mRNA level of two TJ proteins, namely, ZO-1 and occludin, in the ileum of the fluoride-infected mice were also remarkably reduced with gut inflammation enhancement and therefore led to higher levels of DAO activity and D-lactate content in the serum. The tissue of the small intestine contains the highest DAO activity, and serum DAO is derived primarily from the small intestines in many mammalian species <ns0:ref type='bibr' target='#b42'>(Luk, Bayless & Baylin, 1980)</ns0:ref>. Moreover, mammalian species cannot produce D-lactate, and the main source of D-lactate is from the commensal bacteria in the gastrointestinal tract <ns0:ref type='bibr' target='#b60'>(Sun et al, 2001)</ns0:ref>. The metabolism of serum Dlactate is very slow. The increases in serum DAO activity and D-lactate content occurred when the intestinal mucosal integrity was damaged and served as useful plasma markers of mucosal integrity <ns0:ref type='bibr' target='#b18'>(Ewaschuk, Naylor & Zello, 2005;</ns0:ref><ns0:ref type='bibr' target='#b42'>Luk, Bayless & Baylin, 1980)</ns0:ref>. In this study, we found that BS15 effectively improved intestinal permeability as shown by the remarkably lower serum DAO activity and D-lactate concentration in the prob group compared with the F group. The result may be explained in part by the slightly increased TJ proteins in the prob group. Apoptosis is another possible reason that may have caused barrier dysfunction <ns0:ref type='bibr' target='#b55'>(Schulzke et al., 2009)</ns0:ref>. These results suggested that fluoride could cause intestinal inflammation and damage mucosal integrity, which results in enhanced intestinal permeability, and BS15 administration could alleviate these pathological changes.</ns0:p></ns0:div>
<ns0:div><ns0:head>Conclusions</ns0:head><ns0:p>The study deepened our understanding of the link between fluoride neurotoxicity on memory function and gut microenvironment. BS15 exerted beneficial effects against excessive fluoride intake-induced memory impairment, related neural inflammation, and demyelination by improving intestinal inflammation and integrity and increasing apoptosis markers in the hippocampus of mice. Manuscript to be reviewed Manuscript to be reviewed </ns0:p><ns0:note type='other'>Figure 3</ns0:note></ns0:div><ns0:figure xml:id='fig_0'><ns0:head>Figure 1</ns0:head><ns0:label>1</ns0:label><ns0:figDesc>Figure 1</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_1'><ns0:head>Figure 4 mRNA</ns0:head><ns0:label>4</ns0:label><ns0:figDesc>Figure 4</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_2'><ns0:head>Figure 5 mRNA</ns0:head><ns0:label>5</ns0:label><ns0:figDesc>Figure 5</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_3'><ns0:head>Figure 6 mRNA</ns0:head><ns0:label>6</ns0:label><ns0:figDesc>Figure 6</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_4'><ns0:head>Figure 7 mRNA</ns0:head><ns0:label>7</ns0:label><ns0:figDesc>Figure 7</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_5'><ns0:head>Figure 8 mRNA</ns0:head><ns0:label>8</ns0:label><ns0:figDesc>Figure 8</ns0:figDesc></ns0:figure>
<ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='28,42.52,303.41,525.00,309.75' type='bitmap' /></ns0:figure>
<ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='29,42.52,288.81,525.00,453.00' type='bitmap' /></ns0:figure>
<ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='36,42.52,280.87,525.00,142.50' type='bitmap' /></ns0:figure>
<ns0:note place='foot' n='2'>a The primer sequences of c-fos is designed by National Center for Biotechnology Information 3 (NCBI) and the referenced gene ID is 14281. PeerJ reviewing PDF | (2020:02:46283:3:0:CHECK 4 Sep 2020)</ns0:note>
</ns0:body>
" | "Dear Editor,
Thank you very much for your check on our manuscript entitled “Lactobacillus johnsonii BS15 improves intestinal environment against excessive flouride intake-induced memory impairment in mice——a study based on the gut-brain axis hypothesis”. According to your suggestions, we have carefully addressed all technical changes and comments. The following is a detailed list of response to all questions and changes the authors have made. If any question arises, please let us know.
Thank you very much for your consideration.
Sincerely yours,
Ni Xueqin
E-mail: xueqinni@foxmail.com
Replies to editor and reviewers:
1. The description in Result of Abstract is similar to discussion. Please revise it.
We have modified the abstract.
2. Figure-2b,-2c, and -2d are unclear. Please provide pictures with higher magnification.
We have provide pictures with higher magnification.
3. There is counterstaining in Figure 3. The authors didn’t explain how to decrease the effects of counterstaining on the optical density of CREB staining. In addition, similar to Figure 2, figure 3 is also unclear. Please provide pictures with higher magnification.
The color of counterstaining (blue) is significantly different with that of CREB-positive expression (brown). ImageJ and imagePro plus can exactly select an areas of interestig according to the same color standard. We also cite a reference as below to introduce the method of application[1]. In addition, we also provide figure 3 with higher magnification.
4. It is difficult to understand why the authors detected the mRNA and protein levels of certain molecules, but not for other molecules. Please add explanation.
We have already detected a number of parameters by RT-qPCR in our present study to support our hypothesis. From the tested parameters, BDNF, CREB and inflammatory cytokines were selected as representative to test for the protein levels by the method of immunohistochemistry. The time and effort are limited for us to complete all aspects of detections for the present paper. As this study is just part of our project to demonstrate flouride intake-induced memory impairment by gut-brain axis, indeed, more further work should be done in our further study. For the present work, we believe the results are enough to support our hypothesis as a preliminary study. We sincerely appreciate it for your precious comment.
1. Yeen, H., et al., 2015. Image-Pro Plus and ImageJ: Comparison and application in image analysis of biological tissues. Chinese Journal of Stereology and Image Analysis.
2. Mane, D R., et al., 2017. Validation of immunoexpression of tenascin-C in oral precancerous and cancerous tissues using ImageJ analysis with novel immunohistochemistry profiler plugin: An immunohistochemical quantitative analysis. J Oral Maxillofac Pathol 21(2): 211–217.
" | Here is a paper. Please give your review comments after reading it. |
9,765 | "<ns0:abstract xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:p>The size of the organs responsible for emitting and detecting sexual communication signals is a likely target for selection. Communication via bioluminescent signals in synchronous fireflies is a promising model to test hypotheses regarding differences between males and females in the effect of the size of signal emission and detection organs on fitness components. Synchronous firefly species congregate in large numbers during the mating season, displaying bioluminescent signals aimed at potential mates during relatively short nightly periods. Operational sex ratios are male-biased and, thus, the so-called typical sex roles (indiscriminate males and choosy females) are expected to evolve. We studied the synchronous firefly Photinus palaciosi, a species that during the mating season congregates in forests of central Mexico offering a magnificent natural show that attracts numerous tourists. P. palaciosi females have reduced wings (brachyptery) and cannot fly. Our field study tested the hypothesis that the male-biased operational sex ratio and the short daily mating period result in strong male-male competition that selects for males with larger lanterns and larger eyes, and against male mate choice, whereas female-female mate competition is absent and, thus, no selection on lantern or eye size is expected. Even though lantern, eye or body size do not predict the probability of being found in copula for either sex, sexual dimorphism in these features, along with allometric slopes of lantern size and assortative mating in terms of relative lantern size, support not only the hypothesis of intense sexual selection among males, but the possibility of subtle mechanisms of sexual selection among females. Trade-offs between investment in signaling (lanterns) versus detection (eyes) structures, or with pressures different from sexual selection such as those imposed by predators, are also likely to be important in shaping the evolution of sexual signaling in these fireflies.</ns0:p></ns0:div>
</ns0:abstract>
<ns0:body xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:head>INTRODUCTION</ns0:head><ns0:p>Communication between males and females is a fundamental element of the mating biology of most animals <ns0:ref type='bibr'>(Darwin, 1871;</ns0:ref><ns0:ref type='bibr'>Maynard Smith & Harper 2003;</ns0:ref><ns0:ref type='bibr'>Rosenthal, 2017)</ns0:ref>. There is a great variety of organs, newly evolved or specialized via sexual selection, for the emission and reception of sexual signals <ns0:ref type='bibr'>(Darwin, 1871;</ns0:ref><ns0:ref type='bibr'>Rosenthal, 2017;</ns0:ref><ns0:ref type='bibr'>Elgar et al., 2019)</ns0:ref>. A fascinating example of sexual communication involving vision is that of nocturnal fireflies <ns0:ref type='bibr'>(Lloyd, 1979;</ns0:ref><ns0:ref type='bibr'>Lewis, 2016)</ns0:ref>. In these insects, adults of many species possess an organ specialized for the emission of light known as lantern. Typically, males fly searching for females, emitting speciesand sex-specific flashing patterns that are involved in mate choice, while females emit glows or flashes in response <ns0:ref type='bibr'>(Lloyd, 1979;</ns0:ref><ns0:ref type='bibr'>Lewis & Cratsley, 2008;</ns0:ref><ns0:ref type='bibr'>Lewis, 2016;</ns0:ref><ns0:ref type='bibr'>Stanger-Hall et al., 2018)</ns0:ref>. If a successful dialogue is established, the male alights, contacts the female and a closerange courtship ensues <ns0:ref type='bibr'>(Lewis & Cratsley, 2008;</ns0:ref><ns0:ref type='bibr'>Stanger-Hall et al., 2018)</ns0:ref>.</ns0:p><ns0:p>Although female choice has been related to the duration and pulse pattern of male flashes <ns0:ref type='bibr'>(Lewis et al., 2004;</ns0:ref><ns0:ref type='bibr'>Lewis & Cartsley, 2008)</ns0:ref>, it is reasonable to propose that signal intensity, and thus the sizes of the lantern and of the eyes are also important traits influencing the efficiency of sexual communication in fireflies <ns0:ref type='bibr'>(Vencl & Carlson, 1998;</ns0:ref><ns0:ref type='bibr'>Crastley & Lewis, 2003</ns0:ref><ns0:ref type='bibr'>, 2005;</ns0:ref><ns0:ref type='bibr'>Demary et al., 2006;</ns0:ref><ns0:ref type='bibr'>Lau & Meyer-Rochow, 2006)</ns0:ref>. Larger lanterns may increase signal transmission distance <ns0:ref type='bibr'>(Demary et al., 2006)</ns0:ref>, whereas larger eyes are correlated with smaller interommatidial angles that may help improve visual resolution and distance perception <ns0:ref type='bibr'>(Lewis et al., 2004)</ns0:ref>, as well as capture more photons thus helping vision in low-light environments <ns0:ref type='bibr'>(Warrant & Dacke, 2011;</ns0:ref><ns0:ref type='bibr'>Stanger-Hall et al., 2018)</ns0:ref>. Somewhat surprisingly, studies on the relationship between signal-emission organ size and mating success in fireflies are scant and their results are inconsistent. While some studies in non-synchronic fireflies detected an effect of PeerJ reviewing PDF | (2020:04:47496:1:1:NEW 27 Aug 2020)</ns0:p><ns0:p>Manuscript to be reviewed male lantern size on mating success and female responses in a high-density population of Photinus pyralis <ns0:ref type='bibr'>(Vencl & Carlson, 1998)</ns0:ref> and in P. ignitus <ns0:ref type='bibr'>(Crastley & Lewis, 2003</ns0:ref><ns0:ref type='bibr'>, 2005)</ns0:ref>,</ns0:p><ns0:p>another study found that female mating decisions in the non-synchronic P. greeni are not influenced by lantern size <ns0:ref type='bibr'>(Demary et al., 2006)</ns0:ref>. The effects of signal-detection organ (eyes) size on fitness components of both sexes, and of signal emission organ size on female fitness components have not been studied in fireflies (but see <ns0:ref type='bibr'>Crastley & Lewis, 2005)</ns0:ref>.</ns0:p><ns0:p>In this context, the static allometry of sexual characters (i.e. variation of their relative size among the size range of adults in the population) involved in courtship may be of particular interest. Selection is expected to optimize relative size of any organs depending on their reproductive payoff given each individual's body size. Unlike sexually dimorphic characters involved in intra-sexual agonistic signaling (usually among males) in which larger individuals are expected to show disproportionally larger traits (i.e. positive allometry), the size of sexually dimorphic signaling characters involved in courtship are expected to be more frequently either proportional to body size (i.e. isometry) or even relatively smaller in larger individuals (i.e.</ns0:p><ns0:p>negative allometry) <ns0:ref type='bibr'>(Eberhard et al., 2018)</ns0:ref>. The reason to expect these patterns are diverse, but in general the payoff of relatively larger traits for larger individuals is high when they are involved in agonistic interactions in which body size is a good predictor of fight outcome, thus selecting for conflict resolution prior to engaging in costly or dangerous fights. In male-female reproductive interactions, however, the conditions are much more variable. In many cases selection for 'honest' signals accurately reflecting male body size would result in isometry, while in others relatively smaller organs in large males (i.e. negative allometry) result in high payoffs if male quality is not directly related to body size <ns0:ref type='bibr'>(Eberhard et al., 2018)</ns0:ref>. Furthermore, the allometry of sexual organs involved in receiving rather than emitting signals, as well as sexual differences in allometry, have seldom been explored. Synchronous flashing fireflies are good subjects to study hypotheses on the effects of the size and allometry of the organs involved in sexual communication on fitness components because in these species the density of signaling males is very large, nightly mating periods are short and the operational sex ratio is male biased, resulting in intense competition between males, likely absence of female-female mate competition, and plenty of opportunities for female choice <ns0:ref type='bibr'>(Lloyd, 1979;</ns0:ref><ns0:ref type='bibr'>Lewis, 2016)</ns0:ref>. We can hypothesize that intense male-male competition selects for males with larger lanterns that increase signal transmission distance. Selection would also favor males with larger eyes that increase the amount of photons captured in the night and improve the detection distance of the usually faint glows produced by the relatively scarce females. Larger males, but not females, may have a mating advantage over small males due to direct selection on body size (for example, if larger male size is advantageous when several males alight simultaneously and court a female; <ns0:ref type='bibr'>Thornhill & Alcock, 1983)</ns0:ref> or correlative selection (for example, if selection favors larger lantern size and this measure is positively correlated with body size). In contrast to males, females would be selected to emit flashes of just enough intensity to be perceived by the males, thus no selection for increased lantern size is predicted. In fact, it is even possible that females are selected to produce less intense flashes or glows not only reducing costs, but as a female choice mechanism that allows being detected only by particularly sensitive males, an ability they could inherit to their male offspring <ns0:ref type='bibr'>(Eberhard, 1996)</ns0:ref>. On the other hand, selection for an increase in eye size should be relaxed in females due to the fact that they are the limiting sex and the large number of potential mates encountered every night, although larger eyes could be advantageous if they increase the ability to detect and</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:04:47496:1:1:NEW 27 Aug 2020)</ns0:p><ns0:p>Manuscript to be reviewed respond to high quality males. Since larger signaling and receiver structures should benefit larger males despite their costs, but would represent also higher costs but no additional benefits for larger females, we could expect steeper allometric slopes for lantern and eye size in males than in females. In this paper, we present the results of a field study aimed at testing some of the predictions derived from these hypotheses. We studied the synchronous Mexican firefly Photinus palaciosi <ns0:ref type='bibr'>(Zaragoza-Caballero, 2012;</ns0:ref><ns0:ref type='bibr'>Zaragoza-Caballero et al., 2020)</ns0:ref>. Using field-collected data, we estimated the relationship between the probability of being found in copula and signal emission (lantern) and signal detection (eyes) organ size, and used these data to test the following predictions: (1) Males with larger lanterns and eyes have higher probability of being found in copula, whereas lantern and eye size in females are not related to their mating probability. (2) Larger males have a higher probability of being found in copula.</ns0:p><ns0:p>(3) Males have larger lanterns and larger eyes than females.</ns0:p><ns0:p>(4) Lantern and eyes allometric slopes are higher in males than in females, but not higher than 1 since they are not involved in male-male agonistic interactions. Finally, (5) there is no assortative mating for lantern, eye or body size due to the lack of intra-female sexual selection or male mate choice under highly male competitive conditions.</ns0:p></ns0:div>
<ns0:div><ns0:head>MATERIALS AND METHODS</ns0:head></ns0:div>
<ns0:div><ns0:head>Species studied</ns0:head><ns0:p>P. palaciosi lives in pine-oak-fir forests of central Mexico, in the states of Estado de México, Puebla and Tlaxcala, and its reproductive season goes from June to the beginning of August.</ns0:p></ns0:div>
<ns0:div><ns0:head>Mate searching, courtship and mating occur during approximately ninety minutes every night</ns0:head><ns0:p>PeerJ reviewing PDF | (2020:04:47496:1:1:NEW 27 Aug 2020)</ns0:p><ns0:p>Manuscript to be reviewed starting around 20:30 h, although heavy rainfall prevents flying activity. In the study site (Fig. <ns0:ref type='figure' target='#fig_2'>1</ns0:ref>), the municipality of Nanacamilpa de Mariano Arista (Tlaxcala state, México), thousands of males congregate under the canopy of the forest during this period, flying in search of females, frequently synchronizing their flashing and providing a magnificent show that attracts numerous tourists (Acle <ns0:ref type='bibr' target='#b0'>Mena et al., 2018)</ns0:ref>. The females cannot fly because they are brachypterous (i.e.</ns0:p><ns0:p>their wings are extremely reduced; Fig. <ns0:ref type='figure'>2</ns0:ref>) (Zaragoza-Caballero, 2012), and they remain stationary in herbs at heights < 60 cm and glow infrequently during the mating period. The number of sexually receptive females every night is much smaller than that of males and thus the operational sex ratio (OSR) is male biased (personal observation).</ns0:p></ns0:div>
<ns0:div><ns0:head>Sample collection</ns0:head><ns0:p>Samples of males and females found in copula or solitary were collected simultaneously by a team of three researchers during the daily mating period (20:30 -22:00 h GMT-5) in the middle of the 2016 reproduction season (between June 29 and July 14). Signalling males were collected with an entomological net and solitary females and mating couples by hand. Individual mating couples and solitary individuals were kept in Eppendorf vials with absolute alcohol. Captures were made in 10 places (one per night) within a continuous forests in the municipality of Nanacamilpa de Mariano Arista (Table <ns0:ref type='table'>1</ns0:ref>; Fig. <ns0:ref type='figure' target='#fig_2'>1</ns0:ref>). Our collection was made under the SEMARNAT (Mexican Government) permit SGPA/DGVS/06292/16.</ns0:p></ns0:div>
<ns0:div><ns0:head>Measurement of phenotypic traits</ns0:head><ns0:p>We obtained three photographs of each firefly (dorsal view, ventral view and a close up of the eyes) with a digital camera (Canon™ model T3i) mounted on a disection microscope (Olympus™ model SZH10). The phenotypic measurements were obtained with the NIH ImageJ open access software (National Institutes of Health USA, http://rsb.info.nih.gov.ij/). We estimated lantern size by measuring the area covered by the lantern in the ventral-view photographs (Fig. <ns0:ref type='figure'>2</ns0:ref>). Eye size was estimated as the squared difference between maximum eyespan and interocular space (i.e. approximately the sum of the maximum diameter of both eyes) in the eyes close-up photographs (Fig. <ns0:ref type='figure'>2</ns0:ref>). The reason to square this length was to obtain a variable that would covariate linearly with the rest of the body traits since they were all area measurements. Body size was estimated as the area covered by the elytra of the males in the dorsal view photographs, while in females it was estimated as the area covered by the thorax and the abdomen (Fig. <ns0:ref type='figure'>2</ns0:ref>). The area covered by the reduced female elytra was also measured to document brachyptery quantitatively, and as an aditional proxy of body size (see below).</ns0:p></ns0:div>
<ns0:div><ns0:head>Statistical analyses</ns0:head><ns0:p>For predictions 1 and 2, we constructed separate binomial generalized linear models with logit link function for each sex, using mating status (0=captured alone, 1=captured mating) as the binary response variable, and lantern size, eye size (prediction 1), and body size (prediction 2) as explanatory variables, as well as their second order interactions. We simplified the models using backwards stepwise simplification, removing each explanatory variable in order of increasing significance and testing the effect of removing that variable with a chi-squared likelihood ratio test until only terms whose removal leads to worsening of the model remained <ns0:ref type='bibr'>(Crawley, 2013)</ns0:ref>.</ns0:p><ns0:p>A second set of models was constructed and simplified using relative lantern size and relative eye size in order to rule out any effect of colineality among explanatory variables. Relative sizes were estimated dividing trait (lantern or eye) size by body size. A third set of models was constructed and simplified for females in order to rule out the effects of body condition using elytra area as a proxy of body size to construct lantern and eye relative size as explanatory variables. Unlike wing area, abdomen volume (and thus measured abdomen area) may vary</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:04:47496:1:1:NEW 27 Aug 2020)</ns0:p><ns0:p>Manuscript to be reviewed through adulthood depending on body condition, which in turns may depend on water, muscle and fat reserves, and egg load in the case of females <ns0:ref type='bibr'>(Moya-Laraño et al., 2008)</ns0:ref>. Egg load has been shown to be associated to female weight in a congeneric species of brachypterous females <ns0:ref type='bibr'>(Wing, 1989)</ns0:ref>.</ns0:p><ns0:p>For prediction 3, we performed a simple linear model for each morphological trait (body size, elytra size, lantern size, lantern relative size, eye size and eye relative size) as dependent variable, and sex as the only explanatory variable. Although we had no prediction for body size sexual dimorphism, and elytra size dimorphism is so large that does not require a statistical test, we included these variables in the analyses for descriptive purposes. For prediction 4 we estimated the slope of log-log relationships between lantern size or eye size and body size using ordinary least squares (Kilmer & Rodriguez, 2016). As described above for prediction 1, a second set of analyses was performed for females using elytra area instead of body size in order to rule out effects of body condition. In order to compare slopes statistically, 95% confidence intervals were generated for each allometric slope using bootstrap resampling with 10,000 randomizations (Crawley, 2013) For prediction 5, we only used individuals captured in copula and tested assortative mating using linear models relating morphological traits (lantern size, lantern relative size, eye size, eye relative size or body size) of males and their respective female mates. As described above for predictions 1 and 3, female elytra area was included as proxy of female body size in order to rule out effects of body condition. We carried out these analyses in R software, version 3.6.3 (R Core Team, 2020) using the R Studio interface <ns0:ref type='bibr'>(R Studio Team, 2016)</ns0:ref>. The script used for analyses and the databases with all data can be found as supplementary material (Files S1-S3).</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:04:47496:1:1:NEW 27 Aug 2020)</ns0:p><ns0:p>Manuscript to be reviewed</ns0:p></ns0:div>
<ns0:div><ns0:head>RESULTS</ns0:head></ns0:div>
<ns0:div><ns0:head>General observations</ns0:head><ns0:p>We sampled 93 females (51 solitary and 42 in copula) and 113 males (71 solitary and 42 in copula). Two copulating females were not included in the analyses because the posture they had after fixation in alcohol prevented obtaining correct measurements.</ns0:p></ns0:div>
<ns0:div><ns0:head>Phenotypic traits and the probability of capture during mating</ns0:head><ns0:p>We did not find support for predictions 1 or 2. Neither body size nor any of the morphological traits associated to signal emission (lantern size, lantern relative size) or reception (eye size or eye relative size), nor their statistical interactions, showed a significant association with the probability of being captured single or mating for either sex (Table <ns0:ref type='table' target='#tab_1'>2</ns0:ref>).</ns0:p></ns0:div>
<ns0:div><ns0:head>Sexual dimorphism</ns0:head><ns0:p>Males are significantly larger than females, and as could be expected given the brachypterous morphology of females, male wing area is also significantly larger (Table <ns0:ref type='table'>3</ns0:ref>). We found partial support for prediction 3. As predicted, males have significantly larger lanterns than females in terms of absolute and relative size. Interestingly however and contrary to our prediction, they have significantly smaller eyes than females in terms of absolute and relative size.</ns0:p></ns0:div>
<ns0:div><ns0:head>Static allometry</ns0:head><ns0:p>As predicted, the slope of male lantern size allometry was significantly higher than the slope of female lantern size allometry, but not significantly higher than 1 (Table <ns0:ref type='table' target='#tab_3'>4</ns0:ref>, Fig. <ns0:ref type='figure'>3A,C,E</ns0:ref>). In contrast, eye size allometry was not significantly different from 0 in males or females, and its</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:04:47496:1:1:NEW 27 Aug 2020)</ns0:p><ns0:p>Manuscript to be reviewed relative size varied widely along body size ranges in both sexes (Fig. <ns0:ref type='figure'>3B,D,F</ns0:ref>). In other words, lantern relative size is constant along the male size range, but in females lantern relative size decreases somewhat towards larger individuals. On the other hand, eye size does not covary with body size in either sex.</ns0:p></ns0:div>
<ns0:div><ns0:head>Assortative mating</ns0:head><ns0:p>Support for prediction 5 was also partial. We found no correlation between male and female traits in most of the variables (Table <ns0:ref type='table'>5</ns0:ref>), but we did find a significant association between male and female relative lantern size when the effect of body condition was removed using the area of female elytra to estimate relative lantern size (Fig <ns0:ref type='figure'>4</ns0:ref>).</ns0:p></ns0:div>
<ns0:div><ns0:head>DISCUSSION</ns0:head><ns0:p>In this paper we tested the hypothesis that in synchronous fireflies the male biased operational sex ratios and the short nightly mating period result in strong male-male competition that selects for males with larger signal emission (lantern) and signal detection (eyes) organs, as well as larger body sizes. On the other hand, since in these fireflies male mate choice and female-female competition for mates are expected to be absent, no selection on body and lantern size is expected in females, although intersexual selection (female choice) could favor females with larger signal detection organs (eyes). We did not find support for the predictions that body size, lantern size or eye size would be associated with the probability of being found in copula.</ns0:p><ns0:p>However we did find that males are not only larger than females, they have relatively larger lanterns but smaller eyes than females, even adjusting for sexual differences in body size.</ns0:p><ns0:p>Furthermore, the allometric slope of lantern size is steeper in males than in females, but the Manuscript to be reviewed allometric slope for eye size does not differ from 0 in either sex. Finally, and contrary to our predictions, there is some evidence of assortative mating in terms of lantern relative size. Perhaps the best species to compare our results is P. pyralis <ns0:ref type='bibr'>(Vencl & Carlson, 1998)</ns0:ref>, a species resembling P. palaciosi in that there is 'intense competitiveness: aggregations of males regularly attain very high densities', sometimes resulting in several males attempting to mate with the same female <ns0:ref type='bibr'>(Vencl & Carlson, 1998</ns0:ref>), as we have observed in P. palaciosi (personal observations). In contrast to our findings, in P. pyralis the body size (elytral length) and lantern area of males were related to the probability of being found in copula. Interestingly, in this species larger males and males with larger lanterns were more successful when single males courted females (the most common case: 70% of all matings), but smaller males had an advantage when four or more males simultaneously courted a female 'on foot' on her perch (12% of all matings). According to the authors, these contrasting effects 'obscured' the global effect of elytral and lantern length on male mating success <ns0:ref type='bibr'>(Vencl & Carlson, 1998)</ns0:ref>. When we collected many of the copulating pairs there was at least one additional male close to the copulating pair, unfortunately we did not make a record of this fact. However, a trade-off similar to that proposed by Vencl & Carlson (1998) may explain the lack of effects of morphological measures on the probability of being found in copula.</ns0:p><ns0:p>Our results, on the other hand, are similar to those obtained in the non-synchronous P.</ns0:p><ns0:p>greeni, in which the size of lanterns, eyes and body were not related to the probability of males being found alone or in copula <ns0:ref type='bibr'>(Demary et al., 2006)</ns0:ref>. In this species, as well as in other Photinus species <ns0:ref type='bibr' target='#b1'>(Branham & Greenfield, 1996;</ns0:ref><ns0:ref type='bibr'>Crastley & Lewis, 2003;</ns0:ref><ns0:ref type='bibr'>Demary et al., 2006;</ns0:ref><ns0:ref type='bibr'>Lewis, 2016)</ns0:ref>, elements of the flashing pattern are important in determining male mating success.</ns0:p><ns0:p>However, elements of the flashing pattern are also important in Photinus ignites, a non-PeerJ reviewing PDF | (2020:04:47496:1:1:NEW 27 Aug 2020)</ns0:p><ns0:p>Manuscript to be reviewed synchronous firefly in which a significant effect of lantern size and body size on mating success has also been observed <ns0:ref type='bibr'>(Crastley & Lewis, 2003</ns0:ref><ns0:ref type='bibr'>, 2005)</ns0:ref>. A study of the effect of the flashing pattern on mating success, and its possible interaction with lantern size in P. palaciosi remains as an interesting possibility.</ns0:p><ns0:p>The morphology, physiology and behavior of signal detection and emission organs are frequently influenced by selective pressures not related to the sexual communication function <ns0:ref type='bibr'>(Niven & Laughling, 2008;</ns0:ref><ns0:ref type='bibr'>Stöckl et al., 2013;</ns0:ref><ns0:ref type='bibr'>Elgar et al., 2019)</ns0:ref>. Thus, another possible explanation for our results is the existence of additional selective pressures acting in opposite direction to sexual selection or in a more complex way. Although the lantern of adult fireflies is an organ for emitting sexual signals, it can be subject to natural selection <ns0:ref type='bibr' target='#b2'>(Branham & Wenzel, 2003;</ns0:ref><ns0:ref type='bibr'>Woods et al., 2007;</ns0:ref><ns0:ref type='bibr'>Lewis & Cratsley, 2008;</ns0:ref><ns0:ref type='bibr'>Stanger-Hall et al., 2018)</ns0:ref>. For example, a study of two Photinus species determined that flashing increases predation risk and metabolic rate (37% with respect to the basal metabolic rate, even though the experimental setting excluded flight) <ns0:ref type='bibr'>(Woods et al., 2007)</ns0:ref>. In P. palaciosi it is not known if some predator exerts a similar pressure on signaling fireflies and if this possible effect is related to lantern size. We have made non-systematic observations of several unidentified predators (a grasshopper and species of orbwebb spiders) that capture males during the mating period, although light emission seems irrelevant in prey capture at least for orb-webb spiders.</ns0:p><ns0:p>The steeper allometric slope of lantern size in males suggests that the payoff of investing in lanterns proportional to their size may be higher in large males compared to large females.</ns0:p><ns0:p>However, the fact that the slope of females is still higher than 0 suggests that large females also benefit from investing in lanterns somewhat proportional to their size. This is consistent with the finding that there is assortative mating in terms of relative lantern size, suggesting some degree of female-female competition and/or male choice. A likely scenario explaining this pattern may be that among the few females present during a given night, those with larger lanterns could be more detectable to males, or more attractive as the lantern also predicts female size and thus could predict fecundity. In this case, larger females would be the first to mate and, among males competing for these larger females, those with larger lanterns may be detected or selected first by these females. The fact that assortative mating was not found in terms of absolute lantern size but in terms of relative lantern size independently of female body condition is intriguing. At least prior to physical proximity and during the first visual signaling interactions, body size can hardly be assessed by either sex but lantern size could. Relative lantern size may be an honest signal of quality in the case of males as it could show its energetic efficiency independently of body size, or a Fisherian trait, but in the case of females it may be deceiving males if they use it to assess female fecundity as it would not reflect absolute female size or condition, but would still make females with larger lanterns more detectable regardless of their body size. This scenario would not only explain our results in terms of sexual dimorphism, lantern allometry and assortative mating; if males and females throughout the size range of both sexes end up mating, no association between any morphological trait and the probability of being found in copula is expected to arise. Some degree of male mate choice in Photinus fireflies has been suggested before <ns0:ref type='bibr'>(Lewis et al., 2004a, b)</ns0:ref>, especially since the payoff for males of mating with low fecundity females may be negative when a costly nuptial gift is offered, as it is common in this genus. However, although this still unknown for P. palaciosi, female flightlessness as been shown to be associated to loss of spermatophore production in males <ns0:ref type='bibr'>(South et al., 2011)</ns0:ref>. We are currently investigating if P. palaciosi produces nutritious spermatophores.</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:04:47496:1:1:NEW 27 Aug 2020)</ns0:p><ns0:p>Manuscript to be reviewed Not only lantern size was not related to the probability of being found in copula, neither absolute or relative eye size predicted the probability of being found in copula in either sex. As discussed in the case of lantern size, additional selective pressures affecting eye size and acting in opposite or more complex ways could explain these results (see, for example, Lau & <ns0:ref type='bibr'>Meyer-Rochow, 2006)</ns0:ref>. For example, the detection and assessment of visual signals of mate quality <ns0:ref type='bibr'>(Lewis, 2016;</ns0:ref><ns0:ref type='bibr'>Rosenthal, 2017;</ns0:ref><ns0:ref type='bibr'>Elgar et al., 2019;</ns0:ref><ns0:ref type='bibr'>Stanger-Hall et al., 2018)</ns0:ref> suggests that the structure and function of the eyes has evolved influenced by intersexual selection (mate choice).</ns0:p><ns0:p>However, the eyes are also used to navigate through the habitat, find other resources (food, shelter, etc.) and detect natural enemies and, thus, its evolution is also affected by natural selection <ns0:ref type='bibr'>(Elgar et al., 2019)</ns0:ref>. As mentioned above, we have observed several predators that capture males during the mating period and could be significant selective pressures on eye size.</ns0:p><ns0:p>Interestingly, unlike most species of Photinus, eye size was smaller in males than in females, suggesting that in females selection pressures derived from processes such as female choice, predator avoidance and the choice of perch for mate location, could be important to understand the evolution of eye size. The high variation and allometric slopes of eye size in both sexes imply that this trait is unrelated to body size throughout the body size ranges of both males and females, suggesting that, unlike lanterns, there is little or no selection for larger males or females to invest in eyes proportional to their size. Although having large lanterns and eyes may represent selective advantages <ns0:ref type='bibr'>(Lloyd, 1966;</ns0:ref><ns0:ref type='bibr'>Demary et al., 2006)</ns0:ref>, a trade-off may restrict the possibility of investing in both functions (structures). It would seem that both sexes favor investing in signaling (lanterns) proportionally to their size, while investment in reception (eyes) is highly variable and independent of body size.</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:04:47496:1:1:NEW 27 Aug 2020)</ns0:p><ns0:p>Manuscript to be reviewed</ns0:p><ns0:p>Firefly populations worldwide are declining and threatening factors vary in importance for different species and regions <ns0:ref type='bibr'>(Lewis et al., 2020)</ns0:ref>. Although there is a recent and important interest in firefly watching as a tourist attraction, conservation measures and regulation of touristic activities in fireflies 'sanctuaries' need to be based on solid scientific information. Light pollution and tourism are considered important threats for the charismatic synchronous species, such as P. palaciosi, and these factors have their main impact during the mating period.</ns0:p><ns0:p>Unfortunately, mating dynamics have been studied only in a handful of the about 2,000 firefly species described. Manuscript to be reviewed</ns0:p></ns0:div>
<ns0:div><ns0:head>CONCLUSIONS</ns0:head><ns0:note type='other'>Figure 2</ns0:note><ns0:p>Measurements performed in males and females of Photinus palaciosi. Manuscript to be reviewed</ns0:p><ns0:note type='other'>Figure 3</ns0:note><ns0:p>Static allometry of the synchronous firefly Photinus palaciosi.</ns0:p><ns0:p>(A) Female lantern size relative to body size, (B) female eye size relative to body size, (C) female lantern size relative to elytra size, (D) female eye size relative to elytra size, (E) male lantern size relative to body (elytra) size, (F) male eye size relative to body (elytra) size.</ns0:p><ns0:p>Lines in red represent slopes significantly higher than 0 and lines in grey represent slopes non significantly different from 0 (see Manuscript to be reviewed Results of five models evaluating association between probability of being collected copulating rather than alone and trait size. Manuscript to be reviewed Allometric slopes of lantern and eye size for females and males of the firefly Photinus palaciosi.</ns0:p><ns0:p>Two proxies of body size were used for females, body and elytra area, and one, body </ns0:p></ns0:div><ns0:figure xml:id='fig_0'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:04:47496:1:1:NEW 27 Aug 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_1'><ns0:head /><ns0:label /><ns0:figDesc>Sexual dimorphism in lantern and eye size, along with allometric slopes of lantern size, and assortative mating in terms of relative lantern size, support not only the hypothesis of intense sexual selection among males of Photinus palaciosi, but also the possibility of subtle mechanisms of sexual selection among females as well. Trade-offs between investment in signaling (lanterns) versus detection (eyes) structures, or with pressures different from sexual selection such as those imposed by predators, are also likely to be important in shaping the evolution of sexual signaling in this species.</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_2'><ns0:head>Figure 1</ns0:head><ns0:label>1</ns0:label><ns0:figDesc>Figure 1</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_3'><ns0:head>(</ns0:head><ns0:label /><ns0:figDesc>A) Male lantern size area, (B) female lantern size area, (C) maximum eye span, (D) interocular space, (E) male elytra area, (F) female body area and elytra area. PeerJ reviewing PDF | (2020:04:47496:1:1:NEW 27 Aug 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_4'><ns0:head /><ns0:label /><ns0:figDesc>Absolute trait size in males (A), females (B), trait size relative to body (elytra) size in males (C), trait size relative to body size in females (D), and trait size relative to elytra size in females (E) in the synchronous firefly Photinus palaciosi. Parameters from initial models are presented since backwards stepwise simplification resulted in removal of all explanatory variables in all models.PeerJ reviewing PDF | (2020:04:47496:1:1:NEW 27 Aug 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_5'><ns0:head>(</ns0:head><ns0:label /><ns0:figDesc>=elytra) area, for males. m = allometric slope value [95% Bootstrap confidence interval]. Parameters of significant correlations in italics. PeerJ reviewing PDF | (2020:04:47496:1:1:NEW 27 Aug 2020)</ns0:figDesc></ns0:figure>
<ns0:figure type='table' xml:id='tab_0'><ns0:head /><ns0:label /><ns0:figDesc>Table 4 for statistical parameters).</ns0:figDesc><ns0:table><ns0:row><ns0:cell>1</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>Site</ns0:cell><ns0:cell>Date</ns0:cell><ns0:cell>Solitary females</ns0:cell><ns0:cell>Females in copula</ns0:cell><ns0:cell>Solitary males</ns0:cell><ns0:cell>Males in copula</ns0:cell></ns0:row><ns0:row><ns0:cell>1. Centro Turístico Canto del Bosque</ns0:cell><ns0:cell>29/06</ns0:cell><ns0:cell>10</ns0:cell><ns0:cell>1</ns0:cell><ns0:cell>15</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>2. Laguna azul</ns0:cell><ns0:cell>30/06</ns0:cell><ns0:cell>6</ns0:cell><ns0:cell>7</ns0:cell><ns0:cell>9</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>3. El Madroño</ns0:cell><ns0:cell>1/07</ns0:cell><ns0:cell>9</ns0:cell><ns0:cell>2</ns0:cell><ns0:cell>12</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>4. Zona Ecoturística La Granja Salma</ns0:cell><ns0:cell>2/07</ns0:cell><ns0:cell>8</ns0:cell><ns0:cell>5</ns0:cell><ns0:cell>11</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>5. El Posito</ns0:cell><ns0:cell>3/07</ns0:cell><ns0:cell>6</ns0:cell><ns0:cell>4*</ns0:cell><ns0:cell>11</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>6. Rancho Cacaloapan</ns0:cell><ns0:cell>4/07</ns0:cell><ns0:cell>2</ns0:cell><ns0:cell>5*</ns0:cell><ns0:cell>8</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>7. Ejido Miguel Lira y Ortega-La Cañada</ns0:cell><ns0:cell>5/07</ns0:cell><ns0:cell>7</ns0:cell><ns0:cell>5</ns0:cell><ns0:cell>2</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>8. Villas del Bosque Santa Clara</ns0:cell><ns0:cell>6/07</ns0:cell><ns0:cell>3</ns0:cell><ns0:cell>3</ns0:cell><ns0:cell>0</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>9. Bosques Vista Hermosa</ns0:cell><ns0:cell>8/07</ns0:cell><ns0:cell>0</ns0:cell><ns0:cell>8</ns0:cell><ns0:cell>1</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>10. El Encanto de la Luciérnaga-La Obra</ns0:cell><ns0:cell>14/07</ns0:cell><ns0:cell>0</ns0:cell><ns0:cell>4</ns0:cell><ns0:cell>0</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>2</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row></ns0:table><ns0:note>PeerJ reviewing PDF | (2020:04:47496:1:1:NEW 27 Aug 2020)</ns0:note></ns0:figure>
<ns0:figure type='table' xml:id='tab_1'><ns0:head>Table 2 (on next page)</ns0:head><ns0:label>2</ns0:label><ns0:figDesc /><ns0:table /></ns0:figure>
<ns0:figure type='table' xml:id='tab_3'><ns0:head>Table 4 (on next page)</ns0:head><ns0:label>4</ns0:label><ns0:figDesc /><ns0:table /></ns0:figure>
</ns0:body>
" | "Dear Editor;
We are grateful for the opportunity to resubmit our manuscript originally entitled “Size does not matter: No effect of signal detection and signal emission organ size on the mating success of a synchronous firefly” (ref no. 2020:04:47496). We have carried a major revision attending all the valuable comments provided by you and the reviewers. We believe that the manuscript is much more robust now and hope you and the reviewers find it suitable for publication in PeerJ.
In brief, the predictions were restructured along with additional literature on the points raised by the reviewers in the introduction and discussion sections. The main analyses were repeated on the raw data excluding those individuals suggested by one of the reviewers, fortunately some of these excluded individuals were collected in the most isolated sites. Additional analyses on trait allometry were also carried out. All these changes resulted in some major new points addressed in the discussion, and in a new tittle for the manuscript: “The size of signal detection and emission organs in a synchronous firefly: sexual dimorphism, allometry and assortative mating”.
It is also important to point out that we invited a new coauthor, Dr. Rogelio Macías-Ordóñez from the Instituto de Ecología (Veracruz, Mexico). Rogelio is a long time friend and colleague with a research trajectory on arthropod reproductive ecology. He actually was involved and suggested some of the ideas explored in this manuscript at the beginning of this project, and he carried out most of the new analyses and was deeply involved in writing and reviewing the new version upon our invitation to join us in preparing this resubmission.
Find below our reply to each specific comment and suggestion made by you or the reviewers, which were extremely useful to produce this new version.
Best regards,
Carlos Cordero
Corresponding Author
Editor comments:
The novelty of your work needs to be better delivered in the light of previous studies.
R: New references are now addressed and discussed, including those suggested by reviewers (details in replies to reviewers’ comments). We have explained in the Introduction what parts of our study have been rarely, if ever, studied before in fireflies, especially in the 2nd and 3rd paragraphs.
It is perceived that mate choice data are weak and sample sizes are too small.
R: In the new version, we do not imply that our results are necessarily suggestive of mate choice, or lack of. We do not see any specific remark on sample size in the reviewers’ comments, however the sample size used in all our analyses exceeds the recommended minimum sample size for each test and the number of factors in the case of linear models.
Data were collected from different localities that were often several kilometers apart. As these fireflies are not known to be long-distance fliers, and males are known to live only a short period, each capture may represent different populations with no replications. Please address this.
R: All specimens were collected in different sites within a continuous forest (we include a map in the new Figure 1). Furthermore, our population genetics analyses (using Snips) in course, that include four of the Nanacamilpa sites, indicate that they constitute one population.
Please address Reviewer 2’s concern that “the validity of the findings largely hinges on the appropriateness of eye size as the proxy for ability to detect point light sources. This is where the authors need to deliver more information and justification”.
R: We have addressed this point in the introduction using new references; see details in the reply to such comment below.
Please ensure that all R scripts are available to allow replication of your analyses. All aspects of your work should be repeatable.
R: The R script and databases are now included as supplementary material allowing full replication of our analyses
It appears that certain key references are missing. Please add these, as it seems they may help in better framing your hypothesis.
R: The following references have been added and addressed in different sections of the manuscript: Eberhard 1996, Eberhard et al 2108, Kilmer & Rodriguez 2016, Lewis et al 2004, Moya-Laraño et al 2008, Thornhill & Alcock 1983, Warrant & Dacke 2011, South et al 2011, Wing 1989.
Reviewer 1
Basic reporting
L53-55, please specify if these species have synchronous or asynchronous males
R: We have included this information in the second paragraph of the Introduction.
L59, I agree that synchronous species are good subjects due to the density of signaling males. But it would be helpful to discuss the downside which is that males with a smaller lantern might be able to exploit larger neighboring males without paying the severe metabolic costs.
R: Even though we agree with this point and acknowledge that flashing patterns have a clear and better studied role in male-female firefly interaction, this was not the focus of our study so we did not find a suitable place to discuss this point. In order to create such space for discussing it was necessary to enlarge the discussion, which did not seem justified. We’re open to do so if this is still recommended once this new restructured version is reviewed.
Also at line 62 aren’t the authors assuming that there is no pre-copulatory courtship or cryptic female choice by the females, as discussed by Eberhard and others?
R: We were not attempting to assume this but, in any case, the new structure of the introduction should now give a more inclusive perspective of all possible mechanism of sexual selection involved, with special focus on male-female precopulatory interactions.
L72, change for to of
R: Done, this sentence is part of a new section with the correction incorporated.
L82, at first I agreed that selection should favor males with larger eyes for improved detection ability but then I wondered more about the natural history of how far a male can or will fly to try to encounter female: larger eyes should capture more photons meaning a male could detect a female at a greater distance. However under intense competition, by the time he actually arrives, another male closer by may have already gotten there. In synchronous species presumably females are not responding to just one particular male, but to the collective source of light (followed by scramble competition among males to get there first). If so, males be selected to respond via controlled flight to point sources only within a certain range, which may be a constraint on eye size enlargement.
R: Although there are not many studies on this subject in synchronous fireflies, most modern authors consider that females respond to individual males, not to the collective source of light (at least we did not find any recent paper making this suggestion). In fact, most studies indicate the existence, both at long and short range, of bioluminescent “dialogues” between potential pairs. Although it is common to observe more than one males courting a female at close range both in our species and in others referred in our manuscript, it is not rare to observe alighted pairs courting or attempting to mate.
I wonder if the situation might be similar to some deep-sea fishes that, if I recall correctly, have relatively smaller eyes, which is hypothesized to be associated with being relatively weak swimmers (see Warrant and colleagues, Biol Rev 2004). It may be unfair to claim that fireflies are relatively weak flyers but I think the argument is at least worth considering, as it would flip the prediction the authors make.
R: Even though we restructured our predictions and tried to consider this particular argument, we did not find a clear parallel with deep-sea fishes regarding this hypothesis. Rather than weak flyers, the relatively slow flight of fireflies appears to be important to increase the probability that the females detect their signals under the highly competitive conditions, as well as allowing them to detect the faint signals of females.
L121, it was not clear to me how female body size was estimated given size was based on an estimate of elytra length IF the elytra were not reduced, but how can this be known?
R: We have now clarified this further. Two proxies of female size were used: wing area and body area. The reasons for this are explained in the methods section along with a new figure (Figure 2) illustrating how body size was measured in males and females.
L125ff, it would be useful to show that there are, or are not, allometric relationships between body size and eye size, and lantern size.
R: We have added a whole new section, prediction and analyses on allometric slopes for these characters. As a result the discussion now links allometry, sexual dimorphism and assortative mating.
L154, I am not persuaded these data showing female elytra are shorter than those of males are worth presenting since we already know they are brachyopterous.
R: We have kept the analyses (on a revised database as described above) for descriptive reasons. Furthermore, since female wing size is now a second proxy of female body size we thought it was relevant to report it.
L190, in discussing support for their hypothesis, the authors seemingly jump from intra-specific comparisons to inter-specific comparisons. That is, they set up to study by predicting that males with larger eyes and or larger lanterns should be found in copula than males with smaller structures. They found no evidence for this. But then these inferences really are inter-specific, showing that males are larger bodied, and have larger lanterns, than females. To fully substantiate the claim that it is due to sexual selection requires out-group comparisons, showing that the derived species have larger lanterns, for example. This should be clarified (similarly, for female eye size line 194).
R: In our restructured predictions we try to be more conservative. We acknowledged that there are phylogenetic approaches to testing sexual selection hypotheses, but ours is not phylogenetic. We try to derive predictions from theory into what is a single population, and derive our conclusions depending on which predictions are supported and which are not.
L197, here the authors need to again clarify which species are synchronous and which are asynchronous, as I think the predictions differ.
R: In the new version we clarify this when we mention every species discussed.
L204, as above is this species synchronous or asynchronous?
R: R: In the new version we clarify this when we mention every species discussed.
L211, the observation that most meetings involve solitary males contradicts the notion of intense competitiveness described in line 205. Please reconcile.
R: We now mention that other males were present in many mating events, although unfortunately we did not record this in detail. Our notion of intense competitiveness was based on the fact that males seemed to be way more abundant than females, but it is this notion precisely what is being tested in the study and our results suggest that competition may actually not be as intense. In the new version we try to present this notion of intense competition as a useful hypothesis more than as an assumption.
L228, correct spelling of species name
R: done
L 233, fix typo in Laughlin
R: done
L239, change webbing to web (also L 252)
R: done
Reviewer 2
Basic reporting
I have noticed gaps concerning their cited literature works in support of setting up their study and formulating their hypotheses which may potentially question their justification of the study.
R: In this version, we have included all the literature we are aware of related to the questions we ask in our manuscript, independently if they agree or not with our predictions. Several new references have been added and addressed in different sections of the manuscript, including the revised predictions in the introduction. We added specific references: Wing 1989, Lewis et al 2004, Warrant & Dacke 2011, South et al 2011, as well as more general ones: Thornhill & Alcock 1983, Eberhard 1996, Eberhard et al 2108, Kilmer & Rodriguez 2016 and Moya-Laraño et al 2008.
R scripts to replicate their analysis are not supplied which is a shortcoming that must be addressed.
R: The R script and databases are now included as supplementary material allowing full replication of our analyses
Experimental design
the main research question is not well defined and perhaps may not even be meaningful.
R: We have restructured the introduction trying to clarify our research question through 5 specific predictions.
The authors must explain how eye size is linked to detection ability of bioluminescent light flashes. Following Land and Warrant’s extensive body of work on physiological modeling of visual light sensitivity, it is primarily the absolute size of the optical aperture which determines the ability of the eye to detect point light sources. Of course, the size of the aperture will be constrained by the size of the eye and animal, hence one would need to look at absolute aperture size and relative aperture size. Eye size, as used by the authors, may not be meaningful in the detection ability of point light sources such as bioluminescent light flashes.
R: We explain the arguments that support our use of eye size in the second paragraph of the introduction, providing relevant references including papers that hace used eye size in a similar context. In brief, these are the arguments: Larger eyes are correlated with smaller interommatidial angles that may help improve visual resolution and distance perception (Lewis et al., 2004), as well as capture more photons thus helping vision in low-light environments (Warrant & Dacke, 2011; Stanger-Hall et al., 2018).
In addition, the authors must justify the use of lantern size as a proxy of emitted light intensity of the signal and provide evidence that the size of the lantern organ is indeed positively correlated with signal intensity.
R: We would have loved to find evidence of this correlation, but we still believe it is a fair enough assumption that a larger glowing area would emit more light. As cited in the manuscript, other authors have explored this trait under the same assumption and found useful patterns. We acknowledge however that the validity of this and similar studies depends entirely on this assumption.
R scripts were not supplied, which is not conducive to reproducing the statistical analyses. The authors should share their scripts for full transparency.
R: The R script and databases are now included as supplementary material allowing full replication of our analyses
Methods should explicitly address body size as a covariate, hence the analyses should focus on ANCOVAS. ANCOVAS are much more “direct” than PCA and functionally more relevant, because the loadings of the PC represent a mix of the traits. It is therefore difficult to explicitly ascribe a function to the loadings. ANCOVAS also allow for using scaling information as an assessment of selection, such as isometry vs allometry.
R: As suggested by the reviewer, in the new version we have not used PCA scores but raw data on lantern and eye absolute and relative sizes, as well as body size as covariate. Furthermore, a whole new section and analysis on their allometric slopes is included and discussed.
Validity of the findings
The validity of the findings largely hinges on the appropriateness of eye size as the proxy for ability to detect point light sources. This is where the authors need to deliver more information and justification.
As stated above, we have tried to justify further the appropriateness of eye size as the proxy for ability to detect point light sources. However, eye size is only one of the two organs involved in male-female visual communication. We believe that the value of our contribution depends on considering all traits addressed.
Reviewer 3
Basic reporting
This paper could become scientifically sound after rewriting and reanalyzing the data.
R: That’s exactly what we did (and also why it took us so long to revise it), thank you.
A) Authors should be aware of the hypotheses proposed to explain sexual size dimorphism in Lampyrid species (relevant literature listed below).
R: We thank the reviewer for noticing this deficiency in the literature cited. We have considered and contrasted the ideas and findings of the suggested papers, especially regarding nuptial gifts: Wing 1989, Lewis et al. 2004a, b, and South et al. 2011.
B) Correlational data are based on samples captured from different field populations each collected in different times. The data are relevant to describe sexual dimorphism of the species but not relevant to test hypothesis of mate choice.
R: Although the collection sites are within a continuous forest (we include a map in the new Figure 1). Furthermore, our recent population genetics analyses (using Snips) including four of the Nanacamilpa sites, indicate that they constitute one population (this study in progress is part of the doctoral thesis of the first author). The samples were collected in a period of two weeks more or less in the middle of the reproductive season (that goes from June to the beginning of August).
The data also have a limited value as correlation is not causation.
R: In the new version we have tried to be very careful not implying causation from correlations, or lack of.
C) The description of mating system of Photinus palaciosi is lacking.
R: According to classic mating system theory (Emlen & Oring 1977) a description of a mating system should include: (1) the distribution of mating success of males and females, (2) the manner of mate acquisition, (3) the description of pair bonds, (4) and any form of parental care. We would love to know details on all four points for this recently described and virtually unstudied species, but the best we could do was to infer based on similarities in their reproductive behavior with other congeneric species that they present scramble polygynandry with pair bonds that last several hours during mating and no form of parental care. Verifying each of those points would be another study by itself and we believe that several valuable contributions to insect reproductive ecology have not required such description if enough aspects of the mating system may be inferred.
Detailed comments:
Sexual dimorphism and mating system
1.There are several published hypotheses explaining sexual dimorphism in Lampyridae. Male biased sex ratio is not the only explanation to a large male (and lanterns) size. In many species (flightless) females are larger than males and yet male-male competition is intense (see below).
R: We have focused on the hypothesis derived from the biased sex ratio because this is in line with the scant available information regarding this firefly.
Sexual size dimorphism in fireflies has been connected to female flightlessness, evolution of female sexual signals and male nuptial gifts (South et al., 2020, see below).
R: This reference is addressed now in the fifth paragraph of the discussion, thank you.
Is P. palaciosi a nuptial a gift giving species?
Most firefly species are capital breeders and adults do not feed. Thus both females and males have a limited energy for reproduction and maintenance. For example, some Photinus species females may loose seven eggs/day if remain unmated (Wing, 1989).
During mating males of many Photinus species deliver a large spermatophore which increases female fecundity and longevity (Lewis et al., 2011). Male nuptial gift size correlates positively with body size. A large nuptial gift delays female remating and increases his fertilization success. A large nuptial gift also increases female fecundity and longevity. Thus large male size may be selected not because of lantern size but because of spermatophore size. Is this the case in P. palaciosi as well?
R: We agree that knowing this would allow discussing our results better, but we don’t know. We have tried to address some of these possibilities in the new version of the discussion. Considering female brachyptery, it is possible that males do not produce spermatophores. However, a study of this aspect is at the top of our priorities.
Thus testing correlation between lantern size (signal emission organ) with mating success may be irrelevant.
R: In the new version we do not suggest such a direct link between the probability of being found in copula and mating success, although some relationship may be implied. The results are discussed in a more balanced perspective including also allometry, assortative mating, sexual dimorphism and a hypothethical scenario that could explain all our results, including the lack of relationship between lantern and eye size and the probability of being found in copula or any proxy of mating success.
Large male size may be selected because large spermatophores delay female rematings and increase her fecundity and thus his fertilization success. This should be tested in controlled experiment and cannot be detected in field collected data.
R: We agree and acknowledge the limitations of correlative field collected data versus controlled experiments. Our aim, especially stressed in the new version, is to test predictions derived from theory that could be tested using correlative field collected data.
References not found in the ms:
Lewis, S. M., Cratsley, C.K., Rooney. J. A., 2004. Nuptial gifts and
sexual selection in Photinus fireflies. Integr. Comp. Biol. 44:234–237.
South, A., Stanger-Hall, K., Jeng, M.-L., Lewis, S.M., 2011. Correlated evolution of
female neoteny and flightlessness with male spermatophore production in
fireflies (Coleoptera: Lampyridae). Evolution 65, 1099–1113.
Wing, S.R., 1989. Energetic costs of mating in a flightless female firefly, Photinus
collustrans (Coleoptera: Lampyridae). J. Insect Behav. 2:841–847.
R: All three references have been useful and are now included. We also found very useful:
Lewis SM, Cratsley KC, Demaris K. 2004a. Mate recognition and choice in Photinus fireflies. Annales Zoologici Fennici 41:809-821.
2. Male-biased sex ratio
Authors claim “a strong male-biased sex ratio” (line 23). In their own data male biased sex ratio was found in all eight early samples/localities but in nine late samples sex ratio was males biased only in two samples (22% of samples): it was female biased in three samples and equal in four sampels, see ms: supplementary material).
In other Photinus species sex ratio is known to vary temporally. Local sex ratios become female-biased later in mating season (Lewis & Wang,1991). Could that be the case also in P. palaciosi?
R: Our impression of “a strong biased male ratio” is based on observations at any given moment during the time of activity every night, and is based in observations made in more than one year and in different moments of the mating season. Plenty of males flying while females are extremely hard to find. Our sampling was not designed to estimate sex ratios, the limiting factor was usually females and small samples (now excluded from the analyses, following the suggestion of one reviewer) resulted from bad weather conditions (strong rain prevents firefly activity). For the same reasons we can not assess temporal variation in sex ratios throughout the mating season, which would allow us to refine our predictions if such variation existed.
References not cited in ms:
Lewis, S. M., and Wang, O. T. 1991. Reproductive ecology of two species of Photinus fireflies (Coleoptera: Lampyridae). Psyche 98:293–307.
R: We appreciate this suggestion but, although we reviewed it, we found no clear opportunity to address it in our manuscript as these species have flying females and, as stated above, we could not assess or infer seasonal variation in sex ratio.
Experimental design
The work is based on correlative data from field captures. The data are collected from different localities and different dates. The distance between sampling localities are not presented in supplementary material. There is seventeen days between the first and the last sampling. To pool the data from different samples is valid when testing sexual dimorphism in morphological measurements. When testing possible mate choice hypotheses pooled data across sites and localities are not valid. For example, some males (altogether 34) were captured in localities (and on a period) when no females were found. Thus these males could not make any choice or could not be a target for female choice. The data should be reanalysed based only on samples when both solitary and mating individuals are found simultaneously and date (= locality) should be one explanatory factor in the analysis.
R: As stated above, we now include a map of the collection sites and have excluded individuals from isolated sites from the analyses. For the reasons mentioned above, we think all sites as part of a continuous population. We are now more conservative with the implications of our correlative results and discuss them in the context on results from allometry, sexual dimorphism and assortative mating.
Figure 2 and Table 4 should be rewritten respectively.
R: All tables and figures are new.
The relationship between signal detection organ (eyes) and lantern size with mating success should be tested experimentally not sampling field data. The effect of male nuptial gift on male mating success should also be tested.
R: We agree, and in the new version we avoid implying any causal relationship from our results.
Validity of the findings
When the data are reanalysed, irrespective of the results, the findings may be interesting if they are discussed on the light of mating systems of firefly species (South et al. 2011). Even when reanalysis is done, the present data are correlative and not evidence of causality.
R: We agree, and in the new varsion we avoid implying any causal relationship from our results. We try to discuss possible scenarios involving mating strategies suggested by our results.
Comments for the author
The authors should include more information on mating system of the species.
R: As stated above, we included as much as we could, which is not much but should provide beter context to discuss our results.
Sexual size dimorphism should be discussed and tested on the light of published hypotheses.
R: We have tried to provide the theoretical context to sexual dimorphism and other aspects of the study in the new version, by using the references suggested by the reviewers as well as other more general (Eberhard, Thornhill & Alcock).
The data on mate choice should be reanalyzed omitting localities where only individuals of one sex are found.
R: We have followed this advice, and we also decided to exclude sites were very few individuals were collected (see the new Table 1).
Finally, again, we would like to thank you and the three reviewers because your criticism and suggestions helped us to improve our understanding of the implications and limitations of our study. We hope you find this revised version satisfactory.
Best regards
" | Here is a paper. Please give your review comments after reading it. |
9,766 | "<ns0:abstract xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:p>Background. Immunoglobulin superfamily member 10 (IGSF10) is a member of the immunoglobulin superfamily and is broadly expressed in both the gallbladder and ovary.</ns0:p><ns0:p>Currently, the role and possible mechanism of IGSF10 in breast cancer are still unclear.</ns0:p><ns0:p>Method. By applying real-time quantitative polymerase chain reaction (qRT-PCR) and immunohistochemistry (IHC), the expression of IGSF10 in breast cancer cells and tissues was detected. We collected the clinical information of 700 breast cancer patients from The Cancer Genome Atlas (TCGA), and analyzed the relationship between IGSF10 and the clinicopathological features and survival outcomes of these patients. The potential mechanisms and pathways associated with IGSF10 in breast cancer were explored by performing gene set enrichment analysis (GSEA). Results. According to the TCGA data, qRT-PCR and IHC experiments, our data suggested that the mRNA and protein levels of IGSF10 were significantly decreased in breast cancer tissues. The expression of IGSF10 was significantly correlated with age, tumor size, and tumor stage. Moreover, poor overall survival (OS) and relapse-free survival (RFS) were related to lower expression of IGSF10 according to the survival analysis . The multivariate analysis revealed that IGSF10 was an independent prognostic factor for OS (hazard ratio (HR)=1.793, 95% confidence interval (CI): 1.141-2.815, P=0.011) and RFS (HR=2.298, 95% CI: 1.317-4.010, P=0.003) in breast cancer patients. GSEA demonstrated that IGSF10 was involved in DNA repair, cell cycle, and glycolysis. The results indicated that IGSF10 was associated with the PI3K/Akt/mTOR and mTORC1 signaling pathways. Conclusion. This study revealed a clear relationship between IGSF10 and the tumorigenesis of breast cancer for the first time. It is therefore necessary to perform further studies in order to understand the mechanism of IGSF10 in breast cancer.</ns0:p></ns0:div>
</ns0:abstract>
<ns0:body xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:head>Introduction</ns0:head><ns0:p>PeerJ reviewing PDF | (2020:02:45995:1:2:NEW 16 Jul 2020)</ns0:p><ns0:p>Manuscript to be reviewed Breast cancer is a common malignancy that seriously threatens women's health.</ns0:p><ns0:p>Approximately 2.1 million female breast cancer patients were newly diagnosed worldwide in 2018. Breast cancer accounts for one-quarter of all female cancer cases <ns0:ref type='bibr' target='#b4'>(Bray et al., 2018)</ns0:ref>. As a heterogeneous disease, the initiation and development of breast cancer is affected by both genetic and environmental factors <ns0:ref type='bibr' target='#b43'>(Yang et al., 2019)</ns0:ref>. Despite continuous advances made in surgical techniques, biological drugs and targeted therapies, breast cancer remains an arduous clinical problem <ns0:ref type='bibr' target='#b42'>(Woolston, 2015)</ns0:ref>. Therefore, identifying breast cancer biomarkers is crucial for understanding the tumorigenesis and accurate cancer prognosis, as biomarkers may assist the clinical diagnosis and serve as potential tumor therapeutic targets in breast cancer <ns0:ref type='bibr' target='#b7'>(Costa-Pinheiro, Montezuma, Henrique, & Jerónimo, 2015;</ns0:ref><ns0:ref type='bibr' target='#b20'>JR, MA, JT, & medicine, 2012;</ns0:ref><ns0:ref type='bibr' target='#b31'>Qiao et al., 2019)</ns0:ref>.</ns0:p><ns0:p>Immunoglobulin superfamily member 10 (IGSF10) is a gene involved in cell differentiation and developmental processes <ns0:ref type='bibr' target='#b40'>(Thutkawkorapin et al., 2016)</ns0:ref>. A previous study revealed that mutations in IGSF10 delay human puberty <ns0:ref type='bibr' target='#b14'>(Howard, 2018;</ns0:ref><ns0:ref type='bibr' target='#b15'>Howard et al., 2016)</ns0:ref>. Moreover, during embryonic development, mutations in IGSF10 can lead to the dysregulation of gonadotropinreleasing hormone (GnRH)-associated neuronal migration. Increasing evidence supports that IGSF10 deficiency may lead to transient GnRH deficiency and reversible congenital hypogonadotropic hypogonadism <ns0:ref type='bibr' target='#b1'>(Amato et al., 2019;</ns0:ref><ns0:ref type='bibr' target='#b14'>Howard, 2018)</ns0:ref>. Moreover, it has been reported that mutations in IGSF10 likely contribute to an increased risk of rectal and gastric cancers <ns0:ref type='bibr' target='#b40'>(Thutkawkorapin et al., 2016)</ns0:ref>. <ns0:ref type='bibr'>Daino et al.</ns0:ref> showed that IGSF10 is significantly downregulated in alpha-radiation-induced rat osteosarcoma <ns0:ref type='bibr' target='#b8'>(Daino, Ugolin, Altmeyer-Morel, Guilly, & Chevillard, 2009)</ns0:ref>. Other studies have shown that the expression of IGSF10 was downregulated in lung cancer tissues and that the decreased expression of IGSF10 was related to a worse prognosis of lung cancer patients <ns0:ref type='bibr' target='#b28'>(Ling et al., 2020)</ns0:ref>. However, the biological roles of IGSF10 in the majority of cancers have not been investigated, and its role in breast cancer remains largely unknown.</ns0:p><ns0:p>In the present study, the expression of IGSF10 in the collected breast cancer tissues was examined by qRT-PCR and IHC. The clinicopathological features of the disease based on IGSF10 expression and Kaplan-Meier survival curves were analyzed using public data from The Cancer Genome Atlas (TCGA) database. In addition, gene set enrichment analysis (GSEA) was performed to explore the potential mechanisms and signaling pathways, through which IGSF10 may mediate breast cancer tumorigenesis.</ns0:p></ns0:div>
<ns0:div><ns0:head>Materials & Methods</ns0:head></ns0:div>
<ns0:div><ns0:head>Cell culture</ns0:head><ns0:p>The breast cancer cell lines: MDA-MB-231, MCF-7, BT-549, ZR-75-30, SKBR-3, and T47D (ATCC, Manassas, VA, USA) were maintained as previously described <ns0:ref type='bibr'>(Zhang et al., 2019)</ns0:ref>. MCF-10A cells were also maintained as previously described <ns0:ref type='bibr' target='#b9'>(Debnath, Muthuswamy, & Brugge, 2003)</ns0:ref>.</ns0:p><ns0:p>All cell lines were cultured in a humidified incubator at 37°C with 5% CO 2 .</ns0:p></ns0:div>
<ns0:div><ns0:head>Breast cancer patients and tissue samples</ns0:head><ns0:p>The TCGA data were collected as previously described in <ns0:ref type='bibr' target='#b32'>(Qiu, Li, Zeng, Guan, & Li, 2018)</ns0:ref>.</ns0:p><ns0:p>In this study, we included 700 breast cancer patients with complete RNA-seq data and complete clinical information. The clinical information included the following: tumor size, lymph node status, tumor, node, metastasis (TNM) stage, estrogen receptor (ER) status, progesterone receptor (PR) status, human epidermal growth factor receptor 2 (HER2) status, and follow-up information.</ns0:p><ns0:p>Breast cancer tissue samples were collected as previously described <ns0:ref type='bibr' target='#b26'>(Li et al., 2018)</ns0:ref>. Specifically, we collected the breast tumor and adjacent normal tissues from the First Affiliated Hospital of Chongqing Medical University. The collected tissues were used for real-time quantitative polymerase chain reaction (RT-qPCR) and immunohistochemistry (IHC) analyses.</ns0:p><ns0:p>All specimens were stored in liquid nitrogen. The collection and use of the tissues were approved by the Institutional Ethics Committees of the First Affiliated Hospital of Chongqing Medical University. The approval number of this study in the Institutional Ethics Committees is 2017</ns0:p><ns0:p>Research Ethics (2017-012).</ns0:p></ns0:div>
<ns0:div><ns0:head>RNA isolation and RT-qPCR</ns0:head><ns0:p>PeerJ reviewing PDF | (2020:02:45995:1:2:NEW 16 Jul 2020)</ns0:p><ns0:p>Manuscript to be reviewed As previously described <ns0:ref type='bibr' target='#b32'>(Qiu et al., 2018)</ns0:ref>, we followed the manufacturer's instruction and extracted the total RNA using TRIzol reagent (Life Technologies Inc., USA). RT-qPCR was performed in an ABI 7500 Real-Time PCR System (Applied Biosystems) using 21 paired tissues to examine the expression of IGSF10. Relative quantification of mRNA expression levels of IGSF10 were standardized to the expression levels of GAPDH. The primer pairs used were as follows:</ns0:p><ns0:p>Forward primer (IGSF10): 5'-TTGGAGTTTGCCTGATGGAAC-3';</ns0:p><ns0:p>Reverse primer (IGSF10): 5'-CGCTACCCCAACTTTGTTGAAG-3'; Forward primer (GAPDH): 5'-GGAGCGAGATCCCTCCAAAAT-3';</ns0:p><ns0:p>Reverse primer (GAPDH): 5'-GGCTGTTGTCATACTTCTCATGG-3'.</ns0:p></ns0:div>
<ns0:div><ns0:head>IHC</ns0:head><ns0:p>The process of IHC was previously described <ns0:ref type='bibr' target='#b26'>(Li et al., 2018</ns0:ref>). An anti-IGSF10 rabbit polyclonal antibody (ab197671, 1:100, Abcam), a secondary antibody (ZSGB 1:100 SPN9001) and HRP (ZSGB 1:100 SPN9001) were used. In all, 31 paired tissues were subjected to IHC. The IHC staining intensity scoring criteria were as follows: 0, none; 1, weak; 2, medium; 3, strong.</ns0:p><ns0:p>The scoring criteria for the proportion of positive tumor cells were as follows: 0, < 5%; 1, 5%-25%; 2, 26%-50%; 3, 51%-75%; 4, > 75%.</ns0:p></ns0:div>
<ns0:div><ns0:head>Bioinformatics analyses</ns0:head><ns0:p>The expression of IGSF10 in different subtypes of breast cancer was analyzed by UALCAN, which is a web portal that can be used to evaluate gene expression in different tumor subtypes according to various clinicopathologic features in the TCGA database <ns0:ref type='bibr' target='#b5'>(Chandrashekar et al., 2017)</ns0:ref>.</ns0:p><ns0:p>The mRNA expression of IGSF10 in different breast cancer datasets was evaluated using Oncomine gene expression array datasets <ns0:ref type='bibr'>(Rhodes et al.)</ns0:ref>. The cutoff P-value and fold change were defined as 0.01 and 2, respectively.</ns0:p></ns0:div>
<ns0:div><ns0:head>The relationship between IGSF10 expression and prognosis in different breast cancer</ns0:head><ns0:p>PeerJ reviewing <ns0:ref type='table'>PDF | (2020:02:45995:1:2:NEW 16 Jul 2020)</ns0:ref> Manuscript to be reviewed molecular subtypes was analyzed using a Kaplan-Meier plotter (http://kmplot.com/analysis/) <ns0:ref type='bibr'>(A et al., 2016)</ns0:ref>. The Affymetrix probe set ID of IGSF10 is: 230670_at. According to the mean value of the mRNA expression level of IGSF10, patients were automatically stratified into IGSF10-high and IGSF10-low groups.</ns0:p></ns0:div>
<ns0:div><ns0:head>GSEA</ns0:head><ns0:p>This method was previously described in <ns0:ref type='bibr' target='#b19'>(Jiao, Fu, Li, Meng, & Liu, 2018)</ns0:ref>. We performed GSEA (http://software.broadinstitute.org/gsea) to explore the association between IGSF10 expression and biological processes/pathways according to the instructions of the user guide. We performed GSEA using a microarray dataset (GSE1456) and the TCGA microarray dataset.</ns0:p></ns0:div>
<ns0:div><ns0:head>Additional statistical analyses</ns0:head><ns0:p>All statistical analyses were performed using SPSS (version 23.0). OS and RFS were calculated by Kaplan-Meier curves. The differences between two groups were evaluated using Student's t test. Significance was set at P-value less than 0.05.</ns0:p></ns0:div>
<ns0:div><ns0:head>Results</ns0:head></ns0:div>
<ns0:div><ns0:head n='1.'>The expression of IGSF10 in breast cancer and its clinicopathological features</ns0:head><ns0:p>We examined the mRNA expression levels of IGSF10 of 1095 breast cancer patients in the TCGA database. Our results suggested that the mRNA expression level of IGSF10 in adjacent normal tissues was higher than in breast cancer tissues (Figure <ns0:ref type='figure'>A</ns0:ref>). We then detected the differences in IGSF10 expression in 21 paired tissue samples by RT-qPCR. Consistent with the results in the TCGA database, the expression of IGSF10 was drastically downregulated in breast cancer tissues (Figure <ns0:ref type='figure' target='#fig_5'>1B</ns0:ref>) (Supplemental table 1). We collected 31 pairs of breast cancer and corresponding normal tissues and performed IHC. The results showed that the staining scores of the breast cancer tissues were significantly lower compared with that of adjacent normal tissues (Figures <ns0:ref type='figure' target='#fig_5'>1C-1G</ns0:ref>). Finally, we examined the mRNA expression levels of IGSF10 in breast cell lines. We found PeerJ reviewing PDF | (2020:02:45995:1:2:NEW 16 Jul 2020)</ns0:p><ns0:p>Manuscript to be reviewed that the expression of IGSF10 in the normal breast epithelial cell line MCF10A was higher than that in breast cancer cell lines (Figure <ns0:ref type='figure' target='#fig_5'>1E</ns0:ref>) (Supplemental table 2).</ns0:p><ns0:p>To further analyze the clinical correlation between IGSF10 and breast cancer, the TCGA cohort, which included 700 breast cancer patients, was analyzed (Supplemental table <ns0:ref type='table'>3</ns0:ref>). The results showed that the expression of IGSF10 was related to age (P<0.001), tumor size (P=0.003), and TNM stage (P =0.03) (Table <ns0:ref type='table'>1</ns0:ref>).</ns0:p></ns0:div>
<ns0:div><ns0:head n='2.'>High expression of IGSF10 correlated with better prognosis in breast cancer patients</ns0:head><ns0:p>The association of IGSF10 expression with overall survival (OS) and relapse-free survival (RFS) was evaluated using Kaplan-Meier survival curves. Patients were divided by the median IGSF10 mRNA expression level in the TCGA dataset (Supplemental table 4). Patients with a high expression level of IGSF10 were significantly more likely to have a better OS (hazard ratio (HR)=0.63, 95% confidence interval (CI): 0.41-0.97, P<0.05) (Figure <ns0:ref type='figure' target='#fig_2'>2A</ns0:ref>) and a better RFS (HR=0.53, 95% CI: 0.30-0.93, P<0.05) (Figure <ns0:ref type='figure' target='#fig_2'>2B</ns0:ref>) than those with a low expression level of IGSF10. Subsequently, we used the UALCAN database to further evaluate the prognostic value of IGSF10 by stratifying patients into different molecular subtypes. Decreased mRNA levels of IGSF10 were observed in luminal, HER2 positive, and triple-negative breast cancer samples compared with those in normal samples (Figure <ns0:ref type='figure' target='#fig_3'>3A</ns0:ref>). We also found that a low expression level of IGSF10 was significantly related to poor OS in patients with basal (HR =0.44, 95% CI: 0.22-0.86, P=0.013), luminal A (HR=0.47, 95% CI: 0.25-0.88, P=0.017), and HER2+ (HR=0.28, 95% CI: 0.09-0.81, P=0.012) breast cancer subtypes (Figures <ns0:ref type='figure' target='#fig_3'>3B-3E</ns0:ref>). However, no significant relationship was observed between IGSF10 expression and OS in luminal B patients (HR=0.61, 95% CI:0.3-1.23, P=0.17) (Figure <ns0:ref type='figure' target='#fig_3'>3D</ns0:ref>). The multivariate Cox regression analysis in breast cancer patients from the TCGA showed that IGSF10 was an independent prognostic factor for OS (HR=1.793, 95% CI:</ns0:p><ns0:p>1.141-2.815, P=0.011) and RFS (HR=2.298, 95% CI: 1.317-4.010, P=0.003) (Table <ns0:ref type='table'>2</ns0:ref>).</ns0:p></ns0:div>
<ns0:div><ns0:head n='3.'>Potential biological roles and signaling pathways related to IGSF10</ns0:head><ns0:p>PeerJ reviewing PDF | (2020:02:45995:1:2:NEW 16 Jul 2020)</ns0:p><ns0:p>Manuscript to be reviewed Potential mechanisms and signaling pathways that may be related to IGSF10 in regulating the development of breast cancer were explored using GSEA. According to the median value of IGSF10 expression in the microarray dataset (GSE1456) and TCGA datasets, we assigned patients to two groups. We found that nine gene sets were enriched in the GSE1456 dataset and that 16 gene sets were enriched in the TCGA datasets (P<0.05; FDR<0.25) (Figures <ns0:ref type='figure' target='#fig_4'>4A-4B</ns0:ref>) (Supplemental table 5). Interestingly, the results showed that IGSF10 was positively related to several cancer-related biological processes including the DNA repair (HALLMARK_DNA_REPAIR), cell cycle (HALLMARK_G2M_CKECKPOINT), and glycolysis (HALLMARK_GLYCOLYSIS) pathways in both datasets (Figures <ns0:ref type='figure' target='#fig_4'>4C-4E</ns0:ref>). The PI3K/Akt/mTOR and mTORC1 signaling pathways were also associated with IGSF10 (Figures <ns0:ref type='figure' target='#fig_4'>4F-4G</ns0:ref>). Moreover, in the TCGA dataset, the transforming growth factor-β (TGF-β) signaling pathway (HALLMARK_TGF_BETA_SIGNALING), epithelial mesenchymal transition (EMT) (HALLMARK_EPITHELIAL_MESENCHYMAL_TRANSITION) and tumor necrosis factor (TNF) signaling pathway (HALLMARK_TNFA_SIGNALING_VIA_NFKB) were significantly enriched in the IGSF10-low expression group (Supplemental Figure <ns0:ref type='figure' target='#fig_5'>1A</ns0:ref>). These results indicated a possible mechanism for the involvement of IGSF10 in the tumorigenesis of breast cancer.</ns0:p></ns0:div>
<ns0:div><ns0:head>Discussion</ns0:head><ns0:p>In recent years, numerous molecular prognostic biomarkers have been developed and validated in cancers, including breast cancer <ns0:ref type='bibr' target='#b30'>(Nicolini, Ferrari, & Duffy, 2018)</ns0:ref>. In the present study, we demonstrated that IGSF10 may serve as a prognostic biomarker in breast cancer and provided a possible mechanism for its involvement in the tumorigenesis of breast cancer.</ns0:p><ns0:p>In this study, we explored the role of IGSF10 in breast cancer through a TCGA dataset, RT-qPCR and IHC experiments. Our data indicated that IGSF10 was significantly downregulated in breast cancer tissues. Consistent with our results, multiple datasets in the Oncomine database suggested that the expression of IGSF10 was down-regulated in breast cancer (fold change>2) including in the TCGA Breast, Karnoub Breast <ns0:ref type='bibr'>(Karnoub et al., 2007), Zhao Breast (Zhao et al.,</ns0:ref> PeerJ reviewing PDF | (2020:02:45995:1:2:NEW 16 Jul 2020)</ns0:p><ns0:p>Manuscript to be reviewed 2004), Richardson Breast 2 <ns0:ref type='bibr' target='#b34'>(Richardson et al., 2006), and</ns0:ref><ns0:ref type='bibr'>Finak Breast (Finak et al., 2008)</ns0:ref> datasets (Supplemental table <ns0:ref type='table'>6</ns0:ref>). By analyzing the data in the UALCAN database, we found that the expression of IGSF10 was related to the molecular subtype of breast cancer. In addition, our data showed that the expression of IGSF10 was closely associated with age, tumor size, and TNM stage.</ns0:p><ns0:p>Accordingly, IGSF10 may play a crucial role in breast cancer and have the potential to be targeted by anticancer therapy. Moreover, the survival analysis indicated that breast cancer patients with higher expression of IGSF10 had better OS and RFS. The multivariate analysis demonstrated that IGSF10 was an independent prognostic factor for breast cancer patients. Interestingly, according to a subgroup analysis, we found that IGSF10 was significantly related to OS in basal, luminal A and HER2-positive breast cancer patients. These results indicated that IGSF10 may be a prognostic biomarker in breast cancer.</ns0:p><ns0:p>Previous studies have suggested that IGSF10 may exert an important influence on tumorigenesis. Ling and colleagues claimed that IGSF10 knockout promoted the development of lung cancer and that IGSF10 mainly activated the integrin-β1/FAK pathway in lung cancer. <ns0:ref type='bibr' target='#b28'>(Ling et al., 2020)</ns0:ref>. In one family with gastric and colorectal cancers, <ns0:ref type='bibr'>Thutkawkorapin et al. identified 12</ns0:ref> new nonsynonymous single nucleotide variants, which might contribute to an increased cancer risk, in 12 different genes, including IGSF10 <ns0:ref type='bibr' target='#b40'>(Thutkawkorapin et al., 2016)</ns0:ref>. <ns0:ref type='bibr'>Chang et al. identified</ns0:ref> new mutations in endometrial cancer patients in Taiwan by whole-exome sequencing and found that IGSF10, as a passenger gene, may be associated with endometrial cancer <ns0:ref type='bibr' target='#b6'>(Chang, Huang, Yeh, & Chang, 2017)</ns0:ref>. However, to our knowledge, no studies have reported the possible functions and mechanisms of IGSF10 in breast cancer.</ns0:p><ns0:p>During the past decade, growing evidence has shown clear correlations between immunoglobulin superfamily members and human diseases. For instance, studies have reported that the loss-of-function mutations in IGSF1 result in an X-linked syndrome of central hypothyroidism and testicular enlargement. IGSF1 mutations in male patients lead to a late increase in testosterone levels <ns0:ref type='bibr' target='#b15'>(Howard et al., 2016;</ns0:ref><ns0:ref type='bibr'>Roche et al., 2018;</ns0:ref><ns0:ref type='bibr' target='#b38'>Sun et al., 2012)</ns0:ref>.</ns0:p><ns0:p>Significantly better OS was observed for pediatric patients with mixed-lineage leukemia-PeerJ reviewing PDF | (2020:02:45995:1:2:NEW 16 Jul 2020)</ns0:p><ns0:p>Manuscript to be reviewed rearranged acute monoblastic leukemia with t(9; 11) (p22; q23)and high IGSF4 expression than those with low IGSF4expression <ns0:ref type='bibr' target='#b25'>(Kuipers et al., 2011)</ns0:ref>. Wang et al. demonstrated that IGSF8</ns0:p><ns0:p>promoted melanoma proliferation and metastasis by negatively regulating the TGF-β signaling pathway <ns0:ref type='bibr' target='#b41'>(Wang, Sharma, Knoblich, Granter, & Hemler, 2015)</ns0:ref>.</ns0:p><ns0:p>In this study, potential biological roles and signaling pathways that may be related to IGSF10 in breast cancer were analyzed by GSEA. Several biological processes, including DNA repair, the cell cycle, and glycolysis, were found to be associated with IGSF10. Among these processes, the genomic integrity can be maintained through DNA repair pathways. The dysregulation of DNA repair leads to changes in the genome and causes physiological changes in cells that drive tumor initiation <ns0:ref type='bibr' target='#b17'>(Jeggo, Pearl, & Carr, 2016;</ns0:ref><ns0:ref type='bibr' target='#b24'>Khanna, 2015;</ns0:ref><ns0:ref type='bibr' target='#b29'>Mouw, Goldberg, Konstantinopoulos, & D'Andrea, 2017)</ns0:ref>. The cell cycle regulates tumor growth and glycolysis modulates tumor microenvironment heterogeneity. These biological processes were related to tumor progression, metastasis and drug resistance <ns0:ref type='bibr' target='#b16'>(Jahagirdar et al., 2018)</ns0:ref>. Moreover, in human malignancies, the mTORC1 and PI3K/Akt/mTOR signaling pathways are usually abnormally activated and promote the development of malignancies <ns0:ref type='bibr' target='#b13'>(Hare & Harvey, 2017)</ns0:ref>. Previous studies have indicated that mTORC1 promotes cell growth by activating key anabolic processes and the dysregulation of mTORC1 is the basis of many human cancers <ns0:ref type='bibr' target='#b3'>(Ben-Sahra & Manning, 2017;</ns0:ref><ns0:ref type='bibr'>Keppler-Noreuil, Parker, Darling, & Martinez-Agosto, 2016)</ns0:ref>. The PI3K/Akt/mTOR pathway is related to various biological processes in breast cancer, such as tumorigenesis, cellular transformation, tumor progression, and drug resistance <ns0:ref type='bibr' target='#b12'>(Guerrero-Zotano, Mayer, & Arteaga, 2016)</ns0:ref>. Therefore, we speculated that IGSF10 might mechanically regulate the growth of breast cancer cells through the mTORC1 and PI3K/Akt/mTOR signaling pathways. Intriguingly, we also found that IGSF10 was associated with EMT, the TGF-β signaling pathway and the TNF signaling pathway in the TCGA database. The TGF-β signaling pathway was reported to be associated with various tumors and it has been found to regulate the biological processes in multiple cancers, including growth, migration, invasion, apoptosis and EMT <ns0:ref type='bibr' target='#b2'>(Bedi et al., 2012;</ns0:ref><ns0:ref type='bibr' target='#b39'>Tang et al., 2017;</ns0:ref><ns0:ref type='bibr' target='#b45'>Yu et al., 2018;</ns0:ref><ns0:ref type='bibr' target='#b47'>Zhao et al., 2018)</ns0:ref>. EMT plays crucial roles in the metastasis and invasion of breast cancer by regulating Manuscript to be reviewed cell motility and invasiveness <ns0:ref type='bibr' target='#b10'>(Feng et al., 2016)</ns0:ref>. Moreover, TNF-α is highly correlated with inflammation in breast tumors and its elevation is highly associated with relapse and advanced disease <ns0:ref type='bibr' target='#b23'>(Katanov et al., 2015)</ns0:ref>. However, further studies are needed to elucidate the role of IGSF10 in breast cancer and the detailed mechanisms by which IGSF10 modulates these related signaling pathways.</ns0:p></ns0:div>
<ns0:div><ns0:head>Conclusions</ns0:head><ns0:p>In summary, we reported that IGSF10 was expressed at a low level in breast cancer. The expression of IGSF10 was significantly related to age, tumor size, and tumor stage. More importantly, IGSF10 was an independent prognostic factor of better outcomes in breast cancer patients. In addition, the GSEA results showed that IGSF10 was significantly associated with the DNA repair, cell cycle, glycolysis, mTORC1 and PI3K/Akt/mTOR signaling pathways. Overall, we suggested a novel role of IGSF10 in breast cancer. Our data may provide new insights in the identification of potential therapeutic targets in breast cancer patients. Manuscript to be reviewed </ns0:p></ns0:div>
<ns0:div><ns0:head>Figure legends</ns0:head></ns0:div><ns0:figure xml:id='fig_0'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:02:45995:1:2:NEW 16 Jul 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_1'><ns0:head>Figure 1 :</ns0:head><ns0:label>1</ns0:label><ns0:figDesc>Figure 1: (A) Bioinformatics analysis of IGSF10 expression in the TCGA database. (B) The mRNA expression of IGSF10 in BC tissues and matched adjacent normal tissues was evaluated by qRT-PCR (n = 21). (C-F) Representative IHC images of BC specimens and adjacent normal breast tissues. (G) IHC score of IGSF10 protein expression in 31 BC cases (IHC score: 3.12 ± 2.04) and 31 normal samples (IHC score: 4.45 ± 2.13). Data are shown as the Mean ± SD, unpaired t-test, *P < 0.05. (H) qRT-PCR was used to examine IGSF10 expression in human breast cancer cells vs. MCF-10A cells, *P < 0.05, **P < 0.01, ***P < 0.001.</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_2'><ns0:head>Figure 2 :</ns0:head><ns0:label>2</ns0:label><ns0:figDesc>Figure 2: Kaplan-Meier survival curve of TCGA breast cancer patients stratified into the IGSF10high and IGSF10-low groups by the median expression value. P < 0.05 was considered statistically significant. (A) Overall survival curves of breast cancer patients. (B) Relapsefree survival curves of breast cancer patients.</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_3'><ns0:head>Figure 3 :</ns0:head><ns0:label>3</ns0:label><ns0:figDesc>Figure 3: (A) IGSF10 expression in different molecular subtypes of breast cancer in the TCGA database. (B) Basal breast cancer; (C) Luminal A breast cancer; (D) Luminal B breast cancer; (E) HER2+ breast cancer. All the overall survival curves were plotted by the Kaplan-Meier plotter (http://kmplot.com/analysis/). **P < 0.01.</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_4'><ns0:head>Figure 4 :</ns0:head><ns0:label>4</ns0:label><ns0:figDesc>Figure 4: The gene sets with a normal P-value < 0.05 and a false discovery rate (FDR) < 0.25 were considered significant. Gene sets were ranked by normalized enrichment score (NES). (A) The gene sets enriched in the GSE1456 dataset. (B) The gene sets enriched in the TCGA dataset. (C-E) the GSEA enrichment plot showed that IGSF10 was positively associated with DNA repair, cell cycle, and glycolysis. (F-G) The GSEA enrichment plot showed that IGSF10 was positively correlated with the PI3K/Akt/mTOR and mTORC1 signaling pathways.</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_5'><ns0:head>Figure 1</ns0:head><ns0:label>1</ns0:label><ns0:figDesc>Figure 1</ns0:figDesc></ns0:figure>
<ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='22,42.52,301.12,525.00,202.50' type='bitmap' /></ns0:figure>
<ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='24,42.52,70.87,397.24,672.95' type='bitmap' /></ns0:figure>
<ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='26,42.52,70.87,525.00,405.00' type='bitmap' /></ns0:figure>
</ns0:body>
" | "Response to Reviewers Comments
We would like to express our sincere thanks to the reviewers for the constructive and positive comments. Our responses are as follows.
A. Responses to comments of reviewer 1 (Mattheos Bobos)
1. The results from the IGSF10 analyses on their cohort should be presented clearly on the results and discussed thoroughly.
Thank you for this comment. The results and discussion section have been carefully revised, thanks for your suggestion.
2. The findings on their cohort and the TCGA data should be presented on a separate paragraph, on the results section.
Thank you for your suggestion. We have presented the TCGA data on a separate paragraph.
3. Tables presenting their data concerning IGSF10 mRNA and protein expression, as well as the correlations with the clinicopathological data, would be a significant add-on for the study. It could be added as a Supplemental Tables.
Thank you very much for the valuable suggestions. In fact, we previously made statistics on the relationship between the expression of IGSF10 and its clinicopathological. However, the number of cases is too small to get positive results. Therefore, we did not present the data in this study. Thanks again for your suggestions.
4. The authors should present the number of tumour and adjacent normal tissues from breast cancer patients that they used in this study on the methods. The number of the cases is 159 as mentioned on the results (section 3 - Potential biological roles and signalling pathways related to IGSF10)?
Thanks a lot for your valuable suggestions. Microarray dataset (GSE1456) containing 159 breast cancer samples was used for GSEA. The number of tumor and adjacent normal tissues used in this study has been presented on the METHODS section.
5. The authors should add a Remark diagram
Thank you for this comment. We have added the figure legends in the manuscript.
6. Add the source of the secondary antibody and HRP used on the study.
Thank you for this comment. The source of the secondary antibody and HRP has been described in the METHOD (IHC) section.
7. Figure 1, panel A, variable “Tumor”. The number of tumor cases (1102) is probably wrong!
Thank you for this comment. We corrected the number of tumor cases in Figure 1A.
8. Figure 1, panel A, variable “Paratumor”. The authors here denote “normal tissue”. Should be corrected.
Thank you for this comment. We corrected the variable “normal tissue” in Figure 1A.
9. Figure 2. Add the number of the patients included on the KM analysis
Thank you very much for the valuable suggestions. We have added the case number of each group, Hazard Ratio, 95% confidence interval of ratio and P value in Figure 2.
10. The sentence “In Oncomine database, the mRNA expression of IGSF10 was much lower in breast cancer than normal tissues within datasets including TCGA Breast, Karnoub Breast (Karnoub et al., 2007), Zhao Breast (Zhao et al., 2004), Richardson Breast 2 (Richardson et al., 2006), and Finak Breast (Finak et al., 2008) (Table 1).”, should be transferred to the discussion and the table as Supplemental Table 1.
Thanks a lot for your valuable suggestions. The sentence you mentioned has been transferred to the discussion and the Table 1 has been adjusted to Supplemental Table 5.
11. The entitled on the results section “The prognostic value of IGSF10 in breast cancer”, refers not only to prognostic value of IGSF10, but and the correlation of the particular molecule with other clinicopathological variables. The title of the section should be changed accordingly.
Your recommendation is greatly appreciated. We changed the title of the section according to your suggestion.
12. The Table 3 in which cohort refers? The TCGA study? If yes should be mentioned on the legend.
Thank you for this comment. The legend of Table 3 has been changed according to your suggestion.
13. Figure 1, panel D. The data should be presented as box- plot.
Thank you for this comment. Figure 1D has been replaced by box-plot image to better present the protein expression level of IGSF10 staining.
14. Supplemental figures 1-18 should be deleted
Your recommendation is greatly appreciated. These figures have been deleted.
15. The “breast_cancer_tissue”, “CELL_LINE”, “GSEA”, “survival”, “TCGA clinical_data” should be added as Supplemental tables, but it must mentioned on the text.
Your recommendation is greatly appreciated. We added these tables as Supplemental tables and mentioned them in the RESULT section on the text. Thank you again for your suggestive advice, and we hope to learn more from you in the future.
B. Responses to comments of reviewer 2 (Deepak Balamurali)
1. Cell Culture: The authors have used 7 different cell-lines in this study. It is advisable that the authors provide cell-line authenticity certification to ensure that it is the right cell-lines as claimed, due to a lot of cell-line misidentification in recent times. Additionally, this also increases the transparency of the project.
Thank you for this comment and your concern is understandable. We have uploaded the cell-line authenticity certifications of MDA-MB-231, MCF-7 and MCF10A as supplementary materials. The cell-line authenticity certifications of other four cell-lines cannot be provided. But these cell lines have been used in different experiments in our laboratory for many years, and the results obtained have been published in different journals in recent years [1-4]. The authenticity of these cell lines can be guaranteed. Thanks again for your suggestions.
[1] Shi, Y., Luo, X., Li, P., Tan, J., Wang, X., Xiang, T., and Ren, G. (2015). miR-7-5p suppresses cell proliferation and induces apoptosis of breast cancer cells mainly by targeting REGgamma. Cancer Lett 358, 27-36.
[2] Feng, Y., Wu, M., Li, S., He, X., Tang, J., Peng, W., Zeng, B., Deng, C., Ren, G., and Xiang, T. (2018). The epigenetically downregulated factor CYGB suppresses breast cancer through inhibition of glucose metabolism. J Exp Clin Cancer Res 37, 313.
[3] Li, X., Huang, J., Luo, X., Yang, D., Yin, X., Peng, W., Bi, C., Ren, G., and Xiang, T. (2018). Paired box 5 is a novel marker of breast cancers that is frequently downregulated by methylation. Int J Biol Sci 14, 1686-1695.
[4] Li, Y., Huang, J., Zeng, B., Yang, D., Sun, J., Yin, X., Lu, M., Qiu, Z., Peng, W., Xiang, T., Li, H., and Ren, G. (2018). PSMD2 regulates breast cancer cell proliferation and cell cycle progression by modulating p21 and p27 proteasomal degradation. Cancer Lett 430, 109-122.
2. Line 101 – 107: How many tissue samples were collected from patients?
Thank you for this comment. We collected 52 pairs of tumor tissues and adjacent normal tissues from our hospital. In this study, 31 paired breast cancer and normal tissues were used to perform IHC and 21 paired breast cancer and normal tissues were used to perform RT-qPCR.
3. Line 111 – Manufacturer’s instructions
Thank you for this comment. We have described the manufacturer's instructions in the supplementary material: 1. When precipitating RNA from small sample quantities; 2. Add 0.5 mL of 100% isopropanol to the aqueous phase, per 1 mL of TRIzol® Reagent used for homogenization; 3. Incubate at room temperature for 10 minutes; 4. Centrifuge at 12,000 × g for 10 minutes at 4°C; 5. Proceed to RNA wash; 6. Remove the supernatant from the tube, leaving only the RNA pellet; 7. Wash the pellet, with 1 mL of 75% ethanol per 1 mL of TRIzol® Reagent used in the initial homogenization; 8. Vortex the sample briefly, then centrifuge the tube at 7500 × g for 5 minutes at 4°C. Discard the wash; 9. Vacuum or air dry the RNA pellet for 5–10 minutes. Do not dry the pellet by vacuum centrifuge; 10. Resuspend the RNA pellet in RNase-free water or 0.5% SDS solution (20–50 μL) by passing the solution up and down several times through a pipette tip; 11. Incubate in a water bath or heat block set at 55–60°C for 10–15 minutes; 12. Proceed to downstream application, or store at –70°C.
4. Line 151 – 230670_at (Kindly use the correct identifiers)
Thank you for this comment. We corrected the identifier of IGSF10.
5. Supplementary Oncomine figure shows that around 11 datasets were originally considered. Why were the excluded from the final study?
Thank you for this comment. As you mentioned, there are 11 datasets in the ONCOMINE figure, but we chose the datasets with fold change greater than 2 to report. We are sorry that we have not clearly stated this selection criterion. Thanks for your question, we have mentioned our selection criteria in the article.
6. How were the adjacent normal samples confirmed as non-cancerous? Figure 1B shows some cases where the expression difference between adjacent and tumour are not very significant. Perhaps this is due to adjacent normal not being completely “tumour-free” in all aspects?
The adjacent normal samples were obtained from the margin of breast cancer specimen during breast conserving surgery, which was proved to be benign in intraoperative frozen-section examination. For the total breast resection samples, the adjacent normal samples were taken from the tissues more than 5cm away from the edge of the lesions, so as to ensure that the adjacent normal samples were non-cancerous.
7. The expression data from the 21 patients where RT-PCR was performed has not been included in any tables. Perhaps the authors can shed more light on this – what was the distribution like in these 21 samples? How many were IDC/ ILC type? What was the ER/PR/HER2 classification? Though this is a small dataset when compared to TCGA, it will be interesting to see if it matches the data.
Thank you very much for the constructive suggestions. We previously made statistics on the relationship between the expression of IGSF10 and its clinicopathological. However, the number of cases is too small to get positive results. Therefore, we did not present the data in this study. Thanks again for your suggestions.
8. The results presented are indeed interesting. However, it would be too early to consider IGSF-10 as a biomarker with such a limited validation cohort. The authors can also check for associated genes which have correlated expression with IGSF-10 in the specified pathways from GSEA analysis.
Your recommendation is greatly appreciated. We checked for associated genes which have correlated expression with IGSF10 in PI3K/Akt/mTOR signaling pathway and mTORC1 signaling pathway by GSEA according to your good suggestion. There are 16 common related genes in the PI3K/Akt/mTOR signaling pathway and there are 58 common related genes in the mTORC1 signaling pathway. We will do further research to gain more insights in the associated genes of IGSF10 in the future.
9. The results in Table 2 from TCGA cohort indicate that the expression of IGSF 10 decreases with stage-wise progression of BC. Was the same observed in the 21 tissue samples processed for RT-PCR? If not, what are the authors' thoughts on this?
Thank you for this comment. Our results showed that there was no correlation between the progression of breast cancer and the expression of IGSF10 in the 21 tissue samples. We think this is because the number of cases is too small so the results are not statistically significant. In the future, we will do further research to verify this result. Thank you again for your suggestion
10. This manuscript by Wang et al. explores the role of IGSF-10 as a potential biomarker of breast cancer. Though the concept of the paper has been explained well, the manuscript would further benefit from being proofread by a native speaker or a language editor.
Thanks a lot for your comment. We have carefully checked the entire manuscript, and the English language has been improved by a language editing service, and the errors have been corrected.
11. The authors have provided IHC images & some analysis data has supplementary files – but relevant Figure/Table legends have not been provided. It would be easier if the authors can amend this suitably.
We are sorry we did not provide the Figure/Table legends. The relevant Figure/Table legends have been provided. Thanks again for your careful review and comments!
C. Responses to comments of reviewer 3 (Xin Zhang)
1. “Materials & Methods – Patients and tissue samples of breast cancer” Line 105-106. Please provide the approval number of this study in the Institutional Ethics Committees.
Thank you for this comment. The approval number of this study in the Institutional Ethics Committees is 2017 Research Ethics (2017-012). We have also provided a copy in English of the ethical approval document as Supplement Material.
2. “Materials & Methods – Patients and tissue samples of breast cancer” Line 102-104. Please provide the table of the clinical and pathological information for the “Tissue samples of breast cancer” as same as “fully clinical information of TCGA”. In addition, please analyze the correlation between IHC score and the clinicopathological information of breast cancer patients. (Line 124-139, and Line 188-206)
Thank you very much for the valuable suggestions. We previously made statistics on the relationship between the expression of IGSF10 and its clinicopathological. However, the number of cases is too small to get positive results. Therefore, we did not present the data in this study. Thanks again for your suggestions. In the future, we will do further research to verify this result.
3. “Result – 1. The expression of IGSF10 in breast cancer” Line 172-174, Figure 1. “…the mRNA expression level of IGSF10 was significantly down-regulated in breast cancer tissues compared with normal tissues.” There is a difference between tumor adjacent normal tissue and normal tissue. Be sure to use well-defined words throughout the manuscript.
We deeply appreciate your valuable suggestion. We have carefully checked the entire manuscript and corrected these words.
4. “Result – 1. The expression of IGSF10 in breast cancer” Line 180-183, Figure 1C and D. Please provide a larger magnification of the micrograph. From the available images in figure 1C, it appears that the normal breast tissue is closer to hyperplasia.
Thank you very much for your suggestions. We have provided a larger magnification of the IHC image in Figure 1 C.
5. “Result – 1. The expression of IGSF10 in breast cancer” Line 183-186. Please provide the protein expression level of IGSF10 in cell line.
Thank you for this comment. We are very sorry for not verifying the protein expression of IGSF10 in cell lines. Due to the outbreak of COVID-19 in China, our laboratory has been closed for several months and since the time was limited, this part of experiments will certainly be our future project. Thank you again for your suggestive advice.
6. “Result – 2. The prognostic value of IGSF10 in breast cancer” Line 194-195. “Patients with high expression level of IGSF10 were significantly associated with better OS (Figure 2A) and RFS (Figure 2B) than those with low levels of IGSF10.” Please explain how to define the high and low expression groups of IGFS10. In the survival curve of figure 2, the case number of each group, Hazard Ratio, 95% confidence interval of ratio and P value should be indicated.
Thank you very much for the valuable suggestions. Patients were divided into IGSF10-high expression group and IGSF10-low expression group based on the median value of IGSF10 mRNA expression level in the microarray dataset and we have mentioned this in the RESULT section. We also added the case number of each group, Hazard Ratio, 95% confidence interval of ratio and P value in Figure 2.
7. “Result – 3. Potential biological roles and signaling pathways related to IGSF10” Line 208-218. Why GSE1456 was used in GSEA analysis? TCGA dataset can also be used for GSEA analysis. Therefore, I suggest the authors should add the GSEA analysis of the TCGA dataset and display the common pathway appearing in the two datasets (TCGA and GEO1456), so as to increase readers' understanding of the potential functions of IGSF10.
Your recommendation is greatly appreciated. We used TCGA dataset for GSEA analysis according to your suggestions, and the results showed that it was consistent with our GSEA results using GSE1456 dataset (Figure 4B), which increases the credibility of our research. At the same time, we also found other potential biological roles and related pathways of IGSF10, such as EMT, TGF-β and TNF-α signaling pathways. This improves readers' understanding of the potential functions of IGSF10. Thank you again for your suggestive advice, and we hope to learn more from you in the future.
" | Here is a paper. Please give your review comments after reading it. |
9,767 | "<ns0:abstract xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:p>Background. Immunoglobulin superfamily member 10 (IGSF10) is a member of the immunoglobulin superfamily that is expressed at high levels in both the gallbladder and ovary. Currently, the role and possible mechanism of IGSF10 in breast cancer remain unclear. Method. By applying real-time quantitative polymerase chain reaction (qRT-PCR) and immunohistochemistry (IHC), the expression of IGSF10 in breast cancer cells and tissues was detected. We collected the clinical information from 700 patients with breast cancer in The Cancer Genome Atlas (TCGA), and analyzed the relationship between IGSF10 expression and the clinicopathological features and survival outcomes of these patients.</ns0:p><ns0:p>The potential mechanisms and pathways associated with IGSF10 in breast cancer were explored by performing a gene set enrichment analysis (GSEA). Results. According to TCGA data, qRT-PCR and IHC experiments, levels of the IGSF10 mRNA and protein were significantly decreased in breast cancer tissues. IGSF10 expression was significantly correlated with age, tumor size, and tumor stage. Moreover, shorter overall survival (OS) and relapse-free survival (RFS) correlated with lower IGSF10 expression, according to the survival analysis . The multivariate analysis identified that IGSF10 as an independent prognostic factor for the OS (hazard ratio (HR)=1.793, 95% confidence interval (CI):</ns0:p><ns0:p>1.141-2.815, P=0.011) and RFS (HR=2.298, 95% CI: 1.317-4.010, P=0.003) of patients with breast cancer. Based on the GSEA, IGSF10 was involved in DNA repair, cell cycle, and glycolysis. IGSF10 was also associated with the PI3K/Akt/mTOR and mTORC1 signaling pathways. Conclusions. This study revealed a clear relationship between IGSF10 expression and the tumorigenesis of breast cancer for the first time. Therefore, further studies are needed to understand the mechanism of IGSF10 in breast cancer.</ns0:p></ns0:div>
</ns0:abstract>
<ns0:body xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:head>Introduction</ns0:head><ns0:p>PeerJ reviewing PDF | (2020:02:45995:2:0:NEW 8 Sep 2020)</ns0:p><ns0:p>Manuscript to be reviewed Breast cancer is a common malignancy that seriously threatens women's health.</ns0:p><ns0:p>Approximately 2.1 million female patients were newly diagnosed with breast cancer worldwide in 2018. Breast cancer accounts for one-quarter of all female cancer cases <ns0:ref type='bibr' target='#b4'>(Bray et al., 2018)</ns0:ref>. As a heterogeneous disease, the initiation and development of breast cancer are affected by both genetic and environmental factors <ns0:ref type='bibr' target='#b40'>(Yang et al., 2019)</ns0:ref>. Despite continuous advances in surgical techniques, biological drugs and targeted therapies, breast cancer remains an arduous clinical problem <ns0:ref type='bibr' target='#b39'>(Woolston, 2015)</ns0:ref>. Therefore, the identification of breast cancer biomarkers is crucial for obtaining an understanding of the tumorigenesis and accurate cancer prognosis, as biomarkers may assist with the clinical diagnosis and serve as potential tumor therapeutic targets in patients with breast cancer <ns0:ref type='bibr' target='#b7'>(Costa-Pinheiro et al., 2015;</ns0:ref><ns0:ref type='bibr'>John R Prensner, 2012;</ns0:ref><ns0:ref type='bibr' target='#b30'>Qiao et al., 2019)</ns0:ref>.</ns0:p><ns0:p>Immunoglobulin superfamily member 10 (IGSF10) is a gene involved in cell differentiation and developmental processes <ns0:ref type='bibr' target='#b37'>(Thutkawkorapin et al., 2016)</ns0:ref>. Mutations in IGSF10 delay human puberty <ns0:ref type='bibr' target='#b14'>(Howard, 2018;</ns0:ref><ns0:ref type='bibr' target='#b15'>Howard et al., 2016)</ns0:ref>. Moreover, during embryonic development, mutations in IGSF10 lead to the dysregulation of gonadotropin-releasing hormone (GnRH)associated neuronal migration. Based on accumulating evidence, IGSF10 deficiency may lead to a transient GnRH deficiency and reversible congenital hypogonadotropic hypogonadism <ns0:ref type='bibr' target='#b1'>(Amato et al., 2019;</ns0:ref><ns0:ref type='bibr' target='#b14'>Howard, 2018)</ns0:ref>. Moreover, mutations in IGSF10 likely contribute to an increased risk of rectal and gastric cancers <ns0:ref type='bibr' target='#b37'>(Thutkawkorapin et al., 2016)</ns0:ref>. As shown in the study by <ns0:ref type='bibr'>Daino et al.,</ns0:ref> IGSF10 is significantly downregulated in a rat model of alpha-radiation-induced osteosarcoma <ns0:ref type='bibr' target='#b8'>(Daino et al., 2009)</ns0:ref>. The expression of IGSF10 is downregulated in lung cancer tissues, and decreased expression of IGSF10 correlated with a poor prognosis for patients with lung cancer <ns0:ref type='bibr' target='#b27'>(Ling et al., 2020)</ns0:ref>. However, the biological roles of IGSF10 in the majority of cancers have not been investigated, and its role in breast cancer remains largely unknown.</ns0:p><ns0:p>In the present study, the expression of IGSF10 in collected breast cancer tissues was examined using qRT-PCR and IHC. The clinicopathological features of the disease based on IGSF10 expression and Kaplan-Meier survival curves were analyzed using public data from The Cancer Genome Atlas (TCGA) database. In addition, a gene set enrichment analysis (GSEA) was performed to explore the potential mechanisms and signaling pathways by which IGSF10 may mediate breast tumorigenesis.</ns0:p></ns0:div>
<ns0:div><ns0:head>Materials & Methods</ns0:head></ns0:div>
<ns0:div><ns0:head>Cell culture</ns0:head><ns0:p>The breast cancer cell lines: MDA-MB-231, MCF-7, BT-549, ZR-75-30, SKBR-3, and T47D (ATCC, Manassas, VA, USA) were maintained as previously described <ns0:ref type='bibr'>(Zhang et al., 2019)</ns0:ref>. The normal mammary epithelial cell line MCF-10A was also maintained as previously described <ns0:ref type='bibr' target='#b9'>(Debnath, 2003)</ns0:ref>. All cell lines were cultured in a humidified incubator at 37°C with an atmosphere containing 5% CO 2 .</ns0:p></ns0:div>
<ns0:div><ns0:head>Patients with breast cancer and tissue samples</ns0:head><ns0:p>TCGA data were utilized as previously described <ns0:ref type='bibr'>(Qiu, 2018)</ns0:ref>. In the present study, we analyzed IGSF10 expression in 1095 patients with breast cancer in TCGA database. We included 700 patients with breast cancer who had complete RNA-seq data and complete clinical information to analyze the clinical correlation between IGSF10 expression and breast cancer. The following clinical information was collected: age, tumor size, lymph node status, tumor, node, metastasis (TNM) stage, estrogen receptor (ER) status, progesterone receptor (PR) status, human epidermal growth factor receptor 2 (HER2) status, and follow-up information.</ns0:p><ns0:p>Breast cancer tissue samples were collected as previously described <ns0:ref type='bibr' target='#b25'>(Li et al., 2018)</ns0:ref>. Manuscript to be reviewed</ns0:p></ns0:div>
<ns0:div><ns0:head>RNA isolation and RT-qPCR</ns0:head><ns0:p>As described in a previous study <ns0:ref type='bibr' target='#b31'>(Qiu et al., 2018)</ns0:ref>, we followed the manufacturer's instructions and extracted the total RNA using TRIzol reagent (Life Technologies Inc., USA). RT-qPCR of 21 paired tissues was performed with an ABI 7500 Real-Time PCR System (Applied Biosystems) to examine IGSF10 expression. Relative quantification of the expression of the IGSF10 mRNA was standardized to the expression levels of GAPDH. The following primer pairs were used in the present study: Forward primer (IGSF10): 5'-TTGGAGTTTGCCTGATGGAAC-3';</ns0:p><ns0:p>Reverse primer (IGSF10): 5'-CGCTACCCCAACTTTGTTGAAG-3'; Forward primer (GAPDH): 5'-GGAGCGAGATCCCTCCAAAAT-3';</ns0:p><ns0:p>Reverse primer (GAPDH): 5'-GGCTGTTGTCATACTTCTCATGG-3'.</ns0:p></ns0:div>
<ns0:div><ns0:head>IHC</ns0:head><ns0:p>The procedure used for IHC was described in a previous study <ns0:ref type='bibr' target='#b25'>(Li et al., 2018</ns0:ref>). An anti-IGSF10 rabbit polyclonal antibody (ab197671, 1:100, Abcam), a secondary antibody (ZSGB 1:100 SPN9001) and HRP (ZSGB 1:100 SPN9001) were used. Thirty-one paired tissues were subjected to IHC. The IHC staining intensity scoring criteria were as follows: 0, none; 1, weak; 2, medium; and 3, strong. The scoring criteria for the proportion of positive tumor cells were as follows: 0, < 5%; 1, 5%-25%; 2, 26%-50%; 3, 51%-75%; and 4, > 75%. An overall score was derived by multiplying the intensity and percentage scores.</ns0:p></ns0:div>
<ns0:div><ns0:head>Bioinformatics analyses</ns0:head><ns0:p>The expression of IGSF10 in different subtypes of breast cancer was analyzed using UALCAN, a web portal for evaluating gene expression in different tumor subtypes stratified according to the various clinicopathological features of patients in TCGA database <ns0:ref type='bibr' target='#b5'>(Chandrashekar et al., 2017)</ns0:ref>.</ns0:p><ns0:p>The expression of the IGSF10 mRNA in different breast cancer datasets was evaluated using Oncomine gene expression array datasets <ns0:ref type='bibr'>(Rhodes et al., 2004)</ns0:ref>. The cutoff P-value and absolute PeerJ reviewing <ns0:ref type='table' target='#tab_0'>PDF | (2020:02:45995:2:0:NEW 8 Sep 2020)</ns0:ref> Manuscript to be reviewed fold change were defined as 0.01 and 2, respectively.</ns0:p><ns0:p>The relationship between IGSF10 expression and the prognosis of patients with breast cancer presenting different molecular subtypes was analyzed using a Kaplan-Meier plotter (http://kmplot.com/analysis/) <ns0:ref type='bibr' target='#b0'>(András Lánczky et al., 2016)</ns0:ref>. The Affymetrix probe set ID of IGSF10 is 230670_at. Patients were automatically stratified into IGSF10-high and IGSF10-low groups according to the mean expression of the IGSF10 mRNA.</ns0:p></ns0:div>
<ns0:div><ns0:head>GSEA</ns0:head><ns0:p>This method was described in previous study <ns0:ref type='bibr' target='#b19'>(Jiao et al., 2018)</ns0:ref>. We performed a GSEA (http://software.broadinstitute.org/gsea) to explore the association between IGSF10 expression and biological processes/pathways according to the instructions of the user guide. We performed the GSEA using a microarray dataset (GSE1456) and TCGA microarray dataset.</ns0:p></ns0:div>
<ns0:div><ns0:head>Additional statistical analyses</ns0:head><ns0:p>All statistical analyses were performed using SPSS software (version 23.0). OS and RFS were calculated by constructing Kaplan-Meier curves. The differences between two groups were evaluated using Student's t test. Significance was set to a P-value less than 0.05.</ns0:p></ns0:div>
<ns0:div><ns0:head>Results</ns0:head></ns0:div>
<ns0:div><ns0:head n='1.'>The expression of IGSF10 in breast cancer and its clinicopathological features</ns0:head><ns0:p>We examined the expression of the IGSF10 mRNA in 1095 patients with breast cancer in TCGA database. Based on our results, the IGSF10 mRNA was expressed at higher levels in adjacent normal tissues than in breast cancer tissues (Figure <ns0:ref type='figure' target='#fig_5'>1A</ns0:ref>). We then detected the differences in IGSF10 expression in 21 paired tissue samples using RT-qPCR. Consistent with the results from TCGA database, IGSF10 expression was substantially downregulated in breast cancer tissues (Figure <ns0:ref type='figure' target='#fig_5'>1B</ns0:ref> and Supplemental Table <ns0:ref type='table'>1</ns0:ref>). We collected 31 pairs of breast cancer and corresponding normal tissues and performed IHC. The staining scores of the breast cancer tissues were significantly lower than the adjacent normal tissues (Figure <ns0:ref type='figure' target='#fig_5'>1C-1G</ns0:ref>). Finally, we examined the Manuscript to be reviewed expression of the IGSF10 mRNA in breast cell lines. IGSF10 was expressed at higher levels in the normal breast epithelial cell line MCF10A than in the breast cancer cell lines (Figure <ns0:ref type='figure' target='#fig_5'>1E</ns0:ref> and Supplemental Table <ns0:ref type='table' target='#tab_0'>2</ns0:ref>).</ns0:p><ns0:p>Seven hundred patients with breast cancer in the TCGA cohort were analyzed to further confirm the correlation between IGSF10 expression and breast cancer (Supplemental Table <ns0:ref type='table'>3</ns0:ref>). IGSF10 expression correlated with age (P<0.001), tumor size (P=0.003), and TNM stage (P =0.03) (Table <ns0:ref type='table'>1</ns0:ref>).</ns0:p></ns0:div>
<ns0:div><ns0:head n='2.'>High IGSF10 expression correlated with a better prognosis for patients with breast cancer</ns0:head><ns0:p>The associations of IGSF10 expression with overall survival (OS) and relapse-free survival (RFS) were evaluated using Kaplan-Meier survival curves. Patients in TCGA dataset were stratified by the median IGSF10 mRNA expression level (Supplemental Table <ns0:ref type='table'>4</ns0:ref>). Patients with high IGSF10 expression were significantly more likely to experience prolonged OS (hazard ratio (HR)=0.63, 95% confidence interval (CI): 0.41-0.97, P<0.05) (Figure <ns0:ref type='figure' target='#fig_6'>2A</ns0:ref>) and RFS (HR=0.53, 95% CI: 0.30-0.93, P<0.05) (Figure <ns0:ref type='figure' target='#fig_6'>2B</ns0:ref>) than patients with low IGSF10 expression. Subsequently, we used the UALCAN database to further evaluate the prognostic value of IGSF10 by stratifying patients into different molecular subtypes. Decreased levels of the IGSF10 mRNA were observed in luminal, HER2-positive, and triple-negative breast cancer samples compared with normal samples (Figure <ns0:ref type='figure' target='#fig_7'>3A</ns0:ref>). Low IGSF10 expression was significantly correlated with a shorter OS of patients with basal (HR =0.44, 95% CI: 0.22-0.86, P=0.013), luminal A (HR=0.47, 95% CI: 0.25-0.88, P=0.017), and HER2+ (HR=0.28, 95% CI: 0.09-0.81, P=0.012) breast cancer subtypes (Figures <ns0:ref type='figure' target='#fig_7'>3B-3E</ns0:ref>). However, a significant relationship was not observed between the expression of IGSF10 and OS of patients with the luminal B subtype (HR=0.61, 95% CI:0.3-1.23, P=0.17) (Figure <ns0:ref type='figure' target='#fig_7'>3D</ns0:ref>). The multivariate Cox regression analysis of TCGA patients with breast cancer showed that IGSF10 was an independent prognostic factor for OS (HR=1.793, 95% CI: 1.141-2.815, P=0.011) and RFS (HR=2.298, 95% CI: 1.317-4.010, P=0.003) (Table <ns0:ref type='table' target='#tab_0'>2</ns0:ref>).</ns0:p></ns0:div>
<ns0:div><ns0:head n='3.'>Potential biological roles and signaling pathways related to IGSF10</ns0:head><ns0:p>PeerJ reviewing PDF | (2020:02:45995:2:0:NEW 8 Sep 2020)</ns0:p><ns0:p>Manuscript to be reviewed Potential mechanisms and signaling pathways that may be related to the ability of IGSF10 to regulate the development of breast cancer were explored by conducting a GSEA. According to the median value of IGSF10 expression in the microarray dataset (GSE1456) and TCGA dataset, we assigned patients to two groups. Nine gene sets were enriched in the GSE1456 dataset and 16 gene sets were enriched in TCGA dataset (P<0.05; false discovery rate (FDR)<0.25) (Figure <ns0:ref type='figure' target='#fig_8'>4A</ns0:ref>-4B and Supplemental Table <ns0:ref type='table'>5</ns0:ref>). Interestingly, IGSF10 expression was positively correlated with several cancer-related biological processes, including DNA repair (HALLMARK_DNA_REPAIR), cell cycle (HALLMARK_G2M_CKECKPOINT), and glycolysis (HALLMARK_GLYCOLYSIS) pathways in both datasets (Figure <ns0:ref type='figure' target='#fig_8'>4C-4E</ns0:ref>). The PI3K/Akt/mTOR and mTORC1 signaling pathways were also associated with IGSF10 (Figure <ns0:ref type='figure' target='#fig_8'>4F</ns0:ref> </ns0:p></ns0:div>
<ns0:div><ns0:head>Discussion</ns0:head><ns0:p>In recent years, numerous molecular prognostic biomarkers have been identified and validated in cancers, including breast cancer <ns0:ref type='bibr' target='#b29'>(Nicolini, 2018)</ns0:ref>. In the present study, we identified IGSF10 as a potential prognostic biomarker for breast cancer and described a possible mechanism underlying its role in the tumorigenesis of breast cancer.</ns0:p><ns0:p>In the present study, we explored the role of IGSF10 in breast cancer by analyzing TCGA data and performing RT-qPCR and IHC. Our data indicated that IGSF10 expression was significantly downregulated in breast cancer tissues. Consistent with our results, multiple datasets in the Oncomine database suggested that IGSF10 expression was down-regulated in breast cancer tissues (absolute fold change>2) including TCGA Breast, Karnoub Breast <ns0:ref type='bibr' target='#b21'>(Karnoub et al., 2007)</ns0:ref>, Zhao</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:02:45995:2:0:NEW 8 Sep 2020)</ns0:p><ns0:p>Manuscript to be reviewed</ns0:p><ns0:p>Breast <ns0:ref type='bibr' target='#b43'>(Zhao et al., 2004)</ns0:ref>, Richardson Breast 2 <ns0:ref type='bibr' target='#b33'>(Richardson et al., 2006)</ns0:ref>, and Finak Breast <ns0:ref type='bibr' target='#b11'>(Finak et al., 2008)</ns0:ref> datasets (Supplemental Table <ns0:ref type='table'>6</ns0:ref>). Based on the analysis of the data in the UALCAN database, we found that IGSF10 expression correlated with the molecular subtype of breast cancer.</ns0:p><ns0:p>In addition, IGSF10 expression was closely associated with age, tumor size, and TNM stage.</ns0:p><ns0:p>Accordingly, IGSF10 may play a crucial role in breast cancer and have the potential to be targeted by anticancer therapy. Moreover, the survival analysis indicated that patients with breast cancer presenting higher IGSF10 expression experienced prolonged OS and RFS. The multivariate analysis identified IGSF10 as an independent prognostic factor for patients with breast cancer.</ns0:p><ns0:p>Interestingly, in the subgroup analysis, IGSF10 expression was significantly correlated with OS in patients with basal, luminal A and HER2-positive breast cancer. Thus, IGSF10 may be a prognostic biomarker for breast cancer.</ns0:p><ns0:p>IGSF10 may exert an important effect on tumorigenesis. Ling and colleagues claimed that IGSF10 knockout promotes the development of lung cancer and that IGSF10 mainly activates the integrin-β1/FAK pathway in lung cancer <ns0:ref type='bibr' target='#b27'>(Ling et al., 2020)</ns0:ref>. In one family with gastric and colorectal cancers, Thutkawkorapin et al. identified 12 new nonsynonymous single nucleotide variants in 12 different genes, including IGSF10, with potential contributions to an increased cancer risk <ns0:ref type='bibr' target='#b37'>(Thutkawkorapin et al., 2016)</ns0:ref>. Chang et al. identified new mutations in patients with endometrial cancer in Taiwan by performing whole-exome sequencing and identified a potential association between IGSF10, a passenger gene, with endometrial cancer <ns0:ref type='bibr' target='#b6'>(Chang, 2017)</ns0:ref>. However, to our knowledge, no studies have reported the possible functions and mechanisms of IGSF10 in breast cancer.</ns0:p><ns0:p>During the past decade, accumulating evidence has revealed clear correlations between immunoglobulin superfamily members and human diseases. For instance, loss-of-function mutations in IGSF1 result in an X-linked syndrome of central hypothyroidism and testicular enlargement. IGSF1 mutations in male patients lead to a late increase in testosterone levels <ns0:ref type='bibr' target='#b15'>(Howard et al., 2016;</ns0:ref><ns0:ref type='bibr' target='#b34'>Roche et al., 2018;</ns0:ref><ns0:ref type='bibr' target='#b35'>Sun et al., 2012)</ns0:ref>. Significantly prolonged OS was observed in pediatric patients with mixed-lineage leukemia-rearranged acute monoblastic Manuscript to be reviewed leukemia with t(9; 11) (p22; q23) and high IGSF4 expression than in patients with low IGSF4 expression <ns0:ref type='bibr' target='#b24'>(Kuipers et al., 2011)</ns0:ref>. As shown in the study by <ns0:ref type='bibr'>Wang et al.,</ns0:ref><ns0:ref type='bibr'>IGSF8 promotes</ns0:ref> melanoma proliferation and metastasis by negatively regulating the TGF-β signaling pathway <ns0:ref type='bibr' target='#b38'>(Wang et al., 2015)</ns0:ref>.</ns0:p><ns0:p>In the present study, potential biological roles and signaling pathways that may be related to IGSF10 expression in breast cancer were analyzed by conducting a GSEA. Several biological processes, including DNA repair, the cell cycle, and glycolysis, were associated with IGSF10.</ns0:p><ns0:p>Among these processes, the genomic integrity is maintained through DNA repair pathways. The dysregulation of DNA repair leads to changes in the genome and causes physiological changes in cells that drive tumor initiation <ns0:ref type='bibr' target='#b17'>(Jeggo, 2016;</ns0:ref><ns0:ref type='bibr' target='#b23'>Khanna, 2015;</ns0:ref><ns0:ref type='bibr' target='#b28'>Mouw, 2017)</ns0:ref>. The cell cycle regulates tumor growth and glycolysis modulates the heterogeneity of the tumor microenvironment. These biological processes are related to tumor progression, metastasis and drug resistance <ns0:ref type='bibr' target='#b16'>(Jahagirdar et al., 2018)</ns0:ref>. Moreover, in human malignancies, the mTORC1 and PI3K/Akt/mTOR signaling pathways are usually abnormally activated and promote the development of malignancies <ns0:ref type='bibr' target='#b13'>(Hare, 2017)</ns0:ref>. According to previous studies, mTORC1 promotes cell growth by activating key anabolic processes and the dysregulation of mTORC1 is the basis of many human cancers <ns0:ref type='bibr' target='#b3'>(Ben-Sahra, 2017;</ns0:ref><ns0:ref type='bibr'>Keppler-Noreuil, 2016)</ns0:ref>. The PI3K/Akt/mTOR pathway is related to various biological processes in breast cancer, such as tumorigenesis, cellular transformation, tumor progression, and drug resistance <ns0:ref type='bibr' target='#b12'>(Guerrero-Zotano et al., 2016)</ns0:ref>. Therefore, we speculated that IGSF10 might mechanistically regulate the growth of breast cancer cells through the mTORC1 and PI3K/Akt/mTOR signaling pathways. Intriguingly, IGSF10 was associated with EMT, the TGF-β signaling pathway and the TNF signaling pathway in TCGA database. The TGF-β signaling pathway was reported to be associated with various tumors and it regulates the biological processes in multiple cancers, including growth, migration, invasion, apoptosis and the EMT <ns0:ref type='bibr' target='#b2'>(Bedi et al., 2012;</ns0:ref><ns0:ref type='bibr' target='#b36'>Tang et al., 2017;</ns0:ref><ns0:ref type='bibr' target='#b42'>Yu et al., 2018;</ns0:ref><ns0:ref type='bibr' target='#b44'>Zhao et al., 2018)</ns0:ref>.</ns0:p><ns0:p>The EMT plays crucial roles in the metastasis and invasion of breast cancer by regulating cell motility and invasiveness <ns0:ref type='bibr' target='#b10'>(Feng et al., 2016)</ns0:ref>. Moreover, TNF-α is strongly correlated with Manuscript to be reviewed inflammation in breast tumors, and an increase in its expression is strongly correlated with relapse and advanced disease <ns0:ref type='bibr' target='#b22'>(Katanov et al., 2015)</ns0:ref>. However, further studies are needed to elucidate the role of IGSF10 in breast cancer and the detailed mechanisms by which IGSF10 modulates these related signaling pathways.</ns0:p></ns0:div>
<ns0:div><ns0:head>Conclusions</ns0:head><ns0:p>In summary, IGSF10 was expressed at a low level in breast cancer. IGSF10 expression was significantly correlated with age, tumor size, and tumor stage. More importantly, IGSF10 was an independent prognostic factor for better outcomes in patients with breast cancer. In addition, the GSEA results identified significant associations between IGSF10 expression and DNA repair, cell cycle, glycolysis, and the mTORC1 and PI3K/Akt/mTOR signaling pathways. Overall, we suggested a novel role for IGSF10 in breast cancer. Our data may provide new insights into the identification of potential therapeutic targets in patients with breast cancer. Manuscript to be reviewed </ns0:p></ns0:div>
<ns0:div><ns0:head>Figure legends</ns0:head></ns0:div><ns0:figure xml:id='fig_0'><ns0:head /><ns0:label /><ns0:figDesc>Specifically, we collected 52 pairs of breast tumor and adjacent normal tissues from patients with breast cancer during surgery between 2014 and 2016 at The First Affiliated Hospital of Chongqing Medical University. The collected tissues were used for real-time quantitative polymerase chain reaction (RT-qPCR) and immunohistochemistry (IHC) analyses. All specimens were stored in liquid nitrogen. The collection and use of the tissues were approved by the Institutional Ethics Committees of the First Affiliated Hospital of Chongqing Medical University. The approval number allocated to this study by the Institutional Ethics Committees is 2017 Research Ethics (2017-012). PeerJ reviewing PDF | (2020:02:45995:2:0:NEW 8 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_1'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:02:45995:2:0:NEW 8 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_2'><ns0:head /><ns0:label /><ns0:figDesc>-4G). Moreover, in TCGA dataset, signaling pathway (HALLMARK_TNFA_SIGNALING_VIA_NFKB) were significantly enriched in the IGSF10-low group (Supplemental Figure1A). These results indicated a possible mechanism underlying the role of IGSF10 in the tumorigenesis of breast cancer.</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_3'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:02:45995:2:0:NEW 8 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_4'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:02:45995:2:0:NEW 8 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_5'><ns0:head>Figure 1 :</ns0:head><ns0:label>1</ns0:label><ns0:figDesc>Figure 1: (A) Bioinformatics analysis of IGSF10 expression in TCGA database. (B) The expression of the IGSF10 mRNA in BC tissues and matched adjacent normal tissues was evaluated using qRT-PCR (n = 21). (C-F) Representative images of IHC staining in BC specimens and adjacent normal breast tissues. (G) IHC score for the level of the IGSF10 protein expression in 31 BC tissues (IHC score: 3.12 ± 2.04) and 31 normal tissues (IHC score: 4.45 ± 2.13). Data are presented as mean ± SD, unpaired t-test, *P < 0.05. (H) qRT-PCR was used to examine IGSF10 expression in human breast cancer cells and MCF-10A cells; *P < 0.05, **P < 0.01, and ***P < 0.001.</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_6'><ns0:head>Figure 2 :</ns0:head><ns0:label>2</ns0:label><ns0:figDesc>Figure 2: Kaplan-Meier survival curve of TCGA patients with breast cancer stratified into the IGSF10-high and IGSF10-low groups based on the median expression level. P < 0.05 was considered a statistically significant. (A) Curves showing the OS of patients with breast cancer. (B) Curves showing the RFS of patients with breast cancer.</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_7'><ns0:head>Figure 3 :</ns0:head><ns0:label>3</ns0:label><ns0:figDesc>Figure 3: (A) IGSF10 expression in patients with different molecular subtypes of breast cancer in TCGA database. (B) Basal breast cancer, (C) luminal A breast cancer, (D) luminal B breast cancer, and (E) HER2+ breast cancer. All the curves showing OS were plotted using the Kaplan-Meier plotter (http://kmplot.com/analysis/). **P < 0.01.</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_8'><ns0:head>Figure 4 :</ns0:head><ns0:label>4</ns0:label><ns0:figDesc>Figure 4: Gene sets with a normal P-value < 0.05 and an FDR < 0.25 were considered significant. Gene sets were ranked by the normalized enrichment score (NES). (A) Gene sets enriched in the GSE1456 dataset. (B) Gene sets enriched in the TCGA dataset. (C-E) GSEA enrichment plot showing that IGSF10 expression was positively associated with DNA repair, cell cycle, and glycolysis. (F-G) GSEA enrichment plot showing that IGSF10 expression was positively correlated with the PI3K/Akt/mTOR and mTORC1 signaling pathways.</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_9'><ns0:head /><ns0:label /><ns0:figDesc>Figure 1</ns0:figDesc></ns0:figure>
<ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='22,42.52,301.12,525.00,202.50' type='bitmap' /></ns0:figure>
<ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='24,42.52,70.87,397.24,672.95' type='bitmap' /></ns0:figure>
<ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='26,42.52,70.87,525.00,405.00' type='bitmap' /></ns0:figure>
<ns0:figure type='table' xml:id='tab_0'><ns0:head>Table 2 . Univariate and multivariate Cox regression analysis of IGSF10 in TCGA cohort.</ns0:head><ns0:label>2</ns0:label><ns0:figDesc>Abbreviation: OS: overall survival; RFS: relapse-free survival; HR: hazard ratio; CI: confidence interval; p<0.05 was considered statistically significant</ns0:figDesc><ns0:table><ns0:row><ns0:cell>OS</ns0:cell></ns0:row></ns0:table><ns0:note>PeerJ reviewing PDF | (2020:02:45995:2:0:NEW 8 Sep 2020)</ns0:note></ns0:figure>
<ns0:note place='foot'>PeerJ reviewing PDF | (2020:02:45995:2:0:NEW 8 Sep 2020)</ns0:note>
<ns0:note place='foot'>PeerJ reviewing PDF | (2020:02:45995:2:0:NEW 8 Sep 2020)Manuscript to be reviewed</ns0:note>
</ns0:body>
" | "Responses to the Reviewers’ Comments
We would like to express our sincere appreciation to the reviewers for the constructive and positive comments.
A. Responses to the comments provided by reviewer 2 (Deepak Balamurali)
We have revised the manuscript according to your suggestions in the BASIC REPORTING section of your comments. Our responses are listed below.
1. 700 samples were used in this study - were the other samples excluded for any specific reason? What were the inclusion/exclusion criteria?
Answer: Thank you for your comments. The patients with breast cancer included in this study were required to have both complete RNA-seq data and complete clinical information. The clinical information included the tumor size, lymph node status, tumor, node, metastasis stage, ER status, PR status, HER2 status, and follow-up information (METHOD SECTION).
2. Was sample collection during the time of surgery, or during a biopsy? Information not provided. The number of samples collected also is not mentioned here.
Answer: Thank you for your comments. The breast tumor and normal tissues were collected during surgery. We have mentioned this information in the text according your suggestions.
3. How was GSE1456 chosen?
Answer: Thank you for your comments. We chose GSE1456 because this dataset included a sufficient number of patients and compete mRNA expression data. At the same time, it included patients with different pathological types of breast cancer (luminal: 74.8%; HER2+: 13.9%; basal: 11.3%). This proportion was approximately the same as the distribution of clinical patients with breast cancer.
4. Which is the right number of samples? 700 or 1095? Please clarify. There seems to be conflict between the methods and results sections.
Answer: Thank you for your comments. We apologize for failing to clearly describe the number of patients in the METHODS section. We analyzed the expression of the IGSF10 mRNA in 1095 patients in TCGA, and then excluded some patients based on whether complete clinical information and follow-up data were available. After the screen, we finally selected 700 patients for the analysis of the correlation between clinical characteristics and the expression of IGSF10. We have revised the METHODS section to clearly describe this process.
5. The authors mention 21 samples (paired) for qPCR and 31 for IHC. How many samples were used for both? Was the data compared? If not, what was the reason why the same patient sample was not used, especially if collected from a surgery where there will be more tissue availability?
Answer: Thank you for your comments. Normally, the samples collected by our laboratory from the hospital are not used to extract RNA and prepare pathological tissue sections at the same time. Therefore, we did not use the same samples in the two experiments.
6. How many times was the qPCR performed? Were the results consistent across repetitions? What are the values shown in Supplemental Table 1? Are they Ct values? In many cases, there seems to be no difference between the normal and tumor. There is no legend/key provided. Similar issue for S. Table 2 as well.
Answer: Thank you for your comments. We performed qPCR three times. Consistent results were obtained across repetitions. The values presented in Supplemental Table 1 are average Ct values obtained using qPCR. One of the reviewers suggested that we should add these data as supplemental tables in the text. We also added the legends for these tables. Thank you for your suggestions.
7. The authors will have to validate their findings in a larger cohort, especially with an overlap of samples used in IHC and qPCR, to be more confident of the results.
Answer: Thank you for your comments. Currently, we are unable to easily collect tissue samples at our hospital. Moreover, we cannot collect a large number of tissue samples in a short time. Since the time for revision was limited, these experiments will certainly be included in our future studies. Thank you again for your advice.
8.The authors mention in the rebuttal that the manuscript has been proof-read, however, there are still many factual and textual errors in the paper. The general flow in the manuscript also seems erratic.
Answer: Thank you for your comments. We apologize for these errors in the manuscript. We have carefully checked the entire manuscript, and the English language has been improved by a language editing service. The language editing certificate has been uploaded as Supplemental materials.
Responses to the comments provided by reviewer 3 (Xin Zhang)
1.“Response to Major Q4” As can be seen from the pathological picture in Figure 1C to F, there is no typical breast cancer tissue in the micrograph!
Answer: Thank you very much for your suggestions. We have provided a new group of pathological images in Figure 1.
2.“Response to Major Q5” As far as I know, the epidemic in mainland China has been well controlled in May, and I haven't heard of any laboratory that has been closed up to now.
Answer: Thank you for your comments. We apologize for this error. In fact, we wrote the rebuttal letter in May. However, due to the large amount of repetitive text in our manuscript, we had to revise our manuscript. We resubmitted the manuscript and rebuttal letter in July. Your suggestions were very helpful. However, since the time for revision was limited, we will perform further research on this topic in the future. Thank you again for your advice.
3.“Materials & Methods - IHC” Line 120-125. “The IHC staining intensity scoring criteria were as follows: 0, none; 1, weak; 2, medium; 3, strong. The scoring criteria for the proportion of positive tumor cells were as follows: 0, < 5%; 1, 5%–25%; 2, 26%–50%; 3, 51%–75%; 4, > 75%.” The authors should explain how the immunohistochemistry score in the adjacent control tissue was calculated.
Answer: Thank you for your comments. We have revised the METHODS section according to your suggestion. Thank you again, and we hope to learn more from you in the future.
" | Here is a paper. Please give your review comments after reading it. |
9,768 | "<ns0:abstract xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:p>Background. Acquisition of procedures is an important element in health professions education. Traditionally procedures are taught using a 'see one -do one' approach. That is a teacher demonstrates and describes a procedure and afterwards the students practice the procedure. A more recent teaching approach for the acquisition of procedural skills was presented by Walker and Peyton. Peyton's teaching approach is a stepwise teaching approach and consists of the following four steps: demonstration, deconstruction, comprehension and performance. The aims of this study were i) to systematically evaluate the effectiveness of Peyton's 4-step teaching approach on the acquisition of procedural skills in health professions education and ii) to evaluate whether studies with fewer students per teacher showed a larger between group difference than studies with more students per teacher.Methods. We searched in Medline, PsycInfo, Embase and ERIC for eligible studies. Records were screened by two independent reviewers. A random effects meta-analysis was performed to evaluate skill acquisition and time needed to perform the procedures at post-acquisition and retention tests. A meta-regression was used to explore the effect of the number of students per teacher on the estimated effect of the educational interventions.Results. An effect size of 0.45 SMD (95%CI: 0.15; 0.75) at post-acquisition and 0.7 SMD (95%CI: -0.09; 1.49) at retention testing were in favour of Peyton's teaching approach for skill acquisition. The groups using Peyton's teaching approach needed considerably less time to perform the procedure at post-acquisition (SMD: -0.8; 95%CI:</ns0:p><ns0:p>-2.13 to 1.62) and retention (SMD: -2.65; 95%CI: -7.77 to 2.47) testing. The effectiveness of Peyton's teaching approach was less clear in subgroup analyses using peer teachers.</ns0:p><ns0:p>Meta-regression showed that the number of students per teacher was an important moderator variable.Conclusion. Peyton's teaching approach is an effective teaching approach for skill acquisition of procedural skills in health professions education. When peer students or student tutors are used as teachers the effectiveness of Peyton's</ns0:p></ns0:div>
</ns0:abstract>
<ns0:body xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:head>Introduction</ns0:head><ns0:p>Acquisition of procedures is an important element in health professions education <ns0:ref type='bibr' target='#b20'>(Grantcharov & Reznick 2008)</ns0:ref>. Historically, the study of the acquisition of procedural skills was primarily in the field of medical and especially surgical education. However, other health professions such as nursing and physiotherapy education have developed assessment and teaching approaches for these skills as well <ns0:ref type='bibr' target='#b49'>(Oermann et al. 2016;</ns0:ref><ns0:ref type='bibr' target='#b60'>Sattelmayer et al. 2017)</ns0:ref>. Defining procedural skills is challenging. <ns0:ref type='bibr' target='#b43'>Michels et al. (2012)</ns0:ref> reported that there is considerable overlap between the terms clinical skills, psychmotor skills and procedural skills. Traditionally procedures are taught using a 'see one -do one' approach. This means that a teacher demonstrates and describes a procedure and afterwards the students are asked to practice the procedure. This is referred to as Halsted's teaching approach, which is based on the surgeon William Steward <ns0:ref type='bibr' target='#b22'>Halsted (1904)</ns0:ref>. The approach was used as an element to redesign surgical education and create a new system for training young surgeons <ns0:ref type='bibr' target='#b9'>(Cameron 1997</ns0:ref>). Although the 'see one -do one' approach is often used in the training of health professionals, there is criticism of this approach. First, the approach has been used for decades and does not adhere to recent principles of adult learning such as active learner involvement <ns0:ref type='bibr' target='#b42'>(McLeod et al. 2001</ns0:ref>). Furthermore, it was reported that patient safety might be at risk because complex procedures cannot be acquired after a single observation and practice trial <ns0:ref type='bibr' target='#b33'>(Kotsis & Chung 2013)</ns0:ref>. Given the diversity of existing procedures today, others argue that the teaching approach should be modified to 'see many, learn from the result and do many' <ns0:ref type='bibr' target='#b55'>(Rohrich 2006)</ns0:ref>. A more recent teaching approach for the acquisition of procedural skills was presented by <ns0:ref type='bibr' target='#b73'>Walker and Peyton (1998)</ns0:ref>. Peyton's teaching approach is a stepwise teaching approach and consists of the following four steps: i) step 1 refers to the demonstration of the whole procedure in real time ('demonstration'); ii) in step 2 the teacher repeats the demonstration but this time all procedural sub-steps are described ('deconstruction'); iii) during step 3 the student talks the teacher through the procedure. The teacher performs the procedure under the guidance of the student ('comprehension') and iv) in step 4 the students carry out the procedure on their own initiative ('performance'). A similar stepwise teaching approach was presented by George (American College of Surgeons 1997) and later published by <ns0:ref type='bibr' target='#b16'>George and Doto (2001)</ns0:ref>. Originally, it was developed as an educational technique to support the American College of Surgeon's Advanced Trauma Life Support course. In contrast to Peyton's teaching approach George and Doto used five steps. Within Peyton's teaching approach two of the five steps are collated into a single step. George and Doto based their teaching approach on Simpson's taxonomy of the psychomotor domain <ns0:ref type='bibr' target='#b65'>(Simpson 1966)</ns0:ref>. Especially the third step seems to be important in Peyton's teaching approach and was assumed to be beneficial for skill acquisition. The process of guiding the teacher through the procedure requires the student to remember and think about the first two steps before giving the teacher the necessary information <ns0:ref type='bibr' target='#b18'>(Gradl-Dietsch et al. 2016)</ns0:ref>. This process could help students to organise their thoughts and support student-centred learning <ns0:ref type='bibr' target='#b38'>(Lom 2012)</ns0:ref>. Similarly, <ns0:ref type='bibr'>Balafoutas and</ns0:ref> PeerJ reviewing PDF | (2020:07:50815:1:1:NEW 28 Aug 2020)</ns0:p><ns0:p>Manuscript to be reviewed colleagues (2019) argue that students need to manipulate the information stored in their working memory based on the information provided in the first two steps. This could support the transfer of relevant information into the long-term memory. Other authors have argued that recognising the effects of the instructions on the performance could be a valuable source of feedback and might improve metacognitive skills <ns0:ref type='bibr' target='#b24'>(Herrmann-Werner et al. 2013</ns0:ref>). In addition, <ns0:ref type='bibr' target='#b57'>Rossettini et al. (2017)</ns0:ref> mentioned that Peyton's third step involves elements of mental practice. That is, the students have the possibility to develop a mental representation of the movement in absence of an active movement. There exists evidence that mental practice is effective for skill acquisition of procedures in health professions education <ns0:ref type='bibr' target='#b59'>(Sattelmayer et al. 2016)</ns0:ref>. Besides the third step, the fourth step is also of educational importance as in this step the teacher provides feedback to the learner. A systematic review by Issenberg reported that the opportunity to provide feedback is a key component for effective skill acquisition in simulation-based medical education <ns0:ref type='bibr' target='#b31'>(Issenberg et al. 2005)</ns0:ref>. In addition, the fourth step is also supported by Bandura's scaffolding theory <ns0:ref type='bibr' target='#b62'>(Schunk 2012)</ns0:ref>. One of the strengths of Peyton's teaching approach is that it can be effectively combined with other instructional design strategies, which allows the simultaneous delivery of theoretical concepts along with complex procedural skills. For example, <ns0:ref type='bibr' target='#b70'>Tambi et al. (2018)</ns0:ref> and colleagues combined Gagne's instructional model <ns0:ref type='bibr' target='#b15'>(Gagne et al. 2005</ns0:ref>) with Peyton's teaching approach to design a bioinformatics lesson plan for medical students and Ng (2014) combined both teaching approaches for slit-lamp teaching. However, one could assume that the step-by-step approach would require considerably more time for teaching. The traditional teaching approach consist typically of two steps (demonstration and practice). The additional two steps might be assumed to be time-consuming. However, in contrast to this, several authors have reported that not more time was required using Peyton's approach <ns0:ref type='bibr' target='#b35'>(Krautter et al. 2011;</ns0:ref><ns0:ref type='bibr' target='#b57'>Rossettini et al. 2017)</ns0:ref>. Several randomised controlled trials have evaluated the effectiveness of Peyton's teaching approach. The results of these studies are not always consistent. Some trials have reported findings in favour of Peyton's approach (e.g. <ns0:ref type='bibr' target='#b5'>Balafoutas et al. 2019;</ns0:ref><ns0:ref type='bibr' target='#b57'>Rossettini et al. 2017)</ns0:ref>. <ns0:ref type='bibr' target='#b57'>Rossettini et al. (2017)</ns0:ref> showed that acquisition of a cervical mobilisation technique was considerable higher in the Peyton group compared to a standard teaching group. In contrast, <ns0:ref type='bibr' target='#b50'>Orde et al. (2010)</ns0:ref> have reported that Peyton's teaching approach showed only minor differences on skill acquisition regarding insertion of a laryngeal mask airway at post-acquisition and retention testing compared to a traditional teaching approach. Originally Peyton's teaching approach was designed for a student-teacher ratio of 1:1 <ns0:ref type='bibr' target='#b48'>(Nikendei et al. 2014)</ns0:ref>. However, such a ratio is difficult to achieve in educational institutions. Therefore, from a pragmatic point of view it is important to evaluate whether Peytons's teaching approach can be used with more students per teacher. These inconsistencies should be further investigated through a systematic review. Therefore, the aims of this study were i) to systematically evaluate the effectiveness of Peyton's 4 step teaching approach on the acquisition of procedural skills in health professions education and ii) to evaluate whether studies with fewer students per teacher (i.e. the student-teacher ratio) showed a larger between group difference than studies with more students per teacher.</ns0:p></ns0:div>
<ns0:div><ns0:head>Materials & Methods</ns0:head><ns0:p>A protocol of this systematic review was registered in the OSF registries: https://doi.org/10.17605/OSF.IO/5UE7C. To improve clarity of reporting the PRISMA statement was followed <ns0:ref type='bibr' target='#b45'>(Moher et al. 2011)</ns0:ref>.</ns0:p></ns0:div>
<ns0:div><ns0:head>Searches</ns0:head><ns0:p>We searched the following electronic databases for eligible studies: Medline, PsycInfo, Embase and Education Resources Information Center (ERIC). The search was performed by KMS. No restrictions regarding recency or publication language were set. The search strategy was prepared using two blocks. The first block consisted of terms relevant for the identification of the population (i.e. students in health professions education). We searched for keywords and mapped the keywords to relevant subject headings. The second block was designed to identify studies using Peyton's teaching approach. Both search blocks were combined using the Boolean operator 'and'. The search strategy is reported in Appendix 1. In addition, references of included studies were checked for potential eligible studies.</ns0:p></ns0:div>
<ns0:div><ns0:head>Selection criteria</ns0:head><ns0:p>The following selection criteria were applied.</ns0:p></ns0:div>
<ns0:div><ns0:head>Types of studies to be included</ns0:head><ns0:p>Randomised controlled trials were included. If sufficient data was available cross-over studies were eligible as well.</ns0:p></ns0:div>
<ns0:div><ns0:head>Participants</ns0:head><ns0:p>Only studies reporting on students in health professions education were included. Health professions education was used as an umbrella term for medical and allied health profession education (e.g. physiotherapy or nursing education). We included studies reporting on undergraduate and postgraduate students.</ns0:p></ns0:div>
<ns0:div><ns0:head>Interventions</ns0:head><ns0:p>Studies needed to investigate Peyton's 4-step approach for inclusion in at least one study arm (i.e. all 4 steps were used together).</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:07:50815:1:1:NEW 28 Aug 2020)</ns0:p><ns0:p>Manuscript to be reviewed Comparator Studies needed to have a comparator group. The comparator could be a specific educational intervention (e.g. team-based education or peer teaching), educational practice as usual (e.g. a 'see one -do one') or a sham intervention.</ns0:p></ns0:div>
<ns0:div><ns0:head>Outcomes</ns0:head><ns0:p>The primary outcome for this review was the evaluation of procedural skills. These could be evaluated using a performance metric such as a procedure specific checklist or a global rating scale. To be included studies had to report on this outcome. The secondary outcome was the time needed to perform the procedure. If multiple procedures were trained one procedure was selected for inclusion in order to avoid a unit of analysis issue (i.e. in order to avoid including the same participants twice within a single analysis). Means and standard deviations of continuous outcomes were extracted. If standard deviations were not reported we imputed standard deviations based on standard errors or confidence intervals as suggested in the Cochrane Handbook <ns0:ref type='bibr' target='#b27'>(Higgins et al. 2019)</ns0:ref>.</ns0:p></ns0:div>
<ns0:div><ns0:head>Study selection and data extraction</ns0:head><ns0:p>Records were screened by two independent reviewers (RC and KMS). The screening procedure was performed using the Rayyan software <ns0:ref type='bibr' target='#b52'>(Ouzzani et al. 2016)</ns0:ref>. Disagreements were solved by discussion between RC and KMS. If a referee was needed KG was consulted. One reviewer (KMS) extracted relevant data into an electronic database and a second reviewer (KG) controlled the data.</ns0:p></ns0:div>
<ns0:div><ns0:head>Risk of bias assessment</ns0:head><ns0:p>The risk of bias was evaluated using the Cochrane risk of bias tool <ns0:ref type='bibr'>(Higgins et al. 2011)</ns0:ref>. A human reviewer (KMS) evaluated all included studies with respect to these items: sequence generation, allocation concealment, blinding of a) participants and personnel and b) outcome assessors, incomplete outcome data and selective reporting. Evaluations were compared against a machine learning classification of the risk of bias with the application 'RobotReviewer' <ns0:ref type='bibr' target='#b41'>(Marshall et al. 2015)</ns0:ref>. Disagreements were solved by discussion with a third person.</ns0:p></ns0:div>
<ns0:div><ns0:head>Strategy for data synthesis</ns0:head><ns0:p>The primary endpoint for evaluating the effectiveness of the comparisons was at the end of the intervention. A secondary analysis was performed using data from the longest available follow up endpoint.</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:07:50815:1:1:NEW 28 Aug 2020)</ns0:p><ns0:p>Manuscript to be reviewed</ns0:p></ns0:div>
<ns0:div><ns0:head>Data analysis</ns0:head><ns0:p>The analysis was performed using the statistical software package R (R Core Team 2019). A meta-analysis of pairwise comparisons was performed using the meta package <ns0:ref type='bibr' target='#b63'>(Schwarzer 2007)</ns0:ref>. A random effects model was used for the analysis and effectiveness was reported using standardized effect sizes (Hedges' g) and corresponding 95% confidence intervals. The Hartung, Knapp, Sidik, Jonkmann adjustment was applied to achieve robust estimations of the treatment effect <ns0:ref type='bibr' target='#b30'>(IntHout et al. 2014)</ns0:ref>. Effect sizes were interpreted following <ns0:ref type='bibr' target='#b10'>Cohen (1992)</ns0:ref>. This means that an effect size of 0.2 was considered as small, 0.5 as medium and 0.8 as large. Statistical heterogeneity was assessed with I 2 statistics using the guidelines presented in the Cochrane handbook for systematic reviews of interventions <ns0:ref type='bibr' target='#b26'>(Higgins & Green 2011)</ns0:ref>. The following categories were applied: 0-40% might not be important, 30-60% moderate heterogeneity, 50-90% substantial heterogeneity and 75-100% considerable heterogeneity. A mixed effects meta-regression was performed using the meta package <ns0:ref type='bibr' target='#b63'>(Schwarzer 2007)</ns0:ref>. We explored the effect of the students per teacher on the estimated effect of the educational interventions. A mixed effects meta-regression was performed using the meta package <ns0:ref type='bibr' target='#b63'>(Schwarzer 2007)</ns0:ref>. The number of students per teacher during the procedural skills training was used as moderator variable.</ns0:p></ns0:div>
<ns0:div><ns0:head>Results</ns0:head></ns0:div>
<ns0:div><ns0:head>Findings of the search</ns0:head><ns0:p>The electronic search on the databases Medline, PsycInfo, Embase and ERIC identified 482 potential eligible records. In addition, the screening of the abstracts identified 5 further records. After removing 45 duplicates, 442 titles and abstracts were screened. In this phase of the selection process 405 records were excluded. The full-texts of the remaining 37 records were assessed for eligibility and 23 records were excluded with the following reasons: 12 records reported an intervention, which was not eligible for inclusion <ns0:ref type='bibr' target='#b6'>(Bode et al. 2012;</ns0:ref><ns0:ref type='bibr' target='#b8'>Bube et al. 2017;</ns0:ref><ns0:ref type='bibr' target='#b11'>Craven et al. 2018;</ns0:ref><ns0:ref type='bibr' target='#b12'>Custers et al. 1999;</ns0:ref><ns0:ref type='bibr' target='#b23'>Handley & Handley 1998;</ns0:ref><ns0:ref type='bibr' target='#b28'>Hill et al. 2010;</ns0:ref><ns0:ref type='bibr' target='#b29'>Holmes et al. 1998;</ns0:ref><ns0:ref type='bibr' target='#b34'>Krautter et al. 2015;</ns0:ref><ns0:ref type='bibr' target='#b37'>Liu & Hunt 2017;</ns0:ref><ns0:ref type='bibr' target='#b72'>Velmahos et al. 2004;</ns0:ref><ns0:ref type='bibr' target='#b74'>Wirth et al. 2018;</ns0:ref><ns0:ref type='bibr' target='#b76'>Yoganathan et al. 2018)</ns0:ref>; 8 records used a study design, which was not eligible for inclusion <ns0:ref type='bibr' target='#b13'>(Easton et al. 2012;</ns0:ref><ns0:ref type='bibr' target='#b44'>Mishra & Dornan 2003;</ns0:ref><ns0:ref type='bibr' target='#b48'>Nikendei et al. 2014;</ns0:ref><ns0:ref type='bibr' target='#b61'>Schroder et al. 2017;</ns0:ref><ns0:ref type='bibr' target='#b66'>Skrzypek et al. 2018;</ns0:ref><ns0:ref type='bibr' target='#b67'>Smith et al. 2019;</ns0:ref><ns0:ref type='bibr' target='#b68'>Sopka et al. 2012;</ns0:ref><ns0:ref type='bibr' target='#b71'>Tommaso 2016)</ns0:ref>; 2 records were excluded because of missing data <ns0:ref type='bibr' target='#b4'>(Archer et al. 2015;</ns0:ref><ns0:ref type='bibr' target='#b64'>Seymour-Walsh et al. 2015</ns0:ref>) and 1 record did not use the specified primary outcome assessment for procedural skills <ns0:ref type='bibr' target='#b21'>(Greif et al. 2010)</ns0:ref>. Finally, 14 studies were included into this systematic review. An overview of the selection process is presented in Fig. <ns0:ref type='figure' target='#fig_0'>1</ns0:ref>. During the study selection process, 6 conflicts occurred, representing 1.4% of the total decisions. <ns0:ref type='table'>2020:07:50815:1:1:NEW 28 Aug 2020)</ns0:ref> Manuscript to be reviewed</ns0:p></ns0:div>
<ns0:div><ns0:head>Included studies</ns0:head><ns0:p>The 14 included studies in this review were all randomised controlled studies. An overview of included studies and study characteristics is presented Table <ns0:ref type='table'>1</ns0:ref>. Most of the included studies were conducted in Germany (n=10). Four studies with 3 or 4 study arms were included <ns0:ref type='bibr' target='#b19'>(Gradl-Dietsch et al. 2018;</ns0:ref><ns0:ref type='bibr' target='#b24'>Herrmann-Werner et al. 2013;</ns0:ref><ns0:ref type='bibr' target='#b46'>Münster et al. 2016;</ns0:ref><ns0:ref type='bibr' target='#b58'>Ruesseler et al. 2019)</ns0:ref>. In these cases, study arms investigating Peyton's teaching approach or a standard teaching approach were included. Study arms using an intervention not eligible for inclusion were excluded from this review. For example, Gradl-Dietsch et al. ( <ns0:ref type='formula'>2018</ns0:ref>) reported 4 study arms. The study arms peer teaching and peer teaching using Peyton's teaching approach were included. Not included were the study arms team-based learning and video-based learning. All used study arms are presented in Table <ns0:ref type='table'>1</ns0:ref>. The included participants in most studies were within medical education. A range from first year medical students to residents in obstetrics and gynaecology was identified. Two studies used participants from nursing education <ns0:ref type='bibr' target='#b36'>(Lapucci et al. 2018;</ns0:ref><ns0:ref type='bibr' target='#b50'>Orde et al. 2010</ns0:ref>) and one study was conducted with participants from physiotherapy education <ns0:ref type='bibr' target='#b57'>(Rossettini et al. 2017)</ns0:ref>. A broad range of trained procedures has been identified. For example, basic surgical skills <ns0:ref type='bibr' target='#b58'>(Ruesseler et al. 2019)</ns0:ref> <ns0:ref type='bibr' target='#b39'>(Lund et al. 2012)</ns0:ref> or see one, do one <ns0:ref type='bibr' target='#b57'>(Rossettini et al. 2017;</ns0:ref><ns0:ref type='bibr' target='#b58'>Ruesseler et al. 2019)</ns0:ref>. The time allocated to the teaching of the procedural skills was set equal in most included studies. Four studies <ns0:ref type='bibr' target='#b24'>(Herrmann-Werner et al. 2013;</ns0:ref><ns0:ref type='bibr' target='#b35'>Krautter et al. 2011;</ns0:ref><ns0:ref type='bibr' target='#b39'>Lund et al. 2012;</ns0:ref><ns0:ref type='bibr' target='#b57'>Rossettini et al. 2017</ns0:ref>) used this variable as outcome measure. All of them reported that between the groups the same or a similar amount of time was required for teaching. Data to evaluate the following comparisons were available:  Peyton's teaching approach versus a standard teaching approach (PEY vs ST)</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:07:50815:1:1:NEW 28 Aug 2020)</ns0:p><ns0:p>Manuscript to be reviewed  Peyton's teaching approach with peer teaching versus a standard teaching approach with peer teaching (PeerPey vs PeerSt)  Best practice skills lab with peer teaching versus a standard teaching approach with peer teaching (PeerBpsl vs PeerSt)  Media supported Peyton's teaching approach versus a standard teaching approach (MPey-St)  All forms of Peyton's teaching approach versus a standard teaching approach Table <ns0:ref type='table'>1</ns0:ref>. Characteristics of included studies *if multiple procedures or assessments were used in the primary studies the included procedures and assessments within this systematic review are underlined. During the controlling of the data set (https://doi.org/10.6084/m9.figshare.12619151) 7 data entries were flagged and double checked. This corresponded to 2.43% of the data set.</ns0:p></ns0:div>
<ns0:div><ns0:head>Analysis of effectiveness</ns0:head><ns0:p>Below the analysis of effectiveness is presented reporting on two outcomes (i.e. performance and time needed to perform the procedure) at two different endpoints (i.e. after acquisition and after a retention period).</ns0:p></ns0:div>
<ns0:div><ns0:head>Performance -post-acquisition test</ns0:head><ns0:p>Fourteen studies reporting on 17 samples with a total of 970 participants allocated to Peyton's teaching approach and 935 allocated to a standard teaching approach were included for the analysis of the outcome performance at post-acquisition testing. Four different sub-groups were identified. First, 9 studies compared Peyton's teaching approach against a standard teaching approach <ns0:ref type='bibr' target='#b5'>(Balafoutas et al. 2019;</ns0:ref><ns0:ref type='bibr' target='#b18'>Gradl-Dietsch et al. 2016;</ns0:ref><ns0:ref type='bibr' target='#b32'>Jenko et al. 2012;</ns0:ref><ns0:ref type='bibr' target='#b35'>Krautter et al. 2011;</ns0:ref><ns0:ref type='bibr' target='#b36'>Lapucci et al. 2018;</ns0:ref><ns0:ref type='bibr' target='#b39'>Lund et al. 2012;</ns0:ref><ns0:ref type='bibr' target='#b50'>Orde et al. 2010;</ns0:ref><ns0:ref type='bibr' target='#b56'>Romero et al. 2018;</ns0:ref><ns0:ref type='bibr' target='#b57'>Rossettini et al. 2017</ns0:ref>). The analysis showed an effect size of 0.5 SMD (95%CI 0.13 to 0.87) in favour of the Peyton group. Heterogeneity was substantial with an I 2 of 69%. Three studies compared peer or student tutor Peyton's teaching versus peer standard teaching <ns0:ref type='bibr' target='#b17'>(Gradl-Dietsch et al. 2019;</ns0:ref><ns0:ref type='bibr' target='#b19'>Gradl-Dietsch et al. 2018;</ns0:ref><ns0:ref type='bibr' target='#b46'>Münster et al. 2016</ns0:ref>). The effect size was in favour of peer standard teaching with a SMD of -0.15 (95%CI between -0.23 and -0.06). Heterogeneity was not important within this comparison (I 2 : 0%). One study reported on the comparison best practice skills lab (Peyton's teaching approach was part of the intervention) with peer tutors versus standard peer teaching <ns0:ref type='bibr' target='#b24'>(Herrmann-Werner et al. 2013)</ns0:ref>. A large effect in favour of best practice skills lab training was identified (SMD: 1.38; 95%CI between -0.56 and 3.32). The I 2 was 0% for this analysis. The last subgroup compared a media supported Peyton's teaching approach versus standard teaching <ns0:ref type='bibr' target='#b58'>(Ruesseler et al. 2019)</ns0:ref>. A small effect was analysed in favour of the Peyton group with a SMD of 0.24 and a 95%CI between -0.22 and 0.71. The overall model showed a small to moderate effect PeerJ reviewing PDF | (2020:07:50815:1:1:NEW 28 Aug 2020)</ns0:p><ns0:p>Manuscript to be reviewed size in favour of Peyton's teaching approach with an effect size of 0.45 SMD (95%CI between 0.15 and 0.75). Heterogeneity was substantial with an I 2 value of 82%. A prediction interval between -0.6 and 1.5 was analysed (Fig. <ns0:ref type='figure' target='#fig_2'>2</ns0:ref>). </ns0:p></ns0:div>
<ns0:div><ns0:head>Performance -retention test</ns0:head><ns0:p>Five studies were included for the outcome performance at retention testing. The studies reported a total of 169 participants in the Peyton group and 135 in the standard teaching group (Fig. <ns0:ref type='figure'>3</ns0:ref>). It was possible to analyse three different subgroups. First, three studies reported on the comparison Peyton versus standard teaching <ns0:ref type='bibr' target='#b18'>(Gradl-Dietsch et al. 2016;</ns0:ref><ns0:ref type='bibr' target='#b50'>Orde et al. 2010;</ns0:ref><ns0:ref type='bibr' target='#b57'>Rossettini et al. 2017)</ns0:ref>. A small to moderate effect in favour of the Peyton group was identified (SMD: 0.38; with a 95%CI between -0.14 and 0.9). Moderate heterogeneity was analysed (I 2 : 52%). The second subgroup compared peer best practice skills lab teaching with standard peer teaching <ns0:ref type='bibr' target='#b24'>(Herrmann-Werner et al. 2013)</ns0:ref>. A large effect size was analysed in favour of best practice skills lab training SMD: 2.54 (95%CI between 1.75 and 3.33). The third subgroup compared Peyton's peer teaching with standard peer teaching. An SMD of -0.11 with a 95% CI between -0.51 and 0.3 in favour of peer standard teaching was analysed. The random effects model over all subgroups showed a moderate to large effect size in favour of Peyton's teaching approach at retention testing (SMD: 0.7 with a 95%CI between -0.09 and 1.49). The heterogeneity of this analysis was large (I 2 : 90%). The retention period ranged between 1 month <ns0:ref type='bibr' target='#b57'>(Rossettini et al. 2017</ns0:ref>) and 6 months <ns0:ref type='bibr' target='#b18'>(Gradl-Dietsch et al. 2016)</ns0:ref>. Figure <ns0:ref type='figure'>3</ns0:ref>. Forest plot performance -Peyton's 4 step versus standard teaching at retention testing; Pey: Peyton's teaching; St: standard teaching; PeerBpsl: peer best practice skills lab; PeerSt: peer standard teaching; PeerPey: peer Peyton's teaching Time needed for procedure -post-acquisition test Six studies with a total of 657 participants in the Peyton group and 655 in the standard teaching group were included in this analysis (Fig. <ns0:ref type='figure' target='#fig_3'>4</ns0:ref>). Two different subgroups were identified. One study compared peer Peyton's teaching versus peer standard teaching <ns0:ref type='bibr' target='#b17'>(Gradl-Dietsch et al. 2019</ns0:ref>). An effect size of 0.05 SMD (95% CI between -0.07 and 0.18) was analysed. The second subgroup compared Peyton's teaching approach with standard teaching. Five studies were included in this analysis <ns0:ref type='bibr' target='#b35'>(Krautter et al. 2011;</ns0:ref><ns0:ref type='bibr' target='#b39'>Lund et al. 2012;</ns0:ref><ns0:ref type='bibr' target='#b50'>Orde et al. 2010;</ns0:ref><ns0:ref type='bibr' target='#b56'>Romero et al. 2018;</ns0:ref><ns0:ref type='bibr' target='#b57'>Rossettini et al. 2017)</ns0:ref>. Findings were in favour of Peyton's teaching approach with a large effect size of -1.06 SMD and a 95 % CI between -2.77 and 0.65. The overall model showed that participants in the Peyton groups needed considerably less time to perform the procedures at post-acquisition testing. A large effect size of -0.8 SMD (95%CI between -2.13 and 0.53) was associated with this finding. The heterogeneity for this analysis was large with an I 2 of 92%. The prediction interval was between -3.21 and 1.62. For the analysis time needed for the procedure at retention testing two studies were included <ns0:ref type='bibr' target='#b50'>(Orde et al. 2010;</ns0:ref><ns0:ref type='bibr' target='#b57'>Rossettini et al. 2017)</ns0:ref>. Both studies compared Peyton's 4 step teaching approach with a standard teaching approach. A large effect size of -2.65 SMD (95% CI: -7.77 to 2.47) showed that the time needed to perform the procedure was considerable shorter after a training using Peyton's teaching approach. Heterogeneity was large (I 2 : 98%). The retention period ranged between 1 month <ns0:ref type='bibr' target='#b57'>(Rossettini et al. 2017</ns0:ref>) and 2 months <ns0:ref type='bibr' target='#b50'>(Orde et al. 2010)</ns0:ref>.</ns0:p></ns0:div>
<ns0:div><ns0:head>Meta-regression student teacher-ratio -performance post-acquisition</ns0:head><ns0:p>A univariable meta-regression was performed to analyse whether the student-teacher ratio was an independent predictor of performance on post-acquisition tests. All studies from the metaanalysis 'performance -post-acquisition test' with exception of the study of Ordre et al. (2010) (i.e. the authors did not report the student-teacher ratio) were included into the meta-regression. The meta-regression showed that the effectiveness of Peyton's teaching approach was higher in studies with fewer of students per teacher (Fig. <ns0:ref type='figure'>5</ns0:ref>). The overall model explained 58% of the variability of the effect sizes (p: 0.01, r2: 56.86%) and the students per teacher variable showed that for one student more per teacher, the effect size was reduced by 0.08. This association was statistically significant (b1: -0.08 (95% CI: -0.14 to -0.0232), t: -2.96, p: 0.01). Figure <ns0:ref type='figure'>5</ns0:ref>. Scatterplot meta-regression students per teacher as predictor for performance at postacquisition testing</ns0:p></ns0:div>
<ns0:div><ns0:head>Risk of Bias</ns0:head><ns0:p>The risk of bias was low for all studies regarding the item random sequence generation with exception of the study of <ns0:ref type='bibr'>Ruesseler and colleagues (2019)</ns0:ref>, which was classified as unclear. Regarding the allocation concealment most studies were rated as unclear with exception of two studies <ns0:ref type='bibr' target='#b17'>(Gradl-Dietsch et al. 2019;</ns0:ref><ns0:ref type='bibr' target='#b32'>Jenko et al. 2012)</ns0:ref>. Blinding of participants and personnel was rated as high risk of bias in all studies with exception of the study of <ns0:ref type='bibr' target='#b57'>Rossettini et al. (2017)</ns0:ref>.</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:07:50815:1:1:NEW 28 Aug 2020)</ns0:p></ns0:div>
<ns0:div><ns0:head>Manuscript to be reviewed</ns0:head><ns0:p>The authors stated that the participants and teachers were blinded to the aims of the study. The risk of bias regarding outcome assessment was low. Only two studies were rated as unclear regarding this risk of bias item blinding of outcome assessment <ns0:ref type='bibr' target='#b36'>(Lapucci et al. 2018;</ns0:ref><ns0:ref type='bibr' target='#b46'>Münster et al. 2016)</ns0:ref>. One study was assessed as having a high risk of bias regarding incomplete outcome assessment because a relatively high number of study discontinuations were reported <ns0:ref type='bibr' target='#b46'>(Münster et al. 2016)</ns0:ref>. A summary risk of bias plot is presented in Fig. <ns0:ref type='figure'>6</ns0:ref>. Regarding the agreement of the human reviewer and the machine learning algorithm it was possible to compare 48 risk of bias evaluations. No conflicts occurred in 37 (77%) decisions and 11 (23%) decisions resulted in a conflict.</ns0:p></ns0:div>
<ns0:div><ns0:head>Figure 6. Summary risk of bias plot</ns0:head></ns0:div>
<ns0:div><ns0:head>Sensitivity analyses</ns0:head><ns0:p>Findings from a crossover study of Gradl-Dietsch and co-workers (2019) were integrated into the meta-analysis and the study was treated as parallel group trial. In order to address a potential unit of analysis issue a sensitivity analysis was performed. Because data from paired analyses were not available we adjusted the study data based on a method described by <ns0:ref type='bibr' target='#b14'>Elbourne et al. (2002)</ns0:ref>. A correlation coefficient derived from the data of <ns0:ref type='bibr' target='#b39'>Lund et al. (2012)</ns0:ref> was used to calculate an adjusted standard error. For the meta-analysis performance at post-acquisition, the standard error of the study decreased from 0.06 to 0.04. The effect estimate of the analysis peer Peyton versus peer standard teaching remained -0.15 SMD with a slightly changed 95% CI between -0.22 to -0.08. The adjusted standard error had only minimal influence on the meta-regression of the student teacher ratio at post-acquisition. The overall model (p: 0.01, r2: 57.54%) and the students per teacher variable (b1: -0.08 (95% CI: --0.14 to -0.0232), t: -2.96, p: 0.01) remained significantly related to the mean effect size. Within the meta-analysis time needed for the procedure at post-acquisition testing the sensitivity analysis resulted in a slightly smaller standard error of the Gradl-Dietsch et al. ( <ns0:ref type='formula'>2019</ns0:ref>) study. Therefore, the effect estimate of the comparison peer Peyton's teaching versus peer standard teaching changed to 0.05 SMD with a 95% CI between -0.05 and 0.16. The effect estimate of the overall model did not change.</ns0:p></ns0:div>
<ns0:div><ns0:head>Discussion</ns0:head><ns0:p>This systematic review with meta-analysis and integrated meta-regression set out to evaluate the effectiveness of Peyton's teaching approach compared with a standard teaching approach. The primary finding was that Peyton's teaching approach was more effective than a standard teaching approach on the acquisition of procedural skills at post-acquisition testing. A small to moderate effect size was associated with this finding. However, different subgroups of Peytons's teaching approach were analysed and effectiveness differed between subgroups. Two comparisons PeerJ reviewing PDF | (2020:07:50815:1:1:NEW 28 Aug 2020)</ns0:p><ns0:p>Manuscript to be reviewed showed findings in favour of Peyton's teaching approach when the procedure was instructed by teachers or faculty members (i.e. Peyton versus standard teaching and media supported Peyton's teaching approach versus a standard teaching approach). Two comparisons used peers to perform the procedural skills training. Peer Peyton versus peer standard teaching showed inconclusive results with a small effect size in favour of peer standard teaching. In contrast the comparison peer best practice skills lab versus peer standard teaching showed a large effect size in favour of peer best practice skills lab. Therefore, it remains unclear whether Peyton's teaching approach is effective when peers are used as tutors for the outcome skill acquisition. The meta-analysis of skill acquisition at retention testing was in favour of Peyton's teaching approach with a moderate to large effect size. Both subgroups were in favour of Peyton's approach. However, the effect size for the experimental group was considerable smaller compared to the findings at post-acquisition testing. The comparison peer best practice skills lab versus peer standard teaching showed a large effect size. Considerable larger than the effect size at post-acquisition testing. However, only one study reported on this comparison and more studies are needed to confirm this finding. Regarding the outcome time needed to perform the procedure the findings indicated that participants needed considerably less time to perform a procedure if Peyton's teaching approach was instructed by teachers or faculty members. One study showed a very large effect <ns0:ref type='bibr' target='#b57'>(Rossettini et al. 2017</ns0:ref>). This study showed some educational differences to the other studies in the analysis. For example, participants from physiotherapy education were used and the trained procedure was a cervical spine mobilisation. In addition, relatively few students per teacher participated in the teaching events. The potential influence of the different procedures on the effect estimate should be investigated in future studies. An increased effectiveness of Peyton's teaching approach at retention testing was analysed. This was mainly seen in the time needed for procedure outcome. The possible long-term comprehension advantage of Peyton's teaching approach has been previously discussed by <ns0:ref type='bibr' target='#b24'>Herrmann-Werner et al. (2013)</ns0:ref>. The authors showed that Peyton's teaching approach had an increased long-term effect on the acquisition of simple and complex skills. This finding is of educational importance because deterioration of procedural skills is likely after several weeks <ns0:ref type='bibr' target='#b7'>(Bonrath et al. 2012</ns0:ref>) and Peyton's teaching approach could be a useful educational method to reduce this. The meta-regression with the student-teacher ratio as independent predictor showed that Peyton's teaching approach was more effective in groups with fewer students per teacher. This supports the idea that Peyton's teaching approach was designed for a teaching ratio of 1:1 <ns0:ref type='bibr' target='#b48'>(Nikendei et al. 2014)</ns0:ref>. The student-teacher ratio of the analysed studies ranged between 13:1 <ns0:ref type='bibr' target='#b46'>(Münster et al. 2016</ns0:ref>) and several studies using a 1:1 ratio <ns0:ref type='bibr' target='#b5'>(Balafoutas et al. 2019;</ns0:ref><ns0:ref type='bibr' target='#b18'>Gradl-Dietsch et al. 2016;</ns0:ref><ns0:ref type='bibr' target='#b35'>Krautter et al. 2011;</ns0:ref><ns0:ref type='bibr' target='#b56'>Romero et al. 2018)</ns0:ref>. In studies where 9 or more students per teacher were used the treatment effect was close to zero. The highest effect sizes were analysed in studies using a student teacher ratio of 3:1 <ns0:ref type='bibr' target='#b24'>(Herrmann-Werner et al. 2013;</ns0:ref><ns0:ref type='bibr' target='#b57'>Rossettini et al. 2017)</ns0:ref>. This indicates that Peyton's teaching approach should ideally be used in groups with 1 to PeerJ reviewing PDF | (2020:07:50815:1:1:NEW 28 Aug 2020)</ns0:p><ns0:p>Manuscript to be reviewed 3 students per teacher. If this is not possible, it could be argued that group sizes with less than 9 students per teacher are still in favour of Peyton's teaching approach. Furthermore, it should be reported that <ns0:ref type='bibr' target='#b46'>Münster et al. (2016)</ns0:ref> reported a median group size of 13 students with a range between 9 and 13 participants and <ns0:ref type='bibr' target='#b58'>Ruesseler et al. (2019)</ns0:ref> reported a maximum group size of 6 participants per teacher. These summary estimates of the variable were used within the meta-regression, but this might have caused some imprecision. In addition, the variable student-teacher ratio was not reported in the study of <ns0:ref type='bibr' target='#b50'>Orde et al. (2010)</ns0:ref> and therefore the study was not included into the meta-regression. The control intervention in this review was labelled as 'standard teaching' approach. However, the educational approaches used within the control arms presented a source of heterogeneity. A broad range of approaches was identified such as: Halsted teaching, 2-stage teaching approach, Orde's 2-step method, standard instructions, traditional bedside teaching or see one -do one. These educational approaches show considerable similarities but are not exactly the same interventions. However, all of the standard teaching approaches have in common that they did not include the third step of Peyton's teaching approach (i.e. guiding the teacher through the procedure), which is assumed to be beneficial for skill acquisition <ns0:ref type='bibr' target='#b18'>(Gradl-Dietsch et al. 2016;</ns0:ref><ns0:ref type='bibr' target='#b57'>Rossettini et al. 2017</ns0:ref>). To deal with these differences several subgroup analyses were performed. In addition, the meta-analysis was performed using a random effects model. Within the subgroups the statistical heterogeneity was considerable smaller compared to the overall analyses. The overall analyses showed substantial heterogeneity and should therefore be analysed with caution. Eligible outcome assessments for this systematic review were assessments of procedural skills, which could be a procedure specific checklist or a global rating scale. However, when studies reported both types of assessments, the checklists were preferred. This was justified on the basis of the suggested best methods for evaluation by the Accreditation Council for Graduate Medical Education (ACGME) <ns0:ref type='bibr' target='#b1'>(ACGME 2000;</ns0:ref><ns0:ref type='bibr' target='#b69'>Swing 2002</ns0:ref>). Within the guideline, checklists are recommended as 'most desirable' when assessing medical procedures. Rating scales are recommended as 'potentially applicable method'. Therefore, we preferred data based on procedure specific checklists. However, this is a controversial topic and some authors have reported that global rating scales have additional values and should be used when procedural skills are evaluated <ns0:ref type='bibr' target='#b40'>(Ma et al. 2012;</ns0:ref><ns0:ref type='bibr' target='#b54'>Regehr et al. 1998</ns0:ref>).</ns0:p></ns0:div>
<ns0:div><ns0:head>Limitations</ns0:head><ns0:p>Several other potential effect modifiers exist, which were not explored in this study because we did not specify these analyses in the study protocol. First, <ns0:ref type='bibr' target='#b18'>Gradl-Dietsch et al. (2016)</ns0:ref> reported that gender might be considered as potential moderator variable for the effectiveness of Peyton's teaching approach. Within their study the authors suggested that men might benefit more from Peyton's teaching approach compared to women. This could be explained by the results of <ns0:ref type='bibr' target='#b2'>Ali et al. (2015)</ns0:ref>. The authors reported in a systematic review that the acquisition of surgical skills differs between men and women. However, it is difficult to investigate the gender variable with a PeerJ reviewing PDF | (2020:07:50815:1:1:NEW 28 Aug 2020)</ns0:p><ns0:p>Manuscript to be reviewed meta-regression because relatively few studies reported the findings for men and women separately. Second, acquiring simple procedures is different from acquiring complex skills <ns0:ref type='bibr' target='#b75'>(Wulf & Shea 2002)</ns0:ref>. Therefore, the complexity of the procedural skills might affect the effectiveness of Peyton's teaching approach. However, rating the complexity of the included procedures is challenging as procedures from various domains of health professions education were included. Third, the experience of the teacher teaching the procedural skill and the experience of the students learning the skill might affect the effectiveness of Peyton's teaching approach. Findings from a crossover trial of Gradl-Dietsch and co-workers (2019) were integrated into the meta-analysis. Findings from a paired analysis were not available and therefore we used the reported values and treated the study as a parallel group trial. However, when the results of randomised controlled trials and crossover studies are combined, the results of crossover studies should be based on paired analyses <ns0:ref type='bibr' target='#b14'>(Elbourne et al. 2002)</ns0:ref>. If findings from unpaired analyses are used the confidence intervals are likely too wide and this might give rise to a unit of analysis issue <ns0:ref type='bibr' target='#b27'>(Higgins et al. 2019)</ns0:ref>. As a consequence, we performed a sensitivity analysis and adjusted the standard errors using a method described by <ns0:ref type='bibr' target='#b14'>Elbourne et al. (2002)</ns0:ref>. A correlation coefficient derived from the data of <ns0:ref type='bibr' target='#b39'>Lund et al. (2012)</ns0:ref> was used to calculate the adjusted standard errors. Unfortunately, it was only possible to calculate the correlation coefficient using the Lund et al. study. The remaining studies did not provide sufficient data. However, findings remained similar after the sensitivity analysis. The only differences were slightly changed 95% confidence intervals. We have therefore decided to include the study by <ns0:ref type='bibr' target='#b17'>Gradl-Dietsch et al. (2019)</ns0:ref> in the analysis. An additional limitation of this review might be that we did not include studies reporting about the effectiveness of George and Doto's teaching approach (2001). Peyton's and George and Doto's teaching approach are similar regarding their stepwise teaching structure. However, the inclusion of this additional educational intervention would have increased the heterogeneity considerably. In view of the relatively high proportion of analysed heterogeneity within our pairwise analyses, we decided against it. However, in the context of a network meta-analysis future studies could possibly compare these two and other reported teaching approaches for the acquisition of procedural skills.</ns0:p></ns0:div>
<ns0:div><ns0:head>Implications for research</ns0:head><ns0:p>Several implications for research were identified. First, the effectiveness of Peyton's teaching approach on skill acquisition should be explored in various health professions. The included studies reported on the use of Peyton's teaching approach in medical education. Only three studies were found analysing this approach in other health professions. Further studies are therefore needed to investigate this approach in the field of nursing or physiotherapy. Second, the proposed moderator variables gender, skill complexity and level of experience of teacher and students should be further explored. Third, more evidence is needed regarding the use of peer teachers. Fourth, the high effectiveness of the best practice skills lab training should be explored Manuscript to be reviewed in further studies. In addition, future studies should investigate a stabilised learning of motor skills with long-term follow up (during the retention phase). Moreover, there is a need to consider also the assessment of the motor skill acquired in ecological settings (e.g. during internships) suggesting an adequate transfer phase.</ns0:p></ns0:div>
<ns0:div><ns0:head>Implications for practice</ns0:head><ns0:p>Peyton's teaching approach is effective for the acquisition of procedural skills. The evidence is robust for the field of medical education. One might assume that the acquisition of skills in other health professions could also benefit from Peyton's teaching approach. However, this must be further investigated. When Peyton's teaching approach is used the number of students per teacher should be small (e.g. ranging between 1 and 3 students per teacher) to be more effective than a standard teaching approach. Implications for teachers in different healthcare fields (e.g. nursing, physiotherapy or speech and language therapy education) are less robust. However, some procedures within this review are used across healthcare fields. For example, procedures from manual therapy were used in medical education <ns0:ref type='bibr' target='#b18'>(Gradl-Dietsch et al. 2016)</ns0:ref> and in physiotherapy education <ns0:ref type='bibr' target='#b57'>(Rossettini et al. 2017)</ns0:ref>. Educators teaching these procedural skills in different healthcare fields are encouraged to use Peyton's teaching approach (i.e. within the discussed limitations). In addition, given the broad spectrum of included procedures in this review it seems likely that Peyton's teaching approach also applies to procedures in different healthcare fields, but this needs further investigation.</ns0:p></ns0:div>
<ns0:div><ns0:head>Conclusions</ns0:head><ns0:p>Peyton's teaching approach is an effective teaching approach for skill acquisition of procedural skills when faculty members are used as teachers. When peer students or student tutors are used as teachers the effectiveness of Peyton's teaching approach is less clear. Peyton's teaching approach is more effective when small groups with few students per teacher are used. </ns0:p></ns0:div>
<ns0:div><ns0:head>List of abbreviations</ns0:head></ns0:div><ns0:figure xml:id='fig_0'><ns0:head>Figure 1 .</ns0:head><ns0:label>1</ns0:label><ns0:figDesc>Prisma flow diagramPeerJ reviewing PDF | (</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_1'><ns0:head /><ns0:label /><ns0:figDesc>, spine mobilisations (Gradl-Dietsch et al. 2016; Rossettini et al. 2017), musculoskeletal ultrasound (Gradl-Dietsch et al. 2019) or cardiopulmonary resuscitation (Jenko et al. 2012) were used as procedures. Several modified versions of Peyton's teaching approaches were used in the experimental groups. All studies with exception of five studies (Gradl-Dietsch et al. 2019; Gradl-Dietsch et al. 2018; Herrmann-Werner et al. 2013; Münster et al. 2016; Ruesseler et al. 2019) used a standard version of Peytons's teaching approach. The study of Herrmann-Werner et al. (2013) used a best practice skills laboratory, which consisted of structured individual feedback, performance on manikins and Peyton's teaching approach supervised by student tutors. Three studies (Gradl-Dietsch et al. 2019; Gradl-Dietsch et al. 2018; Münster et al. 2016) used peer or student teachers for the teaching events and Ruesseler et al. (2019) used a video 4-step approach. The teaching approach in the control groups was described as traditional Halsted teaching (Balafoutas et al. 2019; Romero et al. 2018), peer teaching or student tutors teaching (Gradl-Dietsch et al. 2019; Gradl-Dietsch et al. 2018; Herrmann-Werner et al. 2013; Münster et al. 2016), 2-stage teaching approach (Jenko et al. 2012), Orde's 2-step method (Lapucci et al. 2018; Orde et al. 2010), standard instructions (Gradl-Dietsch et al. 2016; Krautter et al. 2011), traditional bedside teaching</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_2'><ns0:head>Figure 2 .</ns0:head><ns0:label>2</ns0:label><ns0:figDesc>Forest plot performance -Peyton's 4 step versus standard teaching at post-acquisition testing; Pey: Peyton's teaching; St: standard teaching; PeerPey: peer Peyton's teaching; PeerSt: peer standard teaching; PeerBpsl: peer best practice skills lab; MPey: Media supported Peyton NB. Gradl-Dietsch et al. (2018) and Gradl-Dietsch et al. (2016) are presented as two samples because data for women and men are analysed separately (a: woman, b: men). Data from Herrmann-Werner et al. (2013) are presented as two samples (a: participants with a 3 months follow up, b: participants with a 6 months follow up)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_3'><ns0:head>Figure 4 .</ns0:head><ns0:label>4</ns0:label><ns0:figDesc>Forest plot time needed for procedure -Peyton's 4 step versus standard teaching at post-acquisition testing; PeerPey: peer Peyton's teaching; PeerSt: peer standard teaching; Pey: Peyton's teaching; St: standard teaching Time needed for procedure -retention test</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_4'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:07:50815:1:1:NEW 28 Aug 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_5'><ns0:head /><ns0:label /><ns0:figDesc>multiple procedures or assessments were used in the primary studies the included procedures and assessments within this 5PeerJ reviewing PDF | (2020:07:50815:1:1:NEW 28 Aug 2020)</ns0:figDesc></ns0:figure>
<ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='26,42.52,311.34,525.00,375.00' type='bitmap' /></ns0:figure>
<ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='27,42.52,229.87,525.00,371.25' type='bitmap' /></ns0:figure>
<ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='28,42.52,250.12,525.00,371.25' type='bitmap' /></ns0:figure>
<ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='29,42.52,275.62,525.00,371.25' type='bitmap' /></ns0:figure>
<ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='30,42.52,178.87,525.00,369.75' type='bitmap' /></ns0:figure>
<ns0:note place='foot'>PeerJ reviewing PDF | (2020:07:50815:1:1:NEW 28 Aug 2020)</ns0:note>
</ns0:body>
" | "
Leukerbad, 24 August 2020
Rebuttal letter
The effectiveness of the Peyton’s 4-step teaching approach on skill acquisition of procedures in health professions education: A systematic review and meta-analysis with integrated meta-regression
Dear Professor Sewall,
We thank the reviewers for their helpful comments on the manuscript and have amended the manuscript accordingly.
The amendments are discussed point by point below.
On behalf of the co-authors
Dr Martin Sattelmayer
Associate Professor UAS
HES-SO Valais-Wallis
Reviewer: Dimitrios Balafoutas
Comment 1: Abstract: In the “Background” section both aims of the study are defined. In Line 266 the effectiveness is reported on two outcomes. The authors may choose to include the time-needed-for-procedure variable in the “Results” section of the abstract.
Response: Thank you for this comment. We added the findings of time needed for procedure variable to the abstract-results section. (l: 35-38)
The groups using Peyton’s teaching approach needed considerably less time to perform the procedure at post-acquisition (SMD: -0.8; 95%CI: -2.13 to 1.62) and retention (SMD: -2.65; 95%CI: -7.77 to 2.47) testing.
Comment 2: In the introduction (lines 100-101) the time-for-teaching variable is referred. A comment in the included studies section of the results may state the fact that the time-for-teaching was identical in both groups (i.e. 3h in Grandl-Dietsch 2019, 90 min in Grandl-Dietsch 2016 …).
Response: We agree with you that this information is important. We added the column “time required for teaching” in Table 1. In addition, the following brief statement was added to the results section of the included studies (l: 280-283):
The time allocated to the teaching of the procedural skills was set equal in most included studies. Four studies (Herrmann-Werner et al. 2013; Krautter et al. 2011; Lund et al. 2012; Rossettini et al. 2017) used this variable as outcome measure. All of them reported that between the groups the same or a similar amount of time was required for teaching.
Comment 3: In the Introduction (lines 63-65) a definition is given for the procedural skills. This definition could be omitted, since it neither has an impact in the interpretation of the results, nor is safety an endpoint in the participating studies (only the “correct” performance of the procedure). Alternatively, the authors could state that the procedural skills relate to the parameters examined by each author in his/her field of expertise.
Response: We agree with you that the used definition of procedural skills might not apply to all included studies. Therefore, we have omitted the definition from the manuscript.
Comment 4: (Line 68-69) Sentence is missing the verb
Response: Thank you. We changed the sentences to (66-69):
This means that a teacher demonstrates and describes a procedure and afterwards the students are asked to practice the procedure. This is referred to as Halsted's teaching approach, which is based on the surgeon William Steward Halsted (1904).
Comment 5: Line 184: Hedges’ g is by most authors capital
Response: Thank you. We corrected this (l. 212).
Comment 6: In the discussion there should be a comment on the increased effect of Peyton’s approach in the retention testing (vs post-acquisition testing), mainly in the time-needed-for-procedure variable. The possible long-term comprehension advantage of deconstructive teaching has been previously discussed (see references).
Response: thank you for this advice. We added the following statement to the discussion section (l. 464-471).
An increased effectiveness of Peyton’s teaching approach at retention testing was analysed. This was mainly seen in the time needed for procedure outcome. The possible long-term comprehension advantage of Peyton’s teaching approach has been previously discussed by Herrmann-Werner et al. (2013). The authors showed that Peyton’s teaching approach had an increased long-term effect on the acquisition of simple and complex skills. This finding is of educational importance because deterioration of procedural skills is likely after several weeks (Bonrath et al. 2012) and Peyton’s teaching approach could be a useful educational method to reduce this.
Reviewer: Yajnavalka Banerjee
Comment 1. The authors have rightly discussed the advantages of adopting Peyton’s 4-step approach over traditional pedagogical strategy in the dissemination of complex procedural skills in medical education. However, it would add to the manuscript if they can also touch upon George and Doto's 5-step approach.
Response 1: Thank you for this advice. We added the following section reporting about Doto’s teaching approach to the introduction section (l. 86-92):
A similar stepwise teaching approach was presented by George (American College of Surgeons 1997) and later published by George and Doto (2001). Originally it was developed as an educational technique to support the American College of Surgeon’s Advanced Trauma Life Support course. In contrast to Peyton’s teaching approach George and Doto used five steps. Within Peyton’s teaching approach two of the five steps are collated into a single step. George and Doto based their teaching approach on Simpson’s taxonomy of the psychomotor domain (Simpson 1966).
Response 2: In addition, we added the following statement to the discussion-limitation section (l. 543-550):
An additional limitation of this review might be that we did not include studies reporting about the effectiveness of George and Doto’s teaching approach (2001). Peyton’s and George and Doto’s teaching approach are similar regarding their stepwise teaching structure. However, the inclusion of this additional educational intervention would have increased the heterogeneity considerably. In view of the relatively high proportion of analysed heterogeneity within our pairwise analyses, we decided against it. However, in the context of a network meta-analysis future studies could possibly compare these two and other reported teaching approaches for the acquisition of procedural skills.
Comment 2. Yes, the third step of Peyton is the crucial step, but so is the fourth, as in this step (Elicit performance), the instructor provides feedback to the learner. In fact, a systematic review by Issenberg et al a key component of clinical skill teaching and learning are the opportunity for feedback while practicing the skill (the authors are requested to refer to Features and uses of high-fidelity medical simulations that lead to effective learning: a BEME systematic review, Medical Teacher, 27 (1) (2005), pp. 10 -28 f0r details). Additionally, this step of Peyton’s pedagogical strategy is also supported by Bandura’s scaffolding theory (refer to Learning theories by DH Schunk) . The above aspects should be included in the section of the introduction where the authors have elaborated on the individual steps.
Response: Thank you for this comment. We added the following paragraph (l. 115-119):
Besides the third step, the fourth step is also of educational importance as in this step the teacher provides feedback to the learner. A systematic review by Issenberg reported that the opportunity to provide feedback is a key component for effective skill acquisition in simulation-based medical education (Issenberg et al. 2005). In addition, the fourth step is also supported by Bandura’s scaffolding theory (Schunk 2012).
3. One of the strengths of Peyton instructional design strategy is that it can be effectively combined with other instructional design strategies, which allows the simultaneous delivery of theoretical concepts along with complex procedural skills. In fact, this reviewer hasn’t come across any other pedagogical strategy except Peyton, which has been effectively employed in the design of a blended pedagogical approach. This aspect should be included in the introduction which will strengthen the rationale of the study. (The authors can refer to the following articles: Blending Gagne's Instructional Model with Peyton's Approach to Design an Introductory Bioinformatics Lesson Plan for Medical Students: Proof-of-Concept Study. JMIR Med Educ. 2018;4(2):e11122.; Combining Peyton's four-step approach and Gagne's instructional model in teaching slit-lamp examination.” Perspectives on medical education vol. 3,6 (2014): 480-5.)
Response: Thank you for this important information. The following information was added to the manuscript (l. 120-125):
One of the strengths of Peyton’s teaching approach is that it can be effectively combined with other instructional design strategies, which allows the simultaneous delivery of theoretical concepts along with complex procedural skills. For example, Tambi et al. (2018) and colleagues combined Gagne’s instructional model (Gagne et al. 2005) with Peyton’s teaching approach to design a bioinformatics lesson plan for medical students and Ng (2014) combined both teaching approaches for slit-lamp teaching.
Reviewer 3
Material and methods:
Comment 1:
-Searches: please could you report who performed the search phase? Moreover, could you explode the acronym ERIC as “Education Resources Information Center”
Response: Thank you, we exploded the ERIC acronym and added the requested information (l. 155):
We searched the following electronic databases for eligible studies: Medline, PsycInfo, Embase and Education Resources Information Center (ERIC). The search was performed by KMS.
Comment 2:
-line 130: please consider to change “was build” in “was prepared”.
Response: Thank you. We changed this (l. 156).
Comment 3:
-Study selection, data extractions, risk of bias assessment: please could you report how you have measure the disagreement among authors, thus including a statistical measure (e.g., k, ICC? With 95%IC)
Response 1: We have added the following sentence to the results section for study selection (l. 242-243)
During the study selection process, 6 conflicts occurred, representing 1.4% of the total decisions.
Response 2: We have added the following sentences to the results section for data controlling (l. 299-300)
During the controlling of the data set (https://doi.org/10.6084/m9.figshare.12619151) 7 data entries were flagged and double checked. This corresponded to 2.43% of the data set.
Response 3: We added the following sentences to the results section for the risk of bias evaluation (l. 406-408):
Regarding the agreement of the human reviewer and the machine learning algorithm it was possible to compare 48 risk of bias evaluations. No conflicts occurred in 37 (77%) decisions and 11 (23%) decisions resulted in a conflict.
Comment 4:
Risk of bias:
-line 362-363: I have checked the original paper of Rossettini et al. The authors declared in the abstract: “Participants, teachers and assessors were blinded to the aims of the study”. Moreover, they claim in the material and methods “Teachers were blinded to the study outcomes”, “Each teacher was trained in only one method in order to prevent contamination during the sessions”, and “Both teachers received a detailed sheet regarding students’ learning goals, the duration of the session and the steps required by the approach“. Accordingly, please modify your sentence aimed to be adherent with the original manuscript.
Response: We apologize for this statement. We have reworded the sentence as proposed. We hope that you will agree to this new sentence (l. 393-394):
The authors stated that the participants and teachers were blinded to the aims of the study.
Comment 5:
Discussion
-line 455: please could you explode the acronym “ACGME”?
Response: We are sorry for using only the acronym. We added the full form of ACGME (l. 507):
This was justified on the basis of the suggested best methods for evaluation by the Accreditation Council for Graduate Medical Education (ACGME) (ACGME 2000; Swing 2002).
Comment 6:
Implications for research:
-I suggest authors to add also another implication regard learning (as suggested by Wulf’s research group). We consider an “acquisition phase” when we acquire a motor skill at short term; while a “retention” when there is a follow up and a “transfer phase” when we translate the learning in other ecological settings. Accordingly, future studies should investigate a stabilized learning of motor skills with long-term follow up (during retention phase). Moreover, there is a need to consider also the assessment of the motor skill acquired in ecological settings (e.g., during internship) suggesting an adequate transfer phase.
Response: We agree with you. These are important implications for research. We added these points to the appropriate section (l. 560-563):
In addition, future studies should investigate a stabilised learning of motor skills with long-term follow up (during the retention phase). Moreover, there is a need to consider also the assessment of the motor skill acquired in ecological settings (e.g. during internships) suggesting an adequate transfer phase.
Comment 7:
Conclusion:
The findings of this study are really relevant for educators. I suggest authors also to include other possible implications for inter professional education. Accordingly, how educators involved in different healthcare fields (e.g., nursing, medicine, physiotherapy, speech therapy) can implement your findings in their educational practice?
Response: We agree with you that the implication for practice should be extended. The following section was added (l. 570-578):
Implications for teachers in different healthcare fields (e.g. nursing, physiotherapy or speech and language therapy education) are less robust. However, some procedures within this review are used across healthcare fields. For example, procedures from manual therapy were used in medical education (Gradl-Dietsch et al. 2016) and in physiotherapy education (Rossettini et al. 2017). Educators teaching these procedural skills in different healthcare fields are encouraged to use Peyton’s teaching approach (i.e. within the discussed limitations). In addition, give the broad spectrum of included procedures in this review it seems likely that Peyton’s teaching approach also applies to procedures in different healthcare fields, but this needs further investigation.
Comment 8:
Appendix 1:
I thank you for providing the Appendix 1, however please add also the search strategies for all the databases. This strategy will improve the overall transparency of the manuscript.
Response: We added the search strategies of all databases to Appendix 1.
Comment 9:
References.
In the introduction section please include these fundamental references when you introduce the Peyton’s four approach.
* American College of Surgeons. Advanced Trauma Life Support for Doctors, 6th edn. Chicago, IL: American College of Surgeons 1997.
* George JH, Doto FX. A simple five-step method for teaching clinical skills. Fam Med 2001;33 (8):577–8.
Response: Thank you for this remark. We added both References to the introduction section. In addition, we added a short paragraph and introduced George and Doto’s 5-step method (l. 86-92):
A similar stepwise teaching approach was presented by George (American College of Surgeons 1997) and later published by George and Doto (2001). Originally it was developed as an educational technique to support the American College of Surgeon’s Advanced Trauma Life Support course. In contrast to Peyton’s teaching approach George and Doto used five steps. Within Peyton’s teaching approach two of the five steps are collated into a single step. George and Doto based their teaching approach on Simpson’s taxonomy of the psychomotor domain (Simpson 1966).
References
ACGME. 2000. Suggested best methods for evaluation. ACGME/ABMS Joint Initiative Attachment/Toolbox of Assessment Methods (September 2000).
American College of Surgeons. 1997. Advanced Trauma Life Support for Doctors. Chicago, Il: American College of Surgeons.
Bonrath EM, Weber BK, Fritz M, Mees ST, Wolters HH, Senninger N, and Rijcken E. 2012. Laparoscopic simulation training: testing for skill acquisition and retention. Surgery 152:12-20.
Gagne RM, Wager WW, Golas KC, Keller JM, and Russell JD. 2005. Principles of instructional design. Performance Improvement 44:44-46.
George JH, and Doto FX. 2001. A simple five-step method for teaching clinical skills. Fam Med 33:577-578.
Gradl-Dietsch G, Lübke C, Horst K, Simon M, Modabber A, Sönmez TT, Münker R, Nebelung S, and Knobe M. 2016. Peyton's four-step approach for teaching complex spinal manipulation techniques - a prospective randomized trial. BMC medical education 16:284.
Halsted WS. 1904. The training of the surgeon. Bull Johns Hop Hosp:267-275. {Halsted, 1904 #45}
Herrmann-Werner A, Nikendei C, Keifenheim K, Bosse HM, Lund F, Wagner R, Celebi N, Zipfel S, and Weyrich P. 2013. 'Best Practice' Skills Lab Training vs. a 'see one, do one' Approach in Undergraduate Medical Education: An RCT on Students' Long-Term Ability to Perform Procedural Clinical Skills. PloS one 8.
Issenberg BS, Mcgaghie WC, Petrusa ER, Lee Gordon D, and Scalese RJ. 2005. Features and uses of high-fidelity medical simulations that lead to effective learning: a BEME systematic review. Medical teacher 27:10-28.
Krautter M, Weyrich P, Schultz JH, Buss SJ, Maatouk I, Ünger JJ, and Nikendei C. 2011. Effects of peyton's four-step approach on objective performance measures in technical skills training: A controlled trial. Teaching and learning in medicine 23:244-250.
Lund F, Schultz JH, Maatouk I, Krautter M, Möltner A, Werner A, Weyrich P, Jünger J, and Nikendei C. 2012. Effectiveness of IV cannulation skills laboratory training and its transfer into clinical practice: A randomized, controlled trial. PloS one 7.
Ng JY. 2014. Combining Peyton’s four-step approach and Gagne’s instructional model in teaching slit-lamp examination. Perspectives on medical education 3:480-485.
Rossettini G, Rondoni A, Palese A, Cecchetto S, Vicentini M, Bettale F, Furri L, and Testa M. 2017. Effective teaching of manual skills to physiotherapy students: a randomised clinical trial. Medical education 51:826-838.
Schunk DH. 2012. Learning theories an educational perspective sixth edition: Pearson.
Simpson EJ. 1966. The classification of educational objectives, psychomotor domain.
Swing SR. 2002. Assessing the ACGME general competencies: general considerations and assessment methods. Academic Emergency Medicine 9:1278-1288.
Tambi R, Bayoumi R, Lansberg P, and Banerjee Y. 2018. Blending Gagne’s Instructional Model with Peyton’s Approach to Design an Introductory Bioinformatics Lesson Plan for Medical Students: Proof-of-Concept Study. JMIR medical education 4:e11122.
" | Here is a paper. Please give your review comments after reading it. |
9,769 | "<ns0:abstract xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:p>This study was aimed to assess the relationship between serum uric acid (SUA) level and the clinical, pathological phenotype of IgA nephropathy (IgAN), and to determine the role of SUA level in the progression and prognosis of IgAN.</ns0:p></ns0:div>
<ns0:div><ns0:head>Methods.</ns0:head><ns0:p>A total of 208 patients with IgAN were included in this study, which was classified into the normo-uricemia group and hyperuricemia group according to the SUA level. The clinical data at baseline, IgA Oxford classification scores (MEST-C scoring system), and other pathological features were collected and further analyzed. All patients were followed up and the prognosis was assessed using Kaplan-Meier survival curves. GraphPad Prism 7.0 and SPSS 23.0 were used for statistical analyses.</ns0:p></ns0:div>
<ns0:div><ns0:head>Results.</ns0:head><ns0:p>In clinical indicators, patients with hyperuricemia had the significantly higher proportion of males to females, mean arterial pressure, the levels of total cholesterol, triglyceride, Scr, BUN, 24 hoururine protein, C3, and C4, the lower levels of high-density lipoprotein cholesterol and eGFR than those without(p<0.05). In terms of pathological characteristics, the tubular atrophy/interstitial fibrosis scores, vascular injury scores, and glomerular sclerosis percentage were significantly higher in patients with hyperuricemia compared with those without(p<0.01). There was no significant difference in the scores of mesangial hypercellularity, endocapillary hypercellularity, focal segmental glomerulosclerosis, as well as crescents between the two groups(p>0.05). As for the depositions of immune complexes deposition in IgAN, the hyperuricemia group had less deposition of immunoglobulin G and FRA than the normouricemia group (p<0.05), while the deposition of immunoglobulin A, immunoglobulin M, and complement C3 in the two groups showed no statistical difference. The survival curve suggested that patients in the hyperuricemia group have significantly poorer renal outcome than those in the normo-uricemia group (p= 0.0147). Results also revealed that the SUA level is a valuable predictor of renal outcome in patients with IgAN. The optimal cutoff value was 361.1μmol/L (AUC=0.76±0.08167) and 614 μmol/L (AUC=0.5728±0.2029) for female and male, respectively.</ns0:p></ns0:div>
<ns0:div><ns0:head>Conclusions.</ns0:head><ns0:p>The level of SUA is associated with renal function level and pathological severity of IgAN, and maybe a prognostic indicator of IgAN.</ns0:p></ns0:div>
</ns0:abstract>
<ns0:body xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:head>Introduction</ns0:head><ns0:p>IgA Nephropathy (IgAN), whose feature is the deposition of IgA-dominant in the mesangial area, is the most prevalent primary glomerular disease in the world. Approximately 40% of patients almost completely lose their renal function within 30 years after diagnosis <ns0:ref type='bibr' target='#b16'>(Coppo & D'Amico, 2005;</ns0:ref><ns0:ref type='bibr' target='#b19'>D'Amico, 2004)</ns0:ref>. Previous studies have demonstrated that clinical features, including severe proteinuria, poor renal function, and hypertension at the initial diagnosis are the predictors of poor outcomes in IgAN <ns0:ref type='bibr' target='#b33'>(Le et al., 2012;</ns0:ref><ns0:ref type='bibr' target='#b1'>Barbour & Reich, 2018;</ns0:ref><ns0:ref type='bibr' target='#b43'>Mariani et al., 2018;</ns0:ref><ns0:ref type='bibr' target='#b5'>Barbour & Reich, 2012)</ns0:ref>. Histological features were also identified as crucial predictors of IgAN patients.</ns0:p><ns0:p>The Oxford classification of IgAN, also known as MEST score, including mesangial hypercellularity (M), endocapillary hypercellularity (E), segmental glomerulosclerosis (S), tubular atrophy/interstitial fibrosis (T) has been shown to be of significant value in predicting the prognosis IgAN <ns0:ref type='bibr'>(Barbour et al., 2016;</ns0:ref><ns0:ref type='bibr' target='#b42'>Maixnerova et al., 2016;</ns0:ref><ns0:ref type='bibr' target='#b25'>Herzenberg et al., 2011a;</ns0:ref><ns0:ref type='bibr' /> PeerJ reviewing PDF | (2020:05:48827:1:1:NEW 27 Aug 2020)</ns0:p><ns0:p>Manuscript to be reviewed <ns0:ref type='bibr' target='#b29'>Katafuchi et al., 2011;</ns0:ref><ns0:ref type='bibr'>Coppo et al., 2014a;</ns0:ref><ns0:ref type='bibr' target='#b39'>Lv et al., 2013)</ns0:ref>. Recently, crescents (C) have been proposed to add to the Oxford classification of IgAN to form an updated MEST-C score system, which can provide a more comprehensive pathological prediction for the prognosis of IgAN <ns0:ref type='bibr' target='#b69'>(Trimarchi et al., 2017)</ns0:ref>.</ns0:p><ns0:p>Numerous studies have shown that the level of uric acid can predict the incidence of atherosclerosis <ns0:ref type='bibr' target='#b21'>(Feig 2014;</ns0:ref><ns0:ref type='bibr' target='#b24'>Gustafsson & Unwin 2013;</ns0:ref><ns0:ref type='bibr' target='#b35'>Li et al. 2014)</ns0:ref>, hypertension <ns0:ref type='bibr' target='#b71'>(Wang et al. 2014)</ns0:ref>, and coronary heart disease <ns0:ref type='bibr' target='#b31'>(Kim et al. 2010)</ns0:ref>. Moreover, some studies have emphasized that high levels of serum uric acid (SUA) would form urates crystals that deposit in renal tubules and interstitial, leading to kidney fibrosis and failure <ns0:ref type='bibr' target='#b64'>(Su et al. 2020;</ns0:ref><ns0:ref type='bibr' target='#b70'>Viggiano et al. 2018)</ns0:ref>.</ns0:p><ns0:p>Recently, some studies have found that there is a correlation between the SUA level and the progression of IgAN. A cohort study has shown that hyperuricemia is associated with the progression of IgAN, however, the effect of hyperuricemia on renal pathology was not evaluated <ns0:ref type='bibr' target='#b0'>(Bakan et al. 2015)</ns0:ref> in this study. The study conducted by <ns0:ref type='bibr'>Nagasawa et al. has</ns0:ref> shown that the SUA level is a predictor of IgAN in females but not in males <ns0:ref type='bibr' target='#b51'>(Nagasawa et al. 2016)</ns0:ref>. Similarly, the evaluation of renal pathological changes was not included. <ns0:ref type='bibr'>Moriyama et al.</ns0:ref> have reported that hyperuricemia is a risk factor for the progression of IgAN with CKD stage G3a but not for stage G1, G2, or G3b-4. This study also has shown that, except for glomerulosclerosis percentage, there is no significant difference in the 2009 Oxford classification, crescentic percentage, and focal segmental sclerosis between the hyperuricemia group and the normo-uricemia group <ns0:ref type='bibr' target='#b48'>(Moriyama et al. 2015)</ns0:ref>. Another study suggested that the relationship between hyperuricemia and IgAN progression was not very significant in patients with older age, lower eGFR, or interstitial lesion <ns0:ref type='bibr' target='#b76'>(Zhu et al. 2018)</ns0:ref>. Similarly, this study did not use the updated Oxford classification of IgAN. These studies are partially inconsistent even contradictory and not comprehensive, cannot clearfy the role of the SUA level in the progression and prognosis of IgAN. Therefore, it is imperative to conduct further research. </ns0:p></ns0:div>
<ns0:div><ns0:head>Methods</ns0:head></ns0:div>
<ns0:div><ns0:head>Patients screening and Clinical Data Collection</ns0:head><ns0:p>From January 2015 to May 2016, a total of 240 patients diagnosed with IgAN according to diagnostic criteria at Tongji Hospital were included in the study. The exclusion criteria of this study were listed as below: (1) the secondary IgAN caused by systemic diseases such as autoimmune disorders, chronic hepatitis, tumor, (2) incomplete clinical and pathologic data, or (3) the number of glomeruli in renal biopsy specimen less than eight. According to the above criteria, 32 patients were excluded and 208 patients were eventually involved in our research. The clinical data were collected at IgAN diagnosis. Blood samples were collected in the morning from fasting participants who had been directed to avoid eating any food that might affect the test result. The SUA levels > 420 µmol/L in men and > 360 µmol/L in women was defined as hyperuricemia. Based on the levels of SUA, there were 138 cases with hyperuricemia and 70 cases with normo-uricemia.</ns0:p><ns0:p>This study was approved by the Ethical Committee of Tongji Hospital (No.TJ-IRB20180608), and the Ethics Committee waived the need for informed consent from participants of this study.</ns0:p></ns0:div>
<ns0:div><ns0:head>Histologic Evaluation</ns0:head><ns0:p>Every renal biopsy specimen was scored by two pathologists who did not know the patient's clinical data. All of the renal specimens were classified and graded by five key pathological features: M, E, S, T, and C according to the MEST-C score of Oxford Classification <ns0:ref type='bibr'>(Trimarchi et</ns0:ref> PeerJ reviewing PDF | (2020:05:48827:1:1:NEW 27 Aug 2020)</ns0:p><ns0:p>Manuscript to be reviewed al., 2017). Other pathological features, including glomerular global/ ischemic sclerosis, arteriosclerosis, arteriolar hyalinosis, and immune complex deposition were also described in all specimens. The histopathological grading schema we used was presented in Table <ns0:ref type='table'>1</ns0:ref>. When the two pathologists differed in their assessments, the biopsy specimen must be reviewed again until an agreement was reached.</ns0:p></ns0:div>
<ns0:div><ns0:head>Statistical Analyses</ns0:head><ns0:p>GraphPad Prism 7.0 and SPSS 23.0 were used for statistical analyses. Continuous and categorical data were presented as mean ± SD and number (%), respectively. The comparison between the normo-uricemia group and the hyperuricemia group using the parametric t-test or Mann-Whitney. The overall renal survival rate of IgAN was presented using the Kaplan-Meier curve and the difference between the two groups was compared using the log-rank test. The p < 0.05 was regarded as statistical significance in all tests.</ns0:p></ns0:div>
<ns0:div><ns0:head>Results</ns0:head></ns0:div>
<ns0:div><ns0:head>Demographic and clinical data</ns0:head><ns0:p>The demographic and clinical data of 208 participants in this study were shown in Table <ns0:ref type='table'>2</ns0:ref>. The ages of all patients ranged from 15 to 63 years old, with an average of 33.8 ± 10.7 years old.</ns0:p><ns0:p>Among them, 82 (39.42%) were males, and 126 (68.68%) were females. The mean arterial pressure of patients was 94.0 ± 11.7 mm Hg, among which 23 patients had hypertension previously, 16 (7.69%) had poor blood pressure control, and 22 (10.58%) were taking antihypertensive medication. In this study, 14.42% of the patients had previously received ACEI and/or ARB treatment, and 4.81% underwent tonsillectomy previously. For all patients, the mean levels of serum albumin, 24-h urine protein, SUA, serum creatinine (Scr), and eGFR were 39.48 ± 5.77 g/l, 1.35 ± 1.58 g/d, 348.00 ± 109.10 µmol/l, 89.89 ± 41.49 µmol/l and 82.23 ± 30.22 ml/min/1.73 m 2 , respectively. <ns0:ref type='table'>PDF | (2020:05:48827:1:1:NEW 27 Aug 2020)</ns0:ref> Manuscript to be reviewed</ns0:p></ns0:div>
<ns0:div><ns0:head>PeerJ reviewing</ns0:head></ns0:div>
<ns0:div><ns0:head>The comparison of clinical characteristics between the normo-uricemia group and the hyperuricemia group</ns0:head><ns0:p>The comparison of baseline characteristics between the normo-uricemia group and the hyperuricemia group was presented in Table <ns0:ref type='table'>3</ns0:ref>. There were 70 patients (33.65%) with hyperuricemia, including 36 males (51.43%). These results were similar to the previous study conducted by <ns0:ref type='bibr' target='#b62'>Shi-Y (Shi et al., 2012)</ns0:ref>. Results showed that the mean arterial pressure of patients with hyperuricemia was significantly higher than those without (91.8 ± 10.7 mm Hg vs 98.2 ± 12.5 mm Hg, p = 0.0004). In terms of coagulation indicators, although the FDP level was higher in the hyperuricemia group (3.23 ± 0.74 mg/l vs 3.86 ± 1.38 mg/l, p = 0.0006), the APTT level was similar in both groups (p = 0.7833). As for blood lipids, hyperuricemia patients had an elevated level of total cholesterol, triglyceride, and decreased level of HDL-C (p = 0.026, p = 0.0005, p = 0.0056, respectively). However, the LDL-C levels of the two groups were similar (p > 0.05). This study also revealed the blood urea nitrogen (5.12 ± 1.52 mmol/l vs 6.49 ± 1.95 mmol/l), Scr (80.40 ± 40.58 µmol/l vs 109.00 ± 36.54 µmol/l) and 24-h urine protein (1.05 ± 1.27 g/d vs 1.81 ± 1.67 g/d) in hyperuricemia patients were higher, while the eGFR (98.98 ± 25.12ml/min/1.73m 2 vs 74.50 ± 29.52 ml/min/1.73m 2 ) was lower than those in normo-uricemia patients (all p < 0.05). When comparing immunological indexes of patients between the two groups, no significant difference was found in the levels of IgA, IgG, IgM, and C3 (all p > 0.05).</ns0:p><ns0:p>However, the level of C4 was significantly decreased in the normo-uricemia group (p = 0.003).</ns0:p></ns0:div>
<ns0:div><ns0:head>The comparisons of Oxford classification for IgAN between two groups</ns0:head><ns0:p>Histological features were evaluated using the MEST-C score of the 2016 updated Oxford Classification as Hernan et al. have reported <ns0:ref type='bibr' target='#b69'>(Trimarchi et al., 2017)</ns0:ref>. The comparisons in histological manifestations between the hyperuricemia group and normo-uricemia group were shown in Figure <ns0:ref type='figure'>1</ns0:ref>. The M, E, and S scores in the two groups were similar(p > 0.05). However, hyperuricemia patients had higher T scores than normo-uricemia patients (1.00 ± 0.87 vs 0.64 ± 0.76, p = 0.0023). In the previous Oxford study, crescent formation was not listed as an independent predictor of renal outcomes. However, Hernan Trimarchi and his working group Manuscript to be reviewed supported crescents as a predictor of renal outcome and suggested that crescent (C) should be added to the MEST score. Therefore, we compared the crescent scores in two groups; no significant differences were found (p = 0.4650). When comparing arterial lesions of patients in the two groups, hyperuricemia patients had higher scores of arteriosclerosis (0.99 ± 0.79 vs 0.48 ± 0.71, p < 0.0001) and hyalinosis (1.21 ± 0.87 vs 0.63 ± 0.76, p < 0.0001) than those without.</ns0:p><ns0:p>The glomerular sclerosis percentage in hyperuricemia patients was higher than that in normouricemia patients (0.22 ± 0.20 vs 0.12 ± 0.15, p = 0.0007).</ns0:p></ns0:div>
<ns0:div><ns0:head>The comparison of different types of crescent lesions between normo-uricemia group and hyperuricemia group</ns0:head><ns0:p>In the current investigation, results did not show any significant difference between the normouricemia group and hyperuricemia group in crescent (C) scores. Crescents could be divided into large and small crescents according to size and can be divided into cellular, fibrocellular and fibrous crescents according to crescent component <ns0:ref type='bibr' target='#b26'>(Jennette, 2003)</ns0:ref>. Then we tried to explore whether SUA affected the formation of different types of crescents. As shown in Figure <ns0:ref type='figure' target='#fig_8'>2</ns0:ref>, patients with hyperuricemia had slightly higher percentages of total crescents, large crescents, and small crescents than normo-uricemia, however, there was no statistical difference (0.10 ± 0.14 vs 0.08 ± 0.10, 0.06 ± 0.09 vs 0.05 ± 0.08, 0.04 ± 0.09 vs 0.03 ± 0.05, respectively; all p > 0.05). Our data also showed that participants with hyperuricemia had a slightly higher percentage of fibrocellular (0.05 ± 0.09 vs 0.04 ± 0.07, p = 0.1916) and fibrous crescent (0.04 ± 0.08 vs 0.02 ± 0.05, p = 0.0837), while a significantly lower percentage of cellular (0.005 ± 0.02 vs 0.020 ± 0.05, p = 0.0024) compared to participants without hyperuricemia.</ns0:p></ns0:div>
<ns0:div><ns0:head>The comparison of deposition of immune complexes and complements in renal biopsy between the two groups</ns0:head><ns0:p>The pathogenesis of IgAN was considered to be associated with immunological and or genetic mechanisms. Therefore, we evaluated the deposition of immune complexes and complements in Manuscript to be reviewed the two groups. We classified the immune complex deposition according to the fluorescence intensity (FI) in this study. As shown in Figure <ns0:ref type='figure' target='#fig_9'>3</ns0:ref>, the deposition of IgG and FRA in hyperuricemia group was less than those in normo-uricemia group (0.16 ± 0.44 vs 0.33 ± 0.63, p = 0.0239;0.27 ± 0.51 vs 0.45 ± 0.55, p = 0.0258, respectively). No significant difference was detected in IgA, IgM, and C3 deposition between the two groups (all p > 0.05).</ns0:p></ns0:div>
<ns0:div><ns0:head>The value of the SUA level in predicting renal prognosis</ns0:head><ns0:p>Finally, after an average of 25 months of follow-up, follow-up data were obtained from 193 patients. The endpoint of follow-up was defined as when patients needed dialysis treatment or when creatinine levels doubled. In this study, bout 70% of patients with HUA received uric acidlowering drugs during the follow-up. The renal survival curves according to the uric acid level suggested that patients with hyperuricemia have poorer renal outcome than patients without (p = 0.0147, Figure <ns0:ref type='figure' target='#fig_10'>4A</ns0:ref>). And then, we plotted a ROC curve to estimate the value of SUA in predicting the prognosis of IgAN. For females, the optimal cutoff value was 361.1 µmol/L, with sensitivity, specificity, and area under curve (AUC) of 0.7143, 0.7589, and 0.7679 ± 0.08167, respectively (Figure <ns0:ref type='figure' target='#fig_10'>4B</ns0:ref>). However, for males, the optimal cutoff value was 614 µmol/L, with the sensitivity and specificity of 0.3333 and 0.9718. The AUC for male patients was 0.5728 ± 0.2029 (Figure <ns0:ref type='figure' target='#fig_10'>4C</ns0:ref>). Our results revealed that the SUA level was a valuable indicator in predicting the prognosis of IgAN patients, especially in female patients, which is consistent with the research of Oh et al. <ns0:ref type='bibr' target='#b55'>(Oh et al. 2020)</ns0:ref>.</ns0:p></ns0:div>
<ns0:div><ns0:head>Discussion</ns0:head><ns0:p>IgAN has a high incidence and a variety of clinical manifestations, from asymptomatic hematuria to progress to renal failure rapidly with a few months. Consequently, it is imperative to ascertain its influencing indicator, delay its progression, and prevent ESRD. A few studies have demonstrated that the SUA level is closely related to the progression of IgAN <ns0:ref type='bibr' target='#b62'>(Shi et al., 2012;</ns0:ref><ns0:ref type='bibr' target='#b0'>Bakan et al., 2015;</ns0:ref><ns0:ref type='bibr' target='#b10'>Cheng et al., 2013)</ns0:ref>. However, there are still inconsistencies and even PeerJ reviewing PDF | (2020:05:48827:1:1:NEW 27 Aug 2020)</ns0:p><ns0:p>Manuscript to be reviewed contradictions in some studies. Therefore, predicting the prognosis of IgAN remains difficult due to its diverse clinical features. In this retrospective study of 208 IgAN patients, we comprehensively evaluated the correlation between SUA levels and clinical and histopathological features. Our results revealed that the SUA level can be considered as a predictor for the prognosis of IgAN.</ns0:p><ns0:p>Results showed that patients with hyperuricemia have a poorer renal function and higher blood pressure, which is consistent with Cui et al. researched on 148 patients with IgAN <ns0:ref type='bibr' target='#b18'>(Cui et al., 2011)</ns0:ref>. The results also revealed significant increases in blood pressure, lipid, C3, and C4 levels in patients with hyperuricemia. Previous studies have found that hyperuricemia is closely related to metabolic syndrome. Elevated SUA level has been confirmed to be an independent predictor of the development of diabetes, hypertension, hypertriglyceridemia, and nonalcoholic fatty liver disease <ns0:ref type='bibr' target='#b39'>(Lv et al., 2013;</ns0:ref><ns0:ref type='bibr' target='#b23'>Grayson et al., 2011;</ns0:ref><ns0:ref type='bibr' target='#b32'>Kuwabara et al., 2018;</ns0:ref><ns0:ref type='bibr' target='#b74'>Yamada et al., 2010)</ns0:ref>. Some studies also showed a significant improvement in homeostasis assessment for the IR index after lowering the level of SUA <ns0:ref type='bibr' target='#b38'>(Liu et al. 2015;</ns0:ref><ns0:ref type='bibr' target='#b67'>Takir et al. 2015)</ns0:ref>. Previous researches revealed that the levels of serum C3 and C4 positively correlated with the development of the metabolic syndrome and had been identified as important markers relevant to this disease <ns0:ref type='bibr' target='#b47'>(Meng et al., 2017;</ns0:ref><ns0:ref type='bibr'>Meng et al., 2018;</ns0:ref><ns0:ref type='bibr'>Xin et al., 2018)</ns0:ref>. The increased activation products of C3 can accelerate the uptake of free fatty acids, the synthesis of triacylglycerol, and inhibit hormone-sensitive lipase in several adipocytes, thus contribute to the development of the metabolic syndrome <ns0:ref type='bibr' target='#b57'>(Phieler et al., 2013;</ns0:ref><ns0:ref type='bibr' target='#b6'>Barbu et al., 2015)</ns0:ref>. However, the mechanism of C4 in metabolic syndrome remains unclear.</ns0:p><ns0:p>To evaluate pathological changes, we classified patients using the 2016 update Oxford classification of IgAN as described above. We found that among patients with hyperuricemia, 32.86% had M, 28.57% had S, and 35.71% had E, which was higher than those in normouricemia patients, data were 29.71%, 27.54%, 30.43% respectively. However, no significant difference was discovered in M, E, and S scores between the hyperuricemia group and normouricemia group. As for the tubular atrophy/interstitial fibrosis, this study revealed that the T score PeerJ reviewing PDF | (2020:05:48827:1:1:NEW 27 Aug 2020)</ns0:p><ns0:p>Manuscript to be reviewed of hyperuricemia patients was statistically higher than that of normo-uricemia patients. This result demonstrated that the level of SUA was related to tubular atrophy/interstitial fibrosis, and hyperuricemia can be used as a clinical marker for T, which was consistent with the founding of <ns0:ref type='bibr'>Myllymaki et al. (Myllymaki et al., 2005)</ns0:ref> and Fan et al. <ns0:ref type='bibr' target='#b20'>(Fan et al. 2019)</ns0:ref>. Recently, a study conducted by <ns0:ref type='bibr'>Nigro et al.</ns0:ref> found that the fractal dimension of tubules and the density of tubules negatively correlated the level of SUA and urea. They also suggested that SUA might be a better predictor to identify nephron integrity <ns0:ref type='bibr' target='#b53'>(Nigro et al. 2018)</ns0:ref>. There were different views on the prognostic value of M, E, and S in different studies <ns0:ref type='bibr'>(Coppo et al., 2014b;</ns0:ref><ns0:ref type='bibr'>Herzenberg et al., 2011b;</ns0:ref><ns0:ref type='bibr' target='#b61'>Shi et al., 2011;</ns0:ref><ns0:ref type='bibr' target='#b30'>Kang et al., 2012)</ns0:ref>. Nevertheless, T was recognized as a prognostic indicator of IgAN in these studies. Thus, we regard the level of SUA as a valuable factor in predicting IgAN prognosis. The tubular atrophy/interstitial fibrosis caused by SUA might be explained by the following mechanisms. Urate crystals deposit in the renal tubules can directly damage or block the renal tubules, and also form uric acid renal stones to damage the kidney, which results in renal tubular atrophy and interstitial fibrosis, even renal failure <ns0:ref type='bibr' target='#b70'>(Viggiano et al. 2018)</ns0:ref>. HUA could also promote the production of inflammatory factors such as MCP-1 and TNF-β1, which stimulates the inflammatory response and induce renal tubular injury and renal interstitial fibrosis <ns0:ref type='bibr' target='#b59'>(Romi et al. 2017)</ns0:ref>. Moreover, HUA could decrease E-cadherin expression, increase α-SMA expression, and induced tubular cells epithelial-mesenchymal transition, which results in renal tubulointerstitial injury <ns0:ref type='bibr'>(Liu et al. 2017)</ns0:ref>.</ns0:p><ns0:p>Recently, the updated Oxford classification of IgAN recommends crescents (C) as a pathological predictor of renal outcomes in IgA nephropathy. In this current investigation, no correlation between the SUA level and C score was detected. Then, we compared the percentages of different types of C lesions in patients with hyperuricemia and those with normouricemia and found that the percentages of total crescents, large crescents, small crescents, fibrocellular crescents, and fibrous crescents were slightly higher in hyperuricemia patients, while the percentages of cell crescents were significantly lower. It is well known that the formation of the crescent is caused by the deposition of immune complexes, the activation of Manuscript to be reviewed monocytes, the release of inflammatory cytokines, and the aggregation of fibrocytes. The previous report discovered that uric acid might promote the release of inflammatory cytokines such as TNF-α, IL-1β, IL-6 in IgA patients <ns0:ref type='bibr' target='#b52'>(Nakagawa et al., 2006)</ns0:ref>. Therefore, we speculate that hyperuricemia may be involved in crescent formation by promoting the release of inflammatory cytokines.</ns0:p><ns0:p>As shown in Figure <ns0:ref type='figure' target='#fig_8'>2</ns0:ref>, hyperuricemia patients had significantly high percentages of glomerular sclerosis, including global glomerular sclerosis and ischemic glomerular sclerosis.</ns0:p><ns0:p>Consequently, the level of SUA might indicate the severity of glomerular sclerosis. As reported in previous studies, many factors were related to glomerular sclerosis, including abnormal cytokine expression, podocyte injury or loss, and the activation of the RAS <ns0:ref type='bibr' target='#b58'>(Riser et al., 2000;</ns0:ref><ns0:ref type='bibr' target='#b9'>Brown et al., 2000a;</ns0:ref><ns0:ref type='bibr'>Brown et al., 2000b)</ns0:ref>. As mentioned before, SUA could promote the release of inflammatory cytokines, and persistent inflammation may result in glomerular sclerosis.</ns0:p><ns0:p>Furthermore, hyperuricemia may activate the renin-angiotensin system, cause glomerular hypertension, and reduce perfusion, which ultimately led to the formation of glomerular sclerosis <ns0:ref type='bibr' target='#b60'>(Sanchez-Lozada et al., 2005)</ns0:ref>. Besides, the uric acid can induce cellular oxidation through the xanthine oxidase pathway and then result in podocyte damage, which might be another cause of glomerular sclerosis.</ns0:p><ns0:p>We also evaluated the effect of elevated uric acid on vascular pathological changes, including arteriosclerosis and arteriolar hyalinosis. We found that the vascular lesions were more severe in hyperuricemia patients, and several mechanisms could explain this difference. Firstly, the elevated uric acid levels could inhibit nitric oxide synthase, induce inflammatory reactions, activate the renin-angiotensin system, cause endothelial cells dysfunction, and eventually lead to the lesions of vascular <ns0:ref type='bibr' target='#b7'>(Behradmanesh et al., 2013;</ns0:ref><ns0:ref type='bibr' target='#b17'>Corry et al., 2008)</ns0:ref> Manuscript to be reviewed progression of IgAN, indicating that the activation of complements was closely related to the IgAN pathogenesis <ns0:ref type='bibr' target='#b30'>(Kim et al., 2012)</ns0:ref>. However, our result showed there was no significant difference in the deposition of IgA, IgM, and C3 between patients with hyperuricemia and those without. This could be due to the small sample size or different races of the study.</ns0:p><ns0:p>There are several limitations in this study. On the one hand, the sample size was relatively small, a multicenter, large sample size study should be conducted for further analysis. On the other hand, further studies are required to explore whether lowering uric acid levels could improve the outcome of IgAN.</ns0:p><ns0:p>In conclusion, the SUA level affects renal function and pathophysiology of IgAN. The levels of SUA can be considered as a prognostic indicator for IgAN, which is of great significance in guiding treatment decisions and assessing prognosis. Manuscript to be reviewed Manuscript to be reviewed Manuscript to be reviewed </ns0:p></ns0:div>
<ns0:div><ns0:head>Histopathological features</ns0:head><ns0:note type='other'>Figure 2</ns0:note><ns0:note type='other'>Figure 3</ns0:note><ns0:note type='other'>Figure 4</ns0:note></ns0:div><ns0:figure xml:id='fig_0'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:05:48827:1:1:NEW 27 Aug 2020)Manuscript to be reviewedIn this retrospective study, we first used the updated Oxford classification criteria (MEST-C), vascular injury, glomerulosclerosis, immunofluorescence score, clinical indicators, and prognosis analysis to comprehensively evaluate the relationship between the SUA level and the clinical, pathological characteristics of IgAN, to determine the impact of hyperuricemia on the development and prognosis of IgAN.</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_1'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:05:48827:1:1:NEW 27 Aug 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_2'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:05:48827:1:1:NEW 27 Aug 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_3'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:05:48827:1:1:NEW 27 Aug 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_4'><ns0:head /><ns0:label /><ns0:figDesc>. Secondly, increased serum lipid levels and oxidation of LDL can promote the progression of atherosclerosis and arteriolar hyalinosis. An early study by Kim SJ et al, which included 343 patients with IgAN found that the decreased serum C3 level and the deposition of C3 in mesangial could independently predict the PeerJ reviewing PDF | (2020:05:48827:1:1:NEW 27 Aug 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_6'><ns0:head>Figure 1 Figure 1 .</ns0:head><ns0:label>11</ns0:label><ns0:figDesc>Figure 1</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_7'><ns0:head>(</ns0:head><ns0:label /><ns0:figDesc>A) mesangial hypercellularity scores; (B) endocapillary hypercellularity scores; (C) segmental sclerosis scores; (D) tubular atrophy / interstitial fibrosis scores; (E) crescent scores; (F) MEST-C total scores; (G) arteriosclerosis score; (H) hyalinosis score; (I) the percent of glomerular sclerosis. PeerJ reviewing PDF | (2020:05:48827:1:1:NEW 27 Aug 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_8'><ns0:head>Figure 2 .</ns0:head><ns0:label>2</ns0:label><ns0:figDesc>Figure 2. Percentages of different types of crescents among the IgAN patients with normo-uricemia or hyperuricemia.</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_9'><ns0:head>Figure 3 .</ns0:head><ns0:label>3</ns0:label><ns0:figDesc>Figure 3. The fluorescence intensity scores of immune complex deposits in renal biopsy.</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_10'><ns0:head>Figure 4 .</ns0:head><ns0:label>4</ns0:label><ns0:figDesc>Figure 4. The value of the serum uric acid level in predicting renal prognosis.</ns0:figDesc></ns0:figure>
<ns0:note place='foot'>PeerJ reviewing PDF | (2020:05:48827:1:1:NEW 27 Aug 2020)Manuscript to be reviewed</ns0:note>
</ns0:body>
" | "Dear Dr. Menini,
Thank you very much for your careful review and constructive suggestions regarding our manuscript entitled “Serum uric acid level is correlated with the clinical, pathological progression and prognosis of IgA nephropathy: a retrospective study”. The reviewer’s comments were highly insightful and enabled us to greatly improve the quality of our manuscript. We have tried our best to revise and improve the manuscript, and have provided explanations of each issue in a point by point manner. Revisions are shown in the tracked changes manuscript. And the tracked changes versions of Tables are embedded at the end of the tracked changes manuscript. Please open the tracked changes manuscript at mode of 'Show all revisions as embedded' to display the line number correctly. We hope that the revisions in the manuscript and our accompanying responses will be sufficient to make our manuscript suitable for publication in the PeerJ. Thank you again for your consideration.
Yours sincerely,
Min Han
Reviewer 1:
Basic reporting
Dr. Lu et al. studied the relationship between SUA and the clinical and pathological phenotype of IgA nephropathy. They evaluated 208 patients with IgAN divided in 2 groups, with normo-uricemia and hyper-uricemia, and followed up them for an average of 25 months. Patients with hyperuricemia had higher mean arterial pressure, increased levels of total cholesterol and triglycerides and lower eGFR. When IgAN biopsies were evaluated using MEST-C score, the authors found out that patients with hyperuricemia had a higher scores of tubulointerstitial changes, arteriosclerosis, glomerular sclerosis and a significantly lower percentage of fibrocellular crescents. They concluded that the levels of SUA were associated with renal function level and pathological severity of IgAN and could have a prognostic relevance.
The language is clear and the literature well referenced.
Experimental design
The figures and the tables are well labeled and described and the results are clearly explained. Methods are sufficiently described.
Validity of the findings
Several investigations have demonstrated the detrimental impact of SUA levels on the progression of IgAN, could the authors make clear the novelty of their findings?
Response: Thanks for your crucial reminder. The novelties and significances of this study are as follows.
1) In our study, as for the pathological evaluation of renal biopsy, we used the updated Oxford classification (MEST-C score system) for the first time, combined with the percentage of glomerulosclerosis, vascular injury score, and immunofluorescence score to comprehensively analyze the effect of serum uric acid (SUA) on the pathological characteristics of IgAN. As we all know, since the Oxford Classification of IgAN was published in 2009, MEST scores have been frequently used in clinical practice. However, crescent formation, which was considered as an important pathological predictor for poor outcome, was not included in this classification of IgAN. In 2017, the Oxford Classification of IgAN was updated and the crescent (C) score was recommended to add to the new score system. Cellular/fibrocellular crescents absent, present in at least 1 glomerulus, and present in >25% of glomeruli were scored as C0, C1, and C2, respectively. Our results showed that there was no significant difference in the C score between the normo-uricemia group and the hyperuricemia group. However, we found that the percentages of total crescents, large crescents, small crescents, fibrocellular crescents, and fibrous crescents were slightly higher in hyperuricemia patients, while the percentages of cell crescents were significantly lower. These suggested that hyperuricemia may be involved in the formation of the crescent. Of course, further experiments and clinical studies are needed to confirm.
2) We plotted a ROC curve and calculated the sensitivity, specificity, and area under the curve to estimate the value of the SUA level in predicting the prognosis of IgAN, which was involved in few previous studies. Our results revealed that the SUA was a valuable indicator in predicting the prognosis of IgAN patients, especially in female patients, which is consistent with the research of Oh et al. [1].
3) As we all know, some studies on the role of uric acid in the pathology and prognosis of IgAN are inconsistent even controversial. A cohort study found that the SUA affected the pathophysiology (including global glomerulosclerosis, tubulointerstitial nephritis, and vascular lesions) and prognosis of IgAN positively [2]. However, this study does not accord with some other studies. For example, Moriyama et al. reported that high uric acid level was only a risk factor for the progression of IgAN with CKD stage G3a but not for stage G1, G2, or G3b-4. The study also showed that, except for glomerulosclerosis percentage, there was no significant difference in the 2009 Oxford classification, crescentic percentage, and focal segmental sclerosis between the hyperuricemia group and the normo-uricemia group [3]. Besides, another study conducted by Nagasawa et al. showed that the SUA only predicted the progression of IgAN in females but not in males [4]. Therefore, it is necessary to conduct further research to explore the role of the uric acid level in the pathology and prognosis of IgAN. We have added these points to the second and third paragraphs of the Introduction section in the tracked changes manuscript (see lines 69-93).
4) Moreover, in our study, we found that the SUA was associated with tubular atrophy/interstitial fibrosis, glomerulosclerosis vascular injury, and the prognosis of IgAN. This finding can help predict histopathologic changes and outcomes of IgAN in clinical practice especially for patients not going to or not willing to have a renal biopsy. This result might also raise the concern of doctors, and suggests that hyperuricemia as a potential therapeutic target may be helpful to slow down the IgAN progression.
Comments for the author
1. High blood pressure, increased levels of lipids, and uric acid are some factors associated with metabolic syndrome. What about hyperglycemia and insulin resistance?
Response: Thanks a lot for your attention to this critical issue. Insulin resistance (IR) and hyperglycemia are also a manifestation of metabolic syndrome. Many studies showed that SUA was closely related to insulin resistance (IR) and hyperglycemia. In a systematic review of eight prospective cohort studies, involving 32016 subjects sand 2930 type 2 diabetes mellitus (T2DM) patients, the risk of T2DM was increased by 6% for every 1 mg/dL increase in SUA [5]. A retrospective cohort study involved 1923 male patients found that the risk of new-onset diabetes in hyperuricemia patients was positively correlated with the SUA level. The risk of new-onset diabetes reached 27% when SUA >9mg/dl after adjusting for confounding factors [6]. Some studies showed that the pharmacologic lowering of SUA can improve the IR of patients. In a randomized, parallel controlled study, 176 T2DM patients with asymptomatic hyperuricemia were randomly assigned to either allopurinol treatment group or control group. After 3 years of follow-up, it was found that the homeostasis assessment for insulin resistance index mean value in the allopurinol group was significantly lower than that in the control group [7]. Another prospective cohort study also showed a significant improvement in homeostasis assessment for the IR index, fasting blood glucose, and fasting insulin levels in the allopurinol group [8]. Therefore, SUA is considered to be associated with hyperglycemia and IR. We have added a summary of these observations to the second paragraph of the Discussion section (see lines 224-228 of the tracked changes manuscript).
2. Have the patients received UA lowering agents, during the follow-up?
Response: Thanks for your concern. In our study, about 70% of patients with hyperuricemia received uric acid-lowering drugs during the follow-up (see lines 198-199 in the tracked changes manuscript). Patients with hyperuricemia were first suggested to reduce the intake of high-purine foods as much as possible. If patients still had hyperuricemia after diet control, they were given uric acid-lowering drugs such as febuxostat and allopurinol. Besides, we suggested patients monitor the SUA level closely so that doctors can adjust the dosage of drugs in time. Our further research is to explore whether lowering uric acid levels could improve the outcome of IgAN.
3. T score of patients with hyperuricemia was higher. This finding could be much more discussed.
Response: Thanks for your valuable recommendation. We have added conclusions from some studies to support our findings in the discussion of the effects of hyperuricemia on renal tubular atrophy/interstitial fibrosis (see lines 245-248 in the tracked changes manuscript). Moreover, we also have explained the possible mechanisms of tubular atrophy/interstitial fibrosis induced by SUA in the lines 252-260 of the tracked changes manuscript.
Reviewer 2:
Basic reporting
The employment of the English language is good in general. although it could be improved.
References are adequately addressed, although the format is uneven.
The structure of the paper is adequate.
The results are tuned to the hypothesis.
Experimental design
The design is appropriate.
However, the power of the sample has not been calculated for this sort of hypothesis to be proven. A 200-patient cohort is low.
Response: Thanks a lot for your attention to this critical issue. Your suggestion is very valuable, and we have addressed this deficiency of small sample size in the Discussion section (see lines 299-300 in the tracked changes manuscript). However, due to the strict exclusion criteria and time-consuming pathological evaluation of renal biopsies, we have not yet collected enough cases. Our study showed that there were very significant differences between the normo-uricemia group and hyperuricemia group in pathological indicators such as tubular atrophy/interstitial fibrosis, arterial lesions score, glomerular sclerosis percentage, and outcomes. Besides, there are also some studies whose sample size is not large enough. For example, only 93 patients were screened in a study on the association between serum uric acid (SUA) and IgAN progression [9]. Similarly, Caliskan et al. 's study of the clinical significance of uric acid in the progression of IgAN included only 111 cases [10]. Moreover, this is a pilot study with small sample size. Its purpose is to lay a foundation for the study with a large sample size and to avoid the waste of manpower, material resources, and financial resources caused by the blind study with a large sample size. Therefore, we think that the result of this sample size may be sufficient to prove our hypothesis in the pilot study. However, the larger the sample size, the better. Next, we will conduct a multicenter, large sample size study for further analysis.
Validity of the findings
The impact and novelty are low actually. I explain this topic to the authors.
Conclusions are well stated, but as the impact is low and lacks of novelty, in my opinion, the paper offers nothing interesting.
Response: Thanks for your question. The novelties of this study are as follows.
1) In our study, as for the pathological evaluation of renal biopsy, we used the updated Oxford classification (MEST-C score system) for the first time, combined with the percentage of glomerulosclerosis, vascular injury score, and immunofluorescence score to comprehensively analyze the effect of SUA on the pathological characteristics of IgAN. As we all know, since the Oxford Classification of IgAN was published in 2009, MEST scores have been frequently used in clinical practice. However, crescent formation，which was considered as an important pathological predictor for poor outcome, was not included in this classification of IgAN. In 2017, the Oxford Classification of IgAN was updated and the crescent (C) score was recommended to add to the new score system. Cellular/fibrocellular crescents absent, present in at least 1 glomerulus, and present in >25% of glomeruli were scored as C0, C1, and C2, respectively. Our results showed that there was no significant difference in the C score between the normo-uricemia group and the hyperuricemia group. However, we found that the percentages of total crescents, large crescents, small crescents, fibrocellular crescents, and fibrous crescents were slightly higher in hyperuricemia patients, while the percentages of cell crescents were significantly lower. These suggested that hyperuricemia may be involved in the formation of the crescent. Of course, further experiments and clinical studies are needed to confirm.
2) We plotted a ROC curve and calculated the sensitivity, specificity, and area under the curve to estimate the value of the SUA level in predicting the prognosis of IgAN, which was involved in few previous studies. Our results revealed that the SUA level was a valuable indicator in predicting the prognosis of IgAN patients, especially in female patients, which is consistent with the research of Oh et al.[1]. Therefore, we think the study is novel.
Comments for the author
This paper lacks novelty. Thence, no high impact is expected.
As the authors have demonstrated in a small cohort of 200 patients with biopsy-proven IgAN, high uric acid blood levels correlate well with the chronic variables assessed. Hyperuricemia is a well-known cardiovascular risk factor for chronic situations are hypertension, left ventricular hypertrophy, hypercholesterolemia, CKD. We already know these findings.
With respect to IgAN, it is not unexpected to find that hyperuricemia is, again, correlated with chronic variables, as tubular atrophy and interstitial fibrosis, vascular damage, and glomerulosclerosis.
Hyperuricemia is associated with CKD and is a well-known predictor of renal outcome in CKD, glomerulopathies, and IgAN.
Response: Thank you for your sincere comments. As mentioned above, the novelties of this study have been discussed.
Although hyperuricemia is a risk factor for CKD, its role in pathology and prognosis of IgAN is not very clear. There are still inconsistencies and even contradictions in some studies, so more studies are needed to determine the effect of uric acid levels on the pathology and prognosis of IgAN. For example, a cohort study found that the SUA level affected the pathophysiology (including global glomerulosclerosis, tubulointerstitial nephritis, and vascular lesions) and prognosis of IgAN positively [2]. However, this study does not accord with some other studies. Moriyama et al. reported that high uric acid level was only a risk factor for the progression of IgAN with CKD stage G3a but not for stage G1, G2, or G3b-4. The study also showed that, except for glomerulosclerosis percentage, there was no significant difference in the 2009 Oxford classification, crescentic percentage, and focal segmental sclerosis between the hyperuricemia group and the normo-uricemia group [3]. Besides, another study conducted by Nagasawa et al. showed that the SUA level only predicted the progression of IgAN in females but not in males [4]. Therefore, it is necessary to conduct further research to explore the role of the uric acid level in the pathology and prognosis of IgAN. Moreover, in our study, we found that SUA was associated with tubular atrophy/interstitial fibrosis, glomerulosclerosis vascular injury, and prognosis of IgAN. This finding can help predict histopathologic changes and outcomes of IgAN in clinical practice especially for patients not going to or not willing to have a renal biopsy. This result might also raise the concern of doctors, and suggests that hyperuricemia as a potential therapeutic target may be helpful to slow down the IgAN progression. Based on the above, we think our study is novel and meaningful.
Reviewer 3:
Basic reporting
1. The manuscript is well written. Please check for mistyping errors (seed e.g. table 1 legend 'cells count in per').
Response: Thanks a lot for pointing out the errors in our manuscript, and we are very sorry for that. We have carefully reviewed the manuscript and corrected all the spelling, grammatical errors in the revised manuscript. For example, we have revised ' the mesangial cells count in per mesangial area ' to ' the mesangial cell count in per mesangial area ' in the Notes of Table 1. In the second paragraph of the Results part, “16.49 ± 1.95” and “274.5 ± 29.52” have been revised to “6.49 ± 1.95” and “74.5 ± 29.52” respectively according to the Table 3 (see lines 150,152 in the tracked changes manuscript).
2. References are in part outdated. Morphological measures in IgA nephropathy should be analyzed in terms of modern nephron number techniques (see e.g. PMID: 31534861). The tubular changes also could be interpreted in view of techniques based on fractal analysis (PMID: 29540159). The role of urates could also be discussed taking care of more recent reviews (see e.g. PMID: 29689561).
Response: We thank you a lot for sincere suggestions and we have updated some references and added the above studies to the revised manuscript. For example, we have added the view from PMID: 29689561 to the lines 69-72 and 253-255 of the tracked changes manuscript. Moreover, we also add “Recently, a study conducted by Nigro et al. found that the fractal dimension of tubules and the density of tubules negatively correlated the level of serum uric acid (SUA) and urea. They also suggested that SUA might be a better predictor to identify nephron integrity” based on PMID: 29540159 to the lines 245-248 in the tracked changes manuscript. We were sorry for did not evaluate the effect of hyperuricemia on total nephron count in this study because we haven’t found studies on the relationship between the total number of nephrons and SUA levels. This valuable indicator will be considered for inclusion in our further study.
Experimental design
Aims are clear. Statistics are not adequate: please correct for multiple testing.
Response: Thanks a lot for your attention to this critical issue. The clinical and laboratory data (including the levels of SUA) were collected at the time of IgAN diagnosis in this study. Because some patients with hyperuricemia were given uric acid-lowering drugs and other treatments after diagnosis, which had an impact on SUA levels. So, we chose SUA levels at the time of diagnosis as baseline data for further analysis. In addition, we found that SUA levels at diagnosis or renal biopsy were used as baseline data in many studies, and multiple tests were not involved [4, 11-13].
Validity of the findings
Findings are interesting.
Comments for the author
1. Please clarify in Table 1 notes, what is a 'mesangial area'.
Response: Thanks for your crucial reminder. we are very sorry for not explaining the definition of the mesangial area. The glomerular mesangial area is the area between the glomerular capillary loops and is composed of mesangial cells and mesangial matrix. Each glomerulus has a mesangial area. We have added this explanation to the Notes in the revised Table 1.
2. Table 2: uric acid - please check in parenthesis the units (μmol/l or g/l?).
eGFR: please report in the table the equation used (CKD-Epi or MDRD etc).
Response: Thank you for your careful review and point out the error and defect in Table 2. We were very sorry about this. We have clarified that the unit of uric acid is μmmol/L in the revised Table 2. The eGFR was calculated using the CKD-EPI equation and we have added this to the Notes in revised Table 2 and Table 3.
3. Table 3: please verify the number of decimals reported in the table. For example, it is possibly not useful to report two decimals after the age (years) or to describe the weight (kg) or the MAP (mmHg) when the measure itself has a much lower precision.
Response: Thank you a lot for your sincere suggestions. We previously referred to the article published in PeerJ, in which the values of age and blood pressure were accurate to two decimal places [14]. However, we think your suggestion is very reasonable, because the measures of age (years), weight (kg), and MAP (mmHg) are not accurate enough. Then, we consulted other studies and found that the values of age, weight, and MAP are generally kept to one decimal place. Therefore, we reduced these values to one decimal place in the modified Table 2, Table 3, and the Results part of the revised manuscript (see lines 128,130,143,144 in the tracked changes manuscript).
4. Figure 1 and 3: please use the same layout of Figure 2, reporting the values of each subject rather only the mean.
Response: We thank you a lot and have taken your valuable recommendation to show Figures 1I using values for each subject instead of only the mean in the revised manuscript. However, such as mesangial hypercellularity, endocapillary hypercellularity, and IgA deposition scores are categorical data, which may not suitable for this type of graph. The scatter diagram of the mesangial hypercellularity score similar to Figure 2 is shown below. Therefore, we think it would be better to use bar charts for these kinds of data. The revised layout of Figure 1 is shown below.
Figure 1. The comparisons of Oxford classification and arterial lesions between normo-uricemia and hyperuricemia patients in IgA nephropathy.
(A) mesangial hypercellularity scores; (B) endocapillary hypercellularity scores; (C) segmental sclerosis scores; (D) tubular atrophy/interstitial fibrosis scores; (E) crescent scores; (F) MEST-C total scores; (G) arteriosclerosis score; (H) hyalinosis score; (I) the percent of glomerular sclerosis.
References
1. Oh TR, Choi HS, Kim CS et al. The Effects of Hyperuricemia on the Prognosis of IgA Nephropathy are More Potent in Females. Journal of clinical medicine 2020; 9.
2. Cheng G-y, Liu D-w, Zhang N et al. Clinical and prognostic implications of serum uric acid levels on IgA nephropathy: a cohort study of 348 cases with a mean 5-year follow-up. Clinical nephrology 2013; 80: 40-46.
3. Moriyama T, Itabashi M, Takei T et al. High uric acid level is a risk factor for progression of IgA nephropathy with chronic kidney disease stage G3a. Journal of nephrology 2015; 28: 451-456.
4. Nagasawa Y, Yamamoto R, Shoji T et al. Serum Uric Acid Level Predicts Progression of IgA Nephropathy in Females but Not in Males. PloS one 2016; 11: e0160828.
5. Lv Q, Meng X-F, He F-F et al. High serum uric acid and increased risk of type 2 diabetes: a systemic review and meta-analysis of prospective cohort studies. PloS one 2013; 8: e56864.
6. Krishnan E, Akhras KS, Sharma H et al. Relative and attributable diabetes risk associated with hyperuricemia in US veterans with gout. QJM 2013; 106: 721-729.
7. Liu P, Wang H, Zhang F et al. The Effects of Allopurinol on the Carotid Intima-media Thickness in Patients with Type 2 Diabetes and Asymptomatic Hyperuricemia: A Three-year Randomized Parallel-controlled Study. Intern Med 2015; 54: 2129-2137.
8. Takir M, Kostek O, Ozkok A et al. Lowering Uric Acid With Allopurinol Improves Insulin Resistance and Systemic Inflammation in Asymptomatic Hyperuricemia. J Investig Med 2015; 63: 924-929.
9. Bakan A, Oral A, Elcioglu OC et al. Hyperuricemia is associated with progression of IgA nephropathy. Int Urol Nephrol 2015; 47: 673-678.
10. Caliskan Y, Ozluk Y, Celik D et al. The Clinical Significance of Uric Acid and Complement Activation in the Progression of IgA Nephropathy. Kidney Blood Press Res 2016; 41: 148-157.
11. Fan S, Zhang P, Wang AY et al. Hyperuricemia and its related histopathological features on renal biopsy. BMC Nephrol 2019; 20: 95.
12. Myllymaki J, Honkanen T, Syrjanen J, et al. Uric acid correlates with the severity of histopathological parameters in IgA nephropathy. Nephrol Dial Transplant 2005; 20: 89-95.
13. Syrjanen J, Mustonen J, Pasternack A. Hypertriglyceridaemia and hyperuricaemia are risk factors for progression of IgA nephropathy. Nephrol Dial Transplant 2000; 15: 34-42.
14. Xu Y, Liu X, Sun X, Wang Y. The impact of serum uric acid on the natural history of glomerular filtration rate: a retrospective study in the general population. PeerJ 2016; 4: e1859.
" | Here is a paper. Please give your review comments after reading it. |
9,770 | "<ns0:abstract xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:p>This study was aimed to assess the relationship between serum uric acid (SUA) level and the clinical, pathological phenotype of IgA nephropathy (IgAN), and to determine the role of SUA level in the progression and prognosis of IgAN.</ns0:p></ns0:div>
<ns0:div><ns0:head>Methods.</ns0:head><ns0:p>A total of 208 patients with IgAN were included in this study, which was classified into the normo-uricemia group and hyperuricemia group according to the SUA level. The clinical data at baseline, IgA Oxford classification scores (MEST-C scoring system), and other pathological features were collected and further analyzed. All patients were followed up and the prognosis was assessed using Kaplan-Meier survival curves. GraphPad Prism 7.0 and SPSS 23.0 were used for statistical analyses.</ns0:p></ns0:div>
<ns0:div><ns0:head>Results.</ns0:head><ns0:p>In clinical indicators, patients with hyperuricemia had the significantly higher proportion of males to females, mean arterial pressure, the levels of total cholesterol, triglyceride, Scr, BUN, 24 hoururine protein, C3, and C4, the lower levels of high-density lipoprotein cholesterol and eGFR than those without(p<0.05). In terms of pathological characteristics, the tubular atrophy/interstitial fibrosis scores, vascular injury scores, and glomerular sclerosis percentage were significantly higher in patients with hyperuricemia compared with those without(p<0.01). There was no significant difference in the scores of mesangial hypercellularity, endocapillary hypercellularity, focal segmental glomerulosclerosis, as well as crescents between the two groups(p>0.05). As for the depositions of immune complexes deposition in IgAN, the hyperuricemia group had less deposition of immunoglobulin G and FRA than the normouricemia group (p<0.05), while the deposition of immunoglobulin A, immunoglobulin M, and complement C3 in the two groups showed no statistical difference. The survival curve suggested that patients in the hyperuricemia group have significantly poorer renal outcome than those in the normo-uricemia group (p= 0.0147). Results also revealed that the SUA level is a valuable predictor of renal outcome in patients with IgAN. The optimal cutoff value was 361.1μmol/L (AUC=0.76±0.08167) and 614 μmol/L (AUC=0.5728±0.2029) for female and male, respectively.</ns0:p></ns0:div>
<ns0:div><ns0:head>Conclusions.</ns0:head><ns0:p>The level of SUA is associated with renal function level and pathological severity of IgAN, and maybe a prognostic indicator of IgAN.</ns0:p></ns0:div>
</ns0:abstract>
<ns0:body xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:head>Introduction</ns0:head><ns0:p>IgA Nephropathy (IgAN), whose feature is the deposition of IgA-dominant in the mesangial area, is the most prevalent primary glomerular disease in the world. Approximately 40% of patients almost completely lose their renal function within 30 years after diagnosis <ns0:ref type='bibr' target='#b16'>(Coppo & D'Amico, 2005;</ns0:ref><ns0:ref type='bibr' target='#b19'>D'Amico, 2004)</ns0:ref>. Previous studies have demonstrated that clinical features, including severe proteinuria, poor renal function, and hypertension at the initial diagnosis are the predictors of poor outcomes in IgAN <ns0:ref type='bibr' target='#b33'>(Le et al., 2012;</ns0:ref><ns0:ref type='bibr' target='#b1'>Barbour & Reich, 2018;</ns0:ref><ns0:ref type='bibr' target='#b43'>Mariani et al., 2018;</ns0:ref><ns0:ref type='bibr' target='#b5'>Barbour & Reich, 2012)</ns0:ref>. Histological features were also identified as crucial predictors of IgAN patients.</ns0:p><ns0:p>The Oxford classification of IgAN, also known as MEST score, including mesangial hypercellularity (M), endocapillary hypercellularity (E), segmental glomerulosclerosis (S), tubular atrophy/interstitial fibrosis (T) has been shown to be of significant value in predicting the Manuscript to be reviewed prognosis IgAN <ns0:ref type='bibr'>(Barbour et al., 2016;</ns0:ref><ns0:ref type='bibr' target='#b42'>Maixnerova et al., 2016;</ns0:ref><ns0:ref type='bibr' target='#b25'>Herzenberg et al., 2011a;</ns0:ref><ns0:ref type='bibr' target='#b29'>Katafuchi et al., 2011;</ns0:ref><ns0:ref type='bibr'>Coppo et al., 2014a;</ns0:ref><ns0:ref type='bibr' target='#b39'>Lv et al., 2013)</ns0:ref>. Recently, crescents (C) have been proposed to add to the Oxford classification of IgAN to form an updated MEST-C score system, which can provide a more comprehensive pathological prediction for the prognosis of IgAN <ns0:ref type='bibr' target='#b69'>(Trimarchi et al., 2017)</ns0:ref>.</ns0:p><ns0:p>Numerous studies have shown that the level of uric acid can predict the incidence of atherosclerosis <ns0:ref type='bibr' target='#b21'>(Feig 2014;</ns0:ref><ns0:ref type='bibr' target='#b24'>Gustafsson & Unwin 2013;</ns0:ref><ns0:ref type='bibr' target='#b35'>Li et al. 2014)</ns0:ref>, hypertension <ns0:ref type='bibr' target='#b71'>(Wang et al. 2014)</ns0:ref>, and coronary heart disease <ns0:ref type='bibr' target='#b31'>(Kim et al. 2010)</ns0:ref>. Moreover, some studies have emphasized that high levels of serum uric acid (SUA) would form urates crystals that deposit in renal tubules and interstitial, leading to kidney fibrosis and failure <ns0:ref type='bibr' target='#b64'>(Su et al. 2020;</ns0:ref><ns0:ref type='bibr' target='#b70'>Viggiano et al. 2018)</ns0:ref>.</ns0:p><ns0:p>Recently, some studies have found that there is a correlation between the SUA level and the progression of IgAN. A cohort study has shown that hyperuricemia is associated with the progression of IgAN, however, the effect of hyperuricemia on renal pathology was not evaluated <ns0:ref type='bibr' target='#b0'>(Bakan et al. 2015)</ns0:ref> in this study. The study conducted by <ns0:ref type='bibr'>Nagasawa et al. has</ns0:ref> shown that the SUA level is a predictor of IgAN in females but not in males <ns0:ref type='bibr' target='#b51'>(Nagasawa et al. 2016)</ns0:ref>. Similarly, the evaluation of renal pathological changes was not included. <ns0:ref type='bibr'>Moriyama et al.</ns0:ref> have reported that hyperuricemia is a risk factor for the progression of IgAN with CKD stage G3a but not for stage G1, G2, or G3b-4. This study also has shown that, except for glomerulosclerosis percentage, there is no significant difference in the 2009 Oxford classification, crescentic percentage, and focal segmental sclerosis between the hyperuricemia group and the normo-uricemia group <ns0:ref type='bibr' target='#b48'>(Moriyama et al. 2015)</ns0:ref>. Another study suggested that the relationship between hyperuricemia and IgAN progression was not very significant in patients with older age, lower eGFR, or interstitial lesion <ns0:ref type='bibr' target='#b76'>(Zhu et al. 2018)</ns0:ref>. Similarly, this study did not use the updated Oxford classification of IgAN. These studies are partially inconsistent even contradictory and not comprehensive, cannot clarify the role of the SUA level in the progression and prognosis of IgAN. Therefore, it is imperative to conduct further research. </ns0:p></ns0:div>
<ns0:div><ns0:head>Methods</ns0:head></ns0:div>
<ns0:div><ns0:head>Patients screening and Clinical Data Collection</ns0:head><ns0:p>From January 2015 to May 2016, a total of 240 patients diagnosed with IgAN according to diagnostic criteria at Tongji Hospital were included in the study. The exclusion criteria of this study were listed as below: (1) the secondary IgAN caused by systemic diseases such as autoimmune disorders, chronic hepatitis, tumor, (2) incomplete clinical and pathologic data, or (3) the number of glomeruli in renal biopsy specimen less than eight. According to the above criteria, 32 patients were excluded and 208 patients were eventually involved in our research. The clinical data were collected at IgAN diagnosis. Blood samples were collected in the morning from fasting participants who had been directed to avoid eating any food that might affect the test result. The SUA levels > 420 µmol/L in men and > 360 µmol/L in women was defined as hyperuricemia. Based on the levels of SUA, there were 138 cases with hyperuricemia and 70 cases with normo-uricemia.</ns0:p><ns0:p>This study was approved by the Ethical Committee of Tongji Hospital (No.TJ-IRB20180608), and the Ethics Committee waived the need for informed consent from participants of this study.</ns0:p></ns0:div>
<ns0:div><ns0:head>Histologic Evaluation</ns0:head><ns0:p>Every renal biopsy specimen was scored by two pathologists who did not know the patient's clinical data. All of the renal specimens were classified and graded by five key pathological features: M, E, S, T, and C according to the MEST-C score of Oxford Classification <ns0:ref type='bibr'>(Trimarchi et</ns0:ref> PeerJ reviewing PDF | (2020:05:48827:2:0:NEW 12 Sep 2020)</ns0:p><ns0:p>Manuscript to be reviewed al., 2017). Other pathological features, including glomerular global/ ischemic sclerosis, arteriosclerosis, arteriolar hyalinosis, and immune complex deposition were also described in all specimens. The histopathological grading schema we used was presented in Table <ns0:ref type='table'>1</ns0:ref>. When the two pathologists differed in their assessments, the biopsy specimen must be reviewed again until an agreement was reached.</ns0:p></ns0:div>
<ns0:div><ns0:head>Statistical Analyses</ns0:head><ns0:p>GraphPad Prism 7.0 and SPSS 23.0 were used for statistical analyses. Continuous and categorical data were presented as mean ± SD and number (%), respectively. The comparison between the normo-uricemia group and the hyperuricemia group using the parametric t-test or Mann-Whitney. The overall renal survival rate of IgAN was presented using the Kaplan-Meier curve and the difference between the two groups was compared using the log-rank test. The p < 0.05 was regarded as statistical significance in all tests.</ns0:p></ns0:div>
<ns0:div><ns0:head>Results</ns0:head></ns0:div>
<ns0:div><ns0:head>Demographic and clinical data</ns0:head><ns0:p>The demographic and clinical data of 208 participants in this study were shown in Table <ns0:ref type='table'>2</ns0:ref>. The ages of all patients ranged from 15 to 63 years old, with an average of 33.8 ± 10.7 years old.</ns0:p><ns0:p>Among them, 82 (39.42%) were males, and 126 (68.68%) were females. The mean arterial pressure of patients was 94.0 ± 11.7 mm Hg, among which 23 patients had hypertension previously, 16 (7.69%) had poor blood pressure control, and 22 (10.58%) were taking antihypertensive medication. In this study, 14.42% of the patients had previously received ACEI and/or ARB treatment, and 4.81% underwent tonsillectomy previously. For all patients, the mean levels of serum albumin, 24-h urine protein, SUA, serum creatinine (Scr), and eGFR were 39.48 ± 5.77 g/l, 1.35 ± 1.58 g/d, 348.00 ± 109.10 µmol/l, 89.89 ± 41.49 µmol/l and 82.23 ± 30.22 ml/min/1.73 m 2 , respectively. <ns0:ref type='table'>PDF | (2020:05:48827:2:0:NEW 12 Sep 2020)</ns0:ref> Manuscript to be reviewed</ns0:p></ns0:div>
<ns0:div><ns0:head>PeerJ reviewing</ns0:head></ns0:div>
<ns0:div><ns0:head>The comparison of clinical characteristics between the normo-uricemia group and the hyperuricemia group</ns0:head><ns0:p>The comparison of baseline characteristics between the normo-uricemia group and the hyperuricemia group was presented in Table <ns0:ref type='table'>3</ns0:ref>. There were 70 patients (33.65%) with hyperuricemia, including 36 males (51.43%). These results were similar to the previous study conducted by <ns0:ref type='bibr' target='#b62'>Shi-Y (Shi et al., 2012)</ns0:ref>. Results showed that the mean arterial pressure of patients with hyperuricemia was significantly higher than those without (91.8 ± 10.7 mm Hg vs 98.2 ± 12.5 mm Hg, p = 0.0004). In terms of coagulation indicators, although the FDP level was higher in the hyperuricemia group (3.23 ± 0.74 mg/l vs 3.86 ± 1.38 mg/l, p = 0.0006), the APTT level was similar in both groups (p = 0.7833). As for blood lipids, hyperuricemia patients had an elevated level of total cholesterol, triglyceride, and decreased level of HDL-C (p = 0.026, p = 0.0005, p = 0.0056, respectively). However, the LDL-C levels of the two groups were similar (p > 0.05). This study also revealed the blood urea nitrogen (5.12 ± 1.52 mmol/l vs 6.49 ± 1.95 mmol/l), Scr (80.40 ± 40.58 µmol/l vs 109.00 ± 36.54 µmol/l) and 24-h urine protein (1.05 ± 1.27 g/d vs 1.81 ± 1.67 g/d) in hyperuricemia patients were higher, while the eGFR (98.98 ± 25.12ml/min/1.73m 2 vs 74.50 ± 29.52 ml/min/1.73m 2 ) was lower than those in normo-uricemia patients (all p < 0.05). When comparing immunological indexes of patients between the two groups, no significant difference was found in the levels of IgA, IgG, IgM, and C3 (all p > 0.05).</ns0:p><ns0:p>However, the level of C4 was significantly decreased in the normo-uricemia group (p = 0.003).</ns0:p></ns0:div>
<ns0:div><ns0:head>The comparisons of Oxford classification for IgAN between two groups</ns0:head><ns0:p>Histological features were evaluated using the MEST-C score of the 2016 updated Oxford Classification as Hernan et al. have reported <ns0:ref type='bibr' target='#b69'>(Trimarchi et al., 2017)</ns0:ref>. The comparisons in histological manifestations between the hyperuricemia group and normo-uricemia group were shown in Figure <ns0:ref type='figure'>1</ns0:ref>. The M, E, and S scores in the two groups were similar(p > 0.05). However, hyperuricemia patients had higher T scores than normo-uricemia patients (1.00 ± 0.87 vs 0.64 ± 0.76, p = 0.0023). In the previous Oxford study, crescent formation was not listed as an independent predictor of renal outcomes. However, Hernan Trimarchi and his working group Manuscript to be reviewed supported crescents as a predictor of renal outcome and suggested that crescent (C) should be added to the MEST score. Therefore, we compared the crescent scores in two groups; no significant differences were found (p = 0.4650). When comparing arterial lesions of patients in the two groups, hyperuricemia patients had higher scores of arteriosclerosis (0.99 ± 0.79 vs 0.48 ± 0.71, p < 0.0001) and hyalinosis (1.21 ± 0.87 vs 0.63 ± 0.76, p < 0.0001) than those without.</ns0:p><ns0:p>The glomerular sclerosis percentage in hyperuricemia patients was higher than that in normouricemia patients (0.22 ± 0.20 vs 0.12 ± 0.15, p = 0.0007).</ns0:p></ns0:div>
<ns0:div><ns0:head>The comparison of different types of crescent lesions between the normo-uricemia group and the hyperuricemia group</ns0:head><ns0:p>In the current investigation, results did not show any significant difference between the normouricemia group and hyperuricemia group in crescent (C) scores. Crescents could be divided into large and small crescents according to size and can be divided into cellular, fibrocellular and fibrous crescents according to crescent component <ns0:ref type='bibr' target='#b26'>(Jennette, 2003)</ns0:ref>. Then we tried to explore whether SUA affected the formation of different types of crescents. As shown in Figure <ns0:ref type='figure' target='#fig_10'>2</ns0:ref>, patients with hyperuricemia had slightly higher percentages of total crescents, large crescents, and small crescents than normo-uricemia, however, there was no statistical difference (0.10 ± 0.14 vs 0.08 ± 0.10, 0.06 ± 0.09 vs 0.05 ± 0.08, 0.04 ± 0.09 vs 0.03 ± 0.05, respectively; all p > 0.05). Our data also showed that participants with hyperuricemia had a slightly higher percentage of fibrocellular (0.05 ± 0.09 vs 0.04 ± 0.07, p = 0.1916) and fibrous crescent (0.04 ± 0.08 vs 0.02 ± 0.05, p = 0.0837), while a significantly lower percentage of cellular (0.005 ± 0.02 vs 0.020 ± 0.05, p = 0.0024) compared to participants without hyperuricemia.</ns0:p></ns0:div>
<ns0:div><ns0:head>The comparison of deposition of immune complexes and complements in renal biopsy between the two groups</ns0:head><ns0:p>The pathogenesis of IgAN was considered to be associated with immunological and or genetic mechanisms. Therefore, we evaluated the deposition of immune complexes and complements in Manuscript to be reviewed the two groups. We classified the immune complex deposition according to the fluorescence intensity (FI) in this study. As shown in Figure <ns0:ref type='figure' target='#fig_11'>3</ns0:ref>, the deposition of IgG and FRA in hyperuricemia group was less than those in normo-uricemia group (0.16 ± 0.44 vs 0.33 ± 0.63, p = 0.0239;0.27 ± 0.51 vs 0.45 ± 0.55, p = 0.0258, respectively). No significant difference was detected in IgA, IgM, and C3 deposition between the two groups (all p > 0.05).</ns0:p></ns0:div>
<ns0:div><ns0:head>The value of the SUA level in predicting renal prognosis</ns0:head><ns0:p>Finally, after an average of 25 months of follow-up, follow-up data were obtained from 193 patients. The endpoint of follow-up was defined as when patients needed dialysis treatment or when creatinine levels doubled. In this study, bout 70% of patients with HUA received uric acidlowering drugs during the follow-up. The renal survival curves according to the uric acid level suggested that patients with hyperuricemia have poorer renal outcome than patients without (p = 0.0147, Figure <ns0:ref type='figure' target='#fig_12'>4A</ns0:ref>). And then, we plotted a ROC curve to estimate the value of SUA in predicting the prognosis of IgAN. For females, the optimal cutoff value was 361.1 µmol/L, with sensitivity, specificity, and area under curve (AUC) of 0.7143, 0.7589, and 0.7679 ± 0.08167, respectively (Figure <ns0:ref type='figure' target='#fig_12'>4B</ns0:ref>). However, for males, the optimal cutoff value was 614 µmol/L, with the sensitivity and specificity of 0.3333 and 0.9718. The AUC for male patients was 0.5728 ± 0.2029 (Figure <ns0:ref type='figure' target='#fig_12'>4C</ns0:ref>). Our results revealed that the SUA level was a valuable indicator in predicting the prognosis of IgAN patients, especially in female patients, which is consistent with the research of Oh et al. <ns0:ref type='bibr' target='#b55'>(Oh et al. 2020)</ns0:ref>.</ns0:p></ns0:div>
<ns0:div><ns0:head>Discussion</ns0:head><ns0:p>IgAN has a high incidence and a variety of clinical manifestations, from asymptomatic hematuria to progress to renal failure rapidly with a few months. Consequently, it is imperative to ascertain its influencing indicator, delay its progression, and prevent ESRD. A few studies have demonstrated that the SUA level is closely related to the progression of IgAN <ns0:ref type='bibr' target='#b62'>(Shi et al., 2012;</ns0:ref><ns0:ref type='bibr' target='#b0'>Bakan et al., 2015;</ns0:ref><ns0:ref type='bibr' target='#b10'>Cheng et al., 2013)</ns0:ref>. However, there are still inconsistencies and even Manuscript to be reviewed contradictions in some studies. Therefore, predicting the prognosis of IgAN remains difficult due to its diverse clinical features. In this retrospective study of 208 IgAN patients, we comprehensively evaluated the correlation between SUA levels and clinical and histopathological features. Our results revealed that the SUA level can be considered as a predictor for the prognosis of IgAN.</ns0:p><ns0:p>Results showed that patients with hyperuricemia have a poorer renal function and higher blood pressure, which is consistent with Cui et al. researched on 148 patients with IgAN <ns0:ref type='bibr' target='#b18'>(Cui et al., 2011)</ns0:ref>. The results also revealed significant increases in blood pressure, lipid, C3, and C4 levels in patients with hyperuricemia. Previous studies have found that hyperuricemia is closely related to metabolic syndrome. Elevated SUA level has been confirmed to be an independent predictor of the development of diabetes, hypertension, hypertriglyceridemia, and nonalcoholic fatty liver disease <ns0:ref type='bibr' target='#b39'>(Lv et al., 2013;</ns0:ref><ns0:ref type='bibr' target='#b23'>Grayson et al., 2011;</ns0:ref><ns0:ref type='bibr' target='#b32'>Kuwabara et al., 2018;</ns0:ref><ns0:ref type='bibr' target='#b74'>Yamada et al., 2010)</ns0:ref>. Some studies also showed a significant improvement in homeostasis assessment for the IR index after lowering the level of SUA <ns0:ref type='bibr' target='#b38'>(Liu et al. 2015;</ns0:ref><ns0:ref type='bibr' target='#b67'>Takir et al. 2015)</ns0:ref>. Previous researches revealed that the levels of serum C3 and C4 positively correlated with the development of the metabolic syndrome and had been identified as important markers relevant to this disease <ns0:ref type='bibr' target='#b47'>(Meng et al., 2017;</ns0:ref><ns0:ref type='bibr'>Meng et al., 2018;</ns0:ref><ns0:ref type='bibr'>Xin et al., 2018)</ns0:ref>. The increased activation products of C3 can accelerate the uptake of free fatty acids, the synthesis of triacylglycerol, and inhibit hormone-sensitive lipase in several adipocytes, thus contribute to the development of the metabolic syndrome <ns0:ref type='bibr' target='#b57'>(Phieler et al., 2013;</ns0:ref><ns0:ref type='bibr' target='#b6'>Barbu et al., 2015)</ns0:ref>. However, the mechanism of C4 in metabolic syndrome remains unclear.</ns0:p><ns0:p>To evaluate pathological changes, we classified patients using the 2016 update Oxford classification of IgAN as described above. We found that among patients with hyperuricemia, 32.86% had M, 28.57% had S, and 35.71% had E, which was higher than those in normouricemia patients, data were 29.71%, 27.54%, 30.43% respectively. However, no significant difference was discovered in M, E, and S scores between the hyperuricemia group and normouricemia group. As for the tubular atrophy/interstitial fibrosis, this study revealed that the T score PeerJ reviewing PDF | (2020:05:48827:2:0:NEW 12 Sep 2020)</ns0:p><ns0:p>Manuscript to be reviewed of hyperuricemia patients was statistically higher than that of normo-uricemia patients. This result demonstrated that the level of SUA was related to tubular atrophy/interstitial fibrosis, and hyperuricemia can be used as a clinical marker for T, which was consistent with the founding of <ns0:ref type='bibr'>Myllymaki et al. (Myllymaki et al., 2005)</ns0:ref> and Fan et al. <ns0:ref type='bibr' target='#b20'>(Fan et al. 2019)</ns0:ref>. Recently, a study conducted by <ns0:ref type='bibr'>Nigro et al.</ns0:ref> found that the fractal dimension of tubules and the density of tubules negatively correlated the level of SUA and urea. They also suggested that SUA might be a better predictor to identify nephron integrity <ns0:ref type='bibr' target='#b53'>(Nigro et al. 2018)</ns0:ref>. There were different views on the prognostic value of M, E, and S in different studies <ns0:ref type='bibr'>(Coppo et al., 2014b;</ns0:ref><ns0:ref type='bibr'>Herzenberg et al., 2011b;</ns0:ref><ns0:ref type='bibr' target='#b61'>Shi et al., 2011;</ns0:ref><ns0:ref type='bibr' target='#b30'>Kang et al., 2012)</ns0:ref>. Nevertheless, T was recognized as a prognostic indicator of IgAN in these studies. Thus, we regard the level of SUA as a valuable factor in predicting IgAN prognosis. The tubular atrophy/interstitial fibrosis caused by SUA might be explained by the following mechanisms. Urate crystals deposit in the renal tubules can directly damage or block the renal tubules, and also form uric acid renal stones to damage the kidney, which results in renal tubular atrophy and interstitial fibrosis, even renal failure <ns0:ref type='bibr' target='#b70'>(Viggiano et al. 2018)</ns0:ref>. HUA could also promote the production of inflammatory factors such as MCP-1 and TNF-β1, which stimulates the inflammatory response and induce renal tubular injury and renal interstitial fibrosis <ns0:ref type='bibr' target='#b59'>(Romi et al. 2017)</ns0:ref>. Moreover, HUA could decrease E-cadherin expression, increase α-SMA expression, and induced tubular cells epithelial-mesenchymal transition, which results in renal tubulointerstitial injury <ns0:ref type='bibr'>(Liu et al. 2017)</ns0:ref>.</ns0:p><ns0:p>Recently, the updated Oxford classification of IgAN recommends crescents (C) as a pathological predictor of renal outcomes in IgA nephropathy. In this current investigation, no correlation between the SUA level and C score was detected. Then, we compared the percentages of different types of C lesions in patients with hyperuricemia and those with normouricemia and found that the percentages of total crescents, large crescents, small crescents, fibrocellular crescents, and fibrous crescents were slightly higher in hyperuricemia patients, while the percentages of cell crescents were significantly lower. It is well known that the formation of the crescent is caused by the deposition of immune complexes, the activation of Manuscript to be reviewed monocytes, the release of inflammatory cytokines, and the aggregation of fibrocytes. The previous report discovered that uric acid might promote the release of inflammatory cytokines such as TNF-α, IL-1β, IL-6 in IgA patients <ns0:ref type='bibr' target='#b52'>(Nakagawa et al., 2006)</ns0:ref>. Therefore, we speculate that hyperuricemia may be involved in crescent formation by promoting the release of inflammatory cytokines.</ns0:p><ns0:p>As shown in Figure <ns0:ref type='figure' target='#fig_10'>2</ns0:ref>, hyperuricemia patients had significantly high percentages of glomerular sclerosis, including global glomerular sclerosis and ischemic glomerular sclerosis.</ns0:p><ns0:p>Consequently, the level of SUA might indicate the severity of glomerular sclerosis. As reported in previous studies, many factors were related to glomerular sclerosis, including abnormal cytokine expression, podocyte injury or loss, and the activation of the RAS <ns0:ref type='bibr' target='#b58'>(Riser et al., 2000;</ns0:ref><ns0:ref type='bibr' target='#b9'>Brown et al., 2000a;</ns0:ref><ns0:ref type='bibr'>Brown et al., 2000b)</ns0:ref>. As mentioned before, SUA could promote the release of inflammatory cytokines, and persistent inflammation may result in glomerular sclerosis.</ns0:p><ns0:p>Furthermore, hyperuricemia may activate the renin-angiotensin system, cause glomerular hypertension, and reduce perfusion, which ultimately led to the formation of glomerular sclerosis <ns0:ref type='bibr' target='#b60'>(Sanchez-Lozada et al., 2005)</ns0:ref>. Besides, the uric acid can induce cellular oxidation through the xanthine oxidase pathway and then result in podocyte damage, which might be another cause of glomerular sclerosis.</ns0:p><ns0:p>We also evaluated the effect of elevated uric acid on vascular pathological changes, including arteriosclerosis and arteriolar hyalinosis. We found that the vascular lesions were more severe in hyperuricemia patients, and several mechanisms could explain this difference. Firstly, the elevated uric acid levels could inhibit nitric oxide synthase, induce inflammatory reactions, activate the renin-angiotensin system, cause endothelial cells dysfunction, and eventually lead to the lesions of vascular <ns0:ref type='bibr' target='#b7'>(Behradmanesh et al., 2013;</ns0:ref><ns0:ref type='bibr' target='#b17'>Corry et al., 2008)</ns0:ref> Manuscript to be reviewed progression of IgAN, indicating that the activation of complements was closely related to the IgAN pathogenesis <ns0:ref type='bibr' target='#b30'>(Kim et al., 2012)</ns0:ref>. However, our result showed there was no significant difference in the deposition of IgA, IgM, and C3 between patients with hyperuricemia and those without. This could be due to the small sample size or different races of the study.</ns0:p><ns0:p>There are several limitations in this study. On the one hand, the sample size of this pilot study was relatively small. Next, we will conduct a multicenter, large sample size prospective study for further analysis. On the other hand, further studies are required to explore whether lowering uric acid levels could improve the outcome of IgAN.</ns0:p><ns0:p>In conclusion, the SUA level affects renal function and pathophysiology of IgAN. The levels of SUA can be considered as a prognostic indicator for IgAN, which is of great significance in guiding treatment decisions and assessing prognosis. Manuscript to be reviewed Manuscript to be reviewed Manuscript to be reviewed </ns0:p></ns0:div>
<ns0:div><ns0:head>Histopathological features</ns0:head><ns0:note type='other'>Figure 2</ns0:note><ns0:note type='other'>Figure 3</ns0:note><ns0:note type='other'>Figure 4</ns0:note></ns0:div><ns0:figure xml:id='fig_0'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:05:48827:2:0:NEW 12 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_1'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:05:48827:2:0:NEW 12 Sep 2020) Manuscript to be reviewed In this retrospective study, we first used the updated Oxford classification criteria (MEST-C), vascular injury, glomerulosclerosis, immunofluorescence score, clinical indicators, and prognosis analysis to comprehensively evaluate the relationship between the SUA level and the clinical, pathological characteristics of IgAN, to determine the impact of hyperuricemia on the development and prognosis of IgAN.</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_2'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:05:48827:2:0:NEW 12 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_3'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:05:48827:2:0:NEW 12 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_4'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:05:48827:2:0:NEW 12 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_5'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:05:48827:2:0:NEW 12 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_6'><ns0:head /><ns0:label /><ns0:figDesc>. Secondly, increased serum lipid levels and oxidation of LDL can promote the progression of atherosclerosis and arteriolar hyalinosis. An early study by Kim SJ et al, which included 343 patients with IgAN found that the decreased serum C3 level and the deposition of C3 in mesangial could independently predict the PeerJ reviewing PDF | (2020:05:48827:2:0:NEW 12 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_8'><ns0:head>Figure 1 Figure 1 .</ns0:head><ns0:label>11</ns0:label><ns0:figDesc>Figure 1</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_9'><ns0:head>(</ns0:head><ns0:label /><ns0:figDesc>A) mesangial hypercellularity scores; (B) endocapillary hypercellularity scores; (C) segmental sclerosis scores; (D) tubular atrophy / interstitial fibrosis scores; (E) crescent scores; (F) MEST-C total scores; (G) arteriosclerosis score; (H) hyalinosis score; (I) the percent of glomerular sclerosis. PeerJ reviewing PDF | (2020:05:48827:2:0:NEW 12 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_10'><ns0:head>Figure 2 .</ns0:head><ns0:label>2</ns0:label><ns0:figDesc>Figure 2. Percentages of different types of crescents among the IgAN patients with normo-uricemia or hyperuricemia.</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_11'><ns0:head>Figure 3 .</ns0:head><ns0:label>3</ns0:label><ns0:figDesc>Figure 3. The fluorescence intensity scores of immune complex deposits in renal biopsy.</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_12'><ns0:head>Figure 4 .</ns0:head><ns0:label>4</ns0:label><ns0:figDesc>Figure 4. The value of the serum uric acid level in predicting renal prognosis.</ns0:figDesc></ns0:figure>
<ns0:note place='foot'>PeerJ reviewing PDF | (2020:05:48827:2:0:NEW 12 Sep 2020)Manuscript to be reviewed</ns0:note>
</ns0:body>
" | "Dear Dr. Menini,
Thank you very much for your careful review and constructive suggestions regarding our manuscript entitled “Serum uric acid level is correlated with the clinical, pathological progression and prognosis of IgA nephropathy: a retrospective study”. The reviewer’s comments were highly insightful and enabled us to improve the quality of our manuscript. We have revised the title of this manuscript to “Serum uric acid level is correlated with the clinical, pathological progression and prognosis of IgA nephropathy: an observational retrospective pilot-study” based on the comments of the reviewer 2. Moreover, we have addressed a statement that “On the one hand, the sample size of this pilot study was relatively small. Next, we will conduct a multicenter, large sample size prospective study for further analysis” in the revised manuscript. Revisions are shown in the tracked changes manuscript. Please open the tracked changes manuscript at the mode of 'Show all revisions as embedded' to display the line number correctly. We hope that the revisions in the manuscript will be sufficient to make our manuscript suitable for publication in the PeerJ. Thank you again for your consideration.
Yours sincerely,
Min Han
Reviewer 2:
Comments for the author
In general, the authors have considerabily improved the quality of the paper after the reviewers inputs.
In their letter to the editor and reviewers, after my comment that the number of patents in low, they comment that this study is 'a pliot study', and that they plan to do a prospective trial. Thus, this statement must be addressed in the manuscript. Moreover, the title should be changed to: 'Serum uric acid.....: an observational retrospective pilot-study'.
Response: Thank you for your sincere suggestions. We have revised the title of this manuscript to “Serum uric acid level is correlated with the clinical, pathological progression and prognosis of IgA nephropathy: an observational retrospective pilot-study” based on your comment (see lines 1-4 of the tracked changes manuscript). Moreover, we have addressed a statement that “On the one hand, the sample size of this pilot study was relatively small. Next, we will conduct a multicenter, large sample size prospective study for further analysis” in the eighth paragraphs of the Discussion section in the tracked changes manuscript (see lines 289-291).
" | Here is a paper. Please give your review comments after reading it. |
9,771 | "<ns0:abstract xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:p>Replication studies are essential for evaluating the validity of previous research findings.</ns0:p><ns0:p>However, it has proven challenging to reproduce the results of ecological and evolutionary studies, partly because of the complexity and lability of many of the phenomena being investigated, but also due to small sample sizes, low statistical power and publication bias.</ns0:p><ns0:p>Additionally, replication is often considered too difficult in field settings where many factors are beyond the investigator's control and where spatial and temporal dependencies may be strong. We investigated the feasibility of reproducing original research findings in the field of chemical ecology by performing an exact replication of a previous study of Antarctic fur seals (Arctocephalus gazella). In the original study, skin swabs from 41 mother-offspring pairs from two adjacent breeding colonies on Bird Island, South Georgia, were analysed using gas chromatography-mass spectrometry. Seals from the two colonies differed significantly in their chemical fingerprints, suggesting that colony membership may be chemically encoded, and mothers were also chemically similar to their pups, hinting at the possible involvement of phenotype matching in mother-offspring recognition. In the current study, we generated and analysed chemical data from a nonoverlapping sample of 50 mother-offspring pairs from the same two colonies five years later. The original results were corroborated in both hypothesis testing and estimation contexts, with p-values remaining highly significant and effect sizes, standardized between studies by bootstrapping the chemical data over individuals, being of comparable magnitude. However, exact replication studies are only capable of showing whether a given effect can be replicated in a specific setting. We therefore investigated whether chemical signatures are colony-specific in general by expanding the geographic coverage</ns0:p></ns0:div>
</ns0:abstract>
<ns0:body xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:head>Introduction</ns0:head><ns0:p>Replication studies are fundamental to the scientific process as they are essential for evaluating the correctness of scientific claims and the conclusions of other scientists <ns0:ref type='bibr' target='#b52'>(Schmidt, 2009)</ns0:ref>. Indeed, <ns0:ref type='bibr' target='#b14'>Fisher (1974)</ns0:ref> recommended that a null hypothesis should always be rejected more than once because 'no isolated experiment, however significant in itself, can suffice for the experimental demonstration of any natural phenomenon'. Nevertheless, replication studies are still 'troublingly rare', particularly in fields such as ecology and evolutionary biology <ns0:ref type='bibr' target='#b39'>(Nakagawa & Parker, 2015)</ns0:ref>. <ns0:ref type='bibr' target='#b44'>Palmer (2000)</ns0:ref> argued that we ignore reproducibility at our peril because this perpetuates a 'contract of error' that undermines our understanding of important ecological and evolutionary phenomena.</ns0:p><ns0:p>There has also been debate and confusion over exactly what constitutes reproducible research <ns0:ref type='bibr' target='#b37'>(Mendoza & Garcia, 2017)</ns0:ref>. <ns0:ref type='bibr'>Goodman and colleagues (2016)</ns0:ref> recognized three basic concepts, (i) 'methods reproducibility', which requires that the methodology of a given study be provided in sufficient detail to allow it to be repeated; (ii) 'results reproducibility', often known as 'replication', which is the ability to corroborate previous results using the same experimental methods in a new study; and (iii) 'inferential reproducibility', which relates to whether or not qualitatively similar conclusions are reached on the basis of either an independent replication of a study or a re-analysis of the original data. Furthermore, replication studies can be 'exact', meaning that they show a high degree of fidelity to the original experiment, 'partial', which involves procedural or methodological changes, or 'conceptual', where the same questions are investigated but using different approaches <ns0:ref type='bibr' target='#b28'>(Kelly, 2006)</ns0:ref>. The latter two categories include 'quasi-replication' studies, which extend the scope of the original study beyond the specific system or species in question <ns0:ref type='bibr' target='#b44'>(Palmer, 2000)</ns0:ref>. In general, the closer the replication attempt is to the original study, the more valuable are the results for assessing the validity of the original claims <ns0:ref type='bibr' target='#b39'>(Nakagawa & Parker, 2015)</ns0:ref>. However, quasi and conceptual replications are also important because they can shed light on the generality (also known as 'transportability') of the effects under investigation <ns0:ref type='bibr' target='#b17'>(Goodman, Fanelli & Ioannidis, 2016;</ns0:ref><ns0:ref type='bibr' target='#b10'>Dirnagl, 2019;</ns0:ref><ns0:ref type='bibr' target='#b48'>Piper et al., 2019</ns0:ref>; although see <ns0:ref type='bibr' target='#b28'>Kelly, 2006)</ns0:ref>. Put another way, it is only possible to learn something about the broader significance of a certain effect by probing to what extent it persists in settings that are different from, or which lie outside of the experimental framework of the original study. Quasi and conceptual replications therefore play an important role in increasing the 'external validity' of results <ns0:ref type='bibr' target='#b52'>(Schmidt, 2009)</ns0:ref>.</ns0:p><ns0:p>Another conceptual difficulty relates to the basis on which replication success is judged. Although there is no single standard for evaluating replication outcomes, most replication attempts are deemed successful if a null hypothesis that was rejected in the original study is again rejected <ns0:ref type='bibr' target='#b51'>(Rosenthal, 1991;</ns0:ref><ns0:ref type='bibr' target='#b28'>Kelly, 2006)</ns0:ref>. However, due to the dependence of p-values on sample sizes, success or failure in attaining significance may not always provide a good measure of replication success <ns0:ref type='bibr' target='#b28'>(Kelly, 2006)</ns0:ref>. Consequently, several authors have advocated reporting effect sizes and associated measures of precision, as these allow replication outcomes to be gauged in a continuous manner rather than on the basis of binary significance outcomes <ns0:ref type='bibr' target='#b28'>(Kelly, 2006;</ns0:ref><ns0:ref type='bibr' target='#b17'>Goodman, Fanelli & Ioannidis, 2016;</ns0:ref><ns0:ref type='bibr' target='#b48'>Piper et al., 2019)</ns0:ref>.</ns0:p><ns0:p>In recent years, high-profile failures to reproduce a significant proportion of studies in the medical and social sciences (e.g. <ns0:ref type='bibr' target='#b2'>Begley & Ellis, 2012</ns0:ref>; Open science collaboration, 2015, reviewed by <ns0:ref type='bibr' target='#b29'>Kelly, 2019)</ns0:ref> have led to a crisis of confidence <ns0:ref type='bibr' target='#b1'>(Baker, 2016)</ns0:ref>. The generally poor success of replication studies has been attributed to a 'publish or perish' culture that incentivizes dubious research practices such as selectively reporting significant results, p-value hacking and establishing hypotheses after the results of a study are known <ns0:ref type='bibr' target='#b12'>(Fidler et al., 2017;</ns0:ref><ns0:ref type='bibr' target='#b15'>Fraser et al., 2018)</ns0:ref>. All of these practices increase the risk of false positives and contribute towards publication bias <ns0:ref type='bibr' target='#b25'>(Jennions & Møller, 2002)</ns0:ref>, which undermines the robustness of the scientific literature.</ns0:p><ns0:p>Further issues include poor study design, low statistical power, variability in reagents or the use of specialized techniques that are difficult to repeat, lack of scientific oversight, inadequate reporting of data, methods and results, and insufficient incentives for sharing data and code <ns0:ref type='bibr' target='#b1'>(Baker, 2016;</ns0:ref><ns0:ref type='bibr' target='#b12'>Fidler et al., 2017;</ns0:ref><ns0:ref type='bibr' target='#b48'>Piper et al., 2019)</ns0:ref>.</ns0:p><ns0:p>Despite growing awareness of these issues not being specific to any particular scientific field, ecological and evolutionary studies are seldom replicated, with only around 0.02% of studies having been self-reported as exact replications <ns0:ref type='bibr' target='#b29'>(Kelly, 2019)</ns0:ref>. One reason for this may be the general perception that research in these fields can be difficult to replicate, partly due to the complexity and lability of many of the phenomena under investigation, but also because in many field situations replication may be unfeasible or even unethical <ns0:ref type='bibr' target='#b28'>(Kelly, 2006;</ns0:ref><ns0:ref type='bibr' target='#b39'>Nakagawa & Parker, 2015;</ns0:ref><ns0:ref type='bibr' target='#b12'>Fidler et al., 2017)</ns0:ref>. Furthermore, numerous factors cannot be controlled for in natural settings and environmental variation in particular may confound attempts to reproduce previous results <ns0:ref type='bibr' target='#b28'>(Kelly, 2006)</ns0:ref>. However, these are not valid reasons to neglect replication studies as it is important to understand the extent to which research outcomes hinge upon these and other factors.</ns0:p><ns0:p>The field of chemical ecology provides an interesting case in point. Increasing numbers of studies are using approaches like gas chromatography-mass spectrometry (GC-MS) to characterize the chemical composition of biological samples such as skin swabs or urine. The resulting 'chemical fingerprints', otherwise commonly referred to as 'chemical profiles', 'scent profiles' or 'odour profiles' <ns0:ref type='bibr' target='#b24'>(Hurst & Beynon, 2010)</ns0:ref>, comprise multiple peaks that are separated according to their retention times and which represent different substances. Studies of both captive and wild animal populations have shown that these chemical fingerprints can convey information about species identity <ns0:ref type='bibr' target='#b6'>(Caspers et al., 2009;</ns0:ref><ns0:ref type='bibr' target='#b16'>Fratini et al., 2012;</ns0:ref><ns0:ref type='bibr' target='#b34'>Krause et al., 2014)</ns0:ref>, population membership <ns0:ref type='bibr' target='#b53'>(Schneeberger et al., 2016;</ns0:ref><ns0:ref type='bibr' target='#b60'>Wierucka et al., 2019)</ns0:ref>, sex, age and reproductive state <ns0:ref type='bibr' target='#b8'>(Caspers et al., 2011;</ns0:ref><ns0:ref type='bibr' target='#b27'>Kean, Müller & Chadwick, 2011;</ns0:ref><ns0:ref type='bibr' target='#b59'>Vogt et al., 2016)</ns0:ref>, family membership <ns0:ref type='bibr' target='#b58'>(Sun & Müller-Schwarze, 1998;</ns0:ref><ns0:ref type='bibr'>Müller & Müller, 2016)</ns0:ref>, individual identity <ns0:ref type='bibr' target='#b26'>(Kean, Chadwick & Mueller, 2015;</ns0:ref><ns0:ref type='bibr' target='#b31'>Kohlwey et al., 2016)</ns0:ref>, social status <ns0:ref type='bibr' target='#b5'>(Burgener et al., 2009)</ns0:ref> and genotype <ns0:ref type='bibr' target='#b62'>(Yamazaki et al., 1990;</ns0:ref><ns0:ref type='bibr' target='#b9'>Charpentier, Boulet & Drea, 2008;</ns0:ref><ns0:ref type='bibr' target='#b54'>Setchell et al., 2011)</ns0:ref>. However, concerns have been raised over the small sample sizes of many studies, which afford little statistical power and may ultimately lead to effect sizes being overestimated <ns0:ref type='bibr' target='#b61'>(Wyatt, 2015)</ns0:ref>. Furthermore, GC-MS data are inherently noisy, making peak detection and alignment challenging <ns0:ref type='bibr' target='#b43'>(Ottensmann et al., 2018)</ns0:ref>. The failure to report peak detection and alignment methods in sufficient detail might therefore act as a barrier to the successful replication of chemical studies. Finally, chemical fingerprints are complex and multidimensional, being influenced by a multitude of factors <ns0:ref type='bibr' target='#b24'>(Hurst & Beynon, 2010;</ns0:ref><ns0:ref type='bibr' target='#b56'>Stoffel et al., 2015)</ns0:ref> including both intrinsic (e.g. genes, hormones and metabolic status) and extrinsic (e.g. environmental variation and diet) variables. Consequently, it remains unclear to what extent many chemical patterns will be repeatable, particularly under natural and often highly heterogeneous conditions.</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:04:48370:1:3:NEW 10 Aug 2020)</ns0:p><ns0:p>Manuscript to be reviewed Pinnipeds provide interesting model systems for studying chemical communication as they possess large repertoires of functional olfactory receptor genes <ns0:ref type='bibr' target='#b30'>(Kishida et al., 2007)</ns0:ref> and are sensitive to even the faintest of smells <ns0:ref type='bibr' target='#b33'>(Kowalewsky et al., 2006)</ns0:ref>. Many pinnipeds have a strong musky smell <ns0:ref type='bibr' target='#b19'>(Hamilton, 1956)</ns0:ref>, which has been attributed to facial glands that show hypertrophy during the breeding season <ns0:ref type='bibr' target='#b35'>(Ling, 1974;</ns0:ref><ns0:ref type='bibr' target='#b20'>Hardy et al., 1991)</ns0:ref>, suggesting an important role of olfactory communication during the peak reproductive period. Olfaction may be particularly crucial for mother-offspring recognition because females of many pinniped species accept or reject pups after naso-nasal inspection <ns0:ref type='bibr' target='#b32'>(Kovacs, 1995;</ns0:ref><ns0:ref type='bibr' target='#b11'>Dobson & Jouventin, 2003;</ns0:ref><ns0:ref type='bibr' target='#b47'>Phillips, 2003)</ns0:ref>. For example, a study of Australian sea lions showed that mothers are capable of discriminating their own pups from nonfilial conspecifics based on odour alone <ns0:ref type='bibr' target='#b49'>(Pitcher et al., 2011)</ns0:ref>. This discovery motivated our team to perform a study of Antarctic fur seals (Arctocephalus gazella), in which chemical fingerprints were characterized from skin swabs taken from 41 mother-offspring pairs at two breeding colonies-the special study beach (SSB) and freshwater beach (FWB)-at Bird Island, South Georgia <ns0:ref type='bibr' target='#b56'>(Stoffel et al., 2015)</ns0:ref>. Despite being separated by less than 200m, animals from these two colonies exhibited highly significant chemical differences, while mothers showed greater chemical similarity to their pups than expected by chance.</ns0:p><ns0:p>Although further research is needed, these findings may have implications for the social organization of Antarctic fur seals as well as for individual recognition. On the one hand, chemical differences between animals from different colonies could potentially facilitate colony recognition and thereby help to explain the remarkable natal philopatry and site fidelity of this species <ns0:ref type='bibr' target='#b21'>(Hoffman et al. 2006;</ns0:ref><ns0:ref type='bibr' target='#b23'>Hoffman and Forcada 2012)</ns0:ref>. As a result, it is possible or even likely that chemical communication will influence the local relatedness structure of fur seal breeding colonies with downstream impacts on inbreeding and mate choice <ns0:ref type='bibr' target='#b22'>(Hoffman et al. 2007;</ns0:ref><ns0:ref type='bibr'>Humble et al. 2020</ns0:ref>). On the other hand, chemical similarities between mothers and their pups are consistent with the hypothesis that mother-offspring recognition in this species may involve self-referent phenotype matching, a conceptually simple mechanism whereby an individual's own phenotype is used as a template for the recognition of close relatives <ns0:ref type='bibr' target='#b3'>(Blaustein, 1983)</ns0:ref>.</ns0:p><ns0:p>Here, we attempted to replicate the chemical patterns of colony membership and mother-offspring similarity reported by <ns0:ref type='bibr' target='#b56'>Stoffel et al. (2015)</ns0:ref>. We returned to the same two breeding colonies five years later, collecting and analysing chemical samples from 50 new mother-offspring pairs using virtually identical methodology. Because these two studies were carried out five years apart, none of the individuals overlapped, precluding analysis of the reproducibility of chemical patterns within individuals. Instead, we use the term 'reproducibility' to refer to the extent to which broad chemical patterns, i.e. differences between colonies and similarities between mothers and their offspring, can be replicated with non-overlapping samples from different time points.</ns0:p><ns0:p>In addition, we wanted to know whether chemical differences between animals from SSB and FWB are specific to this particular experimental setting, or whether chemical signatures are colony-specific in general. We therefore analysed chemical samples from an additional 60 pups from four other colonies around Bird Island in order to test for the generality of the colony membership pattern, by which we mean the extent to which chemical differences are more generally present among animals from different colonies. We hypothesised that (i) the originally reported patterns of colony membership and mother-offspring similarity would be repeatable; and</ns0:p><ns0:p>(ii) that animals from different breeding colonies would differ chemically from one another in general. <ns0:ref type='table'>PDF | (2020:04:48370:1:3:NEW 10 Aug 2020)</ns0:ref> Manuscript to be reviewed</ns0:p></ns0:div>
<ns0:div><ns0:head>PeerJ reviewing</ns0:head></ns0:div>
<ns0:div><ns0:head>Materials & Methods</ns0:head></ns0:div>
<ns0:div><ns0:head>Study site and fieldwork</ns0:head><ns0:p>Chemical samples were taken from six Antarctic fur seal breeding colonies on Bird Island, South Georgia (54° 00′ S, 38° 02′ W) during the peak of the 2016 breeding season (November to December; the previous study was conducted during the peak of the 2011 breeding season). A total of 50 mother-offspring pairs (including one pair of twins) were sampled from SSB and FWB as part of annual routine procedures of the long-term monitoring and survey program of the British Antarctic Survey (BAS). Additional samples were collected from a total of 60 pups from four colonies (15 samples each from Johnson Cove, Main Bay, Landing Beach and Natural Arch, Figure <ns0:ref type='figure' target='#fig_3'>1</ns0:ref>). Here, pups were opportunistically sampled from areas of the beach that were easily accessible. Adult females and pups were captured and restrained on land using standard methodology <ns0:ref type='bibr'>(Gentry & Holt, 1982)</ns0:ref>. Chemical samples were obtained by rubbing the cheek underneath the eye and behind the snout with sterile cotton wool swabs, which were stored individually at -20°C in glass vials containing approximately 10 mL of 60%/40% (vol/vol) ethanol/water. All of the chemical samples were collected immediately after capture by the same team of experienced field scientists. The samples were frozen at the latest one hour after collection and were stored for approximately 18 months prior to analysis.</ns0:p></ns0:div>
<ns0:div><ns0:head>GC-MS profiling and data alignment</ns0:head><ns0:p>We first took 2 mL of each sample and allowed the ethanol to evaporate at room temperature for a maximum of 12 hours before resuspending in 2 mL dichloromethane (DCM). After a further evaporation step, in which the DCM was reduced to a final volume of approximately 100 µL, the samples were analysed on a GC with a VF5-MS column (30 m x 0.25 mm inner diameter, 10 m guard column; Agilent Technologies, Santa Clara, USA) connected to a mass spectrometer (GCMS-QP2020, Shimadzu, Kyoto, Japan). One µL of each sample was injected into a deactivated glass-wool-packed liner with an inlet temperature of 225°C. A split ratio of 3.2 was used and the carrier gas (Helium) flow rate was held constant at 1.2 mL/min. The GC run started with three minutes at 60 °C and then ramped up in increments of 10 °C/min to reach a final temperature of 280 °C, which was maintained for 30 minutes. Mass spectra were taken in electron ionization mode with five scans per second in full scan mode (50-600 m/z). The resulting GC-MS data were then processed using OpenChrom (Wenig & Odermatt, 2010) for detection and correction of split peaks. Afterwards, we used GCalignR in R <ns0:ref type='bibr' target='#b43'>(Ottensmann et al., 2018;</ns0:ref><ns0:ref type='bibr' target='#b50'>R Core Team, 2019)</ns0:ref> to align the resulting chromatograms by correcting minor shifts in retention times among samples and maximizing the number of shared components.</ns0:p></ns0:div>
<ns0:div><ns0:head>Data visualisation and statistical analysis</ns0:head><ns0:p>Prior to data analyses, we excluded any compounds that were only observed in a single seal sample.</ns0:p><ns0:p>We then used non-metric multidimensional scaling (NMDS) to visualize the chemical data. This approach reduces dimensionality so that each individual data point can be placed in a 2D scatterplot where ranked between-individual distances are preserved and individuals that are chemically more similar are closer together. NMDS was performed on a log(x+1) transformed relative abundance matrix comprising pairwise Bray-Curtis similarity values. We tested for differences among and between a priori defined groups (i.e. the breeding colonies and motheroffspring pairs) using a non-parametric permutational multivariate analysis of variance (PERMANOVA). PERMANOVA tests whether the centroids of pre-defined groups differ statistically for a chosen distance measure. It compares within-group to among-group variance components and assigns statistical significance based on random permutations of objects within groups. Each PERMANOVA was based on 99,999 permutations, although comparable results were also obtained with 9999, 999 and 99 permutations. To determine whether differences between our pre-defined groups were purely attributable to compositional differences between groups rather than compositional differences within groups, we used the 'betadisper' function in the vegan package in R to analyse the multivariate homogeneity of group dispersions <ns0:ref type='bibr' target='#b41'>(Oksanen et al., 2019)</ns0:ref>. In addition, we performed pairwise PERMANOVAs for different groups within the model strata based on age and colony and Bonferroni corrected the resulting p-values.</ns0:p></ns0:div>
<ns0:div><ns0:head>Quantification of the explained variance</ns0:head><ns0:p>To facilitate a comparison of our effect sizes with those reported by <ns0:ref type='bibr' target='#b56'>Stoffel et al. (2015)</ns0:ref>, we quantified the proportion of the total chemical variance attributable to colony membership and family ID in both studies. The scripts that <ns0:ref type='bibr' target='#b56'>Stoffel et al. (2015)</ns0:ref> used to align their data have now been embedded into GCalignR and are therefore consistent between the two studies. As different chemical datasets will have different optimal parameter settings for the alignment algorithm, we Manuscript to be reviewed did not re-align or adjust the dataframe of <ns0:ref type='bibr' target='#b56'>Stoffel et al. (2015)</ns0:ref>. Enforcing the same parameter settings as in the current study would almost certainly lead to a loss of data quality and result in artificially reduced effect sizes. To standardise effect size estimates between the studies, both chemical datasets were bootstrapped over individuals to generate 5,000 datasets per study, each comprising 15 mother-offspring pairs from SSB and 15 pairs from FWB (i.e. a total of 60 individuals). PERMANOVA was then implemented separately for each dataset and the resulting R 2 values were extracted for each of the predefined groups.</ns0:p></ns0:div>
<ns0:div><ns0:head>Data availability</ns0:head><ns0:p>The raw chemical data generated during this study are available via Github and the data of <ns0:ref type='bibr' target='#b56'>Stoffel et al. (2015)</ns0:ref> can be downloaded from https://github.com/mastoffel/seal_chemical_fingerprints.</ns0:p><ns0:p>All of the code used to analyse the raw data are available as a PDF file written in Rmarkdown (see supplementary information). The full documented data analysis pipeline can be downloaded from our GitHub repository at https://github.com/tebbej/SealScent2020/. Manuscript to be reviewed</ns0:p></ns0:div>
<ns0:div><ns0:head>Results</ns0:head><ns0:p>In order to investigate the reproducibility of chemical patterns of colony membership and motheroffspring similarity in Antarctic fur seals, we analysed chemical data from mother-offspring pairs from SSB and FWB as well as pups from an additional four breeding colonies around Bird Island (Figure <ns0:ref type='figure' target='#fig_3'>1</ns0:ref>). We detected an average of 42 ± 15 s.d. chemicals per sample. No significant differences were found in the number of chemicals between mothers and offspring (unpaired ttest, t = 0.8403, p = 0.403) or among pups from the six breeding colonies (ANOVA, F 5,104 =0.001, p = 0.98).</ns0:p></ns0:div>
<ns0:div><ns0:head>Reproducibility of chemical patterns</ns0:head><ns0:p>Multivariate statistical analysis of the relative proportions of each substance revealed highly significant differences between animals from SSB and FWB (PERMANOVA, p < 0.0001, Figure <ns0:ref type='figure'>2a</ns0:ref>, Table <ns0:ref type='table'>1a</ns0:ref>). A highly significant effect of mother-pup pair ID nested within colony (PERMANOVA, p < 0.0001, Figure <ns0:ref type='figure'>2b</ns0:ref>, Table <ns0:ref type='table'>1a</ns0:ref>) was also found, indicating that mothers and their pups are chemically more similar to one another than expected by chance. A test for multivariate homogeneity of group variances uncovered marginally significant differences among the groups (p = 0.026, Table <ns0:ref type='table'>1a</ns0:ref>), which could potentially indicate the involvement of additional explanatory factors that were not accounted for in the model. We therefore investigated further by splitting the chemical data into four groups, corresponding to mothers and pups from SSB and FWB respectively. Performing PERMANOVAs for all possible pairwise combinations of these groups produced three important outcomes. First, all of the pairwise PERMANOVAs involving groups of animals from the two different colonies were highly significant after table-wide Bonferroni correction for multiple tests (Table <ns0:ref type='table'>S1</ns0:ref>). This indicates that colony membership is chemically encoded irrespective of whether individuals are mothers or pups. Second, both of the pairwise PERMANOVAs involving mothers and pups within colonies were non-significant after Bonferroni correction (Table <ns0:ref type='table'>S1</ns0:ref>). This suggests that mothers and their pups are chemically similar to one another, regardless of the colony in question. Finally, tests for the homogeneity of group variances were not significant for any of the pairwise group comparisons after Bonferroni correction (Table <ns0:ref type='table'>S2</ns0:ref>). This implies that our results are unlikely to be driven by differences in chemical variance among groups.</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:04:48370:1:3:NEW 10 Aug 2020)</ns0:p></ns0:div>
<ns0:div><ns0:head>Manuscript to be reviewed</ns0:head><ns0:p>As p-values cannot be directly compared between studies with different sample sizes, we used the PERMANOVA framework to estimate the effect sizes of colony membership and motheroffspring similarity in both studies. To facilitate direct comparisons while also incorporating uncertainty due to chemical variation among individuals, both datasets were bootstrapped over individuals as described in the Materials and methods. We found that effect size estimates for colony membership and mother-offspring similarity (maximum density R 2 values) differed by only few percent between the two studies (Figure <ns0:ref type='figure'>3</ns0:ref>) and consistently fell within the range of 0.08 < R 2 < 0.15.</ns0:p></ns0:div>
<ns0:div><ns0:head>Generality of chemical patterns</ns0:head><ns0:p>To investigate whether chemical signatures are colony-specific in general, we analysed chemical data from pups sampled from a total of six colonies around Bird Island. PERMANOVA uncovered chemical differences not only between SSB and FWB, but also more generally among colonies (Figure <ns0:ref type='figure' target='#fig_3'>S1</ns0:ref>). These differences were statistically significant both overall (p < 0.0001, Table <ns0:ref type='table'>1b</ns0:ref>) and for the majority of pairwise comparisons after Bonferroni correction (Table <ns0:ref type='table'>2</ns0:ref>).</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:04:48370:1:3:NEW 10 Aug 2020)</ns0:p></ns0:div>
<ns0:div><ns0:head>Manuscript to be reviewed</ns0:head></ns0:div>
<ns0:div><ns0:head>Discussion</ns0:head><ns0:p>A major obstacle to reproducible research in ecology and evolution is the perceived difficulty of replicating original research findings in natural settings where many variables cannot be controlled for and where spatial and temporal dependencies may confound faithful replication attempts <ns0:ref type='bibr' target='#b39'>(Nakagawa & Parker, 2015;</ns0:ref><ns0:ref type='bibr' target='#b12'>Fidler et al., 2017)</ns0:ref>. Although the inherent variability of natural systems undoubtedly poses a challenge to replication studies, our findings suggest that, at least under some circumstances, chemical patterns may be repeatable. Specifically, we found that the effect sizes of patterns of colony membership and mother-offspring similarity in <ns0:ref type='bibr' target='#b56'>Stoffel et al. (2015)</ns0:ref> were of similar magnitude in a new sample of mother-offspring pairs separated by five years. By expanding the geographical scope of our sampling, we could furthermore show that chemical signatures are colony-specific in general. Our results lend further support to the conclusion that colony membership and mother-offspring similarity are chemically encoded in Antarctic fur seals.</ns0:p></ns0:div>
<ns0:div><ns0:head>Motivation and study design</ns0:head><ns0:p>A number of factors motivated the current replication attempt. First, the results of <ns0:ref type='bibr' target='#b56'>Stoffel et al. (2015)</ns0:ref> were based on a modest sample of Antarctic fur seal mother-offspring pairs sampled in a single season. We therefore wanted to safeguard against type I error while also testing for the repeatability of chemical patterns over time. Second, chance results can become highly influential <ns0:ref type='bibr' target='#b28'>(Kelly, 2006)</ns0:ref> and our original study already appears to have motivated comparable investigations in other pinniped species. For example, a recent study of Australian sea lions using a very similar experimental design also reported chemical differences between two breeding colonies, but chemical similarities were not found between mothers and their pups <ns0:ref type='bibr' target='#b60'>(Wierucka et al., 2019)</ns0:ref>.</ns0:p><ns0:p>Although it is not unreasonable to assume that different species might vary in how chemical information is encoded and used in mother-offspring recognition, this point of difference nevertheless encouraged us to revisit our original findings. Finally, being able to confirm and extend our original results strengthens the case for follow-up studies and reduces the risk of time and resources being wasted on chasing up false positives.</ns0:p><ns0:p>Although we acknowledge that no study of a wild population can ever be perfectly replicated <ns0:ref type='bibr' target='#b39'>(Nakagawa & Parker, 2015;</ns0:ref><ns0:ref type='bibr' target='#b12'>Fidler et al., 2017)</ns0:ref>, we believe that our replication study of chemical patterns in Antarctic fur seals is sufficiently close to that of <ns0:ref type='bibr' target='#b56'>Stoffel et al. (2015)</ns0:ref> in terms of both experimental design and implementation to be considered an exact replication. In practice, there were a handful of small differences between the two studies, but these were mainly a consequence of incremental improvements to our methodology and are unlikely to have had a major influence on the final outcome. For example, because replication studies often produce smaller effect sizes than original studies <ns0:ref type='bibr' target='#b55'>(Simonsohn, 2015;</ns0:ref><ns0:ref type='bibr'>Open science collaboration, 2015)</ns0:ref>, we attempted to enlarge our sample size of mother-offspring pairs as far as was practicable. We also improved the standardization and reproducibility of our chemical analysis pipeline by performing peak detection with open source software and by integrating the alignment algorithm of <ns0:ref type='bibr' target='#b56'>Stoffel et al. (2015)</ns0:ref> into an R package <ns0:ref type='bibr' target='#b43'>(Ottensmann et al., 2018)</ns0:ref>. However, these small modifications appear to have been of little consequence as the effect sizes of colony membership and mother-offspring similarity did not differ systematically between the two studies.</ns0:p><ns0:p>Two further methodological differences were beyond our control. First, owing to the fact that the original and replication studies were carried out five years apart, the sampling was conducted by different teams of field biologists. However, we used carefully standardized field protocols in order to minimize any inadvertent experimental variation. Second, the GC-MS machine used by <ns0:ref type='bibr' target='#b56'>Stoffel et al. (2015)</ns0:ref> was subsequently replaced by a newer and more sensitive model. One might have expected this to result in more chemicals being detected in the replication study, which would be expected to provide greater power to detect chemical patterns. If anything, however, fewer chemicals in total were detected in the current study, possibly because of differences in the concentrations of samples or because we used different peak calling software and manually curated the resulting dataset to remove redundant split peaks. Regardless of the exact explanation, the overall similarity of the results of the two studies suggests that patterns of colony membership and mother-offspring similarity in Antarctic fur seals are robust to these minor sources of experimental variation. This robustness would be expected if chemical patterns are influenced by large numbers of compounds and therefore persist independently of minor methodological differences that may influence which subsets of peaks are detected and retained for analysis.</ns0:p></ns0:div>
<ns0:div><ns0:head>Replication outcomes</ns0:head><ns0:p>PeerJ reviewing PDF | (2020:04:48370:1:3:NEW 10 Aug 2020)</ns0:p><ns0:p>Manuscript to be reviewed Successful replication can be defined either in the context of statistical significance <ns0:ref type='bibr' target='#b51'>(Rosenthal, 1991)</ns0:ref> or on the basis of a comparison of effect sizes <ns0:ref type='bibr' target='#b17'>(Goodman, Fanelli & Ioannidis, 2016;</ns0:ref><ns0:ref type='bibr' target='#b48'>Piper et al., 2019)</ns0:ref>. We not only tested for significance but also developed an approach based on PERMANOVA to evaluate the effect sizes of colony membership and mother-offspring similarity in both datasets. Specifically, we extracted R 2 values for the terms in question after bootstrapping both chemical datasets over individuals. This approach controlled for differences in sample size between the two studies while also providing a visual representation of the magnitude of uncertainty associated with the R 2 estimates. We not only found that the patterns reported by <ns0:ref type='bibr' target='#b56'>Stoffel et al. (2015)</ns0:ref> remained highly significant, but also that the effect size estimates of colony membership and mother-offspring similarity in the two studies were more or less similar, varying by at most a few percent. Elsewhere, in a study that attempted to replicate a hundred psychological studies (Open science collaboration, 2015), variation in the strength of the original evidence, such as p-values, was more predictive of replication success than other characteristics such as the experience or expertise of the original and replication teams. This is consistent with the outcome of the current replication exercise given the high significance (p < 0.0001) of the patterns originally reported by <ns0:ref type='bibr' target='#b56'>Stoffel et al. (2015)</ns0:ref>.</ns0:p></ns0:div>
<ns0:div><ns0:head>Generality of the colony membership pattern</ns0:head><ns0:p>We went a step beyond simply repeating our previous study by investigating whether chemical differences between SSB and FWB are specific to these two colonies, or whether chemical signatures are colony-specific in general. Unfortunately, it was not possible to sample mothers from locations other than SSB and FWB due to the difficulty of capturing adult females farther away from the BAS field station where fieldwork on seals is rarely if ever performed. However, the relative ease of capturing pups enabled us to gather a more representative collection of chemical samples from multiple breeding sites around Bird Island. After controlling for the false discovery rate, statistically significant chemical differences were detected in all but two out of 15 pairwise comparisons between colonies. This suggests not only that chemical patterns of colony membership are repeatable over time, but also that they can be generalized over space. Interestingly, we did not find a clear correspondence between chemical similarity and the geographical proximity of colonies. For example, Freshwater Beach and Main Bay were among the most chemically dissimilar colonies despite being only around 500m apart, while Johnson PeerJ reviewing PDF | (2020:04:48370:1:3:NEW 10 Aug 2020)</ns0:p></ns0:div>
<ns0:div><ns0:head>Manuscript to be reviewed</ns0:head><ns0:p>Cove and Natural Arch were among the most chemically similar colonies despite being situated at the opposite extremes of Bird Island. The most probable explanation for this pattern is that chemical differences among colonies are predominantly shaped by as yet unknown environmental factors (see below).</ns0:p></ns0:div>
<ns0:div><ns0:head>Mechanisms encoding chemical information</ns0:head><ns0:p>Relatively little is currently known about the mechanisms by which colony membership and mother-offspring similarity are chemically encoded in Antarctic fur seals. We know that animals from SSB and FWB exhibit chemical differences despite a lack of genetic differentiation <ns0:ref type='bibr' target='#b56'>(Stoffel et al 2015)</ns0:ref>, which implies that environmental drivers play an important role. However, it remains unclear exactly what these drivers might be. Food is unlikely to be an important determinant of colony-specific chemical patterns because breeding female fur seals feed predominantly on Antarctic krill <ns0:ref type='bibr' target='#b4'>(Boyd, Staniland & Martin, 2002)</ns0:ref>. The underlying substrate is also relatively homogenous, with the vast majority of animals occupying cobblestone breeding beaches that show little in the way of obvious differences to the human eye. It is therefore more likely that colonyspecific chemical phenotypes are influenced by differences in local conditions such as temperature, wind or solar radiation, either directly or via alterations to the skin microbiota <ns0:ref type='bibr' target='#b18'>(Grosser et al 2019)</ns0:ref>.</ns0:p><ns0:p>A further possibility could be that chemical differences between colonies reflect differences in microbial communities shaped by social stress. For example, stressful conditions such as high densities of conspecifics can suppress microbial diversity <ns0:ref type='bibr' target='#b0'>(Bailey et al., 2011;</ns0:ref><ns0:ref type='bibr' target='#b57'>Stothart et al., 2016;</ns0:ref><ns0:ref type='bibr' target='#b40'>Noguera et al., 2018;</ns0:ref><ns0:ref type='bibr' target='#b45'>Partrick et al., 2018;</ns0:ref><ns0:ref type='bibr' target='#b63'>Zha et al., 2018)</ns0:ref>. This is consistent with our data, as breeding females on SSB are present at higher density and have chronically elevated levels of the stress hormone cortisol <ns0:ref type='bibr' target='#b36'>(Meise et al., 2016)</ns0:ref>, while skin microbial diversity is also lower in this colony <ns0:ref type='bibr' target='#b18'>(Grosser et al 2019)</ns0:ref>. Investigating the potential linkages between social stress, cortisol, microbial community structure and chemical phenotypes represents a promising avenue for future research.</ns0:p></ns0:div>
<ns0:div><ns0:head>Conclusions</ns0:head><ns0:p>Our study set out to test two hypotheses, namely that chemical patterns of colony membership and mother-offspring similarity in Antarctic fur seals are reproducible over time, and that chemical differences will be present not only between SSB and FWB, but also more generally among Manuscript to be reviewed colonies. Both hypotheses were supported by our data. The overall robustness of chemical patterns of colony membership and mother-offspring similarity in Antarctic fur seals is consistent with the notion that chemical information could be important for social communication in pinnipeds, and lays a solid foundation for future studies of the mechanisms responsible for chemical variation. Finally, as a lack of access to raw data, code and software has been identified as a fundamental obstacle to replication <ns0:ref type='bibr' target='#b12'>(Fidler et al., 2017)</ns0:ref>, we have made the data from both studies as well as the code used to analyze them freely available, while also using maximally transparent, open access software for peak detection and alignment. </ns0:p><ns0:note type='other'>Figure 3</ns0:note><ns0:p>Effect sizes of colony membership and mother-offspring similarity in the original study <ns0:ref type='bibr' target='#b56'>(Stoffel et al. 2015)</ns0:ref> and in this replication study. </ns0:p></ns0:div><ns0:figure xml:id='fig_0'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:04:48370:1:3:NEW 10 Aug 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_1'><ns0:head /><ns0:label /><ns0:figDesc>Ethical statement Samples were collected as part of the Polar Science for Planet Earth program of the British Antarctic Survey under the authorization of the Senior Executive and the Environment Officers of the Government of South Georgia and the South Sandwich Islands (permit no. 2016/013). Samples were collected and retained under Scientific Research Permits for the British Antarctic Survey field activities on South Georgia, and in accordance with the Convention on International Trade in Endangered Species of Wild Fauna and Flora. All field procedures were approved by the British Antarctic Survey Animal Welfare and Ethics Review Body (reference no. PEA6). PeerJ reviewing PDF | (2020:04:48370:1:3:NEW 10 Aug 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_2'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:04:48370:1:3:NEW 10 Aug 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_3'><ns0:head>Figure 1</ns0:head><ns0:label>1</ns0:label><ns0:figDesc>Figure 1</ns0:figDesc></ns0:figure>
<ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='29,42.52,70.87,331.58,672.95' type='bitmap' /></ns0:figure>
<ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='33,42.52,70.87,525.00,364.50' type='bitmap' /></ns0:figure>
<ns0:figure type='table' xml:id='tab_0'><ns0:head>R 2 p-value</ns0:head><ns0:label /><ns0:figDesc /><ns0:table><ns0:row><ns0:cell>Age</ns0:cell><ns0:cell>2.65</ns0:cell><ns0:cell>0.023</ns0:cell><ns0:cell>0.004</ns0:cell></ns0:row><ns0:row><ns0:cell>Colony membership</ns0:cell><ns0:cell>9.07</ns0:cell><ns0:cell>0.076</ns0:cell><ns0:cell><0.0001</ns0:cell></ns0:row><ns0:row><ns0:cell>Family ID nested within colony membership</ns0:cell><ns0:cell>9.02</ns0:cell><ns0:cell>0.153</ns0:cell><ns0:cell><0.0001</ns0:cell></ns0:row><ns0:row><ns0:cell>Test for homogeneity of variance for colony membership</ns0:cell><ns0:cell>5.14</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>0.026</ns0:cell></ns0:row><ns0:row><ns0:cell>Test for homogeneity of variance for age</ns0:cell><ns0:cell>1.47</ns0:cell><ns0:cell /><ns0:cell>0.228</ns0:cell></ns0:row><ns0:row><ns0:cell>Test for homogeneity of variance for age & colony</ns0:cell><ns0:cell>1.91</ns0:cell><ns0:cell /><ns0:cell>0.134</ns0:cell></ns0:row><ns0:row><ns0:cell>membership</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>(b) PERMANOVA of pups from six colonies</ns0:cell><ns0:cell>F</ns0:cell><ns0:cell>R 2</ns0:cell><ns0:cell>p-value</ns0:cell></ns0:row><ns0:row><ns0:cell>Colony membership</ns0:cell><ns0:cell>5.17</ns0:cell><ns0:cell>0.191</ns0:cell><ns0:cell><0.0001</ns0:cell></ns0:row><ns0:row><ns0:cell>Test for homogeneity of variance for colony membership</ns0:cell><ns0:cell>0.50</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>0.778</ns0:cell></ns0:row><ns0:row><ns0:cell>(c)</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row></ns0:table></ns0:figure>
<ns0:figure type='table' xml:id='tab_1'><ns0:head>PERMANOVA of mothers and offspring from two colonies (data from Stoffel et al. 2015).</ns0:head><ns0:label /><ns0:figDesc /><ns0:table><ns0:row><ns0:cell /><ns0:cell>F</ns0:cell><ns0:cell>R²</ns0:cell><ns0:cell>p-value</ns0:cell></ns0:row><ns0:row><ns0:cell>Age</ns0:cell><ns0:cell>0.98</ns0:cell><ns0:cell>0.010</ns0:cell><ns0:cell>0.461</ns0:cell></ns0:row><ns0:row><ns0:cell>Colony membership</ns0:cell><ns0:cell>12.35</ns0:cell><ns0:cell>0.128</ns0:cell><ns0:cell><0.0001</ns0:cell></ns0:row><ns0:row><ns0:cell>Family ID nested within colony membership</ns0:cell><ns0:cell>3.13</ns0:cell><ns0:cell>0.065</ns0:cell><ns0:cell><0.0001</ns0:cell></ns0:row><ns0:row><ns0:cell>Test for homogeneity of variance for colony membership</ns0:cell><ns0:cell>0.22</ns0:cell><ns0:cell /><ns0:cell>0.639</ns0:cell></ns0:row><ns0:row><ns0:cell>Test for homogeneity of variance for age</ns0:cell><ns0:cell>0.35</ns0:cell><ns0:cell /><ns0:cell>0.557</ns0:cell></ns0:row><ns0:row><ns0:cell>Test for homogeneity of variance for age & colony</ns0:cell><ns0:cell>0.21</ns0:cell><ns0:cell /><ns0:cell>0.887</ns0:cell></ns0:row><ns0:row><ns0:cell>membership</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row></ns0:table></ns0:figure>
<ns0:figure type='table' xml:id='tab_2'><ns0:head /><ns0:label /><ns0:figDesc>To quantify the amount of explained variance, we bootstrapped both datasets over individuals and extracted the corresponding R 2 values for each of the predefined groups in separate PERMANOVAs (see Materials and methods for details). The data are presented as sinaplots with overlaid boxplots (centre line = median, bounds of box = 25th and 75th percentiles, upper and lower whiskers = largest and lowest value but no further than 1.5 * inter-quartile range from the hinge) and the grey points represent effect sizes based on the full datasets.</ns0:figDesc><ns0:table><ns0:row><ns0:cell>Pairs</ns0:cell><ns0:cell>F</ns0:cell><ns0:cell>R 2</ns0:cell><ns0:cell>p-value</ns0:cell><ns0:cell>Corrected p-value</ns0:cell></ns0:row><ns0:row><ns0:cell>SSB versus FWB</ns0:cell><ns0:cell cols='3'>6.08 0.110 <0.0001</ns0:cell><ns0:cell><0.001</ns0:cell></ns0:row><ns0:row><ns0:cell>SSB versus Landing Beach</ns0:cell><ns0:cell cols='2'>3.54 0.083</ns0:cell><ns0:cell><0.001</ns0:cell><ns0:cell>0.012</ns0:cell></ns0:row><ns0:row><ns0:cell>SSB versus Main Bay</ns0:cell><ns0:cell cols='3'>6.18 0.137 <0.0001</ns0:cell><ns0:cell><0.0001</ns0:cell></ns0:row><ns0:row><ns0:cell>SSB versus Natural Arch</ns0:cell><ns0:cell cols='2'>4.17 0.097</ns0:cell><ns0:cell><0.001</ns0:cell><ns0:cell>0.002</ns0:cell></ns0:row><ns0:row><ns0:cell>SSB versus Johnson Cove</ns0:cell><ns0:cell cols='3'>4.40 0.104 <0.0001</ns0:cell><ns0:cell><0.001</ns0:cell></ns0:row><ns0:row><ns0:cell>FWB versus Landing Beach</ns0:cell><ns0:cell cols='3'>4.16 0.099 <0.0001</ns0:cell><ns0:cell><0.001</ns0:cell></ns0:row><ns0:row><ns0:cell>FWB versus Main Bay</ns0:cell><ns0:cell cols='3'>7.29 0.161 <0.0001</ns0:cell><ns0:cell><0.001</ns0:cell></ns0:row><ns0:row><ns0:cell>FWB versus Natural Arch</ns0:cell><ns0:cell cols='3'>7.91 0.172 <0.0001</ns0:cell><ns0:cell><0.001</ns0:cell></ns0:row><ns0:row><ns0:cell>FWB versus Johnson Cove</ns0:cell><ns0:cell cols='3'>7.15 0.162 <0.0001</ns0:cell><ns0:cell><0.001</ns0:cell></ns0:row><ns0:row><ns0:cell>Landing Beach versus Main Bay</ns0:cell><ns0:cell cols='2'>3.32 0.106</ns0:cell><ns0:cell>0.002</ns0:cell><ns0:cell>0.030</ns0:cell></ns0:row><ns0:row><ns0:cell>Landing Beach versus Natural Arch</ns0:cell><ns0:cell cols='2'>3.00 0.097</ns0:cell><ns0:cell>0.004</ns0:cell><ns0:cell>0.063</ns0:cell></ns0:row><ns0:row><ns0:cell>Landing Beach versus Johnson Cove</ns0:cell><ns0:cell cols='2'>2.70 0.091</ns0:cell><ns0:cell>0.012</ns0:cell><ns0:cell>0.177</ns0:cell></ns0:row><ns0:row><ns0:cell>Main Bay versus Natural Arch</ns0:cell><ns0:cell cols='3'>5.92 0.175 <0.0001</ns0:cell><ns0:cell><0.001</ns0:cell></ns0:row><ns0:row><ns0:cell>Main Bay versus Johnson Cove</ns0:cell><ns0:cell cols='2'>3.14 0.104</ns0:cell><ns0:cell><0.001</ns0:cell><ns0:cell>0.005</ns0:cell></ns0:row><ns0:row><ns0:cell>Natural Arch versus Johnson Cove</ns0:cell><ns0:cell cols='2'>2.43 0.083</ns0:cell><ns0:cell>0.016</ns0:cell><ns0:cell>0.245</ns0:cell></ns0:row><ns0:row><ns0:cell>1</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row></ns0:table></ns0:figure>
<ns0:note place='foot' n='1'>PeerJ reviewing PDF | (2020:04:48370:1:3:NEW 10 Aug 2020)Manuscript to be reviewed</ns0:note>
</ns0:body>
" | "Editor comments (Jean Clobert)
MAJOR REVISIONS
Dear author,
Both referees recognized the interest of this study and made many interesting comments to improve the flow and clarity of your manuscript.
RESPONSE 1) We are grateful to the referees and the editor for providing detailed feedback on our manuscript. We have addressed all of the methodological criticisms and made substantial changes to all sections of our manuscript to improve clarity and flow.
One referee has however two more serious concerns, one about statistics, and one more general about the mixing between repetition (aim of the manuscript)on the two colonies and the extention to six others non precedenlty sampled.
RESPONSE 2) We have implemented additional analyses in response to the comments raised by the reviewer and have also clarified our methodology. We are even more convinced by our results and hope that reviewer's concerns have now been addressed.
We were extremely surprised to receive the criticism that our manuscript mixes repetition and extension. These two concepts are deeply interconnected in the repeatability literature, and numerous studies have highlighted the importance of both exact and conceptual replications for probing the validity and generality of biological phenomena, respectively (Goodman, Fanelli & Ioannidis, 2016; Dirnagl, 2019; Piper et al., 2019, Schmidt, 2009). To clarify, studies that perform 'exact' replications (e.g. comparing mother-offspring pairs from the same two colonies in different years) can only tell us if a certain effect can be replicated in a specific setting (i.e. SSB and FWB). Conceptual replications go a step further by allowing the claim(s) of the original study to be extended to previously unsampled settings (i.e. other breeding colonies). In our case, chemical data from pups from four additional colonies around Bird Island clearly strengthen our original claim that colony membership is chemically encoded. Consequently, these different forms of replication are both important, and they are complementary. Removing the conceptual replication part of our study would therefore significantly weaken our results and conclusions.
We have re-worked parts of the abstract, introduction and discussion to emphasise the importance of both types of replication and hope that this helps to clarify the motivation of our study and the importance of both sets of results.
There are two further reasons for retaining the pup data. The first one is a statistical one. Referee 1 raised concerns over a marginally significant p-value for the test for homogenity of group variances for the PERMANOVA of mother-pup pairs from SSB and FWB. However, the equivalent value for pups from six colonies is not significant (nor is the value obtained from the original data of Stoffel et al (2015)). Consequently, retaining the pups gives us greater confidence that chemical differences between colonies are not an artefact of differences in within-group variability (see Response 20 for details). Second, Referee 2 expresses a clear interest and curiosity about the pup data (see Response 33). It would therefore be a shame to remove them, as it appears there is a reasonable chance that this will detract from the broader interest of our paper to a general audience.
I am also concerned by this mixture of results especially because I am puzzled by the fact that 1) odors seem to be used for mother-offsrping recognition and 2) that you use colonies differences to emphasize this possibility. are mothers often changing of colonies that colonies differences help recognizing her offspring? I just do not understand the reasoning here.
RESPONSE 3) We have written a new paragraph for the introduction to explain what we believe the significance of these two patterns are. Briefly, colony differences may play a role in colony recognition, which may underpin the impressive site fidelity of this species and have downstream implications for kin structure, inbreeding and mate choice. Conversely, chemical similarities between mothers and their offspring hint at the possible involvement of self-referent phenotype matching in mother-offspring recognition. We at no point meant to link the two, although in principle being able to recognise and return to a specific breeding colony could be an important first step towards identifying young within that colony.
This apparent confusion (to my eyes) might be removed if you dollow the suggestion by the second referee to drop the results concerning the six additional colonies.
RESPONSE 4) As explained above, there are a number of important justifications for not dropping these results. We have therefore restructured parts of the manuscript to explain our motivations for including both types of replication in our study, as well as the importance of both sets of results in terms of reaching robust conclusions about the repeatability of chemical patterns.
This is why I recommend major revisions and expect that you will be able to clarify these points
RESPONSE 5) We hope that we have been able to clarify our position and we are confident that the changes we have made to the manuscript in response to the editor and referee comments have improved the quality of our work.
Reviewer 1
Basic reporting
The manuscript is exceptionally well written and appropriately structured, making it very easy to read and understand. The text flows well and the concepts are thoroughly explained. The knowledge of the authors on the background of the study is evident and the references are appropriate and up to date. Below I have listed a few minor comments about the background and the general outline of the publication that I hope the authors will find the following useful in improving their manuscript:
At the moment, the introduction and therefore the framing of the whole article is focused on replication. 1) Please rephrase the title to reflect this more. After reading the current title, I expected a very different introduction to the article, which was slightly confusing at the beginning.
RESPONSE 6) We do not understand this comment. The title clearly states that chemical patterns are reproducible, and reproducibility of chemical patterns is the focus of the paper. Perhaps the reviewer is confused over specific terminologies such as reproducibility and replication. However, we covered these (as well as related terms) and their meanings in a carefully crafted introduction paragraph. Nevertheless, reading the title, we realised that the last two words were not necessary, and have removed them.
2) I fully support replication studies and consider them an extremely important part of research, as outlined in the introduction. However, the article seems to have two separate goals: a) replicating a subsection of results presented in Stoffel et al. 2015 and b) adding additional information about chemical profile differences among colonies. Adding original results to the study (and their discussion) causes the replication section to lose importance, as it creates the feeling of a side research project that was difficult to publish due to the lack of novelty and was then restructured into a study highlighting its replication aspect.
RESPONSE 7) As outlined in Response 2, the reviewer appears to have misunderstood two key concepts in reproducibility: exact replication and conceptual replication. The former attempts to validate a specific result in a specific setting, while the latter probes it’s generality. Both are important aspects of reproducibility that contribute towards understanding the validity of a given result (Goodman, Fanelli & Ioannidis, 2016; Dirnagl, 2019; Piper et al., 2019, Schmidt, 2009).
Our pup sampling scheme was specifically designed to quantify the extent to which the colony membership pattern can be extended beyond the specific context of the two-colony design of Stoffel et al (2015). At no point did we restructure uninteresting results from a side-study into our replication study. We also disagree with the implication that the pup results are uninteresting – the second reviewer for one seems to be interested in them, given his comments (See Response 33).
Finally, we do not believe that our ability to successfully generalise our results in any way detracts from the importance of the exact replication part of our study. To the contrary, our results from the pups significantly strengthen our original claim that colony membership is chemically encoded, and thereby increase the ‘external validity’ of our findings.
I realise the difficulties of publishing replicated studies and wanting to publish all available results, however I think the paper would benefit if just one, clear aim was presented – the replication study. The analysis is well executed, well explained and provides enough content for a publication and can be used as guide for future studies of this sort. The additional results, presenting chemical profile differences among multiple colonies, are interesting and worth publishing. However, I think it would be better to publish these as a standalone short article or note.
RESPONSE 8) There is absolutely no question of our ‘wanting to publish all available results’. This study was specifically designed from the very start to be a cohesive whole, and we believe that it deserves to be published as such. As explained above, we cannot justify taking a ‘splitting approach’.
L54-62: Rather than providing quotes from the referenced papers, please summarise and rephrase the content, leaving just the reference at the end of the sentence.
RESPONSE 9) We believe it is appropriate, as well as consistent with similar papers, to include Fisher's quote, rather than paraphrasing. We also quoted a small number of specific phrases like ‘contract of error’, which are difficult to effectively reformulate. Given that the second reviewer raised no such concerns about the inclusion of quotes, and that it is perfectly legitimate to include quotations in scientific papers as long as the original articles are cited, we do not see a compelling argument for reformulating this content.
The current introduction contains 5 paragraphs on replication. While most information is relevant to the current study and revisited in the discussion, paragraphs 4 and 5 could be made more concise.
RESPONSE 10) We revisited our introductory paragraphs and have removed a few lines of text to shorten them in places. However, further shortening will only lead to the loss of relevant and important content. We also note that the second reviewer was very positive about the way we wrote the manuscript.
L101: Please remove the abbreviation - it is not used later
RESPONSE 11) We have removed the abbreviation.
L133-140: Please also mention the inconsistencies/lack of appropriate reporting/lack of automated (and therefore reproducible) peak detection and alignment methods. This creates many inconsistencies and difficulties in replicating results using GCMS data.
RESPONSE 12) This is a good point, which we have incorporated. However, it is not strictly correct that there is a lack of automated peak detection and alignment methods. For example, our group produced an R package for peak alignment (GCalignR) and a quick literature search at the time uncovered no fewer than 25 other programs that have been developed for the same purpose! For details, please see S1 file in Ottensmann et al (2018).
Experimental design
The experimental design followed the excellent study of Stoffel et al. 2015 and have no major concerns regarding it. The authors also clearly pointed out how it differed from the original paper and how that could have affected their results, which I greatly appreciated.
The code and raw data from the original as well as the replicated study are made available making it transparent and reproducible.
RESPONSE 13) Thank you. We have also updated the markdown file in accordance with the referee's other suggestions.
Minor comments:
L192-193: Were mother-pup pairs caught at the same time/on the same day/when together? If they were, the mother-offspring similarity found in the study could be a result of animal touching and transferring compounds from one another directly prior to capture and may not be a reflection of ‘stable’ similarity.
RESPONSE 14) The animals were captured when together, so this is indeed possible. However, it is not trivial to resample mother-pup pairs in this species because the pups move away from the beaches and into the tussock grass at a few weeks of age, making it necessary to radio-track the pups in order to relocate them. We hope to be able to look into this question in a follow-up study.
L228: Why was the PERMANOVA based on 99999 permutations? This seems like a lot. The default and what is used in most publications is a lot less. Did lower permutations result in insignificant values?
RESPONSE 15) We ran this number of permutations in order to obtain a more precise estimate of the lower bound of the p-value (i.e. over 10,000 permutations are required to test if a given p-value goes down to < 0.0001). To explore whether the use of a large number of permutations could have affected our results, we repeated these tests with 9999, 999 and 99 permutations. As expected, the corresponding p-values are all significant (see table below) although they are naturally constrained by the smaller number of permutations. We have added a statement to this effect in the relevant section of the methods.
Results for stepwise increase of permutations in PERMANOVAs for mother-offspring pair analysis
PERMANOVA 99 iterations
p-value
Age
0.02
Colony
0.01
Colony:family
0.01
PERMANOVA 999 iterations
Age
0.011
Colony
0.001
Colony:family
0.001
PERMANOVA 9999 iterations
Age
0.0045
Colony
0.0001
Colony:family
0.0001
PERMANOVA 99999 iterations
Age
0.00403
Colony
0.00001
Colony:family
0.00001
Results for stepwise increase of permutations in PERMANOVAs for a colony effect on scent similarity in six pup colonies
Iterations
p-values
99
0.01
999
0.001
9999
0.0001
99999
0.00001
Table 1c: It seems that you re-analysed the raw data of Stoffel et al. 2015 using a PERMANOVA. Please add this information to the text. Did you also realign the raw data using GCalignR (to keep it consistent with your analysis)?
RESPONSE 16) We did not realign the raw data of Stoffel et al, as explained in the main manuscript file (see also Response 17). However, the scripts used to analyse Stoffel et al's (2015) data are embedded within GCalignR and are therefore identical. Hence, our analyses are entirely consistent. We have now added this information to the methods section of the manuscript.
L237: It is not clear what you mean by ‘we did not re-align the dataframe of Stoffel et al.’. Do you mean that you did not use GCalignR and kept the original alignment? If so, please make this clear (as it is not stated that you realigned the data from the original paper for the PERMANOVA in Table 1c (see above)).
RESPONSE 17) As explained in Response 16, Stoffel et al. (2015) used the same scripts as us for aligning their dataframe (although at the time, they were not embedded within GCalignR). In the current paper, we used the original alignment of Stoffel at al. (2015) as opposed to realigning their data with the parameter values that we used for our current dataset. This is because every dataset has an optimal set of parameter values, so using the same parameter values for both datasets would result in a suboptimal alignment for one of the datasets [in this case, the data from Stoffel et al (2015)]. We have clarified this point in our revised methods section.
Supplementary materials: the code that was provided imports data with peaks that are already aligned. It would be useful to provide the code for aligning peaks in the supplementary materials (rather than just the github repository) or briefly describing how this was done in the text.
RESPONSE 18) This is correct. However, all of the parameters that were used as well as the output generated by the alignment procedure are exactly the same as specified in the alignment package. As this seems to have confused the reviewer, we have now updated the markdown file so that it contains a section for peak alignment.
Validity of the findings
As mentioned before, I agree with the authors that replications studies are important and needed. In the field of ecology, this may be particularly difficult, and I commend the authors’ effort of collecting an impressive dataset and conducting this analysis again. The rationale and benefit of this research is clearly stated and the methods are well developed and sound. The discussion is well explained and flows well.
RESPONSE 19) Many thanks.
I have only one major concern with the interpretation of the results – the significant result of the test for homogeneity of group variances (L274-277). This result could mean that the significant results of the PERMANOVA are actually an artefact of differences in within-group variability, rather than between group differences, which is a problem and theoretically puts the entire results into question. In lines 267-278 you mention a post hoc test – this is not included in the code in the supplementary materials. Considering these are the main results of the article, I would like to see more justification and support of why you think significant results of the betadisper analysis do not indicate issues and potential unreliability of the results of the PERMANOVA as well as explanation for considering the post hoc test results more reliable than the original test.
RESPONSE 20) This is an important point that warrants careful clarification:
(1) First of all, even if the significant results of the PERMANOVA were an artefact of differences in within-group variability (which we believe is not the case, based on a number of lines of evidence described below), this would not ‘put the entire results into question’ because this is only relevant to the PERMANOVA of mothers and offspring from the two colonies in the replication study (Table 1a). The test for homogeneity of variance was not significant in the PERMANOVA of pups from six colonies (Table 1b), nor in the PERMANOVA of the original data from Stoffel et al. (2015, Table 1c).
(2) We also do not believe that a single marginally significant p-value (p = 0.026) provides adequate grounds to cast all of our results into doubt, especially given this p-value was not corrected for the total number of tests performed. Unsurprisingly, correcting the p-values in Table 1 for the false discovery rate results in a corresponding p-value that is no longer significant. Consequently, we caution against over-interpreting a single marginally significant p-value. It only takes a cursory glance at the NMDS plots in Figure 2 to see that patterns of colony membership and mother-offspring similarity are very strong, explaining around 8 to 10 percent of the total chemical variance! (Figure 3).
(3) The code for the post-hoc test was actually present in the R markdown file that we supplied with our manuscript. However, we can appreciate how the mismatch between the original and the post-hoc test statistics might confuse the reader. Given that we cannot provide a strong justification for taking one result over another, we have therefore decided to omit the post-hoc test.
(4) Instead, we delved into greater detail by splitting the chemical data into four groups, corresponding to mothers and pups from SSB and FWB respectively, and then performing PERMANOVAs for all possible pairwise combinations of these groups. All of the pairwise PERMANOVAs involving groups from the two different colonies were highly significant after table-wide Bonferroni correction for multiple tests, indicating that the chemical encoding of colony membership is robust to age class. By contrast, pairwise PERMANOVAs involving mothers and pups within colonies were non-significant, indicating that the signal of mother-offspring similarity is robust and present in both colonies. Furthermore, tests for the homogeneity of group variances were not significant for any of the pairwise group comparisons after Bonferroni correction, suggesting that the original marginally significant p-value is probably a type-I error. We have re-written this part of the results section to include these new findings.
A few minor comments:
Since you included age in your model, you may consider running betadisper for scent_factor$age
RESPONSE 21) The results section now includes a statement about the significant effect of age on chemical differences and also about the homogeneity of group variance based on age of the seals, for which there is no significant difference (p = 0.23).
Figure 2B – There are a lot of colours and shapes making it difficult to find pairs. Please make this plot clearer, maybe by adding lines connecting mother-pup pairs. You could then merge 2A and 2B into one plot, by keeping the colour and shapes of 2A and connecting mother-pup pairs with lines.
RESPONSE 22) Our figure was carefully formulated to emphasise two important aspects of our results; panel A emphasises colony differences, while colony B emphasises patterns of mother-offspring similarity. We therefore prefer to present this information as two panels because, in our opinion, merging them is less effective at communicating the two patterns that we wish to emphasise.
At the reviewer's suggestion, we added lines connecting the points in panel 2B (see figure below). However, we do not feel this increases clarity, as the centre of the figure is far more cluttered as a result. In addition, the lines also emphasise the more distant rather than the more close pairs. We have therefore not changed the figure, although we would be happy to do so at the editor's discretion.
L291-296: As mentioned above, I think the replication study is enough for this publication and this section could be left out.
RESPONSE 23) We respectfully disagree, for the reasons outlined in Responses 2 and 7.
L382-387: This statement is too much of a reach. Just because mother-offspring similarity and colony differences persist does not mean that this is related by microbiota. While this may be true and your findings do not contradict this possibility, what you have shown in this study does not relate to this enough to bring this up, please remove.
RESPONSE 24) We originally felt that this was an interesting discussion point given the striking parallels of chemical and microbial data collected from different years. However, we have now removed this paragraph at the request of the reviewer.
L402-403. There are a lot of other factors that may play a role in this and there is no need to make broad speculations. Especially since the main goal of the publication was to replicate a study. Please remove.
RESPONSE 25) We have removed this sentence at the request of the reviewer
L405-425: This was not within the scope of the study and I do not see the need to discuss it here. Considering that the main body and framework of the study focuses on replication of results, the whole section on mechanisms is more appropriate in the discussion of the original paper.
RESPONSE 26) We do not agree that a discussion of the factors and mechanisms responsible for the observed patterns is beyond the scope of the study, particularly in the light of the evident curiosity of the second reviewer about the ‘drivers of chemical similarity or dissimilarity between colonies’ (see Response 33). We have therefore reshaped this part of the discussion to make the relevance of this paragraph more readily evident.
Reviewer 2 (Benjamin Pitcher)
Basic reporting
The manuscript is very well written and a pleasure to read. The authors have provided ample background, both of the issue of replication and of pinniped olfactory signatures. The article is well structured and the methods are clear. The results relate directly to the hypotheses and the conclusions are well supported.
RESPONSE 27) Many thanks.
Specific comments:
The reference to Ling on line 146 appears to have a typo in it.
RESPONSE 28) Thank you for spotting this error. We have corrected this to read 'Ling 1974'.
In the reference list, some of the binomial names are not italicised.
RESPONSE 29) We have corrected this.
Experimental design
The research question is well defined and very relevant. Chemical ecology is a developing field with many different methods being trialled by different labs and the reliability and comparability of those results have not yet been established. This study is an important step in showing that, at least using the methods of Stoffel et al., findings are robust and reproducible.
RESPONSE 30) We agree, and as far as we are aware of, ours is the first attempt to replicate a chemical study, outlining its importance for this field.
One additional detail I would request in the methods is the duration of time from sample collection to freezing, and the length of time samples were stored before analysis. This is not a critique of this study, but a detail that I think is useful to other researchers as the stability of chemical samples post collection is poorly understood and providing this information establishes a benchmark.
RESPONSE 31) We have included this information. The samples took about 1 hour until freezing after collection in the field. Total time of storage (from shipping to processing) of samples was about one and a half year in ethanol.
Validity of the findings
The authors have taken great care with the statistics in this study. The results are clearly presented and the conclusions are supported by the results.
RESPONSE 32) Many thanks, although we have now elaborated on some of the analyses at the request of reviewer 1 (see Responses 15 and 20).
Comments for the Author
Thank you for this very interesting paper. It provides both important findings about pinniped chemical communication, and about methodological approaches. I find it interesting that Johnsons Cove and Natural Arch were similar to each other (as they are the furthest apart), while LB, SSB and FWB were all different yet close to each other. It is probably beyond the speculation of this paper, but it will be interesting to see more investigation in the future into the drivers of similarity (or diversity) between colonies and individuals.
RESPONSE 33) Thanks for raising this point, which is indeed interesting. Given that there is no clear correspondence between geography and odour similarity, and that animals from these colonies are not genetically differentiated, the most obvious explanation for these results is that differences in certain environmental factors such as temperature, wind or solar radiation play a role. These may either influence odour directly or via an indirect link involving the microbiome. We have therefore included a short discussion of the point raised by the second reviewer, and linked this in turn to the paragraph that the first reviewer wanted us to remove (see Response 26), which outlines the factors and mechanisms that might be responsible for shaping these chemical patterns. This both highlights an interesting result that might otherwise be overlooked by some readers, and strengthens the argument for leaving the mechanisms paragraph in the discussion.
Given this study was a replication of Stoffel et al, it isn't appropriate to introduce new methods, but with the potential environmental influence in mind, have you considered testing environmental control samples? Wierucka et al., used environmental controls in an attempt to remove some of the contribution of other odour sources in the colony. Given these methods are relatively new, it would be good to determine if such controls are necessary, and/or if they contribute to our understanding of the sources of diversity among colonies. Again, this is probably beyond this paper, but a potential avenue of future research.
RESPONSE 34) Thank you for this suggestion. We agree that it could be informative to collect and analyse standardised environmental control samples in future studies and will endeavour to do this.
References cited:
Goodman SN, Fanelli D, Ioannidis JPA. 2016. What does research reproducibility mean? Science translational medicine 8.
Dirnagl U. 2019. Rethinking research reproducibility. The EMBO Journal 38:2018–2020. DOI: 10.15252/embj.2018101117.
Piper SK, Grittner U, Rex A, Riedel N, Fischer F, Nadon R, Siegerink B, Dirnagl U. 2019. Exact replication: foundation of science or game of chance? PLoS Biology 17:1–9. DOI: 10.1371/journal.pbio.3000188.
Schmidt S. 2009. Shall we really do it again? The powerful concept of replication is neglected in the social sciences. Review of general psychology 13:90–100.
" | Here is a paper. Please give your review comments after reading it. |
9,772 | "<ns0:abstract xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:p>At the turn of February and March 2020, COVID-19 pandemic reached Europe. Many countries, including Poland imposed lockdown as a method of securing social distance between potentially infected. Stay-at-home orders and movement control within public space not only affected the touristm industry, but also the everyday life of the inhabitants.</ns0:p><ns0:p>The hourly time-lapse from four HD webcams in Cracow (Poland) are used in this study to estimate how pedestrian activity changed during COVID-19 lockdown. The collected data covers the period from June 9, 2016 to April 19, 2020 and comes from various urban zones. One zone is tourist, one is residential and two are mixed. In the first stage of the analysis, state-of-the-art machine learning algorithm (YOLOv3) is used to detect people.</ns0:p><ns0:p>Additionally, a non-standard application of the YOLO method is proposed, oriented to the images from HD webcams. This approach (YOLOtiled) is less prone to pedestrian detection errors with the only drawback being the longer computation time. Splitting the HD image into smaller tiles increases the number of detected pedestrians by over 50%. In the second stage, the analysis of pedestrian activity before and during the COVID-19 lockdown is conducted for hourly, daily and weekly averages. Depending on the type of urban zone, the number of pedestrians decreased from 33% in residential zones to 85% in tourist zones located in the Old Town. The presented method allows for more efficient detection and counting of pedestrians from HD time-lapse webcam images compared to SSD, YOLOv3 and Faster R-CNN. The result of the research is a published database with the detected number of pedestrians from the four-year observation period for four locations in Cracow.</ns0:p></ns0:div>
</ns0:abstract>
<ns0:body xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:head>INTRODUCTION 30</ns0:head><ns0:p>The COVID-19 pandemic that appeared in Europe in early 2020 has a major impact on societies around operation of markets were introduced on March 31.</ns0:p><ns0:p>Patterns of human activity in the urban environment depend on several factors, such as e.g. night lighting <ns0:ref type='bibr' target='#b55'>(Wang et al., 2019)</ns0:ref>, but the impact of formal restrictions on movement in public space is rarely considered and analyzed. In addition to public video surveillance systems, there are also private video monitoring systems that can also be used to detect and count people. Some of them have been operating continuously for several years, enabling comparative studies with pre-pandemic periods.</ns0:p><ns0:p>The study has two main goals: to evaluate the YOLOv3 people detection algorithm on images from HD webcams and application of YOLOv3 to assess changes in pedestrian activity in public space before and during COVID-19 lockdown in Cracow, based on the hourly webcam time-lapse. <ns0:ref type='bibr' target='#b1'>Wellenius et al. (2020)</ns0:ref> used anonymous and aggregated mobility data <ns0:ref type='bibr' target='#b1'>(Aktay et al., 2020)</ns0:ref> to assess social distance in the United States during COVID-19. The impact of the social distance order was very different in each state, from a 36% drop in displacement New Jersey to a 12% drop in Louisiana. The most effective ban was to impose restrictions on the work of bars and restaurants, which was associated with a 25.8% reduction in people's activity. <ns0:ref type='bibr' target='#b1'>Wellenius et al. (2020)</ns0:ref> concludes that public procurement seems to be very effective in encouraging people to stay at home to minimize the risk of COVID-19 transmission.</ns0:p></ns0:div>
<ns0:div><ns0:head>Social distance during COVID-19</ns0:head><ns0:p>In the case of Poland, in COVID-19 Community Mobility Report <ns0:ref type='bibr'>(March 29, 2020)</ns0:ref>, mobility trends in places such as restaurants, cafes, shopping centers, theme parks and museums fell by 78%. In the case of the Lesser Poland Voivodship in which Cracow is located, this decrease is 84% <ns0:ref type='bibr' target='#b1'>(Aktay et al., 2020)</ns0:ref>. Social behavior has a fundamental impact on the dynamics of the spread of infectious diseases <ns0:ref type='bibr' target='#b43'>(Prem et al., 2017)</ns0:ref>. Inhabitants of larger Polish cities are more afraid of overcrowded hospitals and inefficient healthcare than small towns and villages <ns0:ref type='bibr' target='#b22'>(Jarynowski et al., 2020)</ns0:ref>.</ns0:p><ns0:p>The Center for Science and Systems Engineering (CSSE) at Johns Hopkins University provides daily data updates via COVID-19 Data Repository <ns0:ref type='bibr' target='#b13'>(Dong et al., 2020)</ns0:ref>. The first confirmed cases of in Italy and Spain were identified at the end of February 2020 <ns0:ref type='bibr' target='#b51'>(Saglietto et al., 2020)</ns0:ref>. The lockdown has been widely used in Italy since March 8 and in Spain since March 16. Restrictions on citizens' mobility have reduced disease transmission in both countries <ns0:ref type='bibr' target='#b54'>(Tobías, 2020)</ns0:ref>. <ns0:ref type='bibr' target='#b7'>Chinazzi et al. (2020)</ns0:ref> findings indicate that 90% of travel restrictions to and from mainland China only modestly affect the epidemic trajectory unless combined with a 50% or higher reduction of transmission in the community. <ns0:ref type='bibr' target='#b15'>Fang et al. (2020)</ns0:ref> uses the crowd flow model for virus transmission to simulate the spread of the virus caused by close contact during pedestrian traffic. Mobility restrictions are important <ns0:ref type='bibr' target='#b2'>(Arenas et al., 2020;</ns0:ref><ns0:ref type='bibr' target='#b17'>Ferguson et al., 2020)</ns0:ref> or sometimes crucial <ns0:ref type='bibr' target='#b39'>(Mitjà et al., 2020)</ns0:ref>, but as shown by <ns0:ref type='bibr' target='#b38'>Mello (2020)</ns0:ref> the number of people crossing each other can be drastically reduced if one-way traffic is enforced and runners are separated from walkers. To properly quantify the transmission of an epidemic, the spatial distribution of potential disease hazards (e.g. crowd) should be assessed <ns0:ref type='bibr' target='#b40'>(Ng and Wen, 2019;</ns0:ref><ns0:ref type='bibr' target='#b15'>Fang et al., 2020)</ns0:ref>. Webcams can be a potential source of such information.</ns0:p></ns0:div>
<ns0:div><ns0:head>People detection</ns0:head><ns0:p>Object detection is one of the rapidly growing areas of computer vision. Proper detection of people is crucial for autonomous cars, advertising planning and many other industries and public safety. <ns0:ref type='bibr' target='#b24'>Kajabad and Ivanov (2019)</ns0:ref> proposed a method of finding areas more attractive to customers (hot zones) based on people detection. Sometimes, people must be detected in a heavy industry environment <ns0:ref type='bibr' target='#b61'>(Zengeler et al., 2019)</ns0:ref> or in hazy weather <ns0:ref type='bibr' target='#b27'>(Li et al., 2019)</ns0:ref>. A lot of research is being done to detect objects in a variety of environments, but this is not just about detecting people. Computer vision methods are used to count species in environmental research: 1 minute time-lapse for fish passage and abundance in streams <ns0:ref type='bibr' target='#b9'>(Deacy et al., 2016)</ns0:ref>, or 5 minute time-lapse for bears counting <ns0:ref type='bibr' target='#b10'>(Deacy et al., 2019)</ns0:ref>. There are two main approaches to detecting a person or other object in the image. The first approach is based on computer vision techniques, the second on deep learning algorithms. Comprehensive survey on computer vision and deep learning techniques for pedestrian detection and tracking is presented by <ns0:ref type='bibr' target='#b4'>Brunetti et al. (2018)</ns0:ref>.</ns0:p></ns0:div>
<ns0:div><ns0:head>Computer vision</ns0:head><ns0:p>Traditional pedestrian detectors have been known for over two decades. They are based on the representation of the features of objects obtained from computer vision. <ns0:ref type='bibr' target='#b42'>Oren et al. (1997)</ns0:ref> proposed the use of Haara waves in 1997, and Maliniowski in 2005 the use of the Oriented Gradient (HOG) Histogram. Also Local Binary Patterns (LBP) <ns0:ref type='bibr' target='#b41'>(Ojala et al., 2002)</ns0:ref> can be used for pedestrian detection <ns0:ref type='bibr' target='#b65'>(Zheng et al., 2010)</ns0:ref>. Among them, HOG and its variations are considered the most successful hand-engineered features for pedestrian detection <ns0:ref type='bibr' target='#b35'>(Liu et al., 2016a</ns0:ref><ns0:ref type='bibr' target='#b33'>(Liu et al., , 2019b))</ns0:ref>. For visual surveillance applications, background subtraction method can also be used <ns0:ref type='bibr' target='#b37'>(Maddalena and Petrosino, 2008)</ns0:ref>. In hybrid implementation of computer vision methods, pedestrian detection on the basis of 2D/3D LiDAR data and visible images of the same scene are applied <ns0:ref type='bibr' target='#b20'>(Hasfura, 2016;</ns0:ref><ns0:ref type='bibr' target='#b14'>El Ansari et al., 2018)</ns0:ref>.</ns0:p></ns0:div>
<ns0:div><ns0:head>Deep learning</ns0:head><ns0:p>In recent years, several convolutional neural networks (CNN) models for object detection have been proposed <ns0:ref type='bibr' target='#b49'>(Ren et al., 2018)</ns0:ref>: R-CNN in 2014, Fast R-CNN in 2015 and Faster R-CNN in 2015. These two-step detection algorithms divide the problem into two stages: (i) generating region proposals and (ii) classification of candidate regions. But these traditional deep learning algorithms suffer from low speed <ns0:ref type='bibr' target='#b24'>(Kajabad and Ivanov, 2019)</ns0:ref>. To overcome this limitation, <ns0:ref type='bibr' target='#b46'>Redmon et al. (2016)</ns0:ref> proposed a one-step detection algorithm called YOLO (You Only Look Once), enabling easy implementation end-to-end object detection. Further improvements of this algorithm are known as YOLO9000 (or YOLOv2) <ns0:ref type='bibr' target='#b47'>(Redmon and Farhadi, 2017)</ns0:ref> and YOLOv3 <ns0:ref type='bibr' target='#b48'>(Redmon and Farhadi, 2018)</ns0:ref>. Second popular one-stage algorithms is RetinaNet <ns0:ref type='bibr' target='#b30'>(Lin et al., 2017)</ns0:ref>. It deals with the problem of the extreme foreground-background class imbalance encountered during the training of dense detectors and proposes a new solution to this problem.</ns0:p><ns0:p>Third detection algorithm is Single Shot MultiBox Detector (SSD). The core of SSD is predicting category scores and box offsets for a fixed set of default bounding boxes using small convolutional filters applied to feature maps <ns0:ref type='bibr' target='#b36'>(Liu et al., 2016b)</ns0:ref>. To improve model performance for small objects, SSD applies additional data augmentation strategy. All three algorithms achieve state-of-the-art speed and accuracy <ns0:ref type='bibr' target='#b61'>(Zengeler et al., 2019)</ns0:ref>, so they can be used in real-time applications. The CNN-based approaches provide significant improvements over traditional approaches across all datasets <ns0:ref type='bibr' target='#b52'>(Sindagi and Patel, 2018)</ns0:ref>.</ns0:p><ns0:p>The YOLO model applies a single neural network to the complete image. It looks at the whole image at test time so its predictions are based on the global context in the image. This network divides the image into regions and predicts bounding boxes and probabilities for each region. These bounding boxes are weighted by the predicted probabilities. YOLOv3 predicts an objectness score for each bounding box using logistic regression <ns0:ref type='bibr' target='#b48'>(Redmon and Farhadi, 2018)</ns0:ref>. During training, the binary cross-entropy loss is used for class predictions. The original YOLO model trains the classifier network at 224 × 224 and increases the resolution to 448 × 448 for detection <ns0:ref type='bibr' target='#b47'>(Redmon and Farhadi, 2017)</ns0:ref>. Backbone for YOLOv3 is Darknet-53 network, and standard image sizes are 320 × 320. The Darknet-53 network is composed of 53 consecutive 3 × 3 and 1 × 1 convolutional layers. YOLOv3 makes detection at three different scales downsampling the dimensions of the input image by 32, 16, and 8. Darknet architecture is a pre-trained model for classifying 80 different classes. Several improvements to the YOLO model have been proposed for detecting people <ns0:ref type='bibr' target='#b45'>(Putra et al., 2017</ns0:ref><ns0:ref type='bibr' target='#b44'>(Putra et al., , 2018;;</ns0:ref><ns0:ref type='bibr' target='#b25'>Lan et al., 2018;</ns0:ref><ns0:ref type='bibr' target='#b21'>He et al., 2019;</ns0:ref><ns0:ref type='bibr' target='#b27'>Li et al., 2019)</ns0:ref>, but even standard YOLOv3 outperforms traditional computer vision methods and most of deep neural network methods <ns0:ref type='bibr' target='#b18'>(Ghosh and Das, 2019;</ns0:ref><ns0:ref type='bibr' target='#b61'>Zengeler et al., 2019;</ns0:ref><ns0:ref type='bibr' target='#b24'>Kajabad and Ivanov, 2019;</ns0:ref><ns0:ref type='bibr' target='#b60'>Yun et al., 2018)</ns0:ref>.</ns0:p><ns0:p>YOLOv3 achieves 93.8% top-5 score on the COCO dataset <ns0:ref type='bibr' target='#b48'>(Redmon and Farhadi, 2018)</ns0:ref>.</ns0:p><ns0:p>Pre-trained networks for standard image sizes are available in several repositories, enabling fast and relatively easy application of YOLO model. Squeeze YOLO-based People Counting (S-YOLO-PC) proposed by <ns0:ref type='bibr' target='#b49'>Ren et al. (2018)</ns0:ref> can detect and count people with 41 frames per second (FPS) with the Average Precision (AP) of 72%. <ns0:ref type='bibr' target='#b16'>Feng et al. (2019)</ns0:ref> reports YOLOv2 mean Average Precision (mAP) of 76.8%, which is very close to 78.6% reported by authors of YOLO method <ns0:ref type='bibr' target='#b47'>(Redmon and Farhadi, 2017)</ns0:ref>.</ns0:p><ns0:p>However, even the latest version of YOLOv3 has some limitations. If there are two anchor boxes but three objects in the same grid cell, it does not support them correctly, which ultimately leads to missing objects <ns0:ref type='bibr' target='#b24'>(Kajabad and Ivanov, 2019)</ns0:ref>. YOLO achieves about 10% missing detection rate for pedestrian detection <ns0:ref type='bibr' target='#b25'>(Lan et al., 2018)</ns0:ref>. <ns0:ref type='bibr' target='#b60'>Yun et al. (2018)</ns0:ref> reports that YOLOv3 default architecture achieves the mAP of 42.7%.</ns0:p></ns0:div>
<ns0:div><ns0:head>Pedestrian detection</ns0:head></ns0:div>
<ns0:div><ns0:head>Scale problem</ns0:head><ns0:p>Robustly detecting pedestrians with a large variant on sizes and with occlusions remains a challenging problem <ns0:ref type='bibr'>(Liu et al., 2019a,b)</ns0:ref>. Pedestrian detection is limited by image resolution and complexity of the background scene. Effective detector should be able to detect people at different scales. <ns0:ref type='bibr' target='#b31'>Liu et al. (2018)</ns0:ref> presents a method where a large-size pedestrian should be represented by features from deep layers, whereas a small-size pedestrian should be represented by features from shallow layers which are of higher resolutions. <ns0:ref type='bibr' target='#b32'>Liu et al. (2019a)</ns0:ref> propose gated feature extraction framework consisting of</ns0:p></ns0:div>
<ns0:div><ns0:head>3/17</ns0:head><ns0:p>PeerJ reviewing PDF | (2020:05:48857:1:1:NEW 10 Aug 2020)</ns0:p><ns0:p>Manuscript to be reviewed squeeze units, gate units and a concatenation layer which perform feature dimension squeezing, feature elements manipulation and convolutional features combination from multiple CNN layers. The Faster R-CNN is also used as benchmark for detecting occluded pedestrians with results comparable to fine tuned models <ns0:ref type='bibr' target='#b33'>(Liu et al., 2019b;</ns0:ref><ns0:ref type='bibr' target='#b64'>Zhang et al., 2018)</ns0:ref>. However, <ns0:ref type='bibr' target='#b29'>Lin et al. (2020)</ns0:ref> and <ns0:ref type='bibr' target='#b62'>Zhang et al. (2016)</ns0:ref> state that convolutional feature maps of the this classifier are of low resolution for detecting small objects.</ns0:p><ns0:p>Evaluation of average precision and the tuning of the model is usually limited to objects in the 50-100 m range, as in the CityScapes Dataset for Semantic Urban Scene Understanding <ns0:ref type='bibr' target='#b8'>(Cordts et al., 2016)</ns0:ref>. In research by <ns0:ref type='bibr' target='#b12'>Dollar et al. (2011)</ns0:ref> pedestrians represented by 30 pixels or less are treated as distant objects.</ns0:p><ns0:p>Main source of false negative classification according to <ns0:ref type='bibr' target='#b63'>Zhang et al. (2017)</ns0:ref> is mainly the small scale, therefore the authors only consider pedestrians with a height of more than 30 pixels. Scale oriented models such as Scale-Aware Fast R-CNN are developed <ns0:ref type='bibr' target='#b28'>(Li et al., 2017)</ns0:ref>, but even in this case, small objects are about 50 pixels high.</ns0:p></ns0:div>
<ns0:div><ns0:head>Crowd counting</ns0:head><ns0:p>Next issue in urban space or during mass events is the crowd. There are mainly three types of methods to count the number of people in the crowd from video <ns0:ref type='bibr' target='#b49'>(Ren et al., 2018)</ns0:ref>: (i) statistical method to estimate the number of people in a region, (ii) combination of object detection with object tracking and (iii) use of path information of the points, with subsequent cluster analysis of the feature point path. People detection in crowded spaces is the most challenging task, because of the people occlusions <ns0:ref type='bibr' target='#b53'>(Stewart et al., 2016;</ns0:ref><ns0:ref type='bibr' target='#b24'>Kajabad and Ivanov, 2019)</ns0:ref>. Crowd counting requires development of new methods <ns0:ref type='bibr' target='#b53'>(Stewart et al., 2016;</ns0:ref><ns0:ref type='bibr' target='#b26'>Lei et al., 2020)</ns0:ref> like Dynamic Region Division <ns0:ref type='bibr' target='#b21'>(He et al., 2019)</ns0:ref>. <ns0:ref type='bibr' target='#b58'>Yang et al. (2020)</ns0:ref> proposes counting crowds using a scale-distribution-aware network and adaptive human-shaped kernel. Existing crowd counting methods require object location-level annotation or weaker annotations that only know the total count of objects <ns0:ref type='bibr' target='#b26'>(Lei et al., 2020)</ns0:ref>. <ns0:ref type='bibr' target='#b5'>Cheng et al. (2020)</ns0:ref> proposed an FFPM model for the pedestrian detection using body parts (head, shoulders, hands, knees and feet) followed by full body boosting model and a classification layer. This approach works well in a crowded environment with partially obscured pedestrians. Model proposed by <ns0:ref type='bibr' target='#b23'>Jiang et al. (2019)</ns0:ref> combines the classic computer vision approach (HOG+LBP features) with the GA-XGBoost deep learning algorithm.</ns0:p></ns0:div>
<ns0:div><ns0:head>MATERIALS & METHODS</ns0:head></ns0:div>
<ns0:div><ns0:head>Study area</ns0:head><ns0:p>Images from webcams are collected in Cracow, the second largest city in the country and the capital of the Lesser Poland Voivodeship. Cracow is divided into the medieval Old Town, located in the center and the surrounding residential and industrial zones. The Vistula, the largest river in Poland, flows through the city center.</ns0:p><ns0:p>Two of the webcams are located on the Royal Road, going from Wawel Castle through Main Square to north of the city. These webcams are named All Saints Square and Grodzka (Fig. <ns0:ref type='figure' target='#fig_0'>1</ns0:ref>). Grodzka Street has a tourist character, and All Saints Square, being in the tourist zone as shown in Table <ns0:ref type='table' target='#tab_0'>1</ns0:ref>, is also an important communication point in the city. The third webcam (Wawel Castle) is located in the tourist/residential zone. Parking for tourists visiting Wawel Royal Castle is adjacent to the riverside promenade, which is used by residents. The fourth webcam (Podgorze Market Square) is a typical residential zone located on the other side of the Vistula river (Fig. <ns0:ref type='figure' target='#fig_0'>1</ns0:ref>).</ns0:p><ns0:p>Due to the medieval nature of the area, cameras from Royal Road have a very narrow field of view.</ns0:p><ns0:p>The webcam on All Saints Square is located on a small square, so in fact most of the visible pedestrian area belongs to Grodzka Street. This webcam is also in the lowest position among all four, enabling easier detection of pedestrians due to the short distance from the detected objects. The Wawel Castle webcam with probably the most beautiful view from all Cracow webcams has the largest distance to detected pedestrians and is the highest mounted webcam (the sixth floor).</ns0:p><ns0:p>All webcams, shown in Table <ns0:ref type='table' target='#tab_0'>1</ns0:ref>, are publicly available and broadcast live via dedicated websites, but access to the ad-free version is limited due to the commercial nature of the service. Webcamera.pl is probably one of the largest providers of streaming cameras in Poland, with a long history and almost 350 webcams located all over Poland. To make detection results comparable, webcams with a moving field of view were excluded from the analysis, although they are in very good locations, such as the Main Square, the largest medieval town square in Europe (https://krakow4.webcamera.pl/).</ns0:p></ns0:div>
<ns0:div><ns0:head>4/17</ns0:head><ns0:p>PeerJ reviewing PDF | (2020:05:48857:1:1:NEW 10 Aug 2020)</ns0:p><ns0:p>Manuscript to be reviewed </ns0:p></ns0:div>
<ns0:div><ns0:head>Webcam time-lapse</ns0:head><ns0:p>Webcam time-lapse is made and downloaded every hour (Fig. <ns0:ref type='figure' target='#fig_1'>2</ns0:ref>), directly from www.webcamera.pl provider. In this study, approximately 33,800 images were collected and used for each webcam from June 9, 2016 to April 19, 2020. The analysis is based on 1,412 days (201 weeks) of continuous observation.</ns0:p><ns0:p>The total size of the set of hourly time-lapse images for four webcams in this period exceeds 10 GB.</ns0:p></ns0:div>
<ns0:div><ns0:head>Methods</ns0:head><ns0:p>In the first part of these studies, only the problem of the size of the object (person) is considered. Due to the convenient location of the webcams (between the first and sixth floors) and low to moderate density of pedestrians, crowd counting methods can be omitted.</ns0:p><ns0:p>The standard Darknet-53 architecture with the YOLOv3 model is used as the main pedestrian detector.</ns0:p><ns0:p>The interface to the model is built in the Python 3 programming language, in the main script yolo count.py.</ns0:p><ns0:p>All code is available in the public repository (https://gitlab.com/Cracert/pedestrian-count-covid-19) under the MIT license. The OpenCV library <ns0:ref type='bibr' target='#b3'>(Bradski, 2000)</ns0:ref> with built-in support for the Darknet architecture is used in these studies as a machine learning platform. From the pre-trained Darknet architecture, only the first 9 classes (from 80) are saved during calculations. These classes (person, bicycle, car, motorcycle,</ns0:p></ns0:div>
<ns0:div><ns0:head>5/17</ns0:head><ns0:p>PeerJ reviewing PDF | (2020:05:48857:1:1:NEW 10 Aug 2020)</ns0:p><ns0:p>Manuscript to be reviewed The pre-trained YOLOv3 model weights for people detection are available for direct use, so the training phase can be omitted (https://pjreddie.com/media/files/yolov3.weights). One of the standard image resolutions for training is 416 × 416, while the source HD webcam image resolution used in this study is 1280 × 720. As a first approach, YOLOv3 is applied directly to the collected images (Fig. <ns0:ref type='figure' target='#fig_1'>2</ns0:ref>).</ns0:p><ns0:p>The second approach assumes that the right image ratio can improve the average precision of the model.</ns0:p><ns0:p>The input images are divided into 6 almost square tiles 426 × 360 (Fig. <ns0:ref type='figure' target='#fig_2'>3</ns0:ref>). This reduces the image ratio from 1.78 (Fig. <ns0:ref type='figure' target='#fig_1'>2</ns0:ref>) to 1.18 (Fig. <ns0:ref type='figure' target='#fig_2'>3</ns0:ref>) and makes the image's proportions more similar to the training data set.</ns0:p><ns0:p>In addition, not all tiles contain pedestrian areas, so some tiles can be omitted in the calculation, which significantly reduces detection time.</ns0:p></ns0:div>
<ns0:div><ns0:head>6/17</ns0:head><ns0:p>PeerJ reviewing PDF | (2020:05:48857:1:1:NEW 10 Aug 2020)</ns0:p></ns0:div>
<ns0:div><ns0:head>Manuscript to be reviewed</ns0:head><ns0:p>The workflow in the YOLO tiled model can be described with the following steps:</ns0:p><ns0:p>(i) Split the source image into square tiles of a size similar to the image size when training the model.</ns0:p><ns0:p>(ii) For further calculations, only select the tiles in the potential pedestrian zone. Tiles containing only buildings or sky can be excluded.</ns0:p><ns0:p>(iii) Run the YOLO model on each selected tile and aggregate the results. In fact, this method can be applied to any model, not just YOLO.</ns0:p><ns0:p>Number of detected pedestrians is saved in data folder as CSV files. Each file contains a header with the main detected classes and data containing a timestamp (day and hour) with the corresponding number of detected objects. One row corresponds to one hour time-lapse. Further analysis and visualization takes place in Jupyter notebooks using the pandas library. For the purposes of this article, the YOLOv3 method will be named YOLO from this place, and the method of splitting one HD webcam image into six tiles will be named YOLO tiled .</ns0:p><ns0:p>Model performance verification is based on two additional, state-of-the-art deep learning models (SSD, Faster R-CNN), implemented using the GluonCV framework <ns0:ref type='bibr' target='#b19'>(Guo et al., 2020)</ns0:ref>. Pre-trained Resnet50 VOC network was used in both models. Ground truth data was prepared by manually counting pedestrians on webcam hourly snapshots in March 2020 (almost 3,000 images in total). The calculation were made on an AMD Ryzen 5 2600X, 16 GB RAM, running 64 bit GNU/Linux Mint 19.1 system, with the use of CPU only, without use of the GPU. Model performance is assessed by evaluating Mean Absolute Error (MAE) and Root Mean Squared Error (RMSE):</ns0:p><ns0:formula xml:id='formula_0'>MAE = 1 n n−1 ∑ i=0 |y i − ŷi | (1) RMSE = 1 n n−1 ∑ i=0 (y i − ŷi ) 2 (2)</ns0:formula><ns0:p>In both formulas, n is the number of observations (images), y i is the actual number of pedestrians, and ŷi is the predicted number of pedestrians from one webcam image.</ns0:p><ns0:p>The comparison of the YOLO and YOLO tiled methods is based on statistical analysis. The number of pedestrians detected from each time-lapse (hour) enables the identification of extreme and mean differences between the two methods. Cases of extreme differences are examined manually to find problems associated with each method. In addition, the sum of detected pedestrians for the webcam over the entire period is used to detect the overall relative difference between YOLO and YOLO tiled . In the second part of the study, a better method was used to assess the change in pedestrian numbers before and during COVID-19.</ns0:p><ns0:p>The image data provider returns the last image, so if the webcam fails, the same last recorded image is returned, resulting in a constant number of pedestrians over time. By analyzing such anomalies, you can determine the dates of webcam malfunction. The verification of the source image data based on the pedestrian number change analysis can be replicated in the supplied Jupyter notebook (analysispedestrians.ipynb). Doubtful periods are excluded from further analysis.</ns0:p><ns0:p>The webcam observation time is divided into (i) before the COVID-19 period, June 9, 2016 -March 13, 2020 (1,374 days / 196 weeks) and (ii) during the COVID-19 period, March 13, 2020 -April 19, 2020 (38 days / 5 weeks). The number of detected pedestrians for these two periods is aggregated into days and weeks using mean values. This makes it easier to visualize trends and generalize results. The mean number of pedestrians from the hourly period before and during COVID-19 is used for the final evaluation. A change in this value corresponds to changes in pedestrian activity over time. It is assumed that hourly snapshots (time-lapse) from webcams are representative for evaluation of relative change in pedestrian activity. However, the method presented is not suitable for determining the absolute number of pedestrians traveling through the analyzed area.</ns0:p></ns0:div>
<ns0:div><ns0:head>7/17</ns0:head><ns0:p>PeerJ reviewing PDF | (2020:05:48857:1:1:NEW 10 Aug 2020)</ns0:p><ns0:p>Manuscript to be reviewed</ns0:p></ns0:div>
<ns0:div><ns0:head>RESULTS</ns0:head><ns0:p>The overall results of pedestrian detection performance by the three deep learning models and the proposed YOLO tiled model are presented in The mean number of detected pedestrians per image by YOLO tiled model depends on the type of urban zone, with 16.6 pedestrians in the tourist zone and 0.9 pedestrians in the residential zone. The mean difference of detected pedestrians between the two methods is the same, with values exceeding 4.4 in the tourist zone and below 0.4 in the residential zone.</ns0:p><ns0:p>Opposite cases are also reported when YOLO tiled detects fewer pedestrians, but in this case the absolute difference does not exceed 12 pedestrians, as shown in Table <ns0:ref type='table' target='#tab_4'>3</ns0:ref>. The maximum number of detected pedestrians also corresponds to the location of the webcam. Tourist locations in the Old Town (All Saints Square and Grodzka) record up to 80 pedestrians in one image (Fig. <ns0:ref type='figure' target='#fig_3'>4</ns0:ref>), and in residential zones below 35. Simply cutting one large image into six smaller tiles significantly increases the number of correctly detected pedestrians (Fig. <ns0:ref type='figure' target='#fig_3'>4</ns0:ref>). The detection range also increases, but a certain pedestrian detection threshold is clearly visible at the horizontal cutting height of the tiles.</ns0:p><ns0:p>The total sum of detected pedestrians over the entire period (almost 4 years) is from about 4,000 for YOLO from Wawel Castle webcam to over 500,000 for YOLO tiled from All Saints Square. The relative</ns0:p></ns0:div>
<ns0:div><ns0:head>8/17</ns0:head><ns0:p>PeerJ reviewing PDF | (2020:05:48857:1:1:NEW 10 Aug 2020)</ns0:p><ns0:p>Manuscript to be reviewed detection differences between YOLO and YOLO tiled are significant and range from 52% on All Saints</ns0:p><ns0:p>Square to 302% at Wawel Castle. This difference is proportional to the mean distance from pedestrians, as shown in Table <ns0:ref type='table' target='#tab_0'>1</ns0:ref>.</ns0:p><ns0:p>Over long distances YOLO tiled can detect significantly more pedestrian than YOLO. An example of such a case is shown on results from All Saints Square (Fig. <ns0:ref type='figure' target='#fig_3'>4</ns0:ref>) and from Wawel Castle webcam (Fig. <ns0:ref type='figure' target='#fig_4'>5</ns0:ref>).</ns0:p><ns0:p>Wawel Castle webcam has the longest distance from pedestrians, from about 50 m to about 400 m. Also in this case, the detection range of pedestrians does not exceed about 200 m. Pedestrians in Figure <ns0:ref type='figure' target='#fig_4'>5C</ns0:ref>, near the detected boat at the upper part, are also not recognized. Manuscript to be reviewed</ns0:p><ns0:p>The histogram of the difference in pedestrian detection Y OLO tiled − Y OLO is asymmetrical (Fig. <ns0:ref type='figure'>6</ns0:ref>). This also applies to other webcams. The difference of zero is dominant for all webcams, but mean value of 5.70 for the All Saints Square camera, as shown in Table <ns0:ref type='table' target='#tab_4'>3</ns0:ref>, compared to the extreme number of detected pedestrians on one image in the range of 50-80 makes this difference significant. As a result, the YOLO tiled method is selected as a better representation of the actual number of pedestrians on webcam time-lapse. With the awareness that this value is also underestimated in relation to the actual number of pedestrians on one image. Assuming that the detection range for both methods is constant (YOLO and YOLO tiled ), this should not significantly affect the estimation of the relative change in the number of pedestrians.</ns0:p><ns0:p>Figure <ns0:ref type='figure'>6</ns0:ref>. The difference in the number of pedestrians detected (YOLO tiled -YOLO) for All Saints Square webcam.</ns0:p><ns0:p>The YOLO tiled method, which is a better detector, is used as the basis for the second part of research related to estimating pedestrian activity before and during COVID-19. Data analysis enabled the identification of periods during which camera malfunction or camera data transmission was highly likely.</ns0:p><ns0:p>These periods were removed from the dataframe and treated as no data. A detailed analysis with relevant comments can be found in analysis-pedestrians.ipynb Jupyter notebook. A few short periods have been removed from the dataframe for All Saints Square and Grodzka webcams (Fig. <ns0:ref type='figure' target='#fig_5'>7A</ns0:ref>). Another problem was identified in Podgorze Market Square, where two periods are characterized by significantly different average values of detected pedestrians. It was found that in mid-2019 the horizontal angle of the webcam was changed, which changed the field of view. In order to maintain the possibility of comparison with the current period (COVID-19), it was decided to abandon the first part of the dataframe (Fig. <ns0:ref type='figure' target='#fig_5'>7B</ns0:ref>).</ns0:p><ns0:p>The high temporal variability of hourly data makes it difficult to visualize the result. For this reason, daily and weekly data aggregation is used for visual analysis.</ns0:p><ns0:p>Figure <ns0:ref type='figure' target='#fig_5'>7</ns0:ref> contains plots of weekly averages for two types of zones in Cracow. The seasonal cycle in the tourist zone is associated with the summer season, while in a residential zone this seasonal cycle is not visible. In the tourist zone, the mean number of pedestrians does not fall below one person per image (logarithmic scale in Fig. <ns0:ref type='figure' target='#fig_5'>7</ns0:ref>), while in the residential zone the level is lower by an order of magnitude.</ns0:p><ns0:p>There are no visible trends in the number of pedestrians during these four years, but the COVID-19 lockdown is clearly visible in the last weeks of the analyzed period in Figure <ns0:ref type='figure' target='#fig_5'>7</ns0:ref>. The quantitative analysis of this change is presented in the Table <ns0:ref type='table' target='#tab_5'>4</ns0:ref>. Data from the Wawel Castel webcam are more difficult to</ns0:p></ns0:div>
<ns0:div><ns0:head>10/17</ns0:head><ns0:p>PeerJ reviewing PDF | (2020:05:48857:1:1:NEW 10 Aug 2020)</ns0:p><ns0:p>Manuscript to be reviewed </ns0:p></ns0:div>
<ns0:div><ns0:head>DISCUSSION</ns0:head><ns0:p>Performance of state-of-the-art models on ground truth data is consistent with the findings of other On the other hand, in crowded scenes, the standard YOLO method works much better than YOLO tiled . This is visible when comparing Figure <ns0:ref type='figure' target='#fig_1'>2</ns0:ref> with Figure <ns0:ref type='figure' target='#fig_2'>3A</ns0:ref>, or Figure <ns0:ref type='figure' target='#fig_7'>8C</ns0:ref> with Figure <ns0:ref type='figure' target='#fig_7'>8D</ns0:ref> and Figure <ns0:ref type='figure' target='#fig_7'>8E</ns0:ref>. Differences in the number of detected pedestrians, shown in Table <ns0:ref type='table' target='#tab_4'>3</ns0:ref> and Figure <ns0:ref type='figure'>6</ns0:ref>, are often caused by errors related to incorrect classification of objects. Trashcans from Grodzka webcam (Fig. <ns0:ref type='figure' target='#fig_7'>8A</ns0:ref>) or advertisements from Podgorze Market Square (Fig. <ns0:ref type='figure' target='#fig_7'>8B</ns0:ref>) are recognized by YOLO as persons. The same problem is visible in Figure <ns0:ref type='figure' target='#fig_1'>2</ns0:ref>, where the object detected by YOLO as people (above the car) is actually trash container. A potential problem with YOLO tiled may be double detection of a large object (e.g. bus)</ns0:p><ns0:p>split into two tiles. But for pedestrians, this issue is negligible.</ns0:p><ns0:p>The practical range of pedestrian detection with YOLO can be slightly improved using the tiled method, but it is still limited to about 200 m. Beyond this distance, pedestrians are simply too small to be detected. The problem that is difficult to solve with both YOLO and YOLO tiled is the crowd. Methods based on the YOLO algorithm are not oriented to detect people in crowded scenes.</ns0:p><ns0:p>The Figure <ns0:ref type='figure' target='#fig_9'>9</ns0:ref> During COVID-19, pedestrian activity in public spaces fell almost to zero. Observed values changed proportionally, except for two webcams. The number of detected pedestrians from the Grodzka webcam became smaller than the number of pedestrians from Podgorze Market Square (Fig. <ns0:ref type='figure' target='#fig_9'>9</ns0:ref>). This can be explained by a completely different nature of the location (urban zone). Grodzka street is occupied mainly by tourists, while Podgorze Market Square is mainly occupied by residents. This example shows the Manuscript to be reviewed</ns0:p><ns0:p>The properties of the YOLO detector probably allow the assessment of social distance between pedestrians, which may be the next stage of data analysis. By applying a depth map and pedestrian bounding boxes, it could be possible to quantify the social distances from the webcam image. In addition, the goal-oriented tool can mask pedestrian areas, ignoring the others and thus reducing the calculation time.</ns0:p></ns0:div>
<ns0:div><ns0:head>CONCLUSIONS</ns0:head><ns0:p>Detection of pedestrians in urban space can be done using the YOLOv3 method and hourly time-lapse from webcams. A simple split of the HD webcam image into six smaller tiles in the proposed YOLO tiled method can increase the number of detected pedestrians by over 50%. The YOLO tiled method increases the range of pedestrian detection compared to the YOLO method, but only up to a distance estimated in The resulting hourly data with the number of people (pedestrians) for four webcams in Cracow from June 9, 2016 to April 19, 2020 are available for further use as CSV files.</ns0:p></ns0:div><ns0:figure xml:id='fig_0'><ns0:head>Figure 1 .</ns0:head><ns0:label>1</ns0:label><ns0:figDesc>Figure1. Location in Cracow (Poland) and approximate field of view for webcams used in these studies (www.webcamera.pl). Technical details in Table1.</ns0:figDesc><ns0:graphic coords='6,141.73,63.78,413.58,248.01' type='bitmap' /></ns0:figure>
<ns0:figure xml:id='fig_1'><ns0:head>Figure 2 .</ns0:head><ns0:label>2</ns0:label><ns0:figDesc>Figure 2. Sample image from All Saints Square webcam at midnight, with pedestrians and cars detected by YOLO. Timestamp: 2016-10-29 00:00.</ns0:figDesc><ns0:graphic coords='7,141.73,63.78,413.57,232.63' type='bitmap' /></ns0:figure>
<ns0:figure xml:id='fig_2'><ns0:head>Figure 3 .</ns0:head><ns0:label>3</ns0:label><ns0:figDesc>Figure 3. The upper left two tiles from the split image (Fig. 2). Pedestrians (A) undetected and (B) detected by YOLO tiled method.</ns0:figDesc><ns0:graphic coords='7,141.73,402.38,413.56,172.48' type='bitmap' /></ns0:figure>
<ns0:figure xml:id='fig_3'><ns0:head>Figure 4 .</ns0:head><ns0:label>4</ns0:label><ns0:figDesc>Figure 4. Pedestrian detection from All Saints Square webcam by (A) YOLO and (B) YOLO tiled method. Both views are framed to the central part. YOLO tiled view without the upper left tile.</ns0:figDesc><ns0:graphic coords='10,141.73,63.78,413.56,156.06' type='bitmap' /></ns0:figure>
<ns0:figure xml:id='fig_4'><ns0:head>Figure 5 .</ns0:head><ns0:label>5</ns0:label><ns0:figDesc>Figure 5. The biggest difference from the Wawel webcam, with more pedestrians detected by YOLO tiled method. Results from (A) YOLO method, and (B)(C) two bottom left tiles from YOLO tiled method.</ns0:figDesc><ns0:graphic coords='10,141.73,396.13,413.57,298.22' type='bitmap' /></ns0:figure>
<ns0:figure xml:id='fig_5'><ns0:head>Figure 7 .</ns0:head><ns0:label>7</ns0:label><ns0:figDesc>Figure 7. Mean weekly number of pedestrians from hourly time-lapse for (A) two tourist locations and (B) two residential (mixed) locations. Logarithmic scale on both plots for better visualization of annual cycles in tourist locations.</ns0:figDesc><ns0:graphic coords='12,141.73,63.78,413.57,275.72' type='bitmap' /></ns0:figure>
<ns0:figure xml:id='fig_6'><ns0:head /><ns0:label /><ns0:figDesc>authors. The Faster R-CNN model offers the smallest errors, while the YOLO is the fastest. The proposed YOLO tiled model is based on simple adaptation of images to size and ratio of those used in the training phase of the neural network. This operation improved performance of YOLO model by about 20%, even surpassing Faster R-CNN in terms of pedestrian detection.11/17PeerJ reviewing PDF | (2020:05:48857:1:1:NEW 10 Aug 2020)Manuscript to be reviewed Detection of pedestrians using the YOLO algorithm has good accuracy, but you can improve them by simply adjusting the size of the webcam image to the size of the image used for neural network training. The split of the original high resolution image into six smaller images increased the number of detected pedestrians from 52.13% to 302.73%, as shown in Table3. These values are proportional to the visible distance of the webcam. At short distances (All Saints Square), mainly pedestrians near the camera are visible in the field of view. In the case of large distances (Wawel Castle), where the nearest pedestrian is visible at a distance of 100 m, split of images into smaller tiles causes a significant change in the number of detected pedestrians. The cost of better results using the YOLO tiled method is a longer calculation time. The minimum size at which a pedestrian can be detected is approximately 15 pixels of height. So any decrease in this value due to image scaling makes it almost impossible to detect pedestrian. This is probably the main reason for the good performance of proposed new method. Reducing image resolution for large objects may result in better generalization, but for small objects the spatial features of the object may be lost. The proposed YOLO tiled method maintains the aspect ratio of the image compared to the image used during training. It can be therefore assumed that the main source of improvement in performance is the preserved size of the pedestrian.</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_7'><ns0:head>Figure 8 .</ns0:head><ns0:label>8</ns0:label><ns0:figDesc>Figure 8. The biggest difference in pedestrian detection for three webcams -less pedestrians from YOLO tiled method. Error assigning class: (A) trashcan from Grodzka (bottom right), (B) advertisement display from Podgorze (only one real person was detected). Better detection of people in crowd from Wawel webcam (C) by YOLO compared to (D)(E) YOLO tiled method.</ns0:figDesc><ns0:graphic coords='13,141.73,279.64,413.57,129.11' type='bitmap' /></ns0:figure>
<ns0:figure xml:id='fig_8'><ns0:head /><ns0:label /><ns0:figDesc>is the best illustration of changes in pedestrian activity in Cracow before and during COVID-19. Five weeks during lockdown (from the First restrictions) and a few weeks earlier show how significant was the decrease in pedestrian activity in public space. Subsequent restrictions (second and third) did not change the situation. Weekly cycles visible in almost all locations are replaced by flat lines since the first restrictions in mid-March. Until the end of the period under review, this trend remains unchanged.</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_9'><ns0:head>Figure 9 .</ns0:head><ns0:label>9</ns0:label><ns0:figDesc>Figure 9. Mean daily number of pedestrians from hourly time-lapse for four webcams in Cracow, before and during COVID-19. Split time is set on First restrictions (2020-03-13).</ns0:figDesc><ns0:graphic coords='14,141.73,63.78,413.57,275.72' type='bitmap' /></ns0:figure>
<ns0:figure xml:id='fig_10'><ns0:head /><ns0:label /><ns0:figDesc>this study at about 200 m. Pedestrians are not detected at longer distances. The YOLO tiled turned out to be the most efficient model compared to YOLO, Faster R-CNN and SSD.During the COVID-19 pandemic lockdown in Cracow, from March 13, 2020 to April 19, 2020, pedestrian activity decreased by 78-85% in the tourist zone (Old Town) and by 34-55% in the residential zone. The results are very similar to the Google COVID-19 Community Mobility Reports, despite the use of various methods. Polish citizens quickly and responsibly reacted to restrictions related to the social distance, the visible manifestation of which was the limitation of pedestrian traffic in urban space during the COVID-19 pandemic.</ns0:figDesc></ns0:figure>
<ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='11,141.73,182.59,413.57,275.72' type='bitmap' /></ns0:figure>
<ns0:figure type='table' xml:id='tab_0'><ns0:head>Table 1 .</ns0:head><ns0:label>1</ns0:label><ns0:figDesc /><ns0:table><ns0:row><ns0:cell>Webcam name</ns0:cell><ns0:cell>Distance to</ns0:cell><ns0:cell>Pedestrians</ns0:cell><ns0:cell>Urban zone /</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>pedestrians (m)</ns0:cell><ns0:cell>area (ha)</ns0:cell><ns0:cell>URL</ns0:cell></ns0:row><ns0:row><ns0:cell>Wawel Castle</ns0:cell><ns0:cell>50-400</ns0:cell><ns0:cell>0.92</ns0:cell><ns0:cell>touristic / residential</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell>https://krakow2.webcamera.pl/</ns0:cell></ns0:row><ns0:row><ns0:cell>All Saints Square</ns0:cell><ns0:cell>10-150</ns0:cell><ns0:cell>0.32</ns0:cell><ns0:cell>touristic mainly</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell>https://krakow1.webcamera.pl/</ns0:cell></ns0:row><ns0:row><ns0:cell>Grodzka</ns0:cell><ns0:cell>10-100</ns0:cell><ns0:cell>0.14</ns0:cell><ns0:cell>touristic</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell>https://hotel-senacki-krakow.webcamera.pl/</ns0:cell></ns0:row><ns0:row><ns0:cell>Podgorze Market Square</ns0:cell><ns0:cell>30-120</ns0:cell><ns0:cell>0.31</ns0:cell><ns0:cell>residential</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell>https://krakow3.webcamera.pl/</ns0:cell></ns0:row></ns0:table></ns0:figure>
<ns0:figure type='table' xml:id='tab_1'><ns0:head>Table 1 .</ns0:head><ns0:label>1</ns0:label><ns0:figDesc>Webcam visibility range and source image URLs. Pedestrians area refers to the part of the area accessible to pedestrians.</ns0:figDesc><ns0:table /></ns0:figure>
<ns0:figure type='table' xml:id='tab_2'><ns0:head>Table 2 .</ns0:head><ns0:label>2</ns0:label><ns0:figDesc>The performance of the YOLO model compared to the leading Faster R-CNN is 2% worse taking into account Mean Absolute Error and 10% in terms of Root Mean Squared Error. The YOLO tiled model performance is superior to all state-of-the-art pedestrian detection models. Compared to the top Faster R-CNN model, the improvement is 20% for MAE and 13% for the RMSE. Compared to the original YOLO model, the improvement is approximately 22% for both MAE and RMSE. Considering the image processing time, YOLO is twice as fast as the second fastest model (SSD). Dividing the HD webcam image into six tiles for YOLO tiled model for four cameras in</ns0:figDesc><ns0:table><ns0:row><ns0:cell cols='4'>Cracow resulted in 4.4 times longer processing time. However, it is still 2.5 times faster compared to</ns0:cell></ns0:row><ns0:row><ns0:cell>Faster R-CNN.</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>Model</ns0:cell><ns0:cell cols='3'>MAE RMSE Time (s)</ns0:cell></ns0:row><ns0:row><ns0:cell>SSD</ns0:cell><ns0:cell>9.87</ns0:cell><ns0:cell>14.32</ns0:cell><ns0:cell>1.46</ns0:cell></ns0:row><ns0:row><ns0:cell>Faster R −CNN</ns0:cell><ns0:cell>5.38</ns0:cell><ns0:cell>9.16</ns0:cell><ns0:cell>8.35</ns0:cell></ns0:row><ns0:row><ns0:cell>Y OLO</ns0:cell><ns0:cell>5.48</ns0:cell><ns0:cell>10.23</ns0:cell><ns0:cell>0.75</ns0:cell></ns0:row><ns0:row><ns0:cell cols='2'>Y OLO tiled (proposed) 4.28</ns0:cell><ns0:cell>7.96</ns0:cell><ns0:cell>3.28</ns0:cell></ns0:row></ns0:table></ns0:figure>
<ns0:figure type='table' xml:id='tab_3'><ns0:head>Table 2 .</ns0:head><ns0:label>2</ns0:label><ns0:figDesc>Model performance in pedestrian detection with average processing time per image. Image processing time includes reading the file from disc and model prediction.</ns0:figDesc><ns0:table><ns0:row><ns0:cell>Webcam</ns0:cell></ns0:row></ns0:table><ns0:note>The first part of the research focuses on assessing the YOLO pedestrian detection method and comparing with YOLO tiled . The number of detected pedestrians (people) for each hourly image from four webcams in Cracow from 2016-2020 is saved for YOLO and YOLO tiled method.On average, YOLO results are underestimated compared to the YOLO tiled method. Webcams located at Royal Road (All Saints Square and Grodzka) had the highest absolute detected pedestrian differences up to 50 person, as shown in Table3. The other two webcams, located in the residential and mixed zone, had differences of less than 25 people.</ns0:note></ns0:figure>
<ns0:figure type='table' xml:id='tab_4'><ns0:head>Table 3 .</ns0:head><ns0:label>3</ns0:label><ns0:figDesc>Statistics of detected pedestrian number by YOLO and YOLO tiled method.</ns0:figDesc><ns0:table /></ns0:figure>
<ns0:figure type='table' xml:id='tab_5'><ns0:head>Table 4 .</ns0:head><ns0:label>4</ns0:label><ns0:figDesc>Mean number of pedestrians detected by YOLO tiled method from hourly time-lapse, before and during COVID-19.</ns0:figDesc><ns0:table><ns0:row><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell>Webcam</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell /><ns0:cell cols='4'>Wawel Castle All Saints Square Grodzka Podgorze Market Square</ns0:cell></ns0:row><ns0:row><ns0:cell>Before COVID-19</ns0:cell><ns0:cell>0.49</ns0:cell><ns0:cell>16.86</ns0:cell><ns0:cell>7.13</ns0:cell><ns0:cell>1.66</ns0:cell></ns0:row><ns0:row><ns0:cell>During COVID-19</ns0:cell><ns0:cell>0.14</ns0:cell><ns0:cell>2.04</ns0:cell><ns0:cell>0.57</ns0:cell><ns0:cell>0.82</ns0:cell></ns0:row><ns0:row><ns0:cell>Change (%)</ns0:cell><ns0:cell>-54.64</ns0:cell><ns0:cell>-78.41</ns0:cell><ns0:cell>-85.32</ns0:cell><ns0:cell>-33.82</ns0:cell></ns0:row><ns0:row><ns0:cell cols='5'>Before COVID-19, the All Saint Square webcam registered about ten times as many pedestrians</ns0:cell></ns0:row><ns0:row><ns0:cell cols='5'>compared to Podgorze Market Square (Tab. 4). During COVID-19 this ratio changed to 2:1. The largest</ns0:cell></ns0:row><ns0:row><ns0:cell cols='5'>decrease in the number of pedestrians (85%) is observed on the Grodzka camera, which is a typical tourist</ns0:cell></ns0:row><ns0:row><ns0:cell cols='5'>destination, and the lowest on Podgorze Market Square (34%) in the residential zone. Mixed urban zones,</ns0:cell></ns0:row><ns0:row><ns0:cell cols='5'>with tourist and residential activities, report a moderate decrease in pedestrian numbers, from 55% to</ns0:cell></ns0:row><ns0:row><ns0:cell cols='5'>78%. About 1,000 hourly time-lapse images during COVID-19 and 33,000 images before this period for</ns0:cell></ns0:row><ns0:row><ns0:cell cols='4'>each webcam is a long enough time series to draw final conclusions.</ns0:cell><ns0:cell /></ns0:row></ns0:table></ns0:figure>
<ns0:note place='foot' n='2'>/17PeerJ reviewing PDF | (2020:05:48857:1:1:NEW 10 Aug 2020)</ns0:note>
<ns0:note place='foot' n='17'>/17 PeerJ reviewing PDF | (2020:05:48857:1:1:NEW 10 Aug 2020)Manuscript to be reviewed</ns0:note>
</ns0:body>
" | "Cracow University of Technology
Faculty of Environmental and Power Engineering
Warszawska 24, 31-155 Cracow, Poland
https://www.pk.edu.pl
August 3rd, 2020
Dear Editor,
than you very much for your time and valuable comments from you and from two Reviewers.
I have made every effort to address all points raised in the reviews and believe that this version of
the manuscript is suitable for publication in PeerJ Environmental Sciences.
Most of the effort have been focused on preparation of ground truth data (almost 3,000 images) and
the implementation of two additional state-of-the-art models (SSD and Faster R-CNN). This
enabled the model performance to be assessed using standard metrics and confirmed the
effectiveness of the proposed method. The literature review and description of the YOLO model
have been extended in the new version of the paper. The text also adds a more extensive explanation
of the proposed YOLOtiled method. I hope that the accompanying good results of the proposed
model will confirm my contribution to the investigation of pedestrian detection.
The Abstract has also been slightly modified, but the changes made to the reviewed version are not
highlighted by latexdiff. This is probably related to low-level Latex template issues with which I’m
not sure what to do.
Please let me know if any part of the the paper requires further modification.
With kind regards,
Robert Szczepanek, PhD, Asst. Prof.
Reviewer 1
Basic reporting
1. The paper presents a direct application of pedestrians detection in the actual situation of the
world. It compares the number of pedestrians detected before and after COVID 19 lockdown. The
work is very useful and shows important findings even though the contribution is not very
significant.
Main goal of the presented paper was to check possible methods of simple improvement of the
state-of-the-art pre-trained models. Therefore, model training tasks were omitted. I hope that the
added comparison of performance with other models will better show the impact of proposed
method on the final results of pedestrian detection.
2. The paper is easily understood nevertheless we suggest that the authors improve the english used
throughout the paper and correct some typos and grammatical errors such as
-on several factors, such as, for example night lighting
- Python scripts and jupiter notebooks
- Croud counting requires development of new methods
Accepted and revised.
3. The literature review is not sufficient. We suggest to add a section devoted to literature review in
which you can add paraghraph 'People detection' and some other researches aiming at pedestrain
detection using deep learning methods such as
-Jiang, Y., Tong, G., Yin, H. and Xiong, N., 2019. A pedestrian detection method based on genetic
algorithm for optimize XGBoost training parameters. IEEE Access, 7, pp.118310-118321.
-Cheng, E.J., Prasad, M., Yang, J., Khanna, P., Chen, B.H., Tao, X., Young, K.Y. and Lin, C.T.,
2020. A fast fused part-based model with new deep feature for pedestrian detection and security
monitoring. Measurement, 151, p.107081.
Accepted and revised.
The “People detection” section has been reorganized to better emphasize the pedestrian detection
problem. In particular, a new section “Pedestrian detection” has been added. It has been
supplemented with several new publications and a broader pedestrian detection context.
4. The author references the source code in too many places. It is enough to mention the source
code in experimental results or in footnote in order to order to avoid too many hyperlinks.
Accepted and revised.
The information about the source code has been removed from the Conclusions and only left in the
Methods section.
5. Please correct the way you reference tables, figures and sub figures (Fig. 8A).
Accepted and revised.
I really appreciate finding these mistakes. The following references have been improved and
corrected:
[205] Fig. 3AB changed to just Fig. 3
[251] Changed text referencing Figure 4.
[284-5] Changed reference to Figure 7.
[289] Changed reference to Figure 7.
[297] Corrected error. Was (Fig. 3), should be (Tab. 3).
[307] Corrected error. Was (Fig. 2), should be (Tab. 3).
[316] Corrected error. Was (Fig. 5B), should be (Fig. 8B).
Experimental design
6. Please add more details explaining why the YOLOtiled method gives better results compared to
YOLO.
Accepted and revised.
A more detailed explanation of good YOLOtiled performance has been added in the Discussion
section. This can be explained by the optimal detection of extremely small pedestrians 5-50 pixels
high. Any scaling of the source images made in the pre-processing causes additional reduction of
these objects. This is not the case with YOLOtiled, where small objects are preserved in preprocessing.
7. The author did not mention the ground thruth, regarding the true count of pedestrains provided
by human observations, to assess the proposed method.
Accepted and revised.
Indeed, the lack of real number of pedestrians was the main drawback of the previous version of the
paper. One month’s data was manually analyzed by a human to count pedestrians. It consisted of
2976 images from four webcams from March 2020.
8. Please add a table in which you compare YOLOtiled and YOLO in terms of computational time.
Accepted and revised.
Computational time for YOLO, YOLOtiled and two additional state-of-the-art models (SSD, Faster
R-CNN) has been added and shown in new Table 2.
Validity of the findings
9. The results provided by the proposed method compared to YOLOv3 are satisfactory. But it is
highly recommended to assess the performance of the proposed approach with robust scientific
metrics. Moreover, comparison with state-of-the-art approaches, in the same data, will push
forward the paper.
The lack of ground truth data prevented a reliable performance calculation. The preparation of data
on actual number of pedestrians was time consuming, but as suggested, it turned out to be a great
step forward. This made it possible to calculate the MAE and RMSE. A comparison with the two
additional state-of-the art models (SSD and Faster R-CNN) gave a better picture of YOLOtiled
results. I really appreciate this suggestion as the resulting values show the potential benefits of the
proposed approach. In terms of calculation time, the performance of the proposed model is better
compared to other models, including the Faster R-CNN.
10. The conclusion contains unnecessary details especially those taking about source code.
Accepted and revised.
Reviewer 2
Basic reporting
This research applied the YOLO machine learning algorithm for pedestrian detection and track the
changes in pedestrian activity before and during the COVID-19 lockdown in Cracow, Poland.
Experimental design
1) This manuscript is just a case study of using YOLO. Each webcam image was simply divided into
six smaller tiles for YOLO input. I think the authors did not improve or modify the original
algorithm.
My goal was to extract the best pedestrian count information from the images using pre-trained,
state-of-the-art object detection models. I did not modify the original YOLO model. With the image
size information used during network training, I only improved the model's performance by making
optimal use of the size and aspect information of the source images from HD webcams. This
allowed for better performance compared not only to the original YOLO, but even leading Faster RCNN.
2) Please clarify why the similar dimensions/proportions between the input image and the training
image can improve the accuracy of pedestrian detection. The readers expect to understand the
intrinsic nature of this outcome. In Fig4a, will the accuracy be improved if the original image is
cropped to remove buildings on top before dividing it into two tiles (720x720 px)?
Accepted and revised.
A more detailed description of the effect of image size and ratio on model performance has been
added in the Discussion section.
The method proposed in the paper consist of two steps: (1) divide the HD image (1280x720px) to
get tile size close to size used for training the model, (2) only process tiles with potential
pedestrians. I have not researched the opposite approach: (1) crop the image into potential areas, (2)
divide the image to get tile size close to the size used to train the model.
In Fig.4A, after cropping top part of the image and tiling further, the resulting performance will
depend on the size and aspect ratio of the tiles. If the source HD image is 1280x720px and we crop
30% from the top, the final height will be 504px. We get three horizontal tiles 426x504px. The ratio
of such a tile is slightly worse compared to 426x360 if the training images are 416x416px. The tiles
will be reduced in size. So I would say the final result will be a bit worse.
This calculation obviously depends on the scope of the crop and results may vary between
webcams. But this lead us to problem of unpredictability of results. The image size/ratio values will
be different for each webcam, making it difficult to compare. But this alternative approach is
interesting and worth for further research. Thank you for pointing this out.
3) The conclusion is that pedestrian activity decreased by 78-85% during the lockdown. What are
the new things here? I heard similar statistics on media many times.
While looking for similar projects and lockdown data, I found only one scientific publication by the
Google team that reported specific values for the number of pedestrians on a global and local scale.
Aktay, A., Bavadekar, S., Cossoul, G., Davis, J., Desfontaines, D., Fabrikant, A., Gabrilovich, E.,
Gadepalli, K., Gipson, B., Guevara, M., et al. (2020). Google COVID-19 Community Mobility
Reports: Anonymization Process Description (version 1.0). arXiv preprint arXiv:2004.04145.
I used four-year hourly data from four webcams to show a relative change in pedestrian activity. In
the case of Poland, similar data is not available, therefore the presented results may be a valuable
source of quantitative information for other researchers.
Validity of the findings
4) How was the detection accuracy of the methods validated and compared?
What is the novelty of the research? I don't see significant contributions from the manuscript.
False positive results in people detection are a marginal problem with the YOLO model. It was
assumed that the main problem is related to undetected people due to the small object size in highresolution image. However, without ground truth data, the performance of the models could not be
assessed. The revised version of the article already contains such data and has been supplemented
with two additional state-of-the-art models.
The novelty of research is the use of a simple tiling operation to improve the performance of the
YOLOv3 model. The available models are not able to detect objects over a wide range of scales in
standard HD webcam images (1280x720 px). As shown in the revised version of the article, the
proposed YOLOtiled model surpasses other models in terms of the results obtained. The only cost is
the computation time.
Comments for the Author
5) It is strange when the number of pedestrians is described in decimals, for example, 0.14 or 16.86
people. Instead of statistics in hourly, pedestrians can be counted per day for ignoring decimal
usage.
Pedestrians (people) are counted as objects in the image, but when it comes to statistics such as the
average number of pedestrians in the image, decimal numbers are inevitable. We only consider onehour snapshots and not a continuous pedestrian counting, assuming that this temporal resolution is
representative of such a study.
Initially, an estimate of the absolute number of pedestrians crossing the webcam zone per week or
month was taken into account, but such estimates would be burdened with a large error, hampering
an objective assessment. It would be necessary to implement a method of counting pedestrians
based on their traffic. It is not an easy task, because one of the cameras covers the vicinity of the
nightclub (Fig. 2), where people simply stand at night and tourists move during the day. In the end, I
opted to leave the less intuitive but more reliable units – number of pedestrians per image from
hourly webcam snapshots.
6) The principles of the YOLO algorithm should be introduced in more detail. A workflow should be
added.
Accepted and revised.
A more detailed description of YOLO has been added in the Introduction, and a description of the
YOLOtiled model workflow has also been added in the Methods section. As most modern methods
of detecting objects require image resizing during pre-processing, the results presented in this article
should be universal. The results of these studies can probably be applied not only to the YOLO
model, but this issue will be investigated in the future.
" | Here is a paper. Please give your review comments after reading it. |
9,773 | "<ns0:abstract xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:p>Burmese Cretaceous amber (~99 Ma, Myanmar) is famous for the preservation of a wide range of fauna and flora, including representatives of marine, freshwater and terrestrial groups. Here, we report on three ostracod specimens, that came visible as syninclusions to an aquatic isopod. The three specimens represent three different taxa, that were found preserved in a single piece of amber. One of the described specimens was studied using µCT scanning data . On the basis of general carapace morphology we assign all three to the group Podocopida, and (tentatively) its ingroup Cypridocopina. A lack of visibility of more particular diagnostic features such as adductor muscle scars and details of the marginal zone precludes a further identification, but we discuss possible affinities with either the marine-brackish group Pontocypridoidea or the non-marine group Cypridoidea.</ns0:p><ns0:p>The taphonom y indicates that the studied ostracods have been subjects to limited (if any) post-mortem transport, indicating a coastal pal aeoenvironment.</ns0:p></ns0:div>
</ns0:abstract>
<ns0:body xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:head>Introduction</ns0:head><ns0:p>Ostracods (also known as 'seed shrimps') are small crustaceans with a bivalved carapace (e.g., <ns0:ref type='bibr' target='#b6'>Horne et al., 2002)</ns0:ref>. Adult ostracods are typically 0.5-2 mm long, but can also be smaller or much larger, for example the marine Gigantocypris can be more than 30 mm <ns0:ref type='bibr' target='#b16'>(Poulsen, 1962;</ns0:ref><ns0:ref type='bibr' target='#b6'>Horne et al., 2002;</ns0:ref><ns0:ref type='bibr' target='#b0'>Brusca & Brusca, 2003)</ns0:ref>. Ostracods are the most commonly preserved fossil arthropods, which can be dated back to Early Ordovician <ns0:ref type='bibr' target='#b5'>(Horne, 2005)</ns0:ref>. The earliest ostracods are all marine and the first undoubted non-marine representatives of the group are of Early Carboniferous age <ns0:ref type='bibr' target='#b17'>(Rodriguez-Lazaro & Ruiz-Muñoz, 2012)</ns0:ref>. At the present day, non-marine ostracods can be found in most non-marine aquatic ecosystems including freshwater and saline lakes, streams and rivers, springs, wetlands, temporary ponds, and groundwater <ns0:ref type='bibr' target='#b7'>(Horne et al., 2019)</ns0:ref>. They can even be found in semi-terrestrial habitats such as moist soils with leaf litter <ns0:ref type='bibr' target='#b17'>(Rodriguez-Lazaro and Ruiz-Muñoz, 2012)</ns0:ref>. However, most of the relatively few records of ostracods preserved in amber (e.g., <ns0:ref type='bibr' target='#b12'>Keyser & Weitschat, 2005;</ns0:ref><ns0:ref type='bibr' target='#b11'>Keyser & Friedrich, 2017;</ns0:ref><ns0:ref type='bibr' target='#b15'>Matzke-Karasz et al., 2019)</ns0:ref> are from the Cenozoic and, although the number of kinds of organisms trapped in Burmese Cretaceous amber from Myanmar has increased exponentially over the past few years (including flowers, fungi, scorpions, spiders, crabs, frogs, dinosaurs and insects; <ns0:ref type='bibr' target='#b18'>Ross, 2018)</ns0:ref>, only one ostracod (a marine myodocopan) has so far been reported <ns0:ref type='bibr' target='#b30'>(Xing et al., 2018)</ns0:ref>. Here, we report podocopan ostracods from amber of Myanmar ('Burmite'), the age of which has been biostratigraphically constrained to be late Albian-early Cenomanian (latest Early to earliest Late Cretaceous; <ns0:ref type='bibr' target='#b1'>Cruickshank & Ko, 2003)</ns0:ref>. <ns0:ref type='bibr' target='#b27'>Shi et al. (2012)</ns0:ref> further constrained the age geochronologically based on U-Pb zircon ages from the volcanoclastic rock matrix, containing the amber, to lowermost Cenomanian, ~ 99 million years. In this study we describe three fossil remains of Ostracoda in Burmese amber along with a careful interpretation about their systematic position and critically discuss their values as palaeoenvironmental indicators.</ns0:p></ns0:div>
<ns0:div><ns0:head>Materials & Methods</ns0:head><ns0:p>A single piece of amber (26.4 mm feret diameter) is the focus of this study. It was commercially obtained by Mark Pankowski (Rockville, Maryland, USA) and was donated to the collection of the Natural History Museum Vienna ('Naturhistorisches Museum Wien', NHMW) in April 2017 <ns0:ref type='bibr'>(Thomas Nichterl, collection manager at NHMW, pers. comm., 2020)</ns0:ref>. The amber piece is available under the collection number NHMW-2017/0052/0001. The amber piece comes from one of the commercial mining sites in the Hukawng Valley in the Kachin Province of Myanmar. Due to being commercially acquired, stating a more precise provenance is not possible.</ns0:p><ns0:p>Microscopic images were made using a Keyence VHX-6000 digital microscope. The focusstacking function of the digital microscope was used to create in-focus images of threedimensional objects despite the limitation of the depth-of-field. To gather x-ray micro-computer tomography (µCT) data, a Baker Hughes (General Electrics) 'phoenix nanotom m' computer tomograph was used along with the acquisition software 'datos|x'. The µCT imaging was performed at the zoological State collection in Munich. A current of 100 kV was used to scan the object. The amber piece was rotated 360 degrees in 1440 steps. The final reconstruction of the volume data was done using VGStudio MAX 2.2.6.80630 (Volume Graphics). The achieved voxel size of the volume was 2.81295 µm. Drishti 2.6.4 (GNU) was used for volume rendering. In some cases, more than one transfer function was applied to show structures with different xray qualities. Regular (two-dimensional) images and red-cyan stereo anaglyphs were exported. GIMP 2.10 (GNU) was used to optimize the histogram, and enhance colour, brightness and contrast of the final images. Inkscape (versions 0.92.3 and 0.92.4, GNU) was used to create the figure plates. Measurements were performed using ImageJ (FIJI, public domain).</ns0:p></ns0:div>
<ns0:div><ns0:head>Systematic palaeontology</ns0:head><ns0:p>The higher classification draws mainly on schemes published by <ns0:ref type='bibr' target='#b4'>Horne (2002)</ns0:ref>, <ns0:ref type='bibr' target='#b8'>Hou et al. (2002)</ns0:ref> and <ns0:ref type='bibr' target='#b28'>Smith et al. (2015)</ns0:ref>.</ns0:p><ns0:p>Ostracoda <ns0:ref type='bibr'>Latreille, 1806</ns0:ref><ns0:ref type='bibr'>(= Ostrachoda Latreille, 1802)</ns0:ref> Podocopa <ns0:ref type='bibr' target='#b19'>Sars, 1866</ns0:ref><ns0:ref type='bibr'>Podocopida Sars, 1866</ns0:ref><ns0:ref type='bibr'>Cypridocopina Jones, 1901</ns0:ref> All three ostracod taxa are assigned to the Cypridocopina on the basis of their overall carapace morphology, including rounded subtriangular outline in lateral view, compressed fusiform outline in dorsal/ventral view, and smooth or lightly pitted external surface. Neither with light microscopy nor with computer-tomography have we been able to resolve any diagnostic features, such as adductor muscle scars or internal details of the marginal zone, that might lead to a more precise identification. Description. Carapace small, elongate subtriangular in lateral view; slender, fusiform (spindleshaped) with bluntly rounded extremities in dorsal view. Left valve slightly larger than right valve and overlapping right valve along all margins but most markedly on the ventral margin and at the highest point of the dorsal margin. Left valve possibly with alveolus behind a rostrum. Maximum height at one third of length from anterior margin. Anterior cardinal distinct, obtuseangled (approx. 140°), forming the highest point. Posterior cardinal angle weakly rounded and obtuse with about 160°. Anterior margin broad and slightly infracurvate, almost equicurvate with a moderately long, nearly straight dorsal part. Posterior margin distinctly narrower than anterior one, rounded and equicurvate, having a short slightly curved dorsal part. Dorsal margin slightly convex and inclined towards posterior end, about 20°. Ventral margin almost straight, slightly concave at mid-length. Surface smooth. Internal features not observable.</ns0:p><ns0:p>Remarks: The left valve shows shallow indentation near the anterior end of the ventral margin that could be interpreted as an alveolus behind a rostrum, i.e. a 'beak' such as is diagnostic of the group Cypridea (Superfamily Cypridoidea) (Fig. <ns0:ref type='figure' target='#fig_5'>2H</ns0:ref>). However, there is no trace of its equivalent in the left valve, and the feature may actually represent damage to the valve. Based on the gross morphological similarities, taxa A belongs to either Pontocypridoidea or Cypridoidea. However, the strong asymmetry of the right and left valves perhaps favours an assignment to the group Cypridoidea but this is not conclusive.</ns0:p><ns0:p>Taxon B Fig. <ns0:ref type='figure' target='#fig_4'>1B</ns0:ref> Material. One articulated carapace, well-preserved. Remarks: The carapace is similar in lateral outline, but laterally more inflated, than that of Taxon A. Moreover, Taxon A has an apparently smooth exterior while Taxon C is distinctly punctate. Based on its overall shape and distinct punctation it has strong similarities to species of the nonmarine group Harbinia Tsao, 1959 (an ingroup of Cypridoidea). Syninclusions: cf. Psychodidae (Diptera); Alavesia sp. (Diptera); Coleoptera sp.; Euarthropoda sp. (one unidentifiable remain and three isolated legs); Cymothoida sp. (Isopoda) <ns0:ref type='bibr'>(Schädel et al. in review)</ns0:ref>.</ns0:p></ns0:div>
<ns0:div><ns0:head>Discussion</ns0:head><ns0:p>Even though the studied ostracod specimens are well-preserved, their positions within the amber piece do not allow us to study further details, so we are unable to identify them beyond suborder level with confidence. This limitation in turn impedes the palaeoenviromental interpretation of the assemblage. Nevertheless, we can speculate, on the basis of gross morphological similarities, that at least taxa A and B belong to either of two ingroups of Cypridocopina, namely Pontocypridoidea and Cypridoidea. Moreover, as we don't know if the specimens are adults, and the shape of Taxon A could be a juvenile, which tend to have less inflated posterior margins. If we assume Taxon A is an adult, then it resembles a Pontocypridoidea or Cypridoidea (maybe even a Paracypridinae of the Candonidae). Cypridoidea are the predominant group in non-marine environments today, such as fresh water and saline inland water (athalassic environments). Pontocypridoidea, on the other hand, comprises marine and brackish-water taxa <ns0:ref type='bibr' target='#b2'>(Horne, 2003)</ns0:ref>. In the case of Taxon C the general shape and the association with the other two taxa suggest that it, too, is a cypridocopine. Species of the Cretaceous non-marine cypridoidean genus Harbinia Tsao, 1959, have similar carapace shape and punctate/reticulate ornament (e.g. Harbinia hapla <ns0:ref type='bibr' target='#b29'>Tsao, 1959</ns0:ref><ns0:ref type='bibr'>, illustrated by Ye et al. (2003: pl. 27, figs 2a-c)</ns0:ref>. However, we cannot rule out the possibility that it may belong to the group Cytheroidea (which is the dominant marine group today but also has non-marine lineages) without being able to see the characteristic adductor muscle scar patterns that distinguish between cypridoideans and cytheroideans.</ns0:p><ns0:p>All three ostracod specimens in this study lack preserved soft parts but are preserved with articulated carapaces, suggesting limited (if any) post-mortem transport. In view of our uncertainty about the precise systematic interpretation of our specimens we can contribute little to the discussion of the taphonomic and palaeoenvironmental interpretation of the Burmese amber assemblage. Both cypridoidean and pontocypridoidean taxa would be consistent with the mixed marine-freshwater-terrestrial components of the assemblage.</ns0:p><ns0:p>An ammonite shell preserved in Burmese amber argues that the Burmese amber forest was located near a dynamic and shifting coastal environment <ns0:ref type='bibr' target='#b32'>(Yu et al., 2019)</ns0:ref>, a conclusion supported by the occurrence of a marine myodocopan ostracod <ns0:ref type='bibr' target='#b30'>(Xing et al., 2018)</ns0:ref>. Also, semiaquatic and aquatic insects occur in Burmese amber, including Ochteridae (Hemiptera), Heteroceridae (Coleoptera), Chresmododea and Gerridae (Hemiptera), Dytiscidae and Gyrinidae (Coleoptera), adults and larvae of Odonata, larvae of Psephenidae, Trichoptera, and Ephemeroptera <ns0:ref type='bibr' target='#b33'>(Zhang, 2017;</ns0:ref><ns0:ref type='bibr' target='#b30'>Xing et al., 2018;</ns0:ref><ns0:ref type='bibr' target='#b21'>Schädel et al., 2020)</ns0:ref>. The herein presented ostracods are within the same amber piece as a fossil, supposedly aquatic living, isopod <ns0:ref type='bibr'>(Schädel et al. in review)</ns0:ref>. Actuo-palaeontological experiments <ns0:ref type='bibr' target='#b23'>(Schmidt & Dilcher, 2007)</ns0:ref> have demonstrated, that it is easily possible for aquatic organisms to be trapped in submerged bodies of resin. Several records of delicate arthropod remains from groups with supposed aquatic lifestyle <ns0:ref type='bibr' target='#b3'>(Heard et al., 2018;</ns0:ref><ns0:ref type='bibr' target='#b20'>Schädel et al., 2019;</ns0:ref><ns0:ref type='bibr' target='#b21'>Schädel et al. 2020;</ns0:ref><ns0:ref type='bibr' target='#b24'>Serrano-Sánchez et al., 2015;</ns0:ref><ns0:ref type='bibr' target='#b25'>Serrano-Sánchez et al., 2016)</ns0:ref> indicate that the result of in-situ embedment of aquatic organisms is present in many amber sites. Recurrent flows of resin and changing water levels can explain the preservation of aquatic and non-aquatic organisms -such as the dipterans in the herein presented assemblage -in the same amber piece <ns0:ref type='bibr' target='#b30'>(Xing et al., 2018)</ns0:ref>.</ns0:p></ns0:div>
<ns0:div><ns0:head>Conclusions</ns0:head><ns0:p>The three specimens are new records of podocopan ostracods to be reported from Burmese Cretaceous amber, and most likely belong to the Order Podocopida, Suborder Cypridocopina, and either the Superfamily Cypridoidea (indicative of non-marine environments) or the Superfamily Pontocypridoidea (indicative of marine/brackish environments). If the former be true, these would be the oldest non-marine ostracods preserved in amber. Either superfamily assignment would be consistent with previous evidence of a mixed marine-freshwater-terrestrial assemblage deposited in a coastal setting. </ns0:p></ns0:div><ns0:figure xml:id='fig_0'><ns0:head /><ns0:label /><ns0:figDesc>articulated carapace, moderately well preserved.Dimensions. L: 0.57 mm, H: 0.34 mm, W: 0.17 mm.Locality and horizon. Hukawng Valley, Kachin Province, Myanmar (E 96°36′15″, N 26°13′47″, accuracy of about 10 km); lowermost Cenomanian, lowermost Upper Cretaceous.</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_1'><ns0:head /><ns0:label /><ns0:figDesc>Locality and horizon. Hukawng Valley, Kachin Province, Myanmar (E 96°36′15″, N 26°13′47″, accuracy of about 10 km); lowermost Cenomanian, lowermost Upper Cretaceous.Description. Carapace small and fusiform (spindle-shaped) with sharp extremities in dorsal view; surface smooth. Valves approximately equal in size, without evident overlap. Internal features not observable.Remarks: We have been unable to obtain a clear lateral view of this carapace, which appears to differ from those of taxa A and C by having sharper anterior and posterior extremities and no obvious overlap of the valves. We speculate that Taxon B belongs to either Pontocypridoidea or Cypridoidea according to preserved features.Taxon C Fig.1CMaterial. One articulated carapace, well-preserved.Locality and horizon. Hukawng Valley, Kachin Province, Myanmar (E 96°36′15″, N 26°13′47″, accuracy of about 10 km); lowermost Cenomanian, lowermost Upper Cretaceous. Description. Carapace small, rounded subtriangular in lateral view. Left valve slightly overlapping right valve along ventral and posterior margins. Maximum height in front of midlength. Anterior margin broad, almost equicurvate. Posterior margin narrower than anterior one, rounded and equicurvate. Carapace distinctly punctate tending to reticulation. Internal features not observable.</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_2'><ns0:head>Figure explanations Fig</ns0:head><ns0:label>explanations</ns0:label><ns0:figDesc>Figure explanations</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_3'><ns0:head>Fig. 2 :</ns0:head><ns0:label>2</ns0:label><ns0:figDesc>Fig. 2: Taxon A. A−B: light microscopy, mixed translucent and reflected light. A: dorsolateral view. B: left lateral view, side 1; C-L: volume rendering images based on µCT data. C: right lateral view; D: red-cyan stereo anaglyph, posterolateral view; E: posterior view; F: ventral view; G: dorsal view; H: left lateral view; I: red-cyan stereo anaglyph, anterolateral view; J: anterior view; K: ventral view, presumed pyrite crystals in red; L: right lateral, presumed pyrite crystals in red.</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_4'><ns0:head>Figure 1 Fig</ns0:head><ns0:label>1</ns0:label><ns0:figDesc>Figure 1</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_5'><ns0:head>Figure 2</ns0:head><ns0:label>2</ns0:label><ns0:figDesc>Figure 2</ns0:figDesc></ns0:figure>
<ns0:note place='foot'>PeerJ reviewing PDF | (2020:07:50630:1:1:NEW 5 Sep 2020)</ns0:note>
</ns0:body>
" | "Dear editor and reviewers,
Thank you very much for your efforts on this MS, especially for the thoughtful comments. The manuscript has been revised accordingly. Hopefully we have addressed all of your concerns.
Attached please find the revised MS.
Kind regards,
He Wang
Authors’ replies are in red:
Editor (Kenneth De Baets)
1. Dates of Purchase and Donation: Please provide the provenance and dates (as precisely as possible) when the specimen was purchased and when it was donated to the Museum. See also comments by Reviewer 1. This might seem particularly relevant for Burmese amber but should be standard good practice for all fossils.
A more precise description of the provenance was added to the Material &Methods section. The stated date of donation is confirmed by the collection manager in charge. A date of the purchase is not available.
2. Taxonomic uncertainty: you mention that the 3 specimens belong to 3 different species, but I would like to see more explicit information on why these specimens likely belong to different species (which traits speak for this, how do specimens/species differ from one another). Please provide measurements for all specimens (compare comments by Reviewer 1). Are you certain these specimens are adult (see also comment by reviewer 1)? It is a bit confusing as you only compare A and C explicitly with taxa in Cypridoidea in the taxonomic description, while you compare A and B in the discussion. Pontocypridoidea is not mention at all in the taxonomic description. Also, in the case of B - not explicit comparisons are made or similarities are mentioned in the taxonomic description. Please resolve these inconsistencies.
Thanks for the comments. We have to say that we have provided all the information we could get. It’s a little pity that we can not provide more explicit information because of the preservation of these ostracods. Otherwise, we would give a more precise identification. Again, because of the preservation of the studied ostracods, we could only provide the measurements of Taxon A. As the Fig. 1 shows, we could provide only one position of Taxon B and Taxon C. And we are not certain these specimens are adult. Because we could only find groups which are similar to Taxon A and Taxon C and we can not find a group similar to Taxon B according to the limited information, we only compare A and C with similar taxa. For the same reason, we can not provide explicit comparisons of Taxon B. In fact, we don’t compare A and B in the discussion. And we added Pontocypridoidea in the taxonomic description.
3. Taxonomic discrepancies: you are rightfully cautious in your interpretations give the lack of certain traits (compare also with comments by reviewer 1, 2), but there are some inconsistencies between taxonomic assignment/comparison mentioned in the species description and the discussion.
Thanks for the comment. However, we don’t think there are taxonomic discrepancies in this study. We give more probable taxonomic assignments in the remarks which are not conclusive (for A and C). And in the discussion, we further provide all the possible taxonomic assignments.
4. Paleoenvironmental/paleoecological conclusions: please take this opportunity to discuss in a bit more detail the possible constraints given by the preservation (tightly closed valves), possible affinities (and their relationship with swimming abilities) and the associated syninclusions on the environmental to depositional conditions of the amber piece (see particularly the comments by reviewer 1)
About this aspect, we have to say we have discussed the palaeoenvironment as much as possible based on the limited information. And we must say that the comments on the swimming abilities from Reviewer 1 are arbitrary. We think it should be more conservative as such few features are preserved.
5. Figures: there are also some inconsistencies (e.g., left versus right valve enlargement, presence of a shallow indentation not mention in text) between the descriptions and the figures (see comments by both reviewer 1 and 2).
OK. We changed it.
6. Missing Reference: Please cite missing references and primary references rather than review paper where appropriate (see comments by reviewer 1).
OK. We have added them.
7. Please make sure to address these and additional comments/suggestions by the reviewers and annotated pdfs
OK.
Reviewer 1 (Anonymous):
1. I suggest that original sources of data for specific facts should be cited rather than general review chapters on ostracods. For example (line 28–39), “…the marine Gigantocypris can be up to 32 mm (Horne et al., 2002; Brusca & Brusca, 2003).” The original data for the size of Gigantocypris probably came from Poulsen 1962, who recorded a maximum size of 34 mm for Gigantocypris agassizi (credit where credit is due). Poulsen, E. M. 1962. Ostracoda-Myodocopa, 1: Cypridiformes-Cypridinidae. In Dana Report, 57:1-414, 181 figures. Copenhagen: Carlsberg Foundation.
Thanks. We added it.
2. One reference is missing: Line 47–48. “However, most of the relatively few records of ostracods preserved in amber (e.g., Keyser & Friedrich, 2017; Matzke-Karasz et al., 2019)…” There is also Keyser, D. & Weitschat, W. 2005. First record of ostracods (Crustacea) in Baltic amber. Hydrobiologia, 538, 107–114.
OK. We added it.
3. The references need checking through. Smith et al. 2011 is not cited in the text. Smith et al. 2015 (line 82) is not listed in the references. I'm not sure either Smith paper is an appropriate citation for the higher classification of ostracods.
Thanks. We have corrected them.
4. The figures are of good quality, but there is a mistake on the scale bar of Figure 1A. Additionally, the red-cyan stereo anaglyphs (Fig. 2, D and I) didn’t work for me. A photograph of the whole amber piece would also be a good idea.
The scale bar was corrected and an overview image of the amber piece was added. The red-cyan anaglyphs were improved by adjusting the saturation and were tested on color calibrated high and low end monitors. The overall poor quality of the colors was due to bad color management settings on a wide gamut monitor. All color settings have been improved and the style of the scale bars was changed for aesthetic reasons.
5. With the ethical controversy surrounding Burmese amber, further details about how this piece was obtained should be given, such as its provenance, date of purchase by previous owners, and date of donation/purchase by the Natural History Museum of Vienna. Whether this is an ethically sourced specimen needs to ascertained before this paper is published. The Journal of Systematic Palaeontology is now refusing to publish any articles on Burmese amber, but of course, this is a matter for the editor.
See comment above. For ethical considerations we would like to informally refer to Comment on the letter of the Society of Vertebrate Paleontology (SVP) dated April 21, 2020 regarding “Fossils from conflict zones and reproducibility of fossil-based scientific data”: Myanmar amber. doi.org/10.1007/s12542-020-00524-9
6. There does seem to be a small discrepancy about possible taxonomic affinities in the article. The Remarks section of Taxon A concludes “… perhaps favours an assignment to the group Cypridoidea but this is not conclusive.” (line 122).
However, the Discussion (line 170) says: “Nevertheless, we can speculate, on the basis of gross morphological similarities, that at least taxa A and B belong to either of two ingroups of Cypridocopina, namely Pontocypridoidea and Cypridoidea.”
OK, we revised this in the MS. I think we should say “Taxa A and B belong to either of two ingroups of Cypridocopina, namely Pontocypridoidea and Cypridoidea” first, then “… perhaps favours an assignment to the group Cypridoidea but this is not conclusive.”
7. One thing I would mention in the Discussion is that we don’t know if the specimens are adults. They are all quite small, and the shape of Taxon A could be a juvenile, which tend to have less inflated posterior margins. If we assume Taxon A is an adult, then it resembles a Pontocypridoidea or Cypridoidea (maybe even a Paracypridinae of the Candonidae: Cypridoidea).
OK. We added it.
8. The ostracods were living in the same environment as the isopod, so does the isopod give any clues to the palaeo-environment? (e.g. a marine, brackish or freshwater taxon?). The main problem with Burmese amber is the lack of stratigraphic data with the pieces recovered. Some think that Burmese amber may span at least 5 million years (according to Science, 24 May 2019 issue, p. 725). So it is unknown if this piece came from the same time period as the ammonite or the marine myodocopid ostracod. Therefore, previous finds may not be that informative (or even misleading) with respect to the palaeo-environment of isolated amber pieces. I would therefore place more weight on syninclusions; the Diptera and Coleoptera would suggest non-marine for this piece, which in turn nudges the possible taxonomic affinities of the ostracods towards the Cypridoidea.
Because there is a variety of habitats for similarly looking extant isopods and because the systematic position of the isopod is not precisely determinable, the isopod could be a marine, brackish or freshwater taxon. And as what we write in the MS, our amber piece is from the same site as the ammonite’s amber, which should be from the same time period. Based on our discovery alone, it’s a little difficult to discuss the palaeoenvironment as limited information we could get. And we think based on the non-marine syniclusions, we can’t nudges the possible taxonomic affinities, because Burmese amber always have marine and non-marine taxa preserved together.
9. The fact that the ostracods are totally surrounded by amber, with no evidence of a substrate near them, suggests that they were capable of swimming above the substrate and they ‘landed’ on the resin. (It also suggests that the ostracods were alive when trapped as it would not be possible for dead ostracods on a substrate to be cleanly separated out like this.) This would exclude the Cytheroidea, and Candoninae (can’t swim). Could the ostracods have been feeding on, or drawn to, a trapped isopod, and consequently been trapped themselves? I think that one study of all the inclusions in this piece, rather than being split in different publications, would have been a better approach to understanding the palaeo-environment. But ultimately that is up to the authors.
Sorry, we don’t agree with this comment. Without soft parts, it’s arbitrary to say the studied ostracods were capable of swimming. And if these ostracods were alive when trapped, soft parts should be preserved. Please see Matzke-Karasz et al. (2019), a paper about the ostracods with soft parts from Mexican amber. “Could the ostracods have been feeding on, or drawn to, a trapped isopod, and consequently been trapped themselves?” – We would like not to include this because we judge it to be too speculative. In this study, we want to focus only on the ostracods, as few records of ostracods in amber, especially Burmese amber.
10. Other records of ostracods in amber typically have a gape to the valves and appendages protruding. These ostracds don’t, but are instead tightly closed. I don’t know the significance of this, but perhaps it should be noted somewhere.
We mentioned this in the MS.
11. Other small issues with the manuscript style include:
Line 22. The first line of the abstract is a bit limp, and could be changed to something like:“Burmese Cretaceous amber (~99 Ma, Myanmar) is famous for the preservation of a wide range of fauna and flora, including representatives of marine, freshwater and terrestrial groups.
OK. We changed it.
Line 30. “Taphonomic and palaeoenvironmental implications are also discussed.” Rather than just saying they are discussed, the authors should outline their conclusions. I suggest adding that tomography was used to study the ostracods to the abstract.
OK. We revised the abstract accordingly.
Line 39. “Ostracods are also the most common fossil arthropods during the geological history of the group, which can be dated back to Early Ordovician (Horne, 2005).”
A clearer way to say this is perhaps “Ostracods are the most commonly preserved fossil arthropods…”.
OK. We changed it.
Line 61. What is the size of the amber piece?
Is added to to the Material & Methods section.
Line 97. “Fig. 1A; Fig. 2; Suppl. 1” Should note that it is marked by the arrow in fig.1A (not the whole plate).
OK. We added it.
Line 101. There are dimensions for Taxon A, but not for B and C. Is it possible to get a e.g. length, if not all dimensions for the other two?
Sorry. It’s impossible by now.
Line 103. “Locality and horizon.”
Add latitude and longitude (also for the other two taxa).
OK. We added it.
Line 118. “Remarks: The right valve shows shallow indentation near the anterior end of the ventral margin…”
Surely you mean the left valve, as seen in Fig. 2H? An arrow on the figure, and the figure number would help here. This shallow indentation is not mentioned in the description. OK, so you think it is possibly an artefact, but I think it should be included.
OK. We added it.
Line 160. “Alavesia sp. (Dipera);” Should be Diptera.
OK. We corrected it.
Line 177. “However, we cannot rule out the possibility that it may belong to the group Cytheroidea…” Unlikely to be Cytheroidea because they don’t swim. See comment under Validity of the findings.
Again, based on the limited information and without the soft parts, we can not say they don’t swim.
Line 204. Better to write “The three specimens are the first podocopan ostracods to be reported from Burmese Cretaceous amber”
OK. We changed it.
Reviewer 2 (Alan Lord):
1. Taxon A. This carapace is described (lines 107-8) as ‘right valve slightly larger than left valve’, however, from Fig. 2C, a right lateral view, the overlap appears to be LV>RV and this is supported by the description, for example, that maximum valve height is anterior of mid-length. The point is important both from a basic taxonomic perspective and also because it means that the fundamental carapace shape and valve overlap are the same as Taxon C (Fig. 1 C).
OK. We corrected it.
2. Minor comments:
Line 48: ‘…although the number of kinds…’.
Lines 66 and 70: please given the correct German names of the laboratories.
line 66 was correct; institute in line 70 was removed because it represents unnecessary information.
Line 82: Smith et al. (2015) not in References (only Smith et al., 2011).
Line 84: Latreille, 1806 (not 1804).
Lines 84-87: higher systematic authors should be listed in References.
Line 92: ‘have we’ (not ‘we have’).
Line 158: ‘Tsao, 1959’ should be listed in References, even if authors used pictures from other literature.
Lines 165-6: ‘…their positions within the amber do not allow us…’.
Line 176: delete ) after 2003.
Line 195: ‘…that it is easily possible for aquatic organisms to be trapped…’.
Lines 198-199: I suggest ‘…indicate that in-situ embedment of aquatic organisms is present in many amber sites.’.
Line 200: ‘and’ (not ‘an’).
Lines 228-232 (“The fossils at the NHMW are held safely in trust for the benefit of researchers and educators in the world respecting all ethnic groups, ages, sexes, landowners and collectors. Apart from public exhibitions, access is free to all scientists and interested people by prior arrangement during normal working hours and subject to the NHMW laboratory and museum regulations.”): are these two sentences really necessary? (We added these sentences because of the recent and ongoing discussions about the availability of type material. The editor should decide if these statements should be included.)
Line 246: Horne (2003) is run into Cruickshank & Ko.
Line 298: Smith et al. (2011) not seen in text.
Line 307: Dilcher.
Line 314: I see no asterisk on Fig. 1. (The visibility of the asterisk was improved by putting dark background behind it.)
We have revised the manuscript according to all comments above.
" | Here is a paper. Please give your review comments after reading it. |
9,774 | "<ns0:abstract xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:p>Burmese Cretaceous amber (~99 Ma, Myanmar) is famous for the preservation of a wide range of fauna and flora, including representatives of marine, freshwater and terrestrial groups. Here, we report on three ostracod specimens, that came visible as syninclusions to an aquatic isopod. The three specimens represent three different taxa, that were found preserved in a single piece of amber. One of the described specimens was studied using µCT scanning data. On the basis of general carapace morphology we assign all three to the group Podocopida, and (tentatively) its ingroup Cypridocopina. A lack of visibility of more particular diagnostic features such as adductor muscle scars and details of the marginal zone precludes a further identification, but we discuss possible affinities with either the marine-brackish group Pontocypridoidea or the non-marine group Cypridoidea. The taphonomy indicates that the studied ostracods had been subject to limited (if any) postmortem transport, which could be consistent with marginal marine environments.</ns0:p></ns0:div>
</ns0:abstract>
<ns0:body xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:head>Introduction</ns0:head><ns0:p>Ostracods (also known as 'seed shrimps') are small crustaceans with a bivalved carapace (e.g., <ns0:ref type='bibr' target='#b8'>Horne et al., 2002)</ns0:ref>. Adult ostracods are typically 0.5-2 mm long, but can also be smaller or much larger, for example the marine Gigantocypris can be more than 30 mm <ns0:ref type='bibr' target='#b18'>(Poulsen, 1962;</ns0:ref><ns0:ref type='bibr' target='#b8'>Horne et al., 2002;</ns0:ref><ns0:ref type='bibr' target='#b0'>Brusca & Brusca, 2003)</ns0:ref>. Ostracods are the most commonly preserved fossil arthropods, which can be dated back to Early Ordovician <ns0:ref type='bibr' target='#b7'>(Horne, 2005)</ns0:ref>. The earliest ostracods are all marine and the first undoubted non-marine representatives of the group are of Early Carboniferous age <ns0:ref type='bibr' target='#b19'>(Rodriguez-Lazaro & Ruiz-Muñoz, 2012)</ns0:ref>. At the present day, non-marine ostracods can be found in most non-marine aquatic ecosystems including freshwater and saline lakes, streams and rivers, springs, wetlands, temporary ponds, and groundwater <ns0:ref type='bibr' target='#b10'>(Horne et al., 2019)</ns0:ref>. They can even be found in semi-terrestrial habitats such as moist soils with leaf litter <ns0:ref type='bibr' target='#b19'>(Rodriguez-Lazaro and Ruiz-Muñoz, 2012)</ns0:ref>. However, most of the relatively few records of ostracods preserved in amber (e.g., <ns0:ref type='bibr' target='#b14'>Keyser & Weitschat, 2005;</ns0:ref><ns0:ref type='bibr' target='#b13'>Keyser & Friedrich, 2017;</ns0:ref><ns0:ref type='bibr' target='#b17'>Matzke-Karasz et al., 2019)</ns0:ref> are from the Cenozoic and, although the number of kinds of organisms trapped in Burmese Cretaceous amber from Myanmar has increased exponentially over the past few years (including flowers, fungi, scorpions, spiders, crabs, frogs, dinosaurs and insects; <ns0:ref type='bibr' target='#b20'>Ross, 2018)</ns0:ref>, only one ostracod (a marine myodocopan) has so far been reported <ns0:ref type='bibr' target='#b31'>(Xing et al., 2018)</ns0:ref>. Here, we report podocopan ostracods from amber of Myanmar ('Burmite'), the age of which has been biostratigraphically constrained to be late Albian-early Cenomanian (latest Early to earliest Late Cretaceous; <ns0:ref type='bibr' target='#b1'>Cruickshank & Ko, 2003)</ns0:ref>. <ns0:ref type='bibr' target='#b28'>Shi et al. (2012)</ns0:ref> further constrained the age geochronologically based on U-Pb zircon ages from the volcanoclastic rock matrix, containing the amber, to lowermost Cenomanian, ~ 99 million years. In this study we describe three fossil remains of Ostracoda in Burmese amber along with a careful interpretation about their systematic position and critically discuss their values as palaeoenvironmental indicators.</ns0:p></ns0:div>
<ns0:div><ns0:head>Materials & Methods</ns0:head><ns0:p>A single piece of amber (26.4 mm feret diameter) is the focus of this study. It was commercially obtained by Mark Pankowski (Rockville, Maryland, USA) and was donated to the collection of the Natural History Museum Vienna ('Naturhistorisches Museum Wien', NHMW) in April 2017 <ns0:ref type='bibr'>(Thomas Nichterl, collection manager at NHMW, pers. comm., 2020)</ns0:ref>. The amber piece is available under the collection number NHMW-2017/0052/0001. The amber piece comes from one of the commercial mining sites in the Hukawng Valley in the Kachin Province of Myanmar <ns0:ref type='bibr' target='#b2'>(Haug et al., 2020)</ns0:ref>. Due to being commercially acquired, stating a more precise provenance is not possible. Microscopic images were made using a Keyence VHX-6000 digital microscope. The focusstacking function of the digital microscope was used to create in-focus images of threedimensional objects despite the limitation of the depth-of-field. To gather x-ray micro-computer tomography (µCT) data, a Baker Hughes (General Electrics) 'phoenix nanotom m' computer tomograph was used along with the acquisition software 'datos|x'. The µCT imaging was performed at the zoological State collection in Munich. A current of 100 kV was used to scan the object. The amber piece was rotated 360 degrees in 1440 steps. The final reconstruction of the volume data was done using VGStudio MAX 2.2.6.80630 (Volume Graphics). The achieved voxel size of the volume was 2.81295 µm. Drishti 2.6.4 (GNU) was used for volume rendering. In some cases, more than one transfer function was applied to show structures with different xray qualities. Regular (two-dimensional) images and red-cyan stereo anaglyphs were exported. GIMP 2.10 (GNU) was used to optimize the histogram, and enhance colour, brightness and contrast of the final images. Inkscape (versions 0.92.3 and 0.92.4, GNU) was used to create the figure plates. Measurements were performed using ImageJ (FIJI, public domain).</ns0:p></ns0:div>
<ns0:div><ns0:head>Systematic palaeontology</ns0:head><ns0:p>The higher classification draws mainly on schemes published by <ns0:ref type='bibr' target='#b6'>Horne (2002)</ns0:ref>, <ns0:ref type='bibr' target='#b11'>Hou et al. (2002)</ns0:ref> and <ns0:ref type='bibr' target='#b29'>Smith et al. (2015)</ns0:ref>.</ns0:p><ns0:p>Ostracoda <ns0:ref type='bibr'>Latreille, 1806</ns0:ref><ns0:ref type='bibr'>(= Ostrachoda Latreille, 1802)</ns0:ref> Podocopa <ns0:ref type='bibr' target='#b21'>Sars, 1866</ns0:ref><ns0:ref type='bibr'>Podocopida Sars, 1866</ns0:ref><ns0:ref type='bibr'>Cypridocopina Jones, 1901</ns0:ref> All three ostracod taxa are assigned to the Cypridocopina on the basis of their overall carapace morphology, including rounded subtriangular outline in lateral view, compressed fusiform outline in dorsal/ventral view, and smooth or lightly pitted external surface. Neither with light microscopy nor with computer-tomography have we been able to resolve any diagnostic features, such as adductor muscle scars or internal details of the marginal zone, that might lead to a more precise identification. Dimensions. L: 0.57 mm, H: 0.34 mm, W: 0.17 mm. Locality and horizon. Hukawng Valley, Kachin Province, Myanmar (E 96°36′15″, N 26°13′47″, accuracy of about 10 km); lowermost Cenomanian, lowermost Upper Cretaceous.</ns0:p><ns0:p>Description. Carapace small, elongate subtriangular in lateral view; slender, fusiform (spindleshaped) with bluntly rounded extremities in dorsal view. Left valve slightly larger than right valve and overlapping right valve along all margins but most markedly on the ventral margin and at the highest point of the dorsal margin. Left valve possibly with alveolus behind a rostrum. Maximum height at one third of length from anterior margin. Anterior cardinal distinct, obtuseangled (approx. 140°), forming the highest point. Posterior cardinal angle weakly rounded and obtuse with about 160°. Anterior margin broad and slightly infracurvate, almost equicurvate with a moderately long, nearly straight dorsal part. Posterior margin distinctly narrower than anterior one, rounded and equicurvate, having a short slightly curved dorsal part. Dorsal margin slightly convex and inclined towards posterior end, about 20°. Ventral margin almost straight, slightly concave at mid-length. Surface smooth. Internal features not observable.</ns0:p><ns0:p>Remarks: The left valve shows shallow indentation near the anterior end of the ventral margin that could be interpreted as an alveolus behind a rostrum, i.e. a 'beak' such as is diagnostic of the group Cypridea (Superfamily Cypridoidea) (Fig. <ns0:ref type='figure' target='#fig_5'>2H</ns0:ref>). However, there is no trace of its equivalent in the left valve, and the feature may actually represent damage to the valve. Based on the gross morphological similarities, taxa A belongs to either Pontocypridoidea or Cypridoidea. However, the strong asymmetry of the right and left valves perhaps favours an assignment to the group Cypridoidea but this is not conclusive.</ns0:p><ns0:p>Taxon B Fig. <ns0:ref type='figure' target='#fig_4'>1B</ns0:ref> Material. One articulated carapace, well-preserved.</ns0:p><ns0:p>Locality and horizon. Hukawng Valley, Kachin Province, Myanmar (E 96°36′15″, N 26°13′47″, accuracy of about 10 km); lowermost Cenomanian, lowermost Upper Cretaceous.</ns0:p><ns0:p>Description. Carapace small and fusiform (spindle-shaped) with sharp extremities in dorsal view; surface smooth. Valves approximately equal in size, without evident overlap. Internal features not observable.</ns0:p><ns0:p>Remarks: We have been unable to obtain a clear lateral view of this carapace, which appears to differ from those of taxa A and C by having sharper anterior and posterior extremities and no obvious overlap of the valves. We speculate that Taxon B belongs to either Pontocypridoidea or Cypridoidea according to preserved features. Taxon C Fig. <ns0:ref type='figure' target='#fig_4'>1C</ns0:ref> Material. One articulated carapace, well-preserved. Remarks: The carapace is similar in lateral outline, but laterally more inflated, than that of Taxon A. Moreover, Taxon A has an apparently smooth exterior while Taxon C is distinctly punctate. Based on its overall shape and distinct punctation it has strong similarities to species of the nonmarine group Harbinia Tsao, 1959 (an ingroup of Cypridoidea).</ns0:p><ns0:p>Syninclusions: cf. Psychodidae (Diptera); Alavesia sp. (Diptera); Coleoptera sp.; Euarthropoda sp. (one unidentifiable remain and three isolated legs); Cymothoida sp. (Isopoda) <ns0:ref type='bibr'>(Schädel et al. in review)</ns0:ref>.</ns0:p></ns0:div>
<ns0:div><ns0:head>Discussion</ns0:head><ns0:p>Even though the studied ostracod specimens are well-preserved, their positions within the amber piece do not allow us to study further details, so we are unable to identify them beyond suborder level with confidence. This limitation in turn impedes the palaeoenviromental interpretation of the assemblage. Nevertheless, we can speculate, on the basis of gross morphological similarities, that at least taxa A and B belong to either of two ingroups of Cypridocopina, namely Pontocypridoidea and Cypridoidea. Moreover, as we don't know if the specimens are adults, and the shape of Taxon A could be a juvenile, which tend to have less inflated posterior margins. If we assume Taxon A is an adult, then it resembles a Pontocypridoidea or Cypridoidea (maybe even a Paracypridinae of the Candonidae). Cypridoidea are the predominant group in non-marine environments today, such as fresh water and saline inland water (athalassic environments). Pontocypridoidea, on the other hand, comprises marine and brackish-water taxa <ns0:ref type='bibr' target='#b4'>(Horne, 2003)</ns0:ref>. In the case of Taxon C the general shape and the association with the other two taxa suggest that it, too, is a cypridocopine. Species of the Cretaceous non-marine cypridoidean genus Harbinia Tsao, 1959, have similar carapace shape and punctate/reticulate ornament (e.g. Harbinia hapla <ns0:ref type='bibr' target='#b30'>Tsao, 1959</ns0:ref><ns0:ref type='bibr'>, illustrated by Ye et al. (2003: pl. 27, figs 2a-c)</ns0:ref>. However, we cannot rule out the possibility that it may belong to the group Cytheroidea (which is the dominant marine group today but also has non-marine lineages) without being able to see the characteristic adductor muscle scar patterns that distinguish between cypridoideans and cytheroideans.</ns0:p><ns0:p>All three ostracod specimens in this study lack preserved soft parts but are preserved with articulated carapaces, suggesting limited (if any) post-mortem transport. In view of our uncertainty about the precise systematic interpretation of our specimens we can contribute little to the discussion of the taphonomic and palaeoenvironmental interpretation of the Burmese amber assemblage. Both cypridoidean and pontocypridoidean taxa would be consistent with the mixed marine-freshwater-terrestrial components of the assemblage.</ns0:p><ns0:p>An ammonite shell preserved in Burmese amber argues that the Burmese amber forest was located near a dynamic and shifting coastal environment <ns0:ref type='bibr' target='#b33'>(Yu et al., 2019)</ns0:ref>, a conclusion supported by the occurrence of a marine myodocopan ostracod <ns0:ref type='bibr' target='#b31'>(Xing et al., 2018)</ns0:ref>. Also, semiaquatic and aquatic insects occur in Burmese amber, including Ochteridae (Hemiptera), Heteroceridae (Coleoptera), Chresmododea and Gerridae (Hemiptera), Dytiscidae and Gyrinidae (Coleoptera), adults and larvae of Odonata, larvae of Psephenidae, Trichoptera, and Ephemeroptera <ns0:ref type='bibr' target='#b34'>(Zhang, 2017;</ns0:ref><ns0:ref type='bibr' target='#b31'>Xing et al., 2018;</ns0:ref><ns0:ref type='bibr' target='#b23'>Schädel et al., 2020)</ns0:ref>. The herein presented ostracods are within the same amber piece as a fossil, supposedly aquatic living, isopod <ns0:ref type='bibr'>(Schädel et al. in review)</ns0:ref>. Actuo-palaeontological experiments <ns0:ref type='bibr' target='#b25'>(Schmidt & Dilcher, 2007)</ns0:ref> have demonstrated, that it is easily possible for aquatic organisms to be trapped in submerged bodies of resin. Several records of delicate arthropod remains from groups with supposed aquatic lifestyle <ns0:ref type='bibr' target='#b5'>(Heard et al., 2018;</ns0:ref><ns0:ref type='bibr' target='#b22'>Schädel et al., 2019;</ns0:ref><ns0:ref type='bibr' target='#b23'>Schädel et al. 2020;</ns0:ref><ns0:ref type='bibr' target='#b26'>Serrano-Sánchez et al., 2015;</ns0:ref><ns0:ref type='bibr' target='#b27'>Serrano-Sánchez et al., 2016)</ns0:ref> indicate that the result of in-situ embedment of aquatic organisms is present in many amber sites. Recurrent flows of resin and changing water levels can explain the preservation of aquatic and non-aquatic organisms -such as the dipterans in the herein presented assemblage -in the same amber piece <ns0:ref type='bibr' target='#b31'>(Xing et al., 2018)</ns0:ref>.</ns0:p></ns0:div>
<ns0:div><ns0:head>Conclusions</ns0:head><ns0:p>The three specimens are new records of podocopan ostracods to be reported from Burmese Cretaceous amber, and most likely belong to the Order Podocopida, Suborder Cypridocopina, and either the Superfamily Cypridoidea (indicative of non-marine environments) or the Superfamily Pontocypridoidea (indicative of marine/brackish environments). If the former be true, these would be the oldest non-marine ostracods preserved in amber. Either superfamily assignment would be consistent with previous evidence of a mixed marine-freshwater-terrestrial assemblage deposited in a coastal setting. </ns0:p></ns0:div><ns0:figure xml:id='fig_0'><ns0:head /><ns0:label /><ns0:figDesc>articulated carapace, moderately well preserved.</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_1'><ns0:head /><ns0:label /><ns0:figDesc>Locality and horizon. Hukawng Valley, Kachin Province, Myanmar (E 96°36′15″, N 26°13′47″, accuracy of about 10 km); lowermost Cenomanian, lowermost Upper Cretaceous.Description. Carapace small, rounded subtriangular in lateral view. Left valve slightly overlapping right valve along ventral and posterior margins. Maximum height in front of midlength. Anterior margin broad, almost equicurvate. Posterior margin narrower than anterior one, rounded and equicurvate. Carapace distinctly punctate tending to reticulation. Internal features not observable.</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_2'><ns0:head>Figure explanations Fig</ns0:head><ns0:label>explanations</ns0:label><ns0:figDesc>Figure explanations</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_3'><ns0:head>Fig. 2 :</ns0:head><ns0:label>2</ns0:label><ns0:figDesc>Fig. 2: Taxon A. A−B: light microscopy, mixed translucent and reflected light. A: dorsolateral view. B: left lateral view, side 1; C-L: volume rendering images based on µCT data. C: right lateral view; D: red-cyan stereo anaglyph, posterolateral view; E: posterior view; F: ventral view; G: dorsal view; H: left lateral view; I: red-cyan stereo anaglyph, anterolateral view; J: anterior view; K: ventral view, presumed pyrite crystals in red; L: right lateral, presumed pyrite crystals in red.</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_4'><ns0:head>Figure 1 Fig</ns0:head><ns0:label>1</ns0:label><ns0:figDesc>Figure 1</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_5'><ns0:head>Figure 2</ns0:head><ns0:label>2</ns0:label><ns0:figDesc>Figure 2</ns0:figDesc></ns0:figure>
<ns0:note place='foot'>PeerJ reviewing PDF | (2020:07:50630:2:0:NEW 11 Sep 2020)</ns0:note>
</ns0:body>
" | "Dear editor,
Thank you very much for your efforts on this MS, especially for the thoughtful comments. The manuscript has been revised accordingly. Hopefully we have addressed all of your concerns.
Attached please find the revised MS.
Kind regards,
He Wang
Authors’ replies are in red:
Editor (Kenneth De Baets)
1. line 30, “indicate” is replaced by “which could be consistent with”.
OK. We changed it.
2. line 30, I feel a more cautious formulation more in line with what you write in the text would be more appropriate. I also think it is worth to write marginal marine environment rather than coastal paleoenvironment as the former is more general for all kind of environment at the transition from sea to land while the latter is considered more specific by other but not all authors.
Thanks for the comments. We changed it as you suggest.
3. line 65-66, It would be appropriate to cite a reference in this context.
Haug, J.T., Azar, D., Ross, A. et al. Comment on the letter of the Society of Vertebrate Paleontology (SVP) dated April 21, 2020 regarding “Fossils from conflict zones and reproducibility of fossil-based scientific data”: Myanmar amber. PalZ 94, 431–437 (2020). https://doi.org/10.1007/s12542-020-00524-9
OK. We added it.
" | Here is a paper. Please give your review comments after reading it. |
9,775 | "<ns0:abstract xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:p>Background. Flower and fruit development are vital stages of the angiosperm lifecycle. We previously investigated the multi-silique trait in the rapeseed (Brassica napus) line zws-ms on a genomic and transcriptomic level, leading to the identification of two genomic regions and several candidate genes associated with this trait. However, some events on transcriptome level, like alternative splicing, were poorly understood.</ns0:p><ns0:p>Methods. Buds from zws-ms and its near-isogenic line (NIL) zws-217 were sampled and RNA was isolated to perform the transcriptomic sequencing. The numbers and types of alternative splicing (AS) events from the two lines were counted and classified. Genes with AS events and expressed differently between the two lines, as well as genes with AS events which occurred in only one line were emphasized. Their annotations were further studied.</ns0:p><ns0:p>Results. From the plants in Xindu District, an average of 205,496 AS events, which could be sorted into 5 AS types, were identified. zws-ms and zws-217 shared highly similar ratios of each AS type: The alternative 5' and 3' splice site types were the most common, while the exon skipping type was observed least often. Eleven differently expressed AS genes were identified, of which four were upregulated and seven were downregulated in zws-ms. Their annotations implied that five of these genes were directly associated with the multi-silique trait. Additionally, the 205 line-specifically expressed AS genes were analyzed, of which 187 could be annotated, and three were considered to be related to the multi-silique trait.</ns0:p><ns0:p>Discussion. This study provides new insights into the agronomically important multi-silique trait in rapeseed on transcriptome level and screens outs some candidate genes.</ns0:p></ns0:div>
</ns0:abstract>
<ns0:body xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:head>Introduction</ns0:head><ns0:p>Rapeseed (Brassica napus L.), an allotetraploid with a complex genome (AACC, 2n = 38), is the second leading source of vegetable oil globally <ns0:ref type='bibr' target='#b15'>(Liu et al., 2015)</ns0:ref>. The agronomic traits related to rapeseed yield include the pod (silique) number per plant, branch number, and seed weight <ns0:ref type='bibr' target='#b15'>(Liu et al., 2015;</ns0:ref><ns0:ref type='bibr' target='#b44'>Zhang et al., 2006;</ns0:ref><ns0:ref type='bibr' target='#b15'>Li et al., 2015)</ns0:ref>. We previously reported that zws-ms, a multisilique rapeseed line <ns0:ref type='bibr' target='#b3'>(Chai et al., 2019)</ns0:ref>, produces three independent pistils and 9 to 10 stamens on the same receptacle in each flower, which consequently leads to the formation of three independent siliques on each carpopodium rather than the single siliques typically observed.</ns0:p><ns0:p>Moreover, this trait was found to be affected by the environment, with temperature considered to be the factor most likely to switch on/off the formation of multi-silique.</ns0:p><ns0:p>Temperate is a major environmental factor that regulates various aspects of plant morphology, physiology, and biochemistry, affecting germination, growth, development, and flowering <ns0:ref type='bibr' target='#b18'>(Ren et al., 2019)</ns0:ref>. Fertility in crops such as rapeseed <ns0:ref type='bibr' target='#b42'>(Yu et al., 2015)</ns0:ref> and rice (Oryza sativa) <ns0:ref type='bibr' target='#b43'>(Yu et al., 2017)</ns0:ref> is affected by temperature. In winter rapeseed lines, although a period of vernalization under low temperature is necessary to initiate flowering, cold stress inhibits growth and development, disturbs metabolism, and causes wilting or even death. Notably, cold stress also induces alternative splicing (AS) in plants <ns0:ref type='bibr' target='#b13'>(Palusa et al., 2007;</ns0:ref><ns0:ref type='bibr'>Iida, 2004)</ns0:ref>.</ns0:p><ns0:p>AS is defined as the mechanism by which primary transcripts are processed into two or more mature isoforms, which enables a single gene to produce diverse protein products <ns0:ref type='bibr' target='#b14'>(Pan et al., 2008;</ns0:ref><ns0:ref type='bibr' target='#b20'>Sablok et al., 2011)</ns0:ref>. These proteins differ from each other not only in structure but also possibly in function, subcellular localization, and/or stability <ns0:ref type='bibr'>(Huang et al., 2019;</ns0:ref><ns0:ref type='bibr'>Chauhan et al., 2019)</ns0:ref>. AS is common in plants; for example, in Arabidopsis thaliana, more than 60% of intron-containing genes undergo AS <ns0:ref type='bibr' target='#b26'>(Syed et al., 2012)</ns0:ref>. Many environmental factors regulate AS events in plants, including CO 2 concentration <ns0:ref type='bibr'>(Huang et al., 2019</ns0:ref><ns0:ref type='bibr'>), light (Godoy et al., 2019)</ns0:ref>, salt stress <ns0:ref type='bibr'>(Ding et al., 2014)</ns0:ref>, and nutrient deficiencies <ns0:ref type='bibr' target='#b10'>(Nishida et al., 2017)</ns0:ref>. AS not only provides an important source of transcriptomic and proteomic diversity and plasticity for use in PeerJ reviewing PDF | (2019:12:44287:1:1:NEW 5 May 2020)</ns0:p><ns0:p>Manuscript to be reviewed natural selection <ns0:ref type='bibr'>(Labadorf et al., 2010)</ns0:ref>, but it also plays specific roles in the response <ns0:ref type='bibr'>(Chauhan et al., 2019)</ns0:ref> or adaptation to environmental stresses <ns0:ref type='bibr'>(Filichkin et al., 2015)</ns0:ref>. <ns0:ref type='bibr' target='#b37'>Guo et al. (2019)</ns0:ref> identified four splicing variants of two BnCYCD3-1-LIKE genes in B. napus and found evidence that their AS may play an important role in the response to environmental stresses. <ns0:ref type='bibr' target='#b34'>Xia et al. (2017)</ns0:ref> discovered that the AS with intron retention of EARLY MATURITY8 (EAM8) led to early flowering in a barley (Hordeum vulgare) landrace, while in shepherd's purse (Capsella bursapastoris), flowering time varies with changes in the splicing of a FLOWERING LOCUS C (FLC) homolog <ns0:ref type='bibr' target='#b24'>(Slotte et al., 2009)</ns0:ref>. In addition, the heterologous expression of a vacuolar membrane Na + /H + antiporter gene (SsNHX1) AS variant from seepweed (Suaeda salsa) enhances the salt tolerance of Arabidopsis <ns0:ref type='bibr'>(Li et al., 2009)</ns0:ref>.</ns0:p><ns0:p>As mentioned above, low temperatures switch off the multi-silique trait in zws-ms rapeseed.</ns0:p><ns0:p>When zws-ms plants were planted in Xindu, Sichuan Province, China, the multi-silique trait was continuously stable for years; however, when they were grown in Ma'erkang, Sichuan Province, where the annual average temperature is consistently 7.6 °C lower, the multi-silique trait disappeared and all plants displayed normal siliques <ns0:ref type='bibr' target='#b3'>(Chai et al., 2019)</ns0:ref>. We previously investigated the association of chromosomal regions with this trait, at the genomic and transcriptomic levels, selecting potential candidates from the differentially expressed genes (DEGs) between the multi-and single-silique plants. However, the involvement of posttranscriptional modifications and the mechanisms by which temperature regulates this multisilique trait remain unclear. AS is often responsive to cold stress in plants <ns0:ref type='bibr'>(Iida, 2004;</ns0:ref><ns0:ref type='bibr' target='#b13'>Palusa et al., 2007)</ns0:ref> and is a mechanism by which plants perceive temperature fluctuations and modulate the activity of their transcription factors <ns0:ref type='bibr' target='#b22'>(Seo et al., 2013)</ns0:ref>. In view of the above insights, we analyzed AS using transcriptome sequencing (RNA-seq) in this study. High-throughput RNAseq technology is a widely used, highly efficient, and economical strategy for transcriptomic profiling <ns0:ref type='bibr' target='#b28'>(Tong et al., 2013;</ns0:ref><ns0:ref type='bibr' target='#b33'>Wang et al., 2009)</ns0:ref>. It has become increasingly popular because of the following qualities <ns0:ref type='bibr' target='#b7'>(Mortazavi et al., 2008;</ns0:ref><ns0:ref type='bibr' target='#b12'>Ozsolak & Milos, 2011;</ns0:ref><ns0:ref type='bibr' target='#b6'>Marioni et al., 2008;</ns0:ref><ns0:ref type='bibr' target='#b28'>Tong et al., 2013;</ns0:ref><ns0:ref type='bibr' target='#b25'>Sultan et al., 2008)</ns0:ref>: (1) It can be used to detect and quantify the expression of genes, including those expressed at low levels; (2) it can facilitate the annotation of genes and lead to the discovery of novel genes or transcripts; (3) the results are highly reproducible between both technical and biological replicates; and (4) it can detect AS events. We performed transcriptome sequencing (RNA-seq) on the flower buds of zws-ms and its near-isogenic line (NIL), zws-217, which produces normal single siliques. This facilitated the identification of the AS events in both lines and the analysis of the differently expressed AS genes and those with line-specific AS events. Combining these data with Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) annotations, we identified likely candidate genes related to the multi-silique trait. To the best of our knowledge, this is the first time that the regulation of flower/fruit morphology by AS has been investigated in rapeseed, and our results provide insights into this field more generally.</ns0:p></ns0:div>
<ns0:div><ns0:head>Materials & Methods</ns0:head></ns0:div>
<ns0:div><ns0:head>Plant Materials and Growth Conditions</ns0:head><ns0:p>The rapeseed line zws-ms and its NIL, zws-217 <ns0:ref type='bibr' target='#b3'>(Chai et al., 2019)</ns0:ref>, were kept in the Crop Research Institute, Sichuan Academy of Agricultural Sciences, China. Both zws-ms and zws-217 were homozygous for almost all genes, differing from each other only in the multi-silique trait of zws-ms (Figure <ns0:ref type='figure'>1</ns0:ref>). The NILs zws-217 and zws-ms were both grown in an experimental field in the Xindu District of Chengdu in the Sichuan Basin, China, under normal environmental conditions. Additionally, the both lines were also grown in Ma'erkang, a mountainous area in western Sichuan, with a much lower annual average temperature.</ns0:p></ns0:div>
<ns0:div><ns0:head>Total RNA Extraction and Sequencing Library Construction</ns0:head><ns0:p>Three zws-ms plants (samples T01, T02, and T03) and three zws-217 plants (T04, T05, and T06) were selected for RNA isolation, as described previously <ns0:ref type='bibr' target='#b3'>(Chai et al., 2019)</ns0:ref>. Flower buds were detached from each plant at the budding stage (BBCH 57), and their total RNA was extracted using an RNA Isolation Kit (Tiangen, Beijing, China). The quality and concentration of the RNA were determined using a NanoDrop 2000 (Thermo Fisher Scientific, Waltham, MA, USA), and the sequencing libraries were generated using an RNA Library Prep Kit for Illumina (New England Biolabs, Ipswich, MA, USA).</ns0:p></ns0:div>
<ns0:div><ns0:head>Sequencing and Expression Analysis</ns0:head><ns0:p>The samples were sequenced on a HiSeq X Ten platform (Illumina, San Diego, CA, USA) and paired-end reads were generated. Low-quality reads and adaptor sequences were removed, and clean reads were used for the following analysis. TopHat2 <ns0:ref type='bibr'>(Kim et al., 2013)</ns0:ref> was used to map the clean reads onto the Brassica napus reference genome <ns0:ref type='bibr'>(Chalhoub et al., 2014)</ns0:ref>. The number of fragments per kilobase of transcripts per million fragments mapped (FPKM) was calculated to represent the gene expression level, and the DESeq R package <ns0:ref type='bibr' target='#b0'>(Anders & Huber, 2010)</ns0:ref> was used to analyze the differential expression. The P-value was adjusted using Benjamini and Hochberg's approach to control the false discovery rate (FDR). The relative expression levels of each transcript calculated using DESeq were used to define the DEGs, which were defined as having a fold change > 4 and an FDR < 0.01. Pearson's correlation coefficients were determined for the three biological replicates of each line to determine the reliability of the DEGs.</ns0:p></ns0:div>
<ns0:div><ns0:head>AS Event Analysis</ns0:head><ns0:p>The cleaned sequence data were aligned to the reference genome using TopHat with default settings. The resultant gapped alignment data in a binary alignment format were then used as an input for Cufflinks and Cuffcompare, which were run using the default settings to assemble the transcripts and identify splicing junctions from the alignment data. For the AS detection and annotation, the AS events were annotated with ASprofile, which uses Cufflinks and Cuffcompare outputs as input data. Default parameters of the software were used.</ns0:p></ns0:div>
<ns0:div><ns0:head>Annotation of Genes</ns0:head><ns0:p>Gene function was annotated based on the following databases: Nr (NCBI nonredundant protein sequences), Nt (NCBI nonredundant nucleotide sequences), Pfam (Protein family), KOG/COG (Clusters of Orthologous Groups of proteins), Swiss-Prot (a manually annotated and reviewed protein sequence database), KO (KEGG Ortholog database), and GO (Gene Ontology).</ns0:p><ns0:p>The GO enrichment analysis of the DEGs was performed using the GOseq R packages based on a Wallenius noncentral hypergeometric distribution <ns0:ref type='bibr' target='#b41'>(Young et al., 2010)</ns0:ref>, which can adjust for gene length bias in the DEGs.</ns0:p><ns0:p>The KEGG database <ns0:ref type='bibr'>(Kanehisa et al., 2007)</ns0:ref> is a resource used to explore the high-level functions and utilities of the biological system, such as the cell, organism, and ecosystem, from molecular-level information, especially using large-scale molecular datasets generated from genome sequencing and other high-throughput experimental technologies (http://www.genome.jp/kegg/). KOBAS <ns0:ref type='bibr' target='#b5'>(Mao et al., 2005)</ns0:ref> software was used to test the statistical enrichment of the DEGs in the various KEGG pathways. Default parameters were used.</ns0:p></ns0:div>
<ns0:div><ns0:head>Results</ns0:head></ns0:div>
<ns0:div><ns0:head>Transcriptome Sequencing</ns0:head><ns0:p>Flower buds from three plants of both the multi-silique line zws-ms and the single-silique NIL zws-217 (Figure <ns0:ref type='figure'>1</ns0:ref>) were sampled for RNA extraction. The sequencing saturation and cluster analysis of the samples were determined to ensure the validity of the data. In total, 65.6 Gb of clean data were generated, with an average Q30 value of 90.54%. Each sample generated about 36.65 M clean reads with an average GC content of 47.23% (Table <ns0:ref type='table' target='#tab_1'>S1</ns0:ref>). The average proportion of total reads mapped to the reference genome for each sample was 73.72% (Table <ns0:ref type='table'>S2</ns0:ref>).</ns0:p><ns0:p>Validation of transcriptome sequencing data was previously confirmed by qPCR <ns0:ref type='bibr' target='#b3'>(Chai et al., 2019)</ns0:ref>. Similarly, samples from colder area Ma'erkang also generated abundant data (Table <ns0:ref type='table'>S3</ns0:ref> and S4).</ns0:p></ns0:div>
<ns0:div><ns0:head>AS Event Identification and Analysis</ns0:head><ns0:p>According to <ns0:ref type='bibr' target='#b17'>(Reddy, 2007)</ns0:ref>, alternative splicing events were sorted into 5 classes: Alternative 3' splice site, Alternative 5' splice site, Exon Skipping, Intron Retention and Mutually Exclusive</ns0:p><ns0:p>Exons. The six samples grown in Xindu under normal conditions displayed an average of 205,496 AS events (Table <ns0:ref type='table' target='#tab_1'>1</ns0:ref>; Table <ns0:ref type='table'>S5</ns0:ref>). Using TopHat, the proportions of each AS type were analyzed in both zws-ms and zws-217. The two lines shared highly similar ratios of each AS type, with the alternative 5' splice site and alternative 3' splice site types being the most commonly observed, at 43.48% and 42.77% of AS events for both lines, respectively. The mutually exclusive exons type was the next most common (6.61%), followed by the Intron Retention type (5.92%), and the least common types was Exon Skipping, which represented just 1.22% of the AS events (Figure <ns0:ref type='figure'>2</ns0:ref>; Table <ns0:ref type='table' target='#tab_1'>1</ns0:ref>).</ns0:p><ns0:p>As to the plants grown in colder area Ma'erkang, the two lines also shared highly similar ratios of each AS type: the alternative 5' splice site and alternative 3' splice site types represented the greatest proportion, at 42.13% and 41.29%, respectively; while the exon skipping accounted least proportion 1.51% (Table <ns0:ref type='table'>S6</ns0:ref>). The number of each type of AS events were significantly reduced in colder area Ma'erkang, except the Intron Retention.</ns0:p></ns0:div>
<ns0:div><ns0:head>Annotation of the Alternatively Spliced Genes</ns0:head><ns0:p>To study the biological functions of the genes with AS events, GO and KEGG pathway enrichment analyses were performed. The GO annotations included 17 terms involved in biological processes (BP; Figure <ns0:ref type='figure'>3</ns0:ref>), 17 terms associated with cellular components (CC), and 20 terms involved in molecular functions (MF). The most highly enriched BP terms observed in the alternatively spliced genes included 'biological process,' 'cellular process,' and 'metabolic process.' The most common CC categories were 'cell part,' 'cellular component,' and 'intracellular part.' In the MF category, the most enriched terms were 'molecular function', 'binding', and 'catalytic activity'.</ns0:p><ns0:p>These pathways were further classified into five major groups: metabolism, genetic information processing, cellular processes, environmental information processing, and organismal systems (Figure <ns0:ref type='figure'>4</ns0:ref>). Of these, the subgroups 'ribosome,' 'biosynthesis of amino acids,' 'carbon metabolism,' and 'plant hormones signal transduction' contained the highest number of annotated genes.</ns0:p></ns0:div>
<ns0:div><ns0:head>DEGs with AS and Their Arabidopsis Orthologs</ns0:head><ns0:p>DESeq software was used to identify the different expression levels of the AS genes in zws-ms and zws-217. Eleven differently expressed AS genes were identified, of which four were upregulated and seven were downregulated in zws-ms (Table <ns0:ref type='table'>2</ns0:ref>).</ns0:p><ns0:p>The Arabidopsis orthologs of these differentially expressed AS genes were identified using</ns0:p><ns0:p>The Arabidopsis Information Resource (TAIR; https://www.arabidopsis.org; Table <ns0:ref type='table'>3</ns0:ref>). The AT1G10760 encodes an α-glucan, water dikinase (GWD) required for starch degradation.</ns0:p></ns0:div>
<ns0:div><ns0:head>Genes with Line-specific AS Events</ns0:head><ns0:p>Genes with line-specific AS events, defined as those genes with a particular AS event(s) that occurred only in zws-ms or in zws-217, were also identified and analyzed. (AE), either at the 5' end, 3' end, or both; and (12) approximate AE (XAE). In total, 205 linespecifically expressed AS genes were detected, of which 187 could be annotated (Table <ns0:ref type='table'>S7</ns0:ref>). Ten genes related to 'ovule development', 'flower development' and other similar processes were selected for further study (Table <ns0:ref type='table'>4</ns0:ref>, Table <ns0:ref type='table'>S7</ns0:ref>): ( <ns0:ref type='formula'>1</ns0:ref> </ns0:p></ns0:div>
<ns0:div><ns0:head>Discussion</ns0:head><ns0:p>As an important post-transcriptional metabolic event, AS is involved in many plant growth and developmental processes, such as flowering induction <ns0:ref type='bibr'>(Eckardt, 2002;</ns0:ref><ns0:ref type='bibr' target='#b24'>Slotte et al., 2009)</ns0:ref> and the responses to environmental fluctuations and pathogen attacks <ns0:ref type='bibr' target='#b1'>(Barbazuk et al., 2008)</ns0:ref>. To the best of our knowledge, AS events have seldom been reported to regulate the development of flower/fruit morphology in higher plants. This study is the first to analyze the role of AS events in rapeseed flower/fruit development as a whole, let alone those related to the multi-silique trait.</ns0:p><ns0:p>We previously described the morphology and inheritance of the multi-silique trait in B.</ns0:p><ns0:p>napus <ns0:ref type='bibr'>(Jiang et al., 1998)</ns0:ref>, investigating the associated regions of chromosomes at the genomic level and transcriptomically exploring the DEGs in multi-silique and single-silique plants <ns0:ref type='bibr' target='#b3'>(Chai et al., 2019)</ns0:ref>. The multi-silique trait was found to be controlled by three recessive alleles and was significantly affected by environment; however, the mechanisms by which environmental factors affect this trait remained unknown, even if we knew that temperature could switch on/off the multi-silique trait <ns0:ref type='bibr' target='#b3'>(Chai et al., 2019)</ns0:ref>. As mentioned above, AS is a pathway by which the environment could regulate plant physiology, therefore in this study, we analyzed AS events in order to investigate the mechanism by which plants perceive temperature fluctuations.</ns0:p><ns0:p>In this study, we sampled the buds of three individual plants from zws-ms and zws-217 and subjected them to RNA-seq, generating 65.6 Gb of clean data. Of these, 73.72% of the reads could be mapped to the reference rapeseed genome, indicating that the data were high quality and assuring the accuracy of the subsequent analysis.</ns0:p><ns0:p>We identified all of the genes with AS events in the zws-ms and zws-217 plants. Among the genes with AS events, 11 were significantly differently expressed between the multi-silique zwsms line and its NIL, zws-217, which produces normal siliques. We analyzed their annotations and orthologs in Arabidopsis. One such ortholog, AT5G15470 (also known as Galacturonosyltransferase 14, GAUT14), is involved in cell wall pectin biosynthesis <ns0:ref type='bibr' target='#b2'>(Caffall et al., 2009)</ns0:ref>, and the gaut13 gaut14 double mutant was previously shown to be defective in pollen tube growth <ns0:ref type='bibr'>(Wang et al., 2013)</ns0:ref>. AT3G15420 (the ortholog of BnaA02g03080D) and</ns0:p><ns0:p>PeerJ reviewing PDF | (2019:12:44287:1:1:NEW 5 May 2020)</ns0:p><ns0:p>Manuscript to be reviewed AT3G10070 (the ortholog of BnaA09g45000D) encode subunits of the transcription factor complexes TFIIIC and TAF12, respectively. The former does not appear to be substantially involved in plant development; however, some members of the TAF family are involved in the regulation of morphology. The transgenic expression of TAF10 from clustered yellowtops (Flaveria trinervia) in Arabidopsis limited the development of the indeterminate inflorescence and resulted in the production of deformed leaves <ns0:ref type='bibr'>(Furumoto et al., 2005)</ns0:ref>. By contrast, the taf mutant in Arabidopsis has abnormal phyllotaxis and lacks proper vegetative meristem activity <ns0:ref type='bibr' target='#b27'>(Tamada et al., 2007)</ns0:ref>, indicating the important roles played by the TAFs in plant morphological development. Another DEG AS gene, BnaA04g16220D, is not annotated, and its Arabidopsis ortholog AT1G14800 is simply listed as an uncategorized nucleic acid-binding, OB-fold-like protein. The AS gene orthologs AT2G04900 and AT1G15060 encode an unknown protein and an uncategorized alpha/beta hydrolase family protein, respectively, so their roles in the regulation of the multi-silique trait are also currently unclear.</ns0:p><ns0:p>Another DEG AS gene ortholog, AT3G54620, is reported to encode a bZIP transcription factor-like protein. Members of this protein family are typically reported to regulate plant tolerance of environmental stresses. The transgenic expression of the maize (Zea mays) gene ZmbZIP72 in Arabidopsis enhanced its drought and salt tolerance <ns0:ref type='bibr' target='#b40'>(Ying et al., 2012)</ns0:ref>, while BnbZIP3, a ramie (Boehmeria nivea) bZIP transcription factor, also increased the drought, salinity, and heavy metal tolerances of transgenic Arabidopsis <ns0:ref type='bibr'>(Huang et al., 2016)</ns0:ref>. These genes are also involved in the regulation of other processes; for example, the repression of a bZIP transcription factor gene OsABI5 expression in rice resulted in low fertility <ns0:ref type='bibr' target='#b46'>(Zou et al., 2008)</ns0:ref>, while the transgenic expression of tomato (Solanum lycopersicum) SlbZIP2 in tobacco (Nicotiana benthamiana) increased leaf thickness <ns0:ref type='bibr' target='#b23'>(Seong et al., 2016)</ns0:ref>. To date, however, there are no reports of bZIP genes playing a significant role in flower/fruit morphology.</ns0:p><ns0:p>Other AS gene orthologs included AT5G16210, encoding a member of the HEAT repeatcontaining protein family, which are considered to be involved in intracellular transport <ns0:ref type='bibr'>(Hernández-Torres et al., 2014;</ns0:ref><ns0:ref type='bibr' target='#b11'>Oeffinger et al., 2004)</ns0:ref>. Although BnaC06g16950D is not annotated, its ortholog, AT3G59000, was identified as encoding an F-box/RNI-like superfamily protein in Arabidopsis, which typically function in the plant hormone signaling pathways <ns0:ref type='bibr'>(Gao et al., 2009)</ns0:ref>. Similarly, the ortholog AT4G16900 encodes a TIR-NBS-LRR class protein, which are known to be involved in disease resistance <ns0:ref type='bibr' target='#b37'>(Xun et al., 2019)</ns0:ref> and hormonal responses <ns0:ref type='bibr' target='#b21'>(Sarazin et al., 2015)</ns0:ref>. <ns0:ref type='bibr'>Moreover, Hewezi et al. (2006)</ns0:ref> unexpectedly found that these proteins are associated with developmental abnormalities; transgenic sunflowers (Helianthus annuus) expressing the antisense sequence complementing PLFOR48 in sunflower (Helianthus annuus L.), which encodes a TIR-NBS-LRR-type protein, showed stunted growth and a reduction in apical dominance; whereas the pods of transgenic tobacco (N. tabacum) lacking PLFOR48 expression were smaller and showed severe deformations. This indicates that TIR-NBS-LRR-type proteins can regulate the morphology of plants, including fruit morphology, to some extent.</ns0:p><ns0:p>Finally, the AS gene ortholog AT1G10760, which encodes a GWD protein required for starch degradation, is involved in carbohydrate metabolism <ns0:ref type='bibr' target='#b9'>(Nadolska-Orczyk et al., 2017)</ns0:ref>. This gene was also reported to regulate seed size; <ns0:ref type='bibr' target='#b16'>Pirone et al. (2017)</ns0:ref> found that the length and width of the mature seeds were reduced in the gwd1 Arabidopsis mutant, while their density was increased.</ns0:p><ns0:p>To summarize, AT5G15470, AT3G10070, AT3G54620, AT4G16900, and AT1G10760 are all known to be involved in plant development; therefore, their corresponding rapeseed orthologs, BnaA02g02630D, BnaA09g45000D, BnaAnng30260D, BnaC07g33980D, and BnaC08g49610D, the expression levels of which differed significantly between zws-ms and zws-217, are considered to be potential candidate genes regulating the multi-silique trait.</ns0:p><ns0:p>We also explored the line-specific AS genes, which were similarly expressed between zwsms and zws-217, but contained stable and particular AS event(s) that differed between these two lines. These genes are likely to qualitatively regulate the multi-silique trait. In this case, we could obtain better results by fine-classify the AS types into 12 subclasses, rather than 5 classes mentioned above. Because fine classifications could better identify differences between AS types more precisely and subtly. Thus, we found 205 genes of this type, of which 187 could be annotated. Due to the rarity of the multi-silique trait, we did not obtain much useful information from the KEGG pathway analysis. This meant that we were unable to relate this metabolic pathway information to the multi-silique trait directly; however, the GO analysis provided more potential clues. Among these, 10 genes were considered to be associated with flower/carpel/ovule development. BnaC06g32640D is annotated as being involved in the regulation of the vegetative-to-reproductive phase transition in the meristem (GO:0010228) and in ovule development (GO:0048481). Its Arabidopsis ortholog, AT1G71692, is annotated as (GR), which was found to increase the fineness (mass per unit length) and bundle strength of cotton (Gossypium hirsutum) fiber when transgenically expressed <ns0:ref type='bibr' target='#b29'>(Tuttle et al., 2015)</ns0:ref>. Since cotton fibers are single cells initiating from the epidermis of the outer integument of the ovules <ns0:ref type='bibr' target='#b19'>(Ruan et al., 2004)</ns0:ref>, it can be inferred that GR regulates ovule development to some extent.</ns0:p><ns0:formula xml:id='formula_0'>AGAMOUS-LIKE12 (AGL12).</ns0:formula><ns0:p>The line-specific AS gene ortholog AT2G04030 encodes HSP90, a member of the heat shock proteins (HSPs), which are commonly produced in response to heat. The HSPs are molecular chaperones that prevent protein aggregation and mediate the refolding of heatdenatured proteins <ns0:ref type='bibr' target='#b8'>(Murano et al., 2017)</ns0:ref>. Moreover, a HSP was found to be upregulated in the fiber-bearing ovules of cotton (Gossypium hirsutum) <ns0:ref type='bibr'>(Lee et al., 2006)</ns0:ref> implying some unknown function in ovule development. Another ortholog, AT3G28730 (also known as structure-specific recognition protein SSRP1), was also found to regulate floral development, as the ssrp1-2 mutant Arabidopsis produced small and deformed petals with shorter stamens <ns0:ref type='bibr'>(Lolas et al., 2010)</ns0:ref>.</ns0:p><ns0:p>AT4G24560 encodes a ubiquitin-specific protease (UBP), which are known to play critical roles in protein deubiquitination in plants <ns0:ref type='bibr'>(Liu et al., 2008)</ns0:ref>. The UBPs are also involved in plant development; <ns0:ref type='bibr'>Liu et al. (2008)</ns0:ref> found that knocking out UBP15 function in Arabidopsis resulted in the production of smaller flowers and shorter siliques. The final line-specific AS ortholog, AT5G15020, encodes an SIN3-LIKE 2 protein (SNL2) known to be important for seed germination or dormancy <ns0:ref type='bibr' target='#b32'>(Wang et al., 2016;</ns0:ref><ns0:ref type='bibr'>Wang et al., 2013)</ns0:ref>.</ns0:p><ns0:p>To date, there is some evidence to show that the line-specific AS orthologs AT2G20180, AT4G00050, AT3G54660, and AT2G04030 are related to the regulation of flower/fruit morphology, with clear roles reported for AT1G71692, AT3G28730, and AT4G24560.</ns0:p><ns0:p>Consequently, their orthologs in rapeseed, BnaC06g32640D, BnaC07g25280D, and BnaC01g16410D, respectively, are considered to be important candidate genes regulating the multi-silique trait by conferring or removing some specific line-specific AS events.</ns0:p><ns0:p>Some of the genes/loci controlling silique development in Brassica plants have previously been reported. In addition to those regulating traits such as the seed weight and silique length <ns0:ref type='bibr' target='#b15'>(Liu et al., 2015)</ns0:ref> and the number of seeds per silique in B. napus <ns0:ref type='bibr' target='#b15'>(Li et al., 2015)</ns0:ref>, some genes related to silique morphology have been cloned and functionally analyzed. <ns0:ref type='bibr' target='#b36'>Xiao et al. (2013)</ns0:ref> fine-mapped a multi-locular silique gene, Bjln1, to a 208-kb region on chromosome A7 in Brassica juncea and then revealed that it was the mutations in the CDS and promoter of BjuA07.CLV1 gene (equivalent to Bjln1) to cause the multi-locular trait <ns0:ref type='bibr' target='#b35'>(Xiao et al., 2018)</ns0:ref>. Both Manuscript to be reviewed semidominant mutation located on chromosome arm 2DL. Although several studies have explored the multi-pistil trait in wheat, no one has identified any of the specific genes responsible yet.</ns0:p><ns0:p>To sum up, the eight candidate genes mentioned above, including the five differently expressed AS genes of interest and the three genes with line-specific AS events, are therefore hypothesized to regulate the multi-silique trait in rapeseed zws-ms, based on their AS expression levels or line-specific AS events. These findings lay a foundation for further functional analyses in future. The data obtained from plants in colder environment, including the amount and the proportion of each AS type, was generally similar to that under normal conditions, while details are still under further investigation.</ns0:p></ns0:div>
<ns0:div><ns0:head>Conclusions</ns0:head><ns0:p>The utilization of heterosis is a way to increase the yield or improve the quality of crops.</ns0:p><ns0:p>Exploring new germplasm resources and genes, as well as clarifying their inheritance, is the foundation of obtaining of excellent hybrid. This study provides a novel inspection into the multi-silique trait in rapeseed from the transcriptional perspective by AS, deepening the understanding of its molecular mechanism. Further function verifications are now undergoing.</ns0:p><ns0:note type='other'>Figure 1</ns0:note><ns0:p>Multi-silique trait in zws-ms, compared to its near-isogenic line zws-217 with normal siliques.</ns0:p><ns0:p>A: main inflorescences from zws-217; B: main inflorescences from zws-ms; C: close-range of siliques from zws-217; D: close-range of siliques from zws-ms.</ns0:p><ns0:note type='other'>Figure 2</ns0:note><ns0:p>The number of each alternative splicing (AS) categories found in each sample.</ns0:p><ns0:p>A-C: three independent zws-ms plants T01, T02, and T03; D-F: three independent zws-217 plants T04, T05, and T06. Alternative 3' splice site: different-size mRNAs are produced depending on the usage of a proximal or distal 3' splice site; Alternative 5' splice site:</ns0:p><ns0:p>different-size mRNAs are produced depending on the use of a proximal or distal 5' splice site;</ns0:p><ns0:p>Exon Skipping: an exon is either included or excluded from the mRNA; Intron Retention: an intron is either retained or excised in the mRNA, resulting in different-size transcripts;</ns0:p><ns0:p>Mutually Exclusive Exons: adjacent exons are spliced in such a way that only one of them is included at a time in the mRNA.</ns0:p><ns0:note type='other'>Figure 3</ns0:note><ns0:p>Gene ontology (GO) terms associated with the alternatively spliced genes.</ns0:p><ns0:p>GO terms were divided into three categories: biological processes, cellular components, and molecular functions. The x-axis shows the GO categories and subclasses of the alternatively spliced genes. the y-axis shows the number or percentage of annotated alternatively spliced genes.</ns0:p><ns0:p>PeerJ reviewing PDF | (2019:12:44287:1:1:NEW 5 May 2020)</ns0:p><ns0:p>Manuscript to be reviewed</ns0:p><ns0:note type='other'>Figure 4</ns0:note><ns0:p>Classified KEGG pathways associated with the alternatively spliced genes.</ns0:p><ns0:p>The pathways were further classified into five major groups: metabolism, genetic information processing, cellular processes, environmental information processing, and organismal systems. The x-axis shows the number of annotated alternatively spliced genes and the proportion of them corresponding to each pathway category; the y-axis shows the pathway categories that contain more than one alternatively spliced genes.</ns0:p></ns0:div>
<ns0:div><ns0:head>Table 1(on next page)</ns0:head><ns0:p>Numbers of alternative splicing events in six samples.</ns0:p><ns0:p>Note: T01, T02, and T03: Buds of three independent zws-ms plants at the budding stage; T04, T05, and T06: Buds of three independent zws-217 plants at the budding stage.</ns0:p><ns0:p>Alternative 3' splice site: different-size mRNAs are produced depending on the usage of a proximal or distal 3' splice site; Alternative 5' splice site: different-size mRNAs are produced depending on the use of a proximal or distal 5' splice site; Exon Skipping: an exon is either included or excluded from the mRNA; Intron Retention: an intron is either retained or excised in the mRNA, resulting in different-size transcripts; Mutually Exclusive Exons: adjacent exons are spliced in such a way that only one of them is included at a time in the mRNA. 2 Note: T01, T02, and T03: Buds of three independent zws-ms plants at the budding stage; T04, T05, and T06: Buds of three independent zws-217 plants at 3 the budding stage. 4 Alternative 3' splice site: different-size mRNAs are produced depending on the usage of a proximal or distal 3' splice site; Alternative 5' splice site: 5 different-size mRNAs are produced depending on the use of a proximal or distal 5' splice site; Exon Skipping: an exon is either included or excluded from 6 the mRNA; Intron Retention: an intron is either retained or excised in the mRNA, resulting in different-size transcripts; Mutually Exclusive Exons: adjacent 7 exons are spliced in such a way that only one of them is included at a time in the mRNA.</ns0:p></ns0:div><ns0:figure xml:id='fig_0'><ns0:head /><ns0:label /><ns0:figDesc>following orthologs were identified: (1) AT5G15470, the ortholog of BnaA02g02630D, encodes galacturonosyltransferase 14 (GAUT14); (2) AT3G15420 encodes the transcription factor TFIIIC (tau55-related protein); (3) AT1G14800 encodes a nucleic acid-binding, OB-fold-like protein; (4) AT2G04900 encodes an unknown protein; (5) AT3G10070 encodes one of two Arabidopsis proteins with similarity to the TBP-associated factor, TAF12; (6) AT1G15060 encodes an alpha/beta hydrolase family protein; (7) AT3G54620 encodes a bZIP transcription factor-like protein; (8) AT5G16210 encodes a HEAT repeat-containing protein; (9) AT3G59000 encodes an F-box/RNI-like superfamily protein; (10) AT4G16900, the ortholog of BnaC07g33980D, encodes a member of the disease resistance protein (TIR-NBS-LRR class) family; and (11)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_1'><ns0:head>)</ns0:head><ns0:label /><ns0:figDesc>BnaC06g32640D was annotated as 'vegetative to reproductive phase transition of meristem (GO:0010228)' and 'Biological Process: ovule development (GO:0048481)'; (2) BnaC07g00780D was associated with 'reproductive structure development (GO:0048608)'; (3) BnaC04g31460D and (4) BnaC05g34570D were related to 'regulation of flower development (GO:0009909)'; BnaC07g22680D were annotated with 'development (GO:0048481)'; (7) BnaC07g25280D was annotated as 'flower morphogenesis; organ morphogenesis (GO:0009887)' and 'vegetative to reproductive phase transition of meristem (GO:0010228)'; (8) BnaC01g16410D was annotated as 'flower development (GO:0009908)'; (9) BnaC03g32190D was annotated as 'double fertilization forming a zygote and endosperm (GO:0009567)'; and (10) BnaCnng68400D was associated with 'carpel development (GO:0048440).'</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_2'><ns0:head /><ns0:label /><ns0:figDesc><ns0:ref type='bibr' target='#b15'>Peng et al. (2015)</ns0:ref> isolated the BnFUL gene in rapeseed, which isPeerJ reviewing PDF | (2019:12:44287:1:1:NEW 5 May 2020) Manuscript to be reviewed homologous to AGL8 in Arabidopsis. Although BnFUL was hypothesized to be involved in enhancing pod-shattering resistance, when introduced into Arabidopsis, two of the five transgenic plants expressing BnFUL unexpectedly had a multi-silique phenotype. However, the mechanisms by which BnFUL generates this multi-silique phenotype remain elusive thus far, making the AGL12 gene identified in this study a potentially important candidate gene. Other orthologs of the line-specific AS genes include AT2G20180 and AT4G00050, both of which encode phytochrome interacting factors (PIFs). Several transcription factors (AP1, SVP, LFY, AG, and SEP3) involved in the regulation of flowering are known to bind to the PIFs, suggesting a direct link with the reported flowering phenotype of the pif mutants (Leivar & Monte, 2014). AT5G17270 encodes a prenylyltransferase superfamily protein; however, to the best of our knowledge, there have been no reports about its development-related functions. The Arabidopsis ortholog of BnaC05g34570D is AT3G18600, which encodes a P-loop-containing nucleoside triphosphate hydrolase. While few studies have reported the functions of these proteins, Liu et al. (2016) reported that, in sesame (Sesamum indicum), one gene encoding a Ploop-containing nucleoside triphosphate hydrolase showed a reduced expression level in sterile buds, indicating that they may play a role in specifying/determining tapetal fate and development. Another line-specific AS ortholog, AT3G54660, encodes a glutathione reductase</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_3'><ns0:head>Fan</ns0:head><ns0:label /><ns0:figDesc>et al. (2014) and<ns0:ref type='bibr' target='#b38'>Yadava et al. (2014)</ns0:ref> reported that a mutation in BrCLV3, a homologue of CLAVATA3 in Arabidopsis, caused the production of multi-locular siliques in B. rapa. However, the multi-silique (or multi-pistil) phenotype of zws-ms is different from the above-motioned multi-locular trait; zws-ms produces three pods on each carpopodium, rather than multiple loculi per pod.Few studies have investigated this multi-silique trait in rapeseed; however, there have been similar reports of multi-pistil traits in other crops, particularly in wheat (Triticum aestivum):Duan et al. (2015) discovered a male-sterile wheat mutant, dms, with a dwarf status and multipistils, a pleiotropic phenotype found to be controlled by a single recessive gene, which was not identified.<ns0:ref type='bibr' target='#b37'>Guo et al. (2019)</ns0:ref> reported another multi-ovary trait in the wheat line DUOII, which was controlled by a dominant gene, and used a proteomics approach to propose some candidate proteins.<ns0:ref type='bibr' target='#b39'>Yang et al. (2017)</ns0:ref> mapped a gene promoting the formation of three pistils (Pis1) to chromosome 2D and identified some candidate genes according to their annotations, while Zhuet al. (2019) discovered a wheat multi-pistil mutant, 12TP, which was found to contain a PeerJ reviewing PDF | (2019:12:44287:1:1:NEW 5 May 2020)</ns0:figDesc></ns0:figure>
<ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='22,42.52,70.87,333.62,672.95' type='bitmap' /></ns0:figure>
<ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='24,42.52,70.87,525.00,393.00' type='bitmap' /></ns0:figure>
<ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='25,42.52,280.87,525.00,378.75' type='bitmap' /></ns0:figure>
<ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='27,42.52,70.87,525.00,455.25' type='bitmap' /></ns0:figure>
<ns0:figure type='table' xml:id='tab_1'><ns0:head>Table 1 .</ns0:head><ns0:label>1</ns0:label><ns0:figDesc>Numbers of alternative splicing events in six samples.</ns0:figDesc><ns0:table><ns0:row><ns0:cell>1</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>Line</ns0:cell><ns0:cell>Sample ID</ns0:cell><ns0:cell>Alternative 3' splice site</ns0:cell><ns0:cell>Alternative 5' splice site</ns0:cell><ns0:cell>Exon Skipping</ns0:cell><ns0:cell>Intron Retention</ns0:cell><ns0:cell>Mutually Exclusive Exons</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>T01</ns0:cell><ns0:cell>87699</ns0:cell><ns0:cell>89187</ns0:cell><ns0:cell>2333</ns0:cell><ns0:cell>11981</ns0:cell><ns0:cell>12765</ns0:cell></ns0:row><ns0:row><ns0:cell>zws-ms</ns0:cell><ns0:cell>T02</ns0:cell><ns0:cell>87603</ns0:cell><ns0:cell>88889</ns0:cell><ns0:cell>2404</ns0:cell><ns0:cell>11201</ns0:cell><ns0:cell>13100</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>T03</ns0:cell><ns0:cell>88055</ns0:cell><ns0:cell>89623</ns0:cell><ns0:cell>2638</ns0:cell><ns0:cell>12184</ns0:cell><ns0:cell>14060</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>T04</ns0:cell><ns0:cell>87953</ns0:cell><ns0:cell>89443</ns0:cell><ns0:cell>2567</ns0:cell><ns0:cell>12501</ns0:cell><ns0:cell>13887</ns0:cell></ns0:row><ns0:row><ns0:cell>zws-217</ns0:cell><ns0:cell>T05</ns0:cell><ns0:cell>88012</ns0:cell><ns0:cell>89588</ns0:cell><ns0:cell>2581</ns0:cell><ns0:cell>12722</ns0:cell><ns0:cell>13901</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>T06</ns0:cell><ns0:cell>88001</ns0:cell><ns0:cell>89408</ns0:cell><ns0:cell>2549</ns0:cell><ns0:cell>12373</ns0:cell><ns0:cell>13768</ns0:cell></ns0:row></ns0:table></ns0:figure>
<ns0:note place='foot'>PeerJ reviewing PDF | (2019:12:44287:1:1:NEW 5 May 2020)Manuscript to be reviewed</ns0:note>
<ns0:note place='foot'>PeerJ reviewing PDF | (2019:12:44287:1:1:NEW 5 May 2020)</ns0:note>
</ns0:body>
" | "Response to Editors and Reviewers
Dear Editors and Reviewers,
Thanks for your professional recommendations, which helps us to optimize our manuscript. We all look forward to publishing our article on your journal. So if there are still something to revise, please contact us at any time.
Our response to your comments are as following:
Reviewer 1 (Yongxiu Liu)
Basic reporting
no comment
Experimental design
no comment
Validity of the findings
no comment
Comments for the Author
In this manuscript, Chai et al reported an interesting investigation in genome-wide gene alternative splicing in rapeseed for the muti-silique trait. The authors totally identified about 220168 AS which could be sorted into 12 AS types. Eleven differently expressed AS genes were identified, and their annotations implied that five of these genes including BnFUL were directly associated with the multi-silique trait. Additionally, the 205 stably expressed AS genes were analyzed, of which 187 could be annotated, and three candidates including BnaC06g32640D, BnaC07g25280D, and BnaC01g16410D were considered to be related to the multi-silique trait. These findings may be very helpful for our understanding of muti-silique formation in rapeseed.
Some comments:
In the manuscript, authors mentioned that AS is a pathway by which the environment could regulate plant physiology, therefore in this study, we analyzed AS events in order to investigate the mechanism by which plants perceive temperature fluctuations. However, I could not see it from their data because they did RNA-seq using NILs zws-217 and zws-ms growing in the normal environmental conditions. Low temperatures can switch off the multi-silique trait in zws-ms rapeseed, I suggest authors should also do the RNA-seq at different temperature condition using zws-ms rapeseed.
Response:
Thanks for the suggestion. We have already added some relative data about plants from Ma’erkang, which is in highland of Sichuan and the annual average temperature is much lower. Please refer to the revised manuscript.
It's worth mentioning that, in fact, we originally had planned to put that data in our next paper; however, since you asked for the “RNA-seq at different temperature condition using zws-ms rapeseed”, we displayed some basic RNA-seq data and AS-relative data herein, which would not to break integrity of the next coming paper. It is important to investigate the data from colder conditions, but herein we mainly focused on AS events and AS genes. Thanks for your understanding.
Well, on the other hand, if you insist on more data about cold environment, please let us know and we can further discuss. Thanks.
Reviewer 2 (Anonymous)
Basic reporting
This manuscript is to investigate how alternative splicing may contribute to a multi-silique trait in Brassica napus by comparing the transcriptome difference between two NIL lines: one has such a trait, while the other does not. The authors previous found the genomic region and candidate genes associated with this trait. In this manuscript, the authors performed an RNAseq experiment and reported their data analysis in regards to all the transcripts that exhibited alternative splicing in this experiment, differentially expressed genes between the two NIL lines that also exhibited alternative splicing in this experiments, and line-specific alternatively spliced genes which the author defined as “genes with stable AS events”. The introduction provided an adequate overview of prior work that support the rationale of this study.
Experimental design
The RNAseq experiment was performed with three biological replicates of each comparison group, which is standard. However, there is a substantial lack of information and experimental confirmation as listed below:
1. Currently the manuscript does not include sufficient AS analysis results to support the conclusion. Supplementary tables need better title and column headers to provide more detailed information. For example, in Table S3, column G has a header of “event pattern”, while the actual cells only listed the coordinates of each splicing event. This does not provide any information as to the read count of each AS event. Information from all six samples should be analyzed and aggregated into one table to show a comparison across samples that directly support descriptions of AS distributions in Result section. The fact that the authors only described any genes that exhibit AS in any samples without their abundance and analysis on how much changes there may be between the two lines does not provide much insight in how AS may contribute to the multi-silique trait. Authors did mention eleven differentially expressed AS genes. However, this diminishes the value of alternative splicing study because changes in ratios of differently spliced isoforms of the same gene may be important in regulating protein abundance and function without changing total transcript level of that gene.
Response:
(1) We have aggregated original Table S3~S8 into a new table, showing a comparison across samples.
(2) You are right, there some genes which vary their functions or evern regulate some phenotype by changing their spliced isoforms, rather than transcript level. We also described this by citing some references ,such as “Guo et al., 2019”, “Xia et al., 2017” and “Slotte et al., 2009” and so on.
It is important, so we analyzed this point of view indeed: in sections below, we described the 205 “line-specific ASs”, which are the differently spliced isoforms in fact. They had some spliced isoforms only in genes in zws-ms line but not in zws-217 line; or on the contrary, some genes had special spliced isoforms only in zws-217 but not in zws-ms. They did not necessarily change the transcript level, but they produced differently spliced isoforms. We used different statements, but it is exactly the point you mentioned.
AS genes with different expression level and differently spliced isoforms (we described this as line-specific AS events in this study) are two important aspects, and that is why we analyzed them successively in this manuscript. Thanks for your understanding.
2. In addition, the authors reported 205 “stably expressed AS genes” which was later explained as “with a particular AS events that occurred only in zws-ms or in zws-217”. The term “Stably expressed AS genes” is not self-explanatory. This group of events are basically line-specific AS events according to this description. They could be interesting to have a deeper look and validated by rtPCR. Unfortunately, the authors did not validate any of them. Table S9 only provided their annotations without their abundance or any statistical measures.
Response:
(1) Good idea! The term “line-specific AS event” is a better definition. We have replaced the term “stably expressed AS” through the whole manuscript.
(2) We added abundance and statistical measures into the table and designated it as new Table S4, because we had combined original Table S3-S8 to generate a new Table S3, as you suggested in your above Point 1.
3. The description in “AS Event Analysis” and “Annotation of Genes” lacks details regarding statistical parameters, statistical tests used and FDR used, etc.
Response:
Thanks for the advice. We used the default parameters of the software, and we have added this into the manuscript.
4. There was no qPCR validation of differentially expressed AS genes. In table 2, the eleven differentially expressed AS genes are only reported as upregulated or downregulated without any fold change and statistics in there. These should be added.
Response:
(1) As we mentioned in the manuscript, the transcriptome data is based on our previous study “Chai et al. (2019). Identification of genomic regions associated with multi-silique trait in brassica napus. BMC Genomics 20, 304”, which had shown the qPCR validation.
(2) Thanks for the advice. We have already added the information as you required into Table 2.
5. Classes of alternative splicing events have been long established in the past (Reddy. Annu. Rev. Plant Biol. 2007. 58:267-94). The authors should use the existing classification and integrate theirs as subclasses.
Response:
We have already re-sorted the AS events into 5 types (Alternative 3’ splice site, Alternative 5’ splice site, Exon Skipping, Intron Retention and Mutually Exclusive Exons) according to the reference you recommended. Therefore, relative data involved in manuscript, tables and figures were throughout revised, please kindly check them.
There is one more thing to mention, we integrated original 12 subclasses into 5 classes throughout the manuscript except one point: when we identified the 205 line-specific genes, we still used the 12 subclasses, which could precisely identify subtle differences between zws-ms and zws-217; on the other hand, the “5 classes” method would result in some general/ambiguous classification, neglect some subtle differences among “subclasses”, and thus generate too many line-specific genes. For example, some AS events are sorted into the same class according to the general“5 classes” method, but in fact, they had some subtle and fine differences. And the “12 subclasses” method can distinguish them. Accordingly, we added description into Results and Discussion. Thanks for your understanding.
Validity of the findings
Without enough quantitative and validated results and appropriate analyses on these AS events, the conclusions of this manuscript are weak.
Response:
Thank Reviewer 2 for all the professional and patient suggestions above. We have revised and improved our manuscript according to your recommendation: (1) some tables were optimized; (2) we chose better definition for some words; (3) we cleared up some misunderstanding by answering your questions and improving the statement in manuscript; (4) some information was added to provide evidence (such as the part of qPCR); and so on. By doing these, I think the manuscript became more validated and better.
We hope the revised manuscript can meet your requirement.
" | Here is a paper. Please give your review comments after reading it. |
9,776 | "<ns0:abstract xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:p>Background. Flower and fruit development are vital stages of the angiosperm lifecycle.</ns0:p><ns0:p>We previously investigated the multi-silique trait in the rapeseed (Brassica napus) line zwsms on a genomic and transcriptomic level, leading to the identification of two genomic regions and several candidate genes associated with this trait. However, some events on transcriptome level, like alternative splicing, were poorly understood. Methods. Plants from zws-ms and its near-isogenic line (NIL) zws-217 were both grown in Xindu with normal conditions and a colder area Ma'erkang. Buds from the two lines were sampled and RNA was isolated to perform the transcriptomic sequencing. The numbers and types of alternative splicing (AS) events from the two lines were counted and classified. Genes with AS events and expressed differentially between the two lines, as well as genes with AS events which occurred in only one line were emphasized. Their annotations were further studied. Results. From the plants in Xindu District, an average of 205,496 AS events, which could be sorted into 5 AS types, were identified. zws-ms and zws-217 shared highly similar ratios of each AS type: The alternative 5' and 3' splice site types were the most common, while the exon skipping type was observed least often. Eleven differentially expressed AS genes were identified, of which four were upregulated and seven were downregulated in zws-ms. Their annotations implied that five of these genes were directly associated with the multi-silique trait. While samples from colder area Ma'erkang generated generally reduced number of each type of AS events except the Intron Retention; but the number of differentially expressed AS genes increased significantly.</ns0:p><ns0:p>Further analysis found that among the 11 differentially expressed AS genes from Xindu, three of them maintained the same expression models, while the other 8 genes did not showed significant difference between the two lines in expression level. Additionally, the 205 line-specifically expressed AS genes were analyzed, of which 187 could be annotated, and three were considered to be related to the multi-silique trait. Discussion. This study provides new insights into the agronomically important multi-silique trait in rapeseed on transcriptome level and screens outs some candidate genes.</ns0:p></ns0:div>
</ns0:abstract>
<ns0:body xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:head>Introduction</ns0:head><ns0:p>Rapeseed (Brassica napus L.), an allotetraploid with a complex genome (AACC, 2n = 38), is the second leading source of vegetable oil globally <ns0:ref type='bibr' target='#b33'>(Liu et al., 2015)</ns0:ref>. The agronomic traits related to rapeseed yield include the pod (silique) number per plant, branch number, and seed weight <ns0:ref type='bibr' target='#b33'>(Liu et al., 2015;</ns0:ref><ns0:ref type='bibr' target='#b76'>Zhang et al., 2006;</ns0:ref><ns0:ref type='bibr'>Li et al., 2015)</ns0:ref>. We previously reported that zws-ms, a multisilique rapeseed line <ns0:ref type='bibr' target='#b3'>(Chai et al., 2019)</ns0:ref>, produces three independent pistils and 9 to 10 stamens on the same receptacle in each flower, which consequently leads to the formation of three independent siliques on each carpopodium rather than the single siliques typically observed.</ns0:p><ns0:p>Moreover, this trait was found to be affected by the environment, with temperature considered to be the factor most likely to switch on/off the formation of multi-silique.</ns0:p><ns0:p>Temperate is a major environmental factor that regulates various aspects of plant morphology, physiology, and biochemistry, affecting germination, growth, development, and flowering <ns0:ref type='bibr' target='#b50'>(Ren et al., 2019)</ns0:ref>. Fertility in crops such as rapeseed <ns0:ref type='bibr' target='#b74'>(Yu et al., 2015)</ns0:ref> and rice (Oryza sativa) <ns0:ref type='bibr' target='#b75'>(Yu et al., 2017)</ns0:ref> is affected by temperature. In winter rapeseed lines, although a period of vernalization under low temperature is necessary to initiate flowering, cold stress inhibits growth and development, disturbs metabolism, and causes wilting or even death. Notably, cold stress also induces alternative splicing (AS) in plants <ns0:ref type='bibr' target='#b45'>(Palusa et al., 2007;</ns0:ref><ns0:ref type='bibr' target='#b22'>Iida, 2004)</ns0:ref>.</ns0:p><ns0:p>AS is defined as the mechanism by which primary transcripts are processed into two or more mature isoforms, which enables a single gene to produce diverse protein products <ns0:ref type='bibr' target='#b46'>(Pan et al., 2008;</ns0:ref><ns0:ref type='bibr' target='#b52'>Sablok et al., 2011)</ns0:ref>. These proteins differ from each other not only in structure but also possibly in function, subcellular localization, and/or stability <ns0:ref type='bibr' target='#b21'>(Huang et al., 2019;</ns0:ref><ns0:ref type='bibr' target='#b5'>Chauhan et</ns0:ref> PeerJ reviewing PDF | (2019:12:44287:2:0:NEW 4 Jul 2020)</ns0:p><ns0:p>Manuscript to be reviewed <ns0:ref type='bibr'>al., 2019)</ns0:ref>. AS is common in plants; for example, in Arabidopsis thaliana, more than 60% of intron-containing genes undergo AS <ns0:ref type='bibr' target='#b59'>(Syed et al., 2012)</ns0:ref>. Many environmental factors regulate AS events in plants, including CO 2 concentration <ns0:ref type='bibr' target='#b21'>(Huang et al., 2019)</ns0:ref>, light <ns0:ref type='bibr' target='#b15'>(Godoy et al., 2019)</ns0:ref>, salt stress <ns0:ref type='bibr' target='#b7'>(Ding et al., 2014)</ns0:ref>, and nutrient deficiencies <ns0:ref type='bibr' target='#b42'>(Nishida et al., 2017)</ns0:ref>. AS not only provides an important source of transcriptomic and proteomic diversity and plasticity for use in natural selection <ns0:ref type='bibr' target='#b27'>(Labadorf et al., 2010)</ns0:ref>, but it also plays specific roles in the response <ns0:ref type='bibr' target='#b5'>(Chauhan et al., 2019)</ns0:ref> or adaptation to environmental stresses <ns0:ref type='bibr' target='#b11'>(Filichkin et al., 2015)</ns0:ref>. <ns0:ref type='bibr'>Guo et al. (2019)</ns0:ref> identified four splicing variants of two BnCYCD3-1-LIKE genes in B. napus and found evidence that their AS may play an important role in the response to environmental stresses. <ns0:ref type='bibr' target='#b67'>Xia et al. (2017)</ns0:ref> discovered that the AS with intron retention of EARLY MATURITY8 (EAM8) led to early flowering in a barley (Hordeum vulgare) landrace, while in shepherd's purse (Capsella bursapastoris), flowering time varies with changes in the splicing of a FLOWERING LOCUS C (FLC) homolog <ns0:ref type='bibr' target='#b57'>(Slotte et al., 2009)</ns0:ref>. In addition, the heterologous expression of a vacuolar membrane Na + /H + antiporter gene (SsNHX1) AS variant from seepweed (Suaeda salsa) enhances the salt tolerance of Arabidopsis <ns0:ref type='bibr' target='#b31'>(Li et al., 2009)</ns0:ref>.</ns0:p><ns0:p>As mentioned above, low temperatures switch off the multi-silique trait in zws-ms rapeseed.</ns0:p><ns0:p>When zws-ms plants were planted in Xindu, Sichuan Province, China, the multi-silique trait was continuously stable for years; however, when they were grown in Ma'erkang, Sichuan Province, where the annual average temperature is consistently 7.6 °C lower, the multi-silique trait disappeared and all plants displayed normal siliques <ns0:ref type='bibr' target='#b3'>(Chai et al., 2019)</ns0:ref>. We previously investigated the association of chromosomal regions with this trait, at the genomic and transcriptomic levels, selecting potential candidates from the differentially expressed genes (DEGs) between the multi-and single-silique plants. However, the involvement of posttranscriptional modifications and the mechanisms by which temperature regulates this multisilique trait remain unclear. AS is often responsive to cold stress in plants <ns0:ref type='bibr' target='#b22'>(Iida, 2004;</ns0:ref><ns0:ref type='bibr' target='#b45'>Palusa et al., 2007)</ns0:ref> and is a mechanism by which plants perceive temperature fluctuations and modulate the activity of their transcription factors <ns0:ref type='bibr' target='#b54'>(Seo et al., 2013)</ns0:ref>. In view of the above insights, we analyzed AS using transcriptome sequencing (RNA-seq) in this study. High-throughput RNAseq technology is a widely used, highly efficient, and economical strategy for transcriptomic profiling <ns0:ref type='bibr' target='#b61'>(Tong et al., 2013;</ns0:ref><ns0:ref type='bibr' target='#b66'>Wang et al., 2009)</ns0:ref>. It has become increasingly popular because of the following qualities <ns0:ref type='bibr' target='#b38'>(Mortazavi et al., 2008;</ns0:ref><ns0:ref type='bibr' target='#b44'>Ozsolak & Milos, 2011;</ns0:ref><ns0:ref type='bibr' target='#b37'>Marioni et al., 2008;</ns0:ref><ns0:ref type='bibr' /> PeerJ reviewing PDF | (2019:12:44287:2:0:NEW 4 Jul 2020) <ns0:ref type='bibr' target='#b61'>Tong et al., 2013;</ns0:ref><ns0:ref type='bibr' target='#b58'>Sultan et al., 2008)</ns0:ref>: (1) It can be used to detect and quantify the expression of genes, including those expressed at low levels; (2) it can facilitate the annotation of genes and lead to the discovery of novel genes or transcripts; (3) the results are highly reproducible between both technical and biological replicates; and (4) it can detect AS events.</ns0:p><ns0:p>We performed transcriptome sequencing (RNA-seq) on the flower buds of zws-ms and its near-isogenic line (NIL), zws-217, which produces normal single siliques. This facilitated the identification of the AS events in both lines and the analysis of the differentially expressed AS genes and those with line-specific AS events. Combining these data with Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) annotations, we identified likely candidate genes related to the multi-silique trait. To the best of our knowledge, this is the first time that the regulation of flower/fruit morphology by AS has been investigated in rapeseed, and our results provide insights into this field more generally.</ns0:p></ns0:div>
<ns0:div><ns0:head>Materials & Methods</ns0:head></ns0:div>
<ns0:div><ns0:head>Plant Materials and Growth Conditions</ns0:head><ns0:p>The rapeseed line zws-ms and its NIL, zws-217 <ns0:ref type='bibr' target='#b3'>(Chai et al., 2019)</ns0:ref>, were kept in the Crop Research Institute, Sichuan Academy of Agricultural Sciences, China. Both zws-ms and zws-217 were homozygous for almost all genes, differing from each other only in the multi-silique trait of zws-ms (Figure <ns0:ref type='figure'>1</ns0:ref>). The NILs zws-217 and zws-ms were both grown in an experimental field in the Xindu District of Chengdu in the Sichuan Basin, China, under normal environmental conditions. Additionally, the both lines were also grown in Ma'erkang, a mountainous area in western Sichuan, with a much lower annual average temperature. The annual average temperature in Xindu and Ma'erkang is 16.2 °C and 8.6 °C, respectively <ns0:ref type='bibr' target='#b3'>(Chai et al., 2019)</ns0:ref>.</ns0:p></ns0:div>
<ns0:div><ns0:head>Total RNA Extraction and Sequencing Library Construction</ns0:head><ns0:p>Three zws-ms plants (samples T01, T02, and T03) and three zws-217 plants (T04, T05, and T06) were selected for RNA isolation, as described previously <ns0:ref type='bibr' target='#b3'>(Chai et al., 2019)</ns0:ref>. Flower buds were detached from each plant at the budding stage (BBCH 57), and their total RNA was extracted using an RNA Isolation Kit (Tiangen, Beijing, China). The quality and concentration of the RNA were determined using a NanoDrop 2000 (Thermo Fisher Scientific, Waltham, MA, PeerJ reviewing PDF | (2019:12:44287:2:0:NEW 4 Jul 2020)</ns0:p><ns0:p>Manuscript to be reviewed USA), and the sequencing libraries were generated using an RNA Library Prep Kit for Illumina (New England Biolabs, Ipswich, MA, USA).</ns0:p></ns0:div>
<ns0:div><ns0:head>Sequencing and Expression Analysis</ns0:head><ns0:p>The samples were sequenced on a HiSeq X Ten platform (Illumina, San Diego, CA, USA) and paired-end reads were generated. Low-quality reads and adaptor sequences were removed, and clean reads were used for the following analysis. TopHat2 <ns0:ref type='bibr' target='#b26'>(Kim et al., 2013)</ns0:ref> was used to map the clean reads onto the Brassica napus reference genome <ns0:ref type='bibr' target='#b4'>(Chalhoub et al., 2014)</ns0:ref> with default parameters '--read-mismatches 2 --read-edit-dist 2 --library-type fr---max-intron-length 5000000'. The number of fragments per kilobase of transcripts per million fragments mapped (FPKM) was calculated to represent the gene expression level, and the DESeq R package <ns0:ref type='bibr' target='#b0'>(Anders & Huber, 2010)</ns0:ref> was used to analyze the differential expression. The P-value was adjusted using Benjamini and Hochberg's approach to control the false discovery rate (FDR).</ns0:p><ns0:p>The relative expression levels of each transcript calculated using DESeq were used to define the DEGs, which were defined as having a fold change > 4 and an FDR < 0.01. Pearson's correlation coefficients were determined for the three biological replicates of each line to determine the reliability of the DEGs. Moreover, real-time quantitative PCR (qPCR) was performed to validate the transcriptome sequencing. Since the validation for transcriptome sequencing data from plants in Xindu had been confirmed previously <ns0:ref type='bibr' target='#b3'>(Chai et al., 2019)</ns0:ref>, we only validated the data from plants in colder Ma'erkang herein and method was the same as that earlier report.</ns0:p></ns0:div>
<ns0:div><ns0:head>AS Event Analysis</ns0:head><ns0:p>The cleaned sequence data were aligned to the reference genome using TopHat2 <ns0:ref type='bibr' target='#b26'>(Kim et al., 2013)</ns0:ref> with default settings mentioned above. The resultant gapped alignment data in a binary alignment format were then used as an input for Cufflinks and Cuffcompare, which were run using the default settings to assemble the transcripts and identify splicing junctions from the alignment data. For the AS detection and annotation, the AS events were annotated with ASprofile <ns0:ref type='bibr' target='#b12'>(Florea et al., 2013)</ns0:ref>, which uses Cufflinks and Cuffcompare outputs as input data.</ns0:p><ns0:p>Default parameters of the software were used.</ns0:p><ns0:p>PeerJ reviewing PDF | (2019:12:44287:2:0:NEW 4 Jul 2020)</ns0:p><ns0:p>Manuscript to be reviewed</ns0:p></ns0:div>
<ns0:div><ns0:head>Annotation of Genes</ns0:head><ns0:p>Gene function was annotated based on the following databases: Nr (NCBI nonredundant protein sequences), Nt (NCBI nonredundant nucleotide sequences), Pfam (Protein family), KOG/COG (Clusters of Orthologous Groups of proteins), Swiss-Prot (a manually annotated and reviewed protein sequence database), KO (KEGG Ortholog database), and GO (Gene Ontology).</ns0:p><ns0:p>The GO enrichment analysis of the DEGs was performed using the GOseq R packages based on a Wallenius noncentral hypergeometric distribution <ns0:ref type='bibr' target='#b73'>(Young et al., 2010)</ns0:ref>, which can adjust for gene length bias in the DEGs.</ns0:p><ns0:p>The KEGG database <ns0:ref type='bibr' target='#b25'>(Kanehisa et al., 2007)</ns0:ref> is a resource used to explore the high-level functions and utilities of the biological system, such as the cell, organism, and ecosystem, from molecular-level information, especially using large-scale molecular datasets generated from genome sequencing and other high-throughput experimental technologies (http://www.genome.jp/kegg/). KOBAS <ns0:ref type='bibr' target='#b36'>(Mao et al., 2005)</ns0:ref> software was used to test the statistical enrichment of the DEGs in the various KEGG pathways. Default parameters were used.</ns0:p></ns0:div>
<ns0:div><ns0:head>Results</ns0:head></ns0:div>
<ns0:div><ns0:head>Transcriptome Sequencing</ns0:head><ns0:p>Flower buds from three plants of both the multi-silique line zws-ms and the single-silique NIL zws-217 (Figure <ns0:ref type='figure'>1</ns0:ref>) were sampled for RNA extraction. The sequencing saturation and cluster analysis of the samples were determined to ensure the validity of the data. In total, 65.6 Gb of clean data were generated, with an average Q30 value of 90.54%. Each sample generated about 36.65 M clean reads with an average GC content of 47.23% (Table <ns0:ref type='table' target='#tab_0'>S1</ns0:ref>). The average proportion of total reads mapped to the reference genome for each sample was 73.72% (Table <ns0:ref type='table'>S2</ns0:ref>).</ns0:p><ns0:p>Validation of this transcriptome sequencing data was previously confirmed by qPCR <ns0:ref type='bibr' target='#b3'>(Chai et al., 2019)</ns0:ref>. Similarly, samples from colder area Ma'erkang also generated abundant data, which was validated by comparing the relative transcript levels of eight DEGs in zws-ms and zws-217 by qPCR. The qPCR analysis (Figure <ns0:ref type='figure'>S1</ns0:ref>) showed that all genes had similar trends in expression as those observed by transcriptome sequencing (described below). Each sample generated about 22.92M clean reads with an average GC content of 46.27%, and Q30 value of 92.95% (Table <ns0:ref type='table' target='#tab_0'>PeerJ reviewing PDF | (2019:12:44287:2:0:NEW 4 Jul 2020)</ns0:ref> Manuscript to be reviewed S3); average proportion of total reads mapped to the reference genome for each sample was 88.87% (Table <ns0:ref type='table'>S4</ns0:ref>).</ns0:p></ns0:div>
<ns0:div><ns0:head>AS Event Identification and Analysis</ns0:head><ns0:p>According to description by <ns0:ref type='bibr' target='#b49'>Reddy (2007)</ns0:ref>, alternative splicing events were sorted into 5 classes:</ns0:p><ns0:p>Alternative 3' splice site, Alternative 5' splice site, Exon Skipping, Intron Retention and Mutually Exclusive Exons. The six samples grown in Xindu under normal conditions displayed an average of 205,496 AS events (Table <ns0:ref type='table' target='#tab_0'>1</ns0:ref>; Table <ns0:ref type='table'>S5</ns0:ref>). The proportions of each AS type were analyzed in both zws-ms and zws-217. The two lines shared highly similar ratios of each AS type, with the alternative 5' splice site and alternative 3' splice site types being the most commonly observed, at 43.48% and 42.77% of AS events for both lines, respectively. The mutually exclusive exons type was the next most common (6.61%), followed by the Intron Retention type (5.92%), and the least common types was Exon Skipping, which represented just 1.22% of the AS events (Figure <ns0:ref type='figure'>2a</ns0:ref>; Table <ns0:ref type='table' target='#tab_0'>1</ns0:ref>).</ns0:p><ns0:p>As to the plants grown in colder area Ma'erkang, the two lines also shared highly similar ratios of each AS type: the alternative 5' splice site and alternative 3' splice site types represented the greatest proportion, at 42.13% and 41.29%, respectively; while the exon skipping accounted least proportion 1.51% (Figure <ns0:ref type='figure'>2b</ns0:ref>; Table <ns0:ref type='table' target='#tab_0'>1</ns0:ref>). The number of each type of AS events were significantly reduced in colder area Ma'erkang, except the Intron Retention.</ns0:p></ns0:div>
<ns0:div><ns0:head>Annotation of the Alternatively Spliced Genes</ns0:head><ns0:p>To study the biological functions of the genes with AS events, GO and KEGG pathway enrichment analyses were performed. The GO annotations for AS genes from plants in Xindu included 17 terms involved in biological processes (BP; Figure <ns0:ref type='figure'>3a</ns0:ref>), 17 terms associated with cellular components (CC), and 20 terms involved in molecular functions (MF). The most highly enriched BP terms observed in the alternatively spliced genes included 'cellular process', 'single-organism process' and 'metabolic process'. The most common CC categories were 'cell', 'cell part' and 'organelle'. In the MF category, the most enriched terms were 'binding', 'catalytic activity' and 'nucleic acid binding transcription factor activity'. Plants grown in Ma'erkang showed highly similar GO data to that in Xindu: 22 terms involved in BP (Figure <ns0:ref type='figure'>3b</ns0:ref>), Manuscript to be reviewed 15 terms associated with CC, and 15 terms involved in MF. Moreover, data from Ma'erkang and Xindu showed the same top-3 most enriched terms in each category.</ns0:p><ns0:p>These KEGG pathways were classified into five major groups: metabolism, genetic information processing, cellular processes, environmental information processing, and organismal systems. Of these, the subgroups 'biosynthesis of amino acids', 'carbon metabolism', 'ribosome', and 'RNA transport' contained the highest number of annotated genes (Figure <ns0:ref type='figure' target='#fig_9'>4a</ns0:ref>). Data from Ma'erkang showed similar subgroups containing the most of AS genes: 'ribosome', 'carbon metabolism', 'biosynthesis of amino acids' and 'plant hormones signal transduction' contained the highest number of annotated genes (Figure <ns0:ref type='figure' target='#fig_9'>4b</ns0:ref>).</ns0:p></ns0:div>
<ns0:div><ns0:head>DEGs with AS and Their Arabidopsis Orthologs</ns0:head><ns0:p>DESeq software was used to identify the different expression levels of the AS genes in zws-ms and zws-217. From Xindu, eleven differentially expressed AS genes were identified, of which four were upregulated and seven were downregulated in zws-ms (Table <ns0:ref type='table'>2</ns0:ref>).</ns0:p><ns0:p>The Arabidopsis orthologs of these differentially expressed AS genes were identified using</ns0:p><ns0:p>The Arabidopsis Information Resource (TAIR; https://www.arabidopsis.org; Table <ns0:ref type='table'>3</ns0:ref>). The When grown in Ma'erkang, the two lines generated increased number of differentially expressed AS genes significantly to 130 (Table <ns0:ref type='table'>S6</ns0:ref>), including 52 unregulated and 78 down regulated AS genes. Four AS genes were annotated to 'response to cold (GO:0009409)':</ns0:p><ns0:p>BnaAnng17190D, BnaC01g27600D, BnaC08g39130D and BnaC09g53990D; three AS genes, BnaA07g19340D, BnaC01g27600D, BnaC08g39130D, were annotated to 'response to heat</ns0:p><ns0:p>PeerJ reviewing PDF | (2019:12:44287:2:0:NEW 4 Jul 2020)</ns0:p><ns0:p>Manuscript to be reviewed (GO:0009408)'; three were related to 'response to freezing (GO:0050826)', including BnaA05g28590D, BnaC06g15710D and BnaC08g36010D; BnaCnng24040D was found relevant to 'temperature stimulus (GO:0009266)'. Moreover, BnaC08g36010D, BnaC08g39130D and BnaC08g39360D were annotated to 'regulation of flower development (GO:0009909)', 'plant ovule development (GO:0048481)' and 'fruit development (GO:0010154)', respectively (Table <ns0:ref type='table'>S6</ns0:ref>). Compared with the 11 differentially expressed AS genes from normal conditions in Xindu, three of them (BnaA07g04500D, BnaAnng30260D and BnaC06g16950D) maintained the same expression models. In other words, these three genes were upregulated under both normal and colder conditions. While the other 8 genes (BnaA02g02630D, BnaA02g03080D, BnaA04g16220D, BnaA09g45000D, BnaA09g45260D, BnaC02g06440D, BnaC07g33980D and BnaC08g49610D) did not showed significant difference between zms-ms and zws-217 in expression level in Ma'erkang.</ns0:p></ns0:div>
<ns0:div><ns0:head>Genes with Line-specific AS Events</ns0:head><ns0:p>Genes with line-specific AS events, defined as those genes with a particular AS event(s) that occurred only in zws-ms or in zws-217, were also identified and analyzed. Unlike the abovementioned general classifications, we sorted AS events into 12 finer subclasses, in order to identify them more specifically: transcription start site (TSS) and transcription terminal site AS genes were detected, of which 187 could be annotated (Table <ns0:ref type='table'>S7</ns0:ref>). Ten genes related to 'ovule development', 'flower development' and other similar processes were highlighted and considered important for further study in the coming future (Table <ns0:ref type='table'>4</ns0:ref>, Table <ns0:ref type='table'>S7</ns0:ref>): (1)</ns0:p><ns0:p>BnaC06g32640D was annotated as 'vegetative to reproductive phase transition of meristem Manuscript to be reviewed (GO:0009909)';</ns0:p><ns0:p>(5) BnaC04g26180D and (6) BnaC07g22680D were annotated with 'development (GO:0048481)'; (7) BnaC07g25280D was annotated as 'flower morphogenesis;</ns0:p><ns0:p>organ morphogenesis (GO:0009887)' and 'vegetative to reproductive phase transition of meristem (GO:0010228)'; (8) BnaC01g16410D was annotated as 'flower development (GO:0009908)'; (9) BnaC03g32190D was annotated as 'double fertilization forming a zygote and endosperm (GO:0009567)'; and (10) BnaCnng68400D was associated with 'carpel development (GO:0048440).'</ns0:p></ns0:div>
<ns0:div><ns0:head>Discussion</ns0:head><ns0:p>As an important post-transcriptional metabolic event, AS is involved in many plant growth and developmental processes, such as flowering induction <ns0:ref type='bibr' target='#b9'>(Eckardt, 2002;</ns0:ref><ns0:ref type='bibr' target='#b57'>Slotte et al., 2009)</ns0:ref> and the responses to environmental fluctuations and pathogen attacks <ns0:ref type='bibr' target='#b1'>(Barbazuk et al., 2008)</ns0:ref>. To the best of our knowledge, AS events have seldom been reported to regulate the development of flower/fruit morphology in higher plants. This study is the first to analyze the role of AS events in rapeseed flower/fruit development as a whole, let alone those related to the multi-silique trait.</ns0:p><ns0:p>We previously described the morphology and inheritance of the multi-silique trait in B.</ns0:p><ns0:p>napus <ns0:ref type='bibr' target='#b24'>(Jiang et al., 1998)</ns0:ref>, investigating the associated regions of chromosomes at the genomic level and transcriptomically exploring the DEGs in multi-silique and single-silique plants <ns0:ref type='bibr' target='#b3'>(Chai et al., 2019)</ns0:ref>. The multi-silique trait was found to be controlled by three recessive alleles and was significantly affected by environment; however, the mechanisms by which environmental factors affect this trait remained unknown, even if we knew that temperature could switch on/off the multi-silique trait <ns0:ref type='bibr' target='#b3'>(Chai et al., 2019)</ns0:ref>. As mentioned above, AS is a pathway by which the environment could regulate plant physiology, therefore in this study, we analyzed AS events in order to investigate the mechanism by which plants perceive temperature fluctuations.</ns0:p><ns0:p>In this study, we sampled the buds of three individual plants from zws-ms and zws-217 lines in both Xindu and colder area Ma'erkang, and then subjected them to RNA-seq. All of the four groups generated sufficient data, which was validated by qPCR in earlier and present study successively. The samples in Ma'erkang group generated less data, mainly due to the less sequencing depth; nevertheless, this still provided enough data of high quality and assured the accuracy of the subsequent analysis.</ns0:p><ns0:p>PeerJ reviewing PDF | (2019:12:44287:2:0:NEW 4 Jul 2020)</ns0:p></ns0:div>
<ns0:div><ns0:head>Manuscript to be reviewed</ns0:head><ns0:p>We identified all of the genes with AS events in the zws-ms and zws-217 plants. In Xindu group, 11 AS genes were significantly differentially expressed between the multi-silique zws-ms line and its NIL, zws-217, which produces normal siliques. We analyzed their annotations and orthologs in Arabidopsis. One such ortholog, AT5G15470 (also known as Galacturonosyltransferase 14, GAUT14), is involved in cell wall pectin biosynthesis <ns0:ref type='bibr' target='#b2'>(Caffall et al., 2009)</ns0:ref>, and the gaut13 gaut14 double mutant was previously shown to be defective in pollen tube growth <ns0:ref type='bibr'>(Wang et al., 2013)</ns0:ref>. AT3G15420 (the ortholog of BnaA02g03080D) and AT3G10070 (the ortholog of BnaA09g45000D) encode subunits of the transcription factor complexes TFIIIC and TAF12, respectively. The former does not appear to be substantially involved in plant development; however, some members of the TAF family are involved in the regulation of morphology. The transgenic expression of TAF10 from clustered yellowtops (Flaveria trinervia) in Arabidopsis limited the development of the indeterminate inflorescence and resulted in the production of deformed leaves <ns0:ref type='bibr' target='#b13'>(Furumoto et al., 2005)</ns0:ref>. By contrast, the taf mutant in Arabidopsis has abnormal phyllotaxis and lacks proper vegetative meristem activity <ns0:ref type='bibr' target='#b60'>(Tamada et al., 2007)</ns0:ref>, indicating the important roles played by the TAFs in plant morphological development. Another DEG AS gene, BnaA04g16220D, is not annotated, and its Arabidopsis ortholog AT1G14800 is simply listed as an uncategorized nucleic acid-binding, OB-fold-like protein. The AS gene orthologs AT2G04900 and AT1G15060 encode an unknown protein and an uncategorized alpha/beta hydrolase family protein, respectively, so their roles in the regulation of the multi-silique trait are also currently unclear.</ns0:p><ns0:p>Another ortholog for differentially expressed AS gene, AT3G54620, is reported to encode a bZIP transcription factor-like protein. Members of this protein family are typically reported to regulate plant tolerance of environmental stresses. The transgenic expression of the maize (Zea mays) gene ZmbZIP72 in Arabidopsis enhanced its drought and salt tolerance <ns0:ref type='bibr' target='#b72'>(Ying et al., 2012)</ns0:ref>, while BnbZIP3, a ramie (Boehmeria nivea) bZIP transcription factor, also increased the drought, salinity, and heavy metal tolerances of transgenic Arabidopsis <ns0:ref type='bibr' target='#b20'>(Huang et al., 2016)</ns0:ref>. These genes are also involved in the regulation of other processes; for example, the repression of a bZIP transcription factor gene OsABI5 expression in rice resulted in low fertility <ns0:ref type='bibr' target='#b78'>(Zou et al., 2008)</ns0:ref>, while the transgenic expression of tomato (Solanum lycopersicum) SlbZIP2 in tobacco (Nicotiana benthamiana) increased leaf thickness <ns0:ref type='bibr' target='#b55'>(Seong et al., 2016)</ns0:ref>. To date, however, there are no reports of bZIP genes playing a significant role in flower/fruit morphology.</ns0:p><ns0:p>PeerJ reviewing PDF | (2019:12:44287:2:0:NEW 4 Jul 2020)</ns0:p><ns0:p>Manuscript to be reviewed</ns0:p><ns0:p>Other AS gene orthologs included AT5G16210, encoding a member of the HEAT repeatcontaining protein family, which are considered to be involved in intracellular transport <ns0:ref type='bibr' target='#b18'>(Hernández-Torres et al., 2014;</ns0:ref><ns0:ref type='bibr' target='#b43'>Oeffinger et al., 2004)</ns0:ref>. Although BnaC06g16950D is not annotated, its ortholog, AT3G59000, was identified as encoding an F-box/RNI-like superfamily protein in Arabidopsis, which typically function in the plant hormone signaling pathways <ns0:ref type='bibr' target='#b14'>(Gao et al., 2009)</ns0:ref>. Similarly, the ortholog AT4G16900 encodes a TIR-NBS-LRR class protein, which are known to be involved in disease resistance <ns0:ref type='bibr' target='#b70'>(Xun et al., 2019)</ns0:ref> and hormonal responses <ns0:ref type='bibr' target='#b53'>(Sarazin et al., 2015)</ns0:ref>. Moreover, <ns0:ref type='bibr' target='#b19'>Hewezi et al. (2006)</ns0:ref> unexpectedly found that these proteins are associated with developmental abnormalities; transgenic sunflowers (Helianthus annuus) expressing the antisense sequence complementing PLFOR48, which encodes a TIR-NBS-LRR-type protein,</ns0:p><ns0:p>showed stunted growth and a reduction in apical dominance; whereas the pods of transgenic tobacco (N. tabacum) lacking PLFOR48 expression were smaller and showed severe deformations. This indicates that TIR-NBS-LRR-type proteins can regulate the morphology of plants, including fruit morphology, to some extent. Finally, AT1G10760, the ortholog of AS gene BnaC08g49610D, which encodes a GWD protein required for starch degradation, is involved in carbohydrate metabolism <ns0:ref type='bibr' target='#b41'>(Nadolska-Orczyk et al., 2017)</ns0:ref>. This gene was also reported to regulate seed size; <ns0:ref type='bibr' target='#b48'>Pirone et al. (2017)</ns0:ref> found that the length and width of the mature seeds were reduced in the gwd1 Arabidopsis mutant, while their density was increased.</ns0:p><ns0:p>To summarize, AT5G15470, AT3G10070, AT3G54620, AT4G16900, and AT1G10760 are all known to be involved in plant development; therefore, their corresponding rapeseed orthologs, BnaA02g02630D, BnaA09g45000D, BnaAnng30260D, BnaC07g33980D, and BnaC08g49610D, the expression levels of which differed significantly between zws-ms and zws-217, are considered to be potential candidate genes regulating the multi-silique trait.</ns0:p><ns0:p>After that, we continued to investigate data from Ma'erkang, where it is colder and the multi-silique trait in zws-ms line disappeared. Due to the importance of the above-mention 11 differentially expressed AS genes, we first paid attention to them and found that three of them had the same expression models as in Xindu. In other words, they were independent of temperature; in addition, combined with their annotations, they were excluded from the potential candidate genes responding to environmental factors. On the other hand, other 8 AS genes stopped being differentially expressed between zms-ms and zws-217 in Ma'erkang. That is to say, the expression level of these 8 AS genes (BnaA02g02630D, BnaA02g03080D,</ns0:p><ns0:p>PeerJ reviewing PDF | (2019:12:44287:2:0:NEW 4 Jul 2020)</ns0:p><ns0:p>Manuscript to be reviewed BnaA04g16220D, BnaA09g45000D, BnaA09g45260D, BnaC02g06440D, BnaC07g33980D and BnaC08g49610D) were environment-specifc. Besides, BnaC08g36010D, BnaC08g39130D and BnaC08g39360D, which were annotated to flower/ovule/fruit-related terms, were differentially expressed between zws-ms and zws-217 specifically in Ma'erkang. Moreover, we also found 9 temperature-responding AS genes differentially expressed in colder area: BnaAnng17190D, BnaC01g27600D, BnaC08g39130D, BnaC09g53990D, BnaA07g19340D, BnaA05g28590D, BnaC06g15710D, BnaC08g36010D and BnaCnng24040D. The lower temperature motivated more responding genes and this also explained the reason for increased number of AS genes clustered in each KEGG pathway.</ns0:p><ns0:p>We also explored the line-specific AS genes, which were similarly expressed between zwsms and zws-217, but contained stable and particular AS event(s) that differed between these two lines. These genes are likely to qualitatively regulate the multi-silique trait. In this case, we could obtain better results by fine-classify the AS types into 12 subclasses, rather than 5 classes mentioned above. Because fine classifications could better identify differences between AS types more precisely and subtly. Thus, we found 205 genes of this type, of which 187 could be annotated. Due to the rarity of the multi-silique trait, we did not obtain much useful information from the KEGG pathway analysis. This meant that we were unable to relate this metabolic pathway information to the multi-silique trait directly; however, the GO analysis provided more potential clues. Among these, 10 genes were considered to be associated with Manuscript to be reviewed suggesting a direct link with the reported flowering phenotype of the pif mutants <ns0:ref type='bibr' target='#b29'>(Leivar & Monte, 2014)</ns0:ref>. AT5G17270 encodes a prenylyltransferase superfamily protein; however, to the best of our knowledge, there have been no reports about its development-related functions. The Arabidopsis ortholog of BnaC05g34570D is AT3G18600, which encodes a P-loop-containing nucleoside triphosphate hydrolase. While few studies have reported the functions of these proteins, <ns0:ref type='bibr' target='#b32'>Liu et al. (2016)</ns0:ref> reported that, in sesame (Sesamum indicum), one gene encoding a Ploop-containing nucleoside triphosphate hydrolase showed a reduced expression level in sterile buds, indicating that they may play a role in specifying/determining tapetal fate and development. Another line-specific AS ortholog, AT3G54660, encodes a glutathione reductase (GR), which was found to increase the fineness (mass per unit length) and bundle strength of cotton (Gossypium hirsutum) fiber when transgenically expressed <ns0:ref type='bibr' target='#b62'>(Tuttle et al., 2015)</ns0:ref>. Since cotton fibers are single cells initiating from the epidermis of the outer integument of the ovules <ns0:ref type='bibr' target='#b51'>(Ruan et al., 2004)</ns0:ref>, it can be inferred that GR regulates ovule development to some extent.</ns0:p><ns0:p>The line-specific AS gene ortholog AT2G04030 encodes HSP90, a member of the heat shock proteins (HSPs), which are commonly produced in response to heat. The HSPs are molecular chaperones that prevent protein aggregation and mediate the refolding of heatdenatured proteins <ns0:ref type='bibr' target='#b40'>(Murano et al., 2017)</ns0:ref>. Moreover, a HSP was found to be upregulated in the fiber-bearing ovules of cotton (Gossypium hirsutum) <ns0:ref type='bibr' target='#b28'>(Lee et al., 2006)</ns0:ref> implying some unknown function in ovule development. Another ortholog, AT3G28730 (also known as structure-specific recognition protein SSRP1), was also found to regulate floral development, as the ssrp1-2 mutant Arabidopsis produced small and deformed petals with shorter stamens <ns0:ref type='bibr' target='#b35'>(Lolas et al., 2010)</ns0:ref>.</ns0:p><ns0:p>AT4G24560 encodes a ubiquitin-specific protease (UBP), which are known to play critical roles in protein deubiquitination in plants <ns0:ref type='bibr' target='#b34'>(Liu et al., 2008)</ns0:ref>. The UBPs are also involved in plant development; <ns0:ref type='bibr' target='#b34'>Liu et al. (2008)</ns0:ref> found that knocking out UBP15 function in Arabidopsis resulted in the production of smaller flowers and shorter siliques. The final line-specific AS ortholog, AT5G15020, encodes an SIN3-LIKE 2 protein (SNL2) known to be important for seed germination or dormancy <ns0:ref type='bibr' target='#b65'>(Wang et al., 2016;</ns0:ref><ns0:ref type='bibr'>Wang et al., 2013)</ns0:ref>.</ns0:p><ns0:p>To date, there is some evidence to show that the line-specific AS orthologs AT2G20180, AT4G00050, AT3G54660, and AT2G04030 are related to the regulation of flower/fruit morphology, with clear roles reported for AT1G71692, AT3G28730, and AT4G24560.</ns0:p><ns0:p>Consequently, their orthologs in rapeseed, BnaC06g32640D, BnaC07g25280D, and Manuscript to be reviewed BnaC01g16410D, respectively, are considered to be important candidate genes regulating the multi-silique trait by conferring or removing some specific line-specific AS events. Some of the genes/loci controlling silique development in Brassica plants have previously been reported. In addition to those regulating traits such as the seed weight and silique length <ns0:ref type='bibr' target='#b33'>(Liu et al., 2015)</ns0:ref> and the number of seeds per silique in B. napus <ns0:ref type='bibr'>(Li et al., 2015)</ns0:ref>, some genes related to silique morphology have been cloned and functionally analyzed. <ns0:ref type='bibr' target='#b69'>Xiao et al. (2013)</ns0:ref> fine-mapped a multi-locular silique gene, Bjln1, to a 208-kb region on chromosome A7 in Brassica juncea and then revealed that it was the mutations in the CDS and promoter of BjuA07.CLV1 gene (equivalent to Bjln1) to cause the multi-locular trait <ns0:ref type='bibr' target='#b68'>(Xiao et al., 2018)</ns0:ref>. Both To sum up, the eight candidate genes mentioned above, including the five differentially expressed AS genes of interest and the three genes with line-specific AS events, are therefore hypothesized to regulate the multi-silique trait in rapeseed zws-ms, based on their AS expression levels or line-specific AS events. These findings lay a foundation for further functional analyses in future. The data obtained from plants in colder environment, including the amount and the proportion of each AS type, was generally similar to that under normal conditions, while details are still under further investigation.</ns0:p></ns0:div>
<ns0:div><ns0:head>Conclusions</ns0:head><ns0:p>The utilization of heterosis is a way to increase the yield or improve the quality of crops.</ns0:p><ns0:p>Exploring new germplasm resources and genes, as well as clarifying their inheritance, is the foundation of obtaining of excellent hybrid. This study provides a novel inspection into the multi-silique trait in rapeseed from the transcriptional perspective by AS, deepening the understanding of its molecular mechanism. Further function verifications are now undergoing.</ns0:p><ns0:note type='other'>Figure 1</ns0:note><ns0:p>Multi-silique trait in zws-ms, compared with the single siliques of its near-isogenic line zws-217. Manuscript to be reviewed Numbers of alternative splicing events in six samples.</ns0:p><ns0:p>Note: T01, T02, and T03: Buds of three independent zws-ms plants at the budding stage; T04, T05, and T06: Buds of three independent zws-217 plants at the budding stage.</ns0:p><ns0:p>Alternative 3' splice site: different-size mRNAs are produced depending on the usage of a proximal or distal 3' splice site; Alternative 5' splice site: different-size mRNAs are produced depending on the use of a proximal or distal 5' splice site; Exon Skipping: an exon is either included or excluded from the mRNA; Intron Retention: an intron is either retained or excised in the mRNA, resulting in different-size transcripts; Mutually Exclusive Exons: adjacent exons are spliced in such a way that only one of them is included at a time in the mRNA.</ns0:p><ns0:p>PeerJ reviewing PDF | (2019:12:44287:2:0:NEW 4 Jul 2020)</ns0:p><ns0:p>Manuscript to be reviewed</ns0:p></ns0:div><ns0:figure xml:id='fig_0'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2019:12:44287:2:0:NEW 4 Jul 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_1'><ns0:head /><ns0:label /><ns0:figDesc>following orthologs were identified: (1) AT5G15470, the ortholog of BnaA02g02630D, encodes galacturonosyltransferase 14 (GAUT14); (2) AT3G15420 encodes the transcription factor TFIIIC (tau55-related protein); (3) AT1G14800 encodes a nucleic acid-binding, OB-fold-like protein; (4) AT2G04900 encodes an unknown protein; (5) AT3G10070 encodes one of two Arabidopsis proteins with similarity to the TBP-associated factor, TAF12; (6) AT1G15060 encodes an alpha/beta hydrolase family protein; (7) AT3G54620 encodes a bZIP transcription factor-like protein; (8) AT5G16210 encodes a HEAT repeat-containing protein; (9) AT3G59000 encodes an F-box/RNI-like superfamily protein; (10) AT4G16900, the ortholog of BnaC07g33980D, encodes a member of the disease resistance protein (TIR-NBS-LRR class) family; and (11) AT1G10760 encodes an α-glucan, water dikinase (GWD) required for starch degradation.</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_2'><ns0:head>(</ns0:head><ns0:label /><ns0:figDesc>TTS) equaled to original Alternative 5' first exon and Alternative 3' last exon, respectively; Mutually exclusive exons was subdivided into Alternative exon ends (AE) and Approximate AE (XAE); the original Intron retention was subdivided into single Intron retention (IR), Approximate IR (XIR), Multi-IR (MIR) and Approximate MIR (XMIR); the Cassette exon was then was subdivided into single Skipped exon (SKIP), Approximate SKIP (XSKIP), Multi-exon SKIP (MSKIP) and Approximate MSKIP (XMSKIP). In total, 205 line-specifically expressed</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_3'><ns0:head>(</ns0:head><ns0:label /><ns0:figDesc>GO:0010228)' and 'Biological Process: ovule development (GO:0048481)'; (2) BnaC07g00780D was associated with 'reproductive structure development (GO:0048608)'; (3) BnaC04g31460D and (4) BnaC05g34570D were related to 'regulation of flower development PeerJ reviewing PDF | (2019:12:44287:2:0:NEW 4 Jul 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_4'><ns0:head /><ns0:label /><ns0:figDesc>flower/carpel/ovule development. BnaC06g32640D is annotated as being involved in the regulation of the vegetative-to-reproductive phase transition in the meristem (GO:0010228) and in ovule development (GO:0048481). Its Arabidopsis ortholog, AT1G71692, is annotated as AGAMOUS-LIKE12 (AGL12). Peng et al. (2015) isolated the BnFUL gene in rapeseed, which is homologous to AGL8 in Arabidopsis. Although BnFUL was hypothesized to be involved in enhancing pod-shattering resistance, when introduced into Arabidopsis, two of the five transgenic plants expressing BnFUL unexpectedly had a multi-silique phenotype. However, the mechanisms by which BnFUL generates this multi-silique phenotype remain elusive thus far, making the AGL12 gene identified in this study a potentially important candidate gene. Other orthologs of the line-specific AS genes include AT2G20180 and AT4G00050, both of which encode phytochrome interacting factors (PIFs). Several transcription factors (AP1, SVP, LFY, AG, and SEP3) involved in the regulation of flowering are known to bind to the PIFs, PeerJ reviewing PDF | (2019:12:44287:2:0:NEW 4 Jul 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_5'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2019:12:44287:2:0:NEW 4 Jul 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_6'><ns0:head>Fan</ns0:head><ns0:label /><ns0:figDesc>et al. (2014) andYadava et al. (2014) reported that a mutation in BrCLV3, a homologue of CLAVATA3 in Arabidopsis, caused the production of multi-locular siliques in B. rapa. However, the multi-silique (or multi-pistil) phenotype of zws-ms is different from the above-motioned multi-locular trait; zws-ms produces three pods on each carpopodium, rather than multiple loculi per pod.Few studies have investigated this multi-silique trait in rapeseed; however, there have been similar reports of multi-pistil traits in other crops, particularly in wheat (Triticum aestivum):<ns0:ref type='bibr' target='#b8'>Duan et al. (2015)</ns0:ref> discovered a male-sterile wheat mutant, dms, with a dwarf status and multipistils, a pleiotropic phenotype found to be controlled by a single recessive gene, which was not identified.Guo et al. (2019) reported another multi-ovary trait in the wheat line DUOII, which was controlled by a dominant gene, and used a proteomics approach to propose some candidate proteins.<ns0:ref type='bibr' target='#b71'>Yang et al. (2017)</ns0:ref> mapped a gene promoting the formation of three pistils (Pis1) to chromosome 2D and identified some candidate genes according to their annotations, while<ns0:ref type='bibr' target='#b77'>Zhu et al. (2019)</ns0:ref> discovered a wheat multi-pistil mutant, 12TP, which was found to contain a semidominant mutation located on chromosome arm 2DL. Although several studies have explored the multi-pistil trait in wheat, no one has identified any of the specific genes responsible yet.</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_7'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2019:12:44287:2:0:NEW 4 Jul 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_8'><ns0:head>(</ns0:head><ns0:label /><ns0:figDesc>A) Primary inflorescences. (B) Siliques.</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_9'><ns0:head>Figure 4 Classified</ns0:head><ns0:label>4</ns0:label><ns0:figDesc>Figure 4</ns0:figDesc></ns0:figure>
<ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='24,42.52,70.87,333.62,672.95' type='bitmap' /></ns0:figure>
<ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='25,42.52,229.87,525.00,230.25' type='bitmap' /></ns0:figure>
<ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='26,42.52,280.87,525.00,214.50' type='bitmap' /></ns0:figure>
<ns0:figure type='table' xml:id='tab_0'><ns0:head>Table 1 (on next page)</ns0:head><ns0:label>1</ns0:label><ns0:figDesc /><ns0:table /></ns0:figure>
<ns0:note place='foot'>PeerJ reviewing PDF | (2019:12:44287:2:0:NEW 4 Jul 2020)Manuscript to be reviewed</ns0:note>
</ns0:body>
" | "Response to Editors and Reviewers
Dear Editors and Reviewers,
Your 2nd round recommendations helped us to optimize our manuscript further. We have revised the manuscript thoroughly and hope it can meet your requirement. If there are still something to revise, please contact us at any time.
Our response to your comments are as following:
Reviewer 2 (Anonymous)
Basic reporting
The newly added RNA-seq data from a colder condition was not analyzed and discussed.
Response:
Thanks for your suggestions. We have added related information into Abstract, Methods, Results and Discussion Sections, respectively. Please kindly check the new-updated manuscript.
Experimental design
no comments
Validity of the findings
The authors did not validate RNA-seq results.
Response:
We validated the RNA-seq results (from Xindu) in our previously published paper. Herein, we restate it and cite that paper for your information. Moreover, we also validated the RNA-seq results from colder area Ma’erkang and please refer to Figure S1.
Details were in following points.
Comments for the Author
1. It is good that the authors included a different set of data from a colder condition. But this set of data needs to analyzed and discussed in this manuscript. It can’t be just there without any contribution to the manuscript.
Response:
We had added data from a colder condition in response to Reviewer1. Now, thanks for Reviewer2’s advice, we enhanced some analysis and discussion about it in Results and Discussion Sections etc., respectively. Please kindly find it.
2. The authors completely ignored the question on validation of RNA-seq data. Table 2 provides logFC from RNA-seq. That is reporting the same information, not validation.
Response:
Your advices is very important to us and in fact we did not ignore any of your point.
(1) We added FDR and log2FC because of your requirement in last round;
(2) As we described in Results Section, under the subtitle “Transcriptome Sequencing”: “Validation of transcriptome sequencing data was previously confirmed by qPCR (Chai et al., 2019)”. In our previous paper “Chai et al., 2019”, we have provided validation of RNA-seq data by qPCR. For your information, here is the figure displayed in that paper:
And in this manuscript, we described this information and cited that paper as you can see.
Moreover, we provided the validation for the newly added data in new Figure S1.
3. Lines 119-122: Please specify annual average temperatures in both Xindu and Ma’erkang.
Response:
Xindu is located in the Sichuan Basin, which has a humid subtropical monsoon climate (with an annual average temperature of 16.2 °C), whereas Ma’erkang is located in a mountainous area in western Sichuan, where the annual average temperature is only 8.6 °C.
And as you required, we have added this information, as well as the citation information there, please kindly find it.
4. Line 155. Although there is a statement that default parameters were used. It is very cryptic and does not offer any specific information. Please specific statistical tests used and FDR used as asked earlier.
Response:
(1) Parameters for Tophat2 are “--read-mismatches 2 --read-edit-dist 2 --library-type fr- --max-intron-length 5000000”, and we have added this information into Materials & Methods section, under subtitle “Sequencing and Expression Analysis”.
(2) When you use software like ASprofile, you only need to input your sequences. The software is easy to use and parameters are in-built and default. Moreover, these tools are widely used, so we added the citation information in this round and we think “default parameters” will provide clear and enough information to the readers.
Many published papers (for example, “Zhu et al. 2017. Proteogenomic analysis reveals alternative splicing and translation as part of the abscisic acid response in Arabidopsis seedlings. The Plant Journal 91:518-533”) also described like “default parameters”. This description is clear and readable enough to readers. Thanks for your understanding.
5. There is no Table 5, while it was referred to on line 190. Is it supposed to be Table S5?
Response:
Dear Reviewer2, we checked this edition carefully and confirmed that we did not mention “Table 5” at any part of the manuscript.
6. Line 220: It is differentially expressed, not differently expressed. This mistake appeared again later in the manuscript (Line 315).
Response:
Thanks for reminding and we have replaced them.
7. Lines 263-273: Since new classifications of AS events are created here, there need to be some sort of illustration to visualize the difference between the establish classification and this new one. Since Table S7 is mentioned here. What is the FDR for in Table S7? Column headers need explanations for what they represent in all tables.
Response:
(1) Transcription start site (TSS) and transcription terminal site (TTS) equaled to original Alternative 5’ first exon and Alternative 3’ last exon, respectively; Mutually exclusive exons was subdivided into Alternative exon ends (AE) and Approximate AE (XAE); the original Intron retention was subdivided into single Intron retention (IR), Approximate IR (XIR), Multi-IR (MIR) and Approximate MIR (XMIR); the Cassette exon was then was subdivided into single Skipped exon (SKIP), Approximate SKIP(XSKIP), Multi-exon SKIP (MSKIP) and Approximate MSKIP (XMSKIP).
This description was added to the manuscript, in the Results Section, under the “Genes with Line-specific AS Events” subtitle.
(2) The FDR in Table S7 is for fold change (FC) of expression level. We have revised it in the table.
8. Lines 275-276: There is a statement saying that 10 genes were selected for further study. But there is no information on what this further study is and what the results of the further study are.
Response:
Dear Reviewer2, when we mentioned “further study” there, we meant some future experiments like function analyses etc. Since they are still in the planning/preparation stage, we do not have any result about them at this moment. We originally had placed those sentences there in order to arouse readers’ interest to constantly follow our researches.
Anyhow, in response to your points, we improved the sentences there.
We hope our revision can meet your requirement.
" | Here is a paper. Please give your review comments after reading it. |
9,777 | "<ns0:abstract xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:p>Background. Flower and fruit development are vital stages of the angiosperm lifecycle.</ns0:p><ns0:p>We previously investigated the multi-silique trait in the rapeseed (Brassica napus) line zwsms on a genomic and transcriptomic level, leading to the identification of two genomic regions and several candidate genes associated with this trait. However, some events on transcriptome level, like alternative splicing, were poorly understood. Methods. Plants from zws-ms and its near-isogenic line (NIL) zws-217 were both grown in Xindu with normal conditions and a colder area Ma'erkang. Buds from the two lines were sampled and RNA was isolated to perform the transcriptomic sequencing. The numbers and types of alternative splicing (AS) events from the two lines were counted and classified. Genes with AS events and expressed differentially between the two lines, as well as genes with AS events which occurred in only one line were emphasized. Their annotations were further studied. Results. From the plants in Xindu District, an average of 205,496 AS events, which could be sorted into 5 AS types, were identified. zws-ms and zws-217 shared highly similar ratios of each AS type: The alternative 5' and 3' splice site types were the most common, while the exon skipping type was observed least often. Eleven differentially expressed AS genes were identified, of which four were upregulated and seven were downregulated in zws-ms. Their annotations implied that five of these genes were directly associated with the multi-silique trait. While samples from colder area Ma'erkang generated generally reduced number of each type of AS events except the Intron Retention; but the number of differentially expressed AS genes increased significantly.</ns0:p><ns0:p>Further analysis found that among the 11 differentially expressed AS genes from Xindu, three of them maintained the same expression models, while the other 8 genes did not showed significant difference between the two lines in expression level. Additionally, the 205 line-specifically expressed AS genes were analyzed, of which 187 could be annotated, and three were considered to be related to the multi-silique trait. Discussion. This study provides new insights into the agronomically important multi-silique trait in rapeseed on</ns0:p></ns0:div>
</ns0:abstract>
<ns0:body xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:head>Introduction</ns0:head><ns0:p>Rapeseed (Brassica napus L.), an allotetraploid with a complex genome (AACC, 2n = 38), is the second leading source of vegetable oil globally <ns0:ref type='bibr' target='#b15'>(Liu et al., 2015)</ns0:ref>. The agronomic traits related to rapeseed yield include the pod (silique) number per plant, branch number, and seed weight <ns0:ref type='bibr' target='#b15'>(Liu et al., 2015;</ns0:ref><ns0:ref type='bibr' target='#b43'>Zhang et al., 2006;</ns0:ref><ns0:ref type='bibr' target='#b15'>Li et al., 2015)</ns0:ref>. We previously reported that zws-ms, a multisilique rapeseed line <ns0:ref type='bibr' target='#b3'>(Chai et al., 2019)</ns0:ref>, produces three independent pistils and 9 to 10 stamens on the same receptacle in a flower, which consequently leads to the formation of three independent siliques on a carpopodium rather than the single siliques typically observed.</ns0:p><ns0:p>Moreover, this trait was found to be affected by the environment, with temperature considered to be the factor most likely to switch on/off the formation of multi-silique.</ns0:p><ns0:p>Temperate is a major environmental factor that regulates various aspects of plant morphology, physiology, and biochemistry, affecting germination, growth, development, and flowering <ns0:ref type='bibr' target='#b18'>(Ren et al., 2019)</ns0:ref>. Fertility in crops such as rapeseed <ns0:ref type='bibr' target='#b41'>(Yu et al., 2015)</ns0:ref> and rice (Oryza sativa) <ns0:ref type='bibr' target='#b42'>(Yu et al., 2017)</ns0:ref> is affected by temperature. In winter rapeseed lines, although a period of vernalization under low temperature is necessary to initiate flowering, cold stress inhibits growth and development, disturbs metabolism, and causes wilting or even death. Notably, cold stress also induces alternative splicing (AS) in plants <ns0:ref type='bibr' target='#b13'>(Palusa et al., 2007;</ns0:ref><ns0:ref type='bibr'>Iida, 2004)</ns0:ref>.</ns0:p><ns0:p>AS is defined as the mechanism by which primary transcripts are processed into two or more mature isoforms, which enables a single gene to produce diverse protein products <ns0:ref type='bibr' target='#b14'>(Pan et al., 2008;</ns0:ref><ns0:ref type='bibr' target='#b20'>Sablok et al., 2011)</ns0:ref>. These proteins differ from each other not only in structure but also possibly in function, subcellular localization, and/or stability <ns0:ref type='bibr'>(Huang et al., 2019;</ns0:ref><ns0:ref type='bibr'>Chauhan et</ns0:ref> PeerJ reviewing PDF | (2019:12:44287:3:0:NEW 7 Aug 2020)</ns0:p><ns0:p>Manuscript to be reviewed <ns0:ref type='bibr'>al., 2019)</ns0:ref>. AS is common in plants; for example, in Arabidopsis thaliana, more than 60% of intron-containing genes undergo AS <ns0:ref type='bibr' target='#b26'>(Syed et al., 2012)</ns0:ref>. Many environmental factors regulate AS events in plants, including CO 2 concentration <ns0:ref type='bibr'>(Huang et al., 2019</ns0:ref><ns0:ref type='bibr'>), light (Godoy et al., 2019)</ns0:ref>, salt stress <ns0:ref type='bibr'>(Ding et al., 2014)</ns0:ref>, and nutrient deficiencies <ns0:ref type='bibr' target='#b10'>(Nishida et al., 2017)</ns0:ref>. AS not only provides an important source of transcriptomic and proteomic diversity and plasticity for use in natural selection <ns0:ref type='bibr'>(Labadorf et al., 2010)</ns0:ref>, but it also plays specific roles in the response <ns0:ref type='bibr'>(Chauhan et al., 2019)</ns0:ref> or adaptation to environmental stresses <ns0:ref type='bibr'>(Filichkin et al., 2015)</ns0:ref>. <ns0:ref type='bibr' target='#b37'>Guo et al. (2019)</ns0:ref> identified four splicing variants of two BnCYCD3-1-LIKE genes in B. napus and found evidence that their AS may play an important role in the response to environmental stresses. <ns0:ref type='bibr' target='#b34'>Xia et al. (2017)</ns0:ref> discovered that the AS with intron retention of EARLY MATURITY8 (EAM8) led to early flowering in a barley (Hordeum vulgare) landrace; while in shepherd's purse (Capsella bursapastoris), flowering time varies with changes in the splicing of a FLOWERING LOCUS C (FLC) homolog <ns0:ref type='bibr' target='#b24'>(Slotte et al., 2009)</ns0:ref>. In addition, the heterologous expression of a vacuolar membrane Na + /H + antiporter gene (SsNHX1) AS variant from seepweed (Suaeda salsa) enhances the salt tolerance of Arabidopsis <ns0:ref type='bibr'>(Li et al., 2009)</ns0:ref>.</ns0:p><ns0:p>As mentioned above, low temperatures switch off the multi-silique trait in zws-ms rapeseed.</ns0:p><ns0:p>When zws-ms plants were planted in Xindu, Sichuan Province, China, the multi-silique trait was continuously stable for years; however, when they were grown in Ma'erkang, Sichuan Province, where the annual average temperature is consistently 7.6 °C lower, the multi-silique trait disappeared and all plants displayed normal siliques <ns0:ref type='bibr' target='#b3'>(Chai et al., 2019)</ns0:ref>. We previously investigated the association of chromosomal regions with this trait, at the genomic and transcriptomic levels, selecting potential candidates from the differentially expressed genes (DEGs) between the multi-and single-silique plants. However, the involvement of posttranscriptional modifications and the mechanisms by which temperature regulates this multisilique trait remain unclear. AS is often responsive to cold stress in plants <ns0:ref type='bibr'>(Iida, 2004;</ns0:ref><ns0:ref type='bibr' target='#b13'>Palusa et al., 2007)</ns0:ref> and is a mechanism by which plants perceive temperature fluctuations and modulate the activity of their transcription factors <ns0:ref type='bibr' target='#b22'>(Seo et al., 2013)</ns0:ref>. In view of the above insights, we analyzed AS using transcriptome sequencing (RNA-seq) in this study. High-throughput RNAseq technology is a widely used, highly efficient, and economical strategy for transcriptomic profiling <ns0:ref type='bibr' target='#b28'>(Tong et al., 2013;</ns0:ref><ns0:ref type='bibr' target='#b33'>Wang et al., 2009)</ns0:ref>. It has become increasingly popular because of the following qualities <ns0:ref type='bibr' target='#b7'>(Mortazavi et al., 2008;</ns0:ref><ns0:ref type='bibr' target='#b12'>Ozsolak & Milos, 2011;</ns0:ref><ns0:ref type='bibr' target='#b6'>Marioni et al., 2008;</ns0:ref><ns0:ref type='bibr' target='#b28'>Tong et al., 2013;</ns0:ref><ns0:ref type='bibr' target='#b25'>Sultan et al., 2008)</ns0:ref>: (1) It can be used to detect and quantify the expression of genes, including those expressed at low levels; (2) it can facilitate the annotation of genes and lead to the discovery of novel genes or transcripts; (3) the results are highly reproducible between both technical and biological replicates; and (4) it can detect AS events.</ns0:p><ns0:p>We performed transcriptome sequencing (RNA-seq) on the flower buds of zws-ms and its near-isogenic line (NIL), zws-217, which produces normal single siliques. This facilitated the identification of the AS events in both lines and the analysis of the differentially expressed AS genes and those with line-specific AS events. Combining these data with Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) annotations, we identified likely candidate genes related to the multi-silique trait. To the best of our knowledge, this is the first time that the regulation of flower/fruit morphology by AS has been investigated in rapeseed, and our results provide insights into this field more generally.</ns0:p></ns0:div>
<ns0:div><ns0:head>Materials & Methods</ns0:head></ns0:div>
<ns0:div><ns0:head>Plant Materials and Growth Conditions</ns0:head><ns0:p>The rapeseed line zws-ms and its NIL, zws-217 <ns0:ref type='bibr' target='#b3'>(Chai et al., 2019)</ns0:ref>, were kept in the Crop Research Institute, Sichuan Academy of Agricultural Sciences, China. Both zws-ms and zws-217 were homozygous for almost all genes, differing from each other only in the multi-silique trait of zws-ms (Figure <ns0:ref type='figure'>1</ns0:ref>). The NILs zws-217 and zws-ms were both grown in an experimental field in the Xindu District of Chengdu in the Sichuan Basin, China, under normal environmental conditions. Additionally, the both lines were also grown in Ma'erkang, a mountainous area in western Sichuan, with a much lower annual average temperature. The annual average temperature in Xindu and Ma'erkang is 16.2 °C and 8.6 °C, respectively <ns0:ref type='bibr' target='#b3'>(Chai et al., 2019)</ns0:ref>.</ns0:p></ns0:div>
<ns0:div><ns0:head>Total RNA Extraction and Sequencing Library Construction</ns0:head><ns0:p>Three zws-ms plants (samples T01, T02, and T03) and three zws-217 plants (T04, T05, and T06) were selected for RNA isolation, as described previously <ns0:ref type='bibr' target='#b3'>(Chai et al., 2019)</ns0:ref>. Flower buds were detached from each plant at the budding stage (BBCH 57), and their total RNA was extracted using an RNA Isolation Kit (Tiangen, Beijing, China). The quality and concentration of the RNA were determined using a NanoDrop 2000 (Thermo Fisher Scientific, Waltham, MA, USA), and the sequencing libraries were generated using an RNA Library Prep Kit for Illumina (New England Biolabs, Ipswich, MA, USA).</ns0:p></ns0:div>
<ns0:div><ns0:head>Sequencing and Expression Analysis</ns0:head><ns0:p>The samples were sequenced on a HiSeq X Ten platform (Illumina, San Diego, CA, USA) and paired-end reads were generated. Low-quality reads and adaptor sequences were removed, and clean reads were used for the following analysis. TopHat2 <ns0:ref type='bibr'>(Kim et al., 2013)</ns0:ref> was used to map the clean reads onto the Brassica napus reference genome <ns0:ref type='bibr'>(Chalhoub et al., 2014)</ns0:ref> with default parameters '--read-mismatches 2 --read-edit-dist 2 --library-type fr---max-intron-length 5000000'. The number of fragments per kilobase of transcripts per million fragments mapped (FPKM) was calculated to represent the gene expression level, and the DESeq R package <ns0:ref type='bibr' target='#b0'>(Anders & Huber, 2010)</ns0:ref> was used to analyze the differential expression. The P-value was adjusted using Benjamini and Hochberg's approach to control the false discovery rate (FDR).</ns0:p><ns0:p>The relative expression levels of each transcript calculated using DESeq were used to define the DEGs, which were defined as having a fold change > 4 and an FDR < 0.01. Pearson's correlation coefficients were determined for the three biological replicates of each line to determine the reliability of the DEGs. Moreover, real-time quantitative PCR (qPCR) was performed to validate the transcriptome sequencing. Since the validation for transcriptome sequencing data from plants in Xindu had been confirmed previously <ns0:ref type='bibr' target='#b3'>(Chai et al., 2019)</ns0:ref>, we only validated the data from plants in colder Ma'erkang herein. Amplification reactions were performed on iQ™5 Real-Time PCR System (Biorad) as follows: an initial denaturation step at 95 °C for 3 min, 39 cycles at 95 °C for 10 s, 60 °C for 30 s, and 72 °C for 30 s. After each run, a melt curve was acquired to check for amplification specificity by heating the samples from 60 °C to 95 °C. Three biological replicates were applied.</ns0:p></ns0:div>
<ns0:div><ns0:head>AS Event Analysis</ns0:head><ns0:p>The cleaned sequence data were aligned to the reference genome using TopHat2 <ns0:ref type='bibr'>(Kim et al., 2013)</ns0:ref> with default settings mentioned above. The resultant gapped alignment data in a binary alignment format were then used as an input for Cufflinks and Cuffcompare, which were run using the default settings to assemble the transcripts and identify splicing junctions from the alignment data. For the AS detection and annotation, the AS events were annotated with <ns0:ref type='bibr'>ASprofile (Florea et al., 2013)</ns0:ref>, which uses Cufflinks and Cuffcompare outputs as input data.</ns0:p><ns0:p>Default parameters of the software were used.</ns0:p></ns0:div>
<ns0:div><ns0:head>Annotation of Genes</ns0:head><ns0:p>Gene function was annotated based on the following databases: Nr (NCBI nonredundant protein sequences), Nt (NCBI nonredundant nucleotide sequences), Pfam (Protein family), KOG/COG (Clusters of Orthologous Groups of proteins), Swiss-Prot (a manually annotated and reviewed protein sequence database), KO (KEGG Ortholog database), and GO (Gene Ontology).</ns0:p><ns0:p>The GO enrichment analysis of the DEGs was performed using the GOseq R packages based on a Wallenius noncentral hypergeometric distribution <ns0:ref type='bibr' target='#b40'>(Young et al., 2010)</ns0:ref>, which can adjust for gene length bias in the DEGs.</ns0:p><ns0:p>The KEGG database <ns0:ref type='bibr'>(Kanehisa et al., 2007)</ns0:ref> is a resource used to explore the high-level functions and utilities of the biological system, such as the cell, organism, and ecosystem, from molecular-level information, especially using large-scale molecular datasets generated from genome sequencing and other high-throughput experimental technologies (http://www.genome.jp/kegg/). KOBAS <ns0:ref type='bibr' target='#b5'>(Mao et al., 2005)</ns0:ref> software was used to test the statistical enrichment of the DEGs in the various KEGG pathways. Default parameters were used.</ns0:p></ns0:div>
<ns0:div><ns0:head>Results</ns0:head></ns0:div>
<ns0:div><ns0:head>Transcriptome Sequencing</ns0:head><ns0:p>Flower buds from three plants of both the multi-silique line zws-ms and the single-silique NIL zws-217 (Figure <ns0:ref type='figure'>1</ns0:ref>) were sampled for RNA extraction. The sequencing saturation and cluster analysis of the samples were determined to ensure the validity of the data. In total, 65.6 Gb of clean data were generated, with an average Q30 value of 90.54%. Each sample generated about 36.65 M clean reads with an average GC content of 47.23% (Table <ns0:ref type='table' target='#tab_1'>S1</ns0:ref>). The average proportion of total reads mapped to the reference genome for each sample was 73.72% (Table <ns0:ref type='table'>S2</ns0:ref>).</ns0:p><ns0:p>Validation of this transcriptome sequencing data was previously confirmed by qPCR <ns0:ref type='bibr' target='#b3'>(Chai et al., 2019)</ns0:ref>. Similarly, samples from colder area Ma'erkang also generated abundant data, which was validated by comparing the relative transcript levels of eight DEGs in zws-ms and zws-217 by qPCR. The qPCR analysis (Figure <ns0:ref type='figure'>S1</ns0:ref>) showed that all genes had similar trends in expression as those observed by transcriptome sequencing (described below). Each sample generated about 22.92M clean reads with an average GC content of 46.27%, and Q30 value of 92.95% (Table <ns0:ref type='table'>S3</ns0:ref>); average proportion of total reads mapped to the reference genome for each sample was 88.87% (Table <ns0:ref type='table'>S4</ns0:ref>).</ns0:p></ns0:div>
<ns0:div><ns0:head>AS Event Identification and Analysis</ns0:head><ns0:p>According to description by <ns0:ref type='bibr' target='#b17'>Reddy (2007)</ns0:ref>, alternative splicing events were sorted into 5 classes:</ns0:p><ns0:p>Alternative 3' splice site, Alternative 5' splice site, Exon Skipping, Intron Retention and Mutually Exclusive Exons. The six samples grown in Xindu under normal conditions displayed an average of 205,496 AS events (Table <ns0:ref type='table' target='#tab_1'>1</ns0:ref>; Table <ns0:ref type='table'>S5</ns0:ref>). The proportions of each AS type were analyzed in both zws-ms and zws-217. The two lines shared highly similar ratios of each AS type, with the alternative 5' splice site and alternative 3' splice site types being the most commonly observed, at 43.48% and 42.77% of AS events for both lines, respectively. The mutually exclusive exons type was the next most common (6.61%), followed by the Intron Retention type (5.92%), and the least common types was Exon Skipping, which represented just 1.22% of the AS events (Figure <ns0:ref type='figure'>2a</ns0:ref>; Table <ns0:ref type='table' target='#tab_1'>1</ns0:ref>).</ns0:p><ns0:p>As to the plants grown in colder area Ma'erkang, the two lines also shared highly similar ratios of each AS type: the alternative 5' splice site and alternative 3' splice site types represented the greatest proportion, at 42.13% and 41.29%, respectively; while the exon skipping accounted least proportion 1.51% (Figure <ns0:ref type='figure'>2b</ns0:ref>; Table <ns0:ref type='table' target='#tab_1'>1</ns0:ref>). The number of each type of AS events were significantly reduced in colder area Ma'erkang, except the Intron Retention.</ns0:p></ns0:div>
<ns0:div><ns0:head>Annotation of the Alternatively Spliced Genes</ns0:head><ns0:p>To study the biological functions of the genes with AS events, GO and KEGG pathway enrichment analyses were performed. The GO annotations for AS genes from plants in Xindu included 17 terms involved in biological processes (BP; Figure <ns0:ref type='figure'>3a</ns0:ref>), 17 terms associated with cellular components (CC), and 20 terms involved in molecular functions (MF). The most highly enriched BP terms observed in the alternatively spliced genes included 'cellular process', 'single-organism process' and 'metabolic process'. The most common CC categories were 'cell', 'cell part' and 'organelle'. In the MF category, the most enriched terms were 'binding', 'catalytic activity' and 'nucleic acid binding transcription factor activity'. Plants grown in Ma'erkang showed highly similar GO data to that in Xindu: 22 terms involved in BP (Figure <ns0:ref type='figure'>3b</ns0:ref>), 15 terms associated with CC, and 15 terms involved in MF. Moreover, data from Ma'erkang and Xindu showed the same top-3 most enriched terms in each category.</ns0:p><ns0:p>These KEGG pathways were classified into five major groups: metabolism, genetic information processing, cellular processes, environmental information processing, and organismal systems. Of these, the subgroups 'biosynthesis of amino acids', 'carbon metabolism', 'ribosome', and 'RNA transport' contained the highest number of annotated genes (Figure <ns0:ref type='figure' target='#fig_5'>4a</ns0:ref>). Data from Ma'erkang showed similar subgroups containing the most of AS genes: 'ribosome', 'carbon metabolism', 'biosynthesis of amino acids' and 'plant hormones signal transduction' contained the highest number of annotated genes (Figure <ns0:ref type='figure' target='#fig_5'>4b</ns0:ref>).</ns0:p></ns0:div>
<ns0:div><ns0:head>DEGs with AS and Their Arabidopsis Orthologs</ns0:head><ns0:p>DESeq software was used to identify the different expression levels of the AS genes in zws-ms and zws-217. From Xindu, eleven differentially expressed AS genes were identified, of which four were upregulated and seven were downregulated in zws-ms (Table <ns0:ref type='table'>2</ns0:ref>).</ns0:p><ns0:p>The Arabidopsis orthologs of these differentially expressed AS genes were identified using</ns0:p><ns0:p>The Arabidopsis Information Resource (TAIR; https://www.arabidopsis.org; Table <ns0:ref type='table'>3</ns0:ref>). The AT1G10760 encodes an α-glucan, water dikinase (GWD) required for starch degradation.</ns0:p><ns0:p>When grown in Ma'erkang, the two lines generated increased number of differentially expressed AS genes significantly to 130 (Table <ns0:ref type='table'>S6</ns0:ref>), including 52 unregulated and 78 down</ns0:p><ns0:p>PeerJ reviewing PDF | (2019:12:44287:3:0:NEW 7 Aug 2020)</ns0:p><ns0:p>Manuscript to be reviewed regulated AS genes. Four AS genes were annotated to 'response to cold (GO:0009409)':</ns0:p><ns0:p>BnaAnng17190D, BnaC01g27600D, BnaC08g39130D and BnaC09g53990D; three AS genes, BnaA07g19340D, BnaC01g27600D, BnaC08g39130D, were annotated to 'response to heat (GO:0009408)'; three were related to 'response to freezing (GO:0050826)', including BnaA05g28590D, BnaC06g15710D and BnaC08g36010D; BnaCnng24040D was found relevant to 'temperature stimulus (GO:0009266)'. Moreover, BnaC08g36010D, BnaC08g39130D and BnaC08g39360D were annotated to 'regulation of flower development (GO:0009909)', 'plant ovule development (GO:0048481)' and 'fruit development (GO:0010154)', respectively (Table <ns0:ref type='table'>S6</ns0:ref>). Compared with the 11 differentially expressed AS genes from normal conditions in Xindu, three of them (BnaA07g04500D, BnaAnng30260D and BnaC06g16950D) maintained the same expression models. In other words, these three genes were upregulated under both normal and colder conditions. While the other 8 genes (BnaA02g02630D, BnaA02g03080D, BnaA04g16220D, BnaA09g45000D, BnaA09g45260D, BnaC02g06440D, BnaC07g33980D and BnaC08g49610D) did not show significant difference between zms-ms and zws-217 in expression level in Ma'erkang.</ns0:p></ns0:div>
<ns0:div><ns0:head>Genes with Line-specific AS Events</ns0:head><ns0:p>Genes with line-specific AS events, defined as those genes with a particular AS event(s) that occurred only in zws-ms or in zws-217, were also identified and analyzed. Unlike the abovementioned general classifications, we sorted AS events into 12 finer subclasses, in order to identify them more specifically: transcription start site (TSS) and transcription terminal site AS genes were detected, of which 187 could be annotated (Table <ns0:ref type='table'>S7</ns0:ref>). Ten genes related to 'ovule development', 'flower development' and other similar processes were highlighted and considered important for further study in the coming future (Table <ns0:ref type='table'>4</ns0:ref>, Table <ns0:ref type='table'>S7</ns0:ref>): (1)</ns0:p><ns0:p>BnaC06g32640D was annotated as 'vegetative to reproductive phase transition of meristem (5) BnaC04g26180D and (6) BnaC07g22680D were annotated with 'development (GO:0048481)'; (7) BnaC07g25280D was annotated as 'flower morphogenesis;</ns0:p><ns0:p>organ morphogenesis (GO:0009887)' and 'vegetative to reproductive phase transition of meristem (GO:0010228)'; (8) BnaC01g16410D was annotated as 'flower development (GO:0009908)'; (9) BnaC03g32190D was annotated as 'double fertilization forming a zygote and endosperm (GO:0009567)'; and (10) BnaCnng68400D was associated with 'carpel development (GO:0048440).'</ns0:p></ns0:div>
<ns0:div><ns0:head>Discussion</ns0:head><ns0:p>As an important post-transcriptional metabolic event, AS is involved in many plant growth and developmental processes, such as flowering induction <ns0:ref type='bibr'>(Eckardt, 2002;</ns0:ref><ns0:ref type='bibr' target='#b24'>Slotte et al., 2009)</ns0:ref> and the responses to environmental fluctuations and pathogen attacks <ns0:ref type='bibr' target='#b1'>(Barbazuk et al., 2008)</ns0:ref>. To the best of our knowledge, AS events have seldom been reported to regulate the development of flower/fruit morphology in higher plants. This study is the first to analyze the role of AS events in rapeseed flower/fruit development as a whole, let alone those related to the multi-silique trait.</ns0:p><ns0:p>We previously described the morphology and inheritance of the multi-silique trait in B.</ns0:p><ns0:p>napus <ns0:ref type='bibr'>(Jiang et al., 1998)</ns0:ref>, investigating the associated regions of chromosomes at the genomic level and transcriptomically exploring the DEGs in multi-silique and single-silique plants <ns0:ref type='bibr' target='#b3'>(Chai et al., 2019)</ns0:ref>. The multi-silique trait was found to be controlled by three recessive alleles and was significantly affected by environment; however, the mechanisms by which environmental factors affect this trait remained unknown, even if we knew that temperature could switch on/off the multi-silique trait <ns0:ref type='bibr' target='#b3'>(Chai et al., 2019)</ns0:ref>. As mentioned above, AS is a pathway by which the environment could regulate plant physiology, therefore in this study, we analyzed AS events in order to investigate the mechanism by which plants perceive temperature fluctuations.</ns0:p><ns0:p>In this study, we sampled the buds of three individual plants from zws-ms and zws-217 lines in both Xindu and colder area Ma'erkang, and then subjected them to RNA-seq. All of the four groups generated sufficient data, which was validated by qPCR in earlier and present study successively. The samples in Ma'erkang group generated less data, mainly due to the less sequencing depth; nevertheless, this still provided enough data of high quality and assured the accuracy of the subsequent analysis.</ns0:p><ns0:p>We identified all of the genes with AS events in the zws-ms and zws-217 plants. In Xindu group, 11 AS genes were significantly differentially expressed between the multi-silique zws-ms line and its NIL, zws-217, which produces normal siliques. We analyzed their annotations and orthologs in Arabidopsis. One such ortholog, AT5G15470 (also known as Galacturonosyltransferase 14, GAUT14), is involved in cell wall pectin biosynthesis <ns0:ref type='bibr' target='#b2'>(Caffall et al., 2009)</ns0:ref>, and the gaut13 gaut14 double mutant was previously shown to be defective in pollen tube growth <ns0:ref type='bibr'>(Wang et al., 2013)</ns0:ref>. AT3G15420 (the ortholog of BnaA02g03080D) and AT3G10070 (the ortholog of BnaA09g45000D) encode subunits of the transcription factor complexes TFIIIC and TAF12, respectively. The former does not appear to be substantially involved in plant development; however, some members of the TAF family are involved in the regulation of morphology. The transgenic expression of TAF10 from clustered yellowtops (Flaveria trinervia) in Arabidopsis limited the development of the indeterminate inflorescence and resulted in the production of deformed leaves <ns0:ref type='bibr'>(Furumoto et al., 2005)</ns0:ref>. By contrast, the taf mutant in Arabidopsis has abnormal phyllotaxis and lacks proper vegetative meristem activity <ns0:ref type='bibr' target='#b27'>(Tamada et al., 2007)</ns0:ref>, indicating the important roles played by the TAFs in plant morphological development. Another DEG AS gene, BnaA04g16220D, is not annotated, and its Arabidopsis ortholog AT1G14800 is simply listed as an uncategorized nucleic acid-binding, OB-fold-like protein. The AS gene orthologs AT2G04900 and AT1G15060 encode an unknown protein and an uncategorized alpha/beta hydrolase family protein, respectively, so their roles in the regulation of the multi-silique trait are also currently unclear.</ns0:p><ns0:p>Another ortholog for differentially expressed AS gene, AT3G54620, is reported to encode a bZIP transcription factor-like protein. Members of this protein family are typically reported to regulate plant tolerance of environmental stresses. The transgenic expression of the maize (Zea mays) gene ZmbZIP72 in Arabidopsis enhanced its drought and salt tolerance <ns0:ref type='bibr' target='#b39'>(Ying et al., 2012)</ns0:ref>, while BnbZIP3, a ramie (Boehmeria nivea) bZIP transcription factor, also increased the drought, salinity, and heavy metal tolerances of transgenic Arabidopsis <ns0:ref type='bibr'>(Huang et al., 2016)</ns0:ref>. These genes are also involved in the regulation of other processes; for example, the repression of a bZIP transcription factor gene OsABI5 expression in rice resulted in low fertility <ns0:ref type='bibr' target='#b45'>(Zou et al., 2008)</ns0:ref>, while the transgenic expression of tomato (Solanum lycopersicum) SlbZIP2 in tobacco (Nicotiana benthamiana) increased leaf thickness <ns0:ref type='bibr' target='#b23'>(Seong et al., 2016)</ns0:ref>. To date, however, there are no reports of bZIP genes playing a significant role in flower/fruit morphology.</ns0:p><ns0:p>Other AS gene orthologs included AT5G16210, encoding a member of the HEAT repeatcontaining protein family, which are considered to be involved in intracellular transport <ns0:ref type='bibr'>(Hernández-Torres et al., 2014;</ns0:ref><ns0:ref type='bibr' target='#b11'>Oeffinger et al., 2004)</ns0:ref>. Although BnaC06g16950D is not annotated, its ortholog, AT3G59000, was identified as encoding an F-box/RNI-like superfamily protein in Arabidopsis, which typically function in the plant hormone signaling pathways <ns0:ref type='bibr'>(Gao et al., 2009)</ns0:ref>. Similarly, the ortholog AT4G16900 encodes a TIR-NBS-LRR class protein, which are known to be involved in disease resistance <ns0:ref type='bibr' target='#b37'>(Xun et al., 2019)</ns0:ref> and hormonal responses <ns0:ref type='bibr' target='#b21'>(Sarazin et al., 2015)</ns0:ref>. <ns0:ref type='bibr'>Moreover, Hewezi et al. (2006)</ns0:ref> unexpectedly found that these proteins are associated with developmental abnormalities; transgenic sunflowers (Helianthus annuus) expressing the antisense sequence complementing PLFOR48, which encodes a TIR-NBS-LRR-type protein,</ns0:p><ns0:p>showed stunted growth and a reduction in apical dominance; whereas the pods of transgenic tobacco (N. tabacum) lacking PLFOR48 expression were smaller and showed severe deformations. This indicates that TIR-NBS-LRR-type proteins can regulate the morphology of plants, including fruit morphology, to some extent. Finally, AT1G10760, the ortholog of AS gene BnaC08g49610D, which encodes a GWD protein required for starch degradation, is involved in carbohydrate metabolism <ns0:ref type='bibr' target='#b9'>(Nadolska-Orczyk et al., 2017)</ns0:ref>. This gene was also reported to regulate seed size; <ns0:ref type='bibr' target='#b16'>Pirone et al. (2017)</ns0:ref> found that the length and width of the mature seeds were reduced in the gwd1 Arabidopsis mutant, while their density was increased.</ns0:p><ns0:p>To summarize, AT5G15470, AT3G10070, AT3G54620, AT4G16900, and AT1G10760 are all known to be involved in plant development; therefore, their corresponding rapeseed orthologs, BnaA02g02630D, BnaA09g45000D, BnaAnng30260D, BnaC07g33980D, and BnaC08g49610D, the expression levels of which differed significantly between zws-ms and zws-217, are considered to be potential candidate genes regulating the multi-silique trait.</ns0:p><ns0:p>After that, we continued to investigate data from Ma'erkang, where it is colder and the multi-silique trait in zws-ms line disappeared. Due to the importance of the above-mention 11 differentially expressed AS genes, we first paid attention to them and found that three of them had the same expression models as in Xindu. In other words, they were independent of temperature; in addition, combined with their annotations, they were excluded from the potential PeerJ reviewing PDF | (2019:12:44287:3:0:NEW 7 Aug 2020)</ns0:p><ns0:p>Manuscript to be reviewed candidate genes responding to environmental factors. On the other hand, other 8 AS genes stopped being differentially expressed between zms-ms and zws-217 in Ma'erkang. That is to say, the expression level of these 8 AS genes (BnaA02g02630D, BnaA02g03080D, BnaA04g16220D, BnaA09g45000D, BnaA09g45260D, BnaC02g06440D, BnaC07g33980D and BnaC08g49610D) were environment-specifc. Besides, BnaC08g36010D, BnaC08g39130D and BnaC08g39360D, which were annotated to flower/ovule/fruit-related terms, were differentially expressed between zws-ms and zws-217 specifically in Ma'erkang. Moreover, we also found 9 temperature-responding AS genes differentially expressed in colder area: BnaAnng17190D, BnaC01g27600D, BnaC08g39130D, BnaC09g53990D, BnaA07g19340D, BnaA05g28590D, BnaC06g15710D, BnaC08g36010D and BnaCnng24040D. The lower temperature motivated more responding genes and this also explained the reason for increased number of AS genes clustered in each KEGG pathway.</ns0:p><ns0:p>We also explored the line-specific AS genes, which were similarly expressed between zwsms and zws-217, but contained stable and particular AS event(s) that differed between these two lines. These genes are likely to qualitatively regulate the multi-silique trait. In this case, we could obtain better results by fine-classify the AS types into 12 subclasses, rather than 5 classes mentioned above. Because fine classifications could better identify differences between AS types more precisely and subtly. Thus, we found 205 genes of this type, of which 187 could be annotated. Due to the rarity of the multi-silique trait, we did not obtain much useful information from the KEGG pathway analysis. This meant that we were unable to relate this metabolic pathway information to the multi-silique trait directly; however, the GO analysis provided more potential clues. Among these, 10 genes were considered to be associated with flower/carpel/ovule development. BnaC06g32640D is annotated as being involved in the regulation of the vegetative-to-reproductive phase transition in the meristem (GO:0010228) and in ovule development (GO:0048481). Its Arabidopsis ortholog, AT1G71692, is annotated as (GR), which was found to increase the fineness (mass per unit length) and bundle strength of cotton (Gossypium hirsutum) fiber when transgenically expressed <ns0:ref type='bibr' target='#b29'>(Tuttle et al., 2015)</ns0:ref>. Since cotton fibers are single cells initiating from the epidermis of the outer integument of the ovules <ns0:ref type='bibr' target='#b19'>(Ruan et al., 2004)</ns0:ref>, it can be inferred that GR regulates ovule development to some extent.</ns0:p><ns0:formula xml:id='formula_0'>AGAMOUS-LIKE12 (AGL12).</ns0:formula><ns0:p>The line-specific AS gene ortholog AT2G04030 encodes HSP90, a member of the heat shock proteins (HSPs), which are commonly produced in response to heat. The HSPs are molecular chaperones that prevent protein aggregation and mediate the refolding of heatdenatured proteins <ns0:ref type='bibr' target='#b8'>(Murano et al., 2017)</ns0:ref>. Moreover, a HSP was found to be upregulated in the fiber-bearing ovules of cotton (Gossypium hirsutum) <ns0:ref type='bibr'>(Lee et al., 2006)</ns0:ref> implying some unknown function in ovule development. Another ortholog, AT3G28730 (also known as structure-specific recognition protein SSRP1), was also found to regulate floral development, as the ssrp1-2 mutant Arabidopsis produced small and deformed petals with shorter stamens <ns0:ref type='bibr'>(Lolas et al., 2010)</ns0:ref>.</ns0:p><ns0:p>AT4G24560 encodes a ubiquitin-specific protease (UBP), which are known to play critical roles in protein deubiquitination in plants <ns0:ref type='bibr'>(Liu et al., 2008)</ns0:ref>. The UBPs are also involved in plant development; <ns0:ref type='bibr'>Liu et al. (2008)</ns0:ref> found that knocking out UBP15 function in Arabidopsis resulted in the production of smaller flowers and shorter siliques. The final line-specific AS ortholog, AT5G15020, encodes an SIN3-LIKE 2 protein (SNL2) known to be important for seed germination or dormancy <ns0:ref type='bibr' target='#b32'>(Wang et al., 2016;</ns0:ref><ns0:ref type='bibr'>Wang et al., 2013)</ns0:ref>.</ns0:p><ns0:p>To date, there is some evidence to show that the line-specific AS orthologs AT2G20180, AT4G00050, AT3G54660, and AT2G04030 are related to the regulation of flower/fruit morphology, with clear roles reported for AT1G71692, AT3G28730, and AT4G24560.</ns0:p><ns0:p>Consequently, their orthologs in rapeseed, BnaC06g32640D, BnaC07g25280D, and BnaC01g16410D, respectively, are considered to be important candidate genes regulating the multi-silique trait by conferring or removing some specific line-specific AS events.</ns0:p><ns0:p>Some of the genes/loci controlling silique development in Brassica plants have previously been reported. In addition to those regulating traits such as the seed weight and silique length <ns0:ref type='bibr' target='#b15'>(Liu et al., 2015)</ns0:ref> and the number of seeds per silique in B. napus <ns0:ref type='bibr' target='#b15'>(Li et al., 2015)</ns0:ref>, some genes related to silique morphology have been cloned and functionally analyzed. <ns0:ref type='bibr' target='#b36'>Xiao et al. (2013)</ns0:ref> fine-mapped a multi-locular silique gene, Bjln1, to a 208-kb region on chromosome A7 in Brassica juncea and then revealed that it was the mutations in the CDS and promoter of BjuA07.CLV1 gene (equivalent to Bjln1) to cause the multi-locular trait <ns0:ref type='bibr' target='#b35'>(Xiao et al., 2018)</ns0:ref>. Both Manuscript to be reviewed</ns0:p><ns0:p>To sum up, the eight candidate genes mentioned above, including the five differentially expressed AS genes of interest and the three genes with line-specific AS events, are therefore hypothesized to regulate the multi-silique trait in rapeseed zws-ms, based on their AS expression levels or line-specific AS events. These findings lay a foundation for further functional analyses in future. The data obtained from plants in colder environment, including the amount and the proportion of each AS type, was generally similar to that under normal conditions, while details are still under further investigation.</ns0:p></ns0:div>
<ns0:div><ns0:head>Conclusions</ns0:head><ns0:p>The utilization of heterosis is a way to increase the yield or improve the quality of crops.</ns0:p><ns0:p>Exploring new germplasm resources and genes, as well as clarifying their inheritance, is the foundation of obtaining of excellent hybrid. This study provides a novel inspection into the multi-silique trait in rapeseed from the transcriptional perspective by AS, deepening the understanding of its molecular mechanism. Further function verifications are now undergoing.</ns0:p><ns0:note type='other'>Figure 1</ns0:note><ns0:p>Multi-silique trait in zws-ms, compared with the single siliques of its near-isogenic line zws-217. Manuscript to be reviewed Numbers of alternative splicing events in six samples.</ns0:p><ns0:p>Note: T01, T02, and T03: Buds of three independent zws-ms plants at the budding stage; T04, T05, and T06: Buds of three independent zws-217 plants at the budding stage.</ns0:p><ns0:p>Alternative 3' splice site: different-size mRNAs are produced depending on the usage of a proximal or distal 3' splice site; Alternative 5' splice site: different-size mRNAs are produced depending on the use of a proximal or distal 5' splice site; Exon Skipping: an exon is either included or excluded from the mRNA; Intron Retention: an intron is either retained or excised in the mRNA, resulting in different-size transcripts; Mutually Exclusive Exons: adjacent exons are spliced in such a way that only one of them is included at a time in the mRNA.</ns0:p><ns0:p>PeerJ reviewing PDF | (2019:12:44287:3:0:NEW 7 Aug 2020)</ns0:p><ns0:p>Manuscript to be reviewed</ns0:p></ns0:div><ns0:figure xml:id='fig_0'><ns0:head /><ns0:label /><ns0:figDesc>following orthologs were identified: (1) AT5G15470, the ortholog of BnaA02g02630D, encodes galacturonosyltransferase 14 (GAUT14); (2) AT3G15420 encodes the transcription factor TFIIIC (tau55-related protein); (3) AT1G14800 encodes a nucleic acid-binding, OB-fold-like protein; (4) AT2G04900 encodes an unknown protein; (5) AT3G10070 encodes one of two Arabidopsis proteins with similarity to the TBP-associated factor, TAF12; (6) AT1G15060 encodes an alpha/beta hydrolase family protein; (7) AT3G54620 encodes a bZIP transcription factor-like protein; (8) AT5G16210 encodes a HEAT repeat-containing protein; (9) AT3G59000 encodes an F-box/RNI-like superfamily protein; (10) AT4G16900, the ortholog of BnaC07g33980D, encodes a member of the disease resistance protein (TIR-NBS-LRR class) family; and (11)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_1'><ns0:head>(</ns0:head><ns0:label /><ns0:figDesc>TTS) equaled to original Alternative 5' first exon and Alternative 3' last exon, respectively; Mutually exclusive exons was subdivided into Alternative exon ends (AE) and Approximate AE (XAE); the original Intron retention was subdivided into single Intron retention (IR), Approximate IR (XIR), Multi-IR (MIR) and Approximate MIR (XMIR); the Cassette exon was then was subdivided into single Skipped exon (SKIP), Approximate SKIP (XSKIP), Multi-exon SKIP (MSKIP) and Approximate MSKIP (XMSKIP). In total, 205 line-specifically expressed</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_2'><ns0:head>(</ns0:head><ns0:label /><ns0:figDesc>GO:0010228)' and 'Biological Process: ovule development (GO:0048481)'; (2) BnaC07g00780D was associated with 'reproductive structure development (GO:0048608)'; (3) BnaC04g31460D and (4) BnaC05g34570D were related to 'regulation of flower development (GO:0009909)';</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_3'><ns0:head>Fan</ns0:head><ns0:label /><ns0:figDesc>et al. (2014) andYadava et al. (2014) reported that a mutation in BrCLV3, a homologue of CLAVATA3 in Arabidopsis, caused the production of multi-locular siliques in B. rapa. However, the multi-silique (or multi-pistil) phenotype of zws-ms is different from the above-motioned multi-locular trait; zws-ms produces three pods on each carpopodium, rather than multiple loculi per pod.Few studies have investigated this multi-silique trait in rapeseed; however, there have been similar reports of multi-pistil traits in other crops, particularly in wheat (Triticum aestivum):Duan et al. (2015) discovered a male-sterile wheat mutant, dms, with a dwarf status and multipistils, a pleiotropic phenotype found to be controlled by a single recessive gene, which was not identified.<ns0:ref type='bibr' target='#b37'>Guo et al. (2019)</ns0:ref> reported another multi-ovary trait in the wheat line DUOII, which was controlled by a dominant gene, and used a proteomics approach to propose some candidate proteins.<ns0:ref type='bibr' target='#b38'>Yang et al. (2017)</ns0:ref> mapped a gene promoting the formation of three pistils (Pis1) to chromosome 2D and identified some candidate genes according to their annotations, while<ns0:ref type='bibr' target='#b44'>Zhu et al. (2019)</ns0:ref> discovered a wheat multi-pistil mutant, 12TP, which was found to contain a semidominant mutation located on chromosome arm 2DL. Although several studies have explored the multi-pistil trait in wheat, no one has identified any of the specific genes responsible yet.PeerJ reviewing PDF | (2019:12:44287:3:0:NEW 7 Aug 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_4'><ns0:head>(</ns0:head><ns0:label /><ns0:figDesc>A) Primary inflorescences. (B) Siliques.</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_5'><ns0:head>Figure 4 Classified</ns0:head><ns0:label>4</ns0:label><ns0:figDesc>Figure 4</ns0:figDesc></ns0:figure>
<ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='25,42.52,70.87,333.62,672.95' type='bitmap' /></ns0:figure>
<ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='26,42.52,229.87,525.00,230.25' type='bitmap' /></ns0:figure>
<ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='27,42.52,280.87,525.00,214.50' type='bitmap' /></ns0:figure>
<ns0:figure type='table' xml:id='tab_0'><ns0:head /><ns0:label /><ns0:figDesc><ns0:ref type='bibr' target='#b15'>Peng et al. (2015)</ns0:ref> isolated the BnFUL gene in rapeseed, which is homologous to AGL8 in Arabidopsis. Although BnFUL was hypothesized to be involved in enhancing pod-shattering resistance, when introduced into Arabidopsis, two of the five transgenic plants expressing BnFUL unexpectedly had a multi-silique phenotype. However, the mechanisms by which BnFUL generates this multi-silique phenotype remain elusive thus far, making the AGL12 gene identified in this study a potentially important candidate gene.Other orthologs of the line-specific AS genes include AT2G20180 and AT4G00050, both of which encode phytochrome interacting factors (PIFs). Several transcription factors (AP1, SVP, LFY, AG, and SEP3) involved in the regulation of flowering are known to bind to the PIFs, suggesting a direct link with the reported flowering phenotype of the pif mutants (Leivar &</ns0:figDesc><ns0:table><ns0:row><ns0:cell>Monte, 2014). AT5G17270 encodes a prenylyltransferase superfamily protein; however, to the</ns0:cell></ns0:row><ns0:row><ns0:cell>best of our knowledge, there have been no reports about its development-related functions. The</ns0:cell></ns0:row><ns0:row><ns0:cell>Arabidopsis ortholog of BnaC05g34570D is AT3G18600, which encodes a P-loop-containing</ns0:cell></ns0:row><ns0:row><ns0:cell>nucleoside triphosphate hydrolase. While few studies have reported the functions of these</ns0:cell></ns0:row><ns0:row><ns0:cell>proteins, Liu et al. (2016) reported that, in sesame (Sesamum indicum), one gene encoding a P-</ns0:cell></ns0:row><ns0:row><ns0:cell>loop-containing nucleoside triphosphate hydrolase showed a reduced expression level in sterile</ns0:cell></ns0:row><ns0:row><ns0:cell>buds, indicating that they may play a role in specifying/determining tapetal fate and</ns0:cell></ns0:row><ns0:row><ns0:cell>development. Another line-specific AS ortholog, AT3G54660, encodes a glutathione reductase</ns0:cell></ns0:row></ns0:table></ns0:figure>
<ns0:figure type='table' xml:id='tab_1'><ns0:head>Table 1 (on next page)</ns0:head><ns0:label>1</ns0:label><ns0:figDesc /><ns0:table /></ns0:figure>
<ns0:note place='foot'>PeerJ reviewing PDF | (2019:12:44287:3:0:NEW 7 Aug 2020)Manuscript to be reviewed</ns0:note>
</ns0:body>
" | "Response to Editors and Reviewers
Dear Editors and Reviewers,
Thanks for your 3nd round recommendations. We have revised the manuscript thoroughly and hope it can meet your requirement.
Our response to your comments are as following:
Reviewer 2 (Anonymous)
Basic reporting
The authors have made significant improvement on the manuscript. I am fine with the manuscript with two minor revisions:
1. the manuscript still has grammar errors, such as line 79 'did not showed'. The authors should do read the entire manuscript thoroughly and correct all language errors.
Response:
We have corrected the grammar errors you mentioned and checked entire manuscript thoroughly.
2. For the added qPCR results in Figure S1, there is no mentioning of basic experimental conditions, such as the use of biological replicates and how many? the description of methods in manuscript should have sufficient information to allow readers to understand the design of the experiments. It should not be assumed that every paper uses the same conditions.
Response:
Thanks for your suggestions. We have added the details of basic experimental conditions into the manuscript. Please kindly check it.
" | Here is a paper. Please give your review comments after reading it. |
9,778 | "<ns0:abstract xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:p>Background. Flower and fruit development are vital stages of the angiosperm lifecycle. We previously investigated the multi-silique trait in the rapeseed (Brassica napus) line zws-ms on a genomic and transcriptomic level, leading to the identification of two genomic regions and several candidate genes associated with this trait. However, some events on transcriptome level, like alternative splicing, were poorly understood.</ns0:p><ns0:p>Methods. Plants from zws-ms and its near-isogenic line (NIL) zws-217 were both grown in Xindu with normal conditions and a colder area Ma'erkang. Buds from the two lines were sampled and RNA was isolated to perform the transcriptomic sequencing. The numbers and types of alternative splicing (AS) events from the two lines were counted and classified. Genes with AS events and expressed differentially between the two lines, as well as genes with AS events which occurred in only one line were emphasized. Their annotations were further studied.</ns0:p></ns0:div>
<ns0:div><ns0:head>Results.</ns0:head><ns0:p>From the plants in Xindu District, an average of 205,496 AS events, which could be sorted into 5 AS types, were identified. zws-ms and zws-217 shared highly similar ratios of each AS type: The alternative 5' and 3' splice site types were the most common, while the exon skipping type was observed least often. Eleven differentially expressed AS genes were identified, of which four were upregulated and seven were downregulated in zws-ms. Their annotations implied that five of these genes were directly associated with the multi-silique trait. While samples from colder area Ma'erkang generated generally reduced number of each type of AS events except the Intron Retention; but the number of differentially expressed AS genes increased significantly. Further analysis found that among the 11 differentially expressed AS genes from Xindu, three of them maintained the same expression models, while the other 8 genes did not show significant difference between the two lines in expression level. Additionally, the 205 line-specifically expressed AS genes were analyzed, of which 187 could be annotated, and two were considered to be important.</ns0:p><ns0:p>Discussion. This study provides new insights into the molecular mechanism of the agronomically important multi-silique trait in rapeseed on transcriptome level and screens out some environmentresponding candidate genes.</ns0:p></ns0:div>
</ns0:abstract>
<ns0:body xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:head>Introduction</ns0:head><ns0:p>Rapeseed (Brassica napus L.), an allotetraploid with a complex genome (AACC, 2n = 38), is the second leading source of vegetable oil globally <ns0:ref type='bibr' target='#b30'>(Liu et al., 2015)</ns0:ref>. The agronomic traits related to rapeseed yield include the pod (silique) number per plant, branch number, and seed weight <ns0:ref type='bibr' target='#b30'>(Liu et al., 2015;</ns0:ref><ns0:ref type='bibr' target='#b69'>Zhang et al., 2006;</ns0:ref><ns0:ref type='bibr'>Li et al., 2015)</ns0:ref>. We previously reported that zws-ms, a multisilique rapeseed line <ns0:ref type='bibr' target='#b3'>(Chai et al., 2019)</ns0:ref>, produces three independent pistils and 9 to 10 stamens on the same receptacle in a flower, which consequently leads to the formation of three independent siliques on a carpopodium rather than the single siliques typically observed.</ns0:p><ns0:p>Moreover, this trait was found to be affected by the environment, with temperature considered to be the factor most likely to switch on/off the formation of multi-silique.</ns0:p><ns0:p>Temperate is a major environmental factor that regulates various aspects of plant morphology, physiology, and biochemistry, affecting germination, growth, development, and flowering <ns0:ref type='bibr' target='#b44'>(Ren et al., 2019)</ns0:ref>. Fertility in crops such as rapeseed <ns0:ref type='bibr' target='#b67'>(Yu et al., 2015)</ns0:ref> and rice (Oryza sativa) <ns0:ref type='bibr' target='#b68'>(Yu et al., 2017)</ns0:ref> is affected by temperature. In winter rapeseed lines, although a period of vernalization under low temperature is necessary to initiate flowering, cold stress inhibits growth and development, disturbs metabolism, and causes wilting or even death. Notably, cold stress also induces alternative splicing (AS) in plants <ns0:ref type='bibr' target='#b39'>(Palusa et al., 2007;</ns0:ref><ns0:ref type='bibr' target='#b21'>Iida, 2004)</ns0:ref>.</ns0:p><ns0:p>AS is defined as the mechanism by which primary transcripts are processed into two or more mature isoforms, which enables a single gene to produce diverse protein products <ns0:ref type='bibr' target='#b40'>(Pan et al., 2008;</ns0:ref><ns0:ref type='bibr' target='#b46'>Sablok et al., 2011)</ns0:ref>. These proteins differ from each other not only in structure but also possibly in function, subcellular localization, and/or stability <ns0:ref type='bibr' target='#b20'>(Huang et al., 2019;</ns0:ref><ns0:ref type='bibr' target='#b5'>Chauhan et al., 2019)</ns0:ref>. AS is common in plants; for example, in Arabidopsis thaliana, more than 60% of intron-containing genes undergo AS <ns0:ref type='bibr' target='#b52'>(Syed et al., 2012)</ns0:ref>. Many environmental factors regulate AS events in plants, including CO 2 concentration <ns0:ref type='bibr' target='#b20'>(Huang et al., 2019)</ns0:ref>, light <ns0:ref type='bibr' target='#b14'>(Godoy et al., 2019)</ns0:ref>, salt stress <ns0:ref type='bibr' target='#b6'>(Ding et al., 2014)</ns0:ref>, and nutrient deficiencies <ns0:ref type='bibr' target='#b36'>(Nishida et al., 2017)</ns0:ref>. AS not only provides an important source of transcriptomic and proteomic diversity and plasticity for use in natural selection <ns0:ref type='bibr' target='#b25'>(Labadorf et al., 2010)</ns0:ref>, but it also plays specific roles in the response <ns0:ref type='bibr' target='#b5'>(Chauhan et al., 2019)</ns0:ref> or adaptation to environmental stresses <ns0:ref type='bibr' target='#b10'>(Filichkin et al., 2015)</ns0:ref>. <ns0:ref type='bibr'>Guo et al. (2019)</ns0:ref> identified four splicing variants of two BnCYCD3-1-LIKE genes in B. napus and found evidence that their AS may play an important role in the response to environmental stresses. <ns0:ref type='bibr' target='#b60'>Xia et al. (2017)</ns0:ref> discovered that the AS with intron retention of EARLY MATURITY8 (EAM8) led to early flowering in a barley (Hordeum vulgare) landrace; while in shepherd's purse (Capsella bursapastoris), flowering time varies with changes in the splicing of a FLOWERING LOCUS C (FLC) homolog <ns0:ref type='bibr' target='#b50'>(Slotte et al., 2009)</ns0:ref>. In addition, the heterologous expression of a vacuolar membrane Na + /H + antiporter gene (SsNHX1) AS variant from seepweed (Suaeda salsa) enhances the salt tolerance of Arabidopsis <ns0:ref type='bibr' target='#b28'>(Li et al., 2009)</ns0:ref>.</ns0:p><ns0:p>As mentioned above, low temperatures switch off the multi-silique trait in zws-ms rapeseed.</ns0:p><ns0:p>When zws-ms plants were planted in Xindu, Sichuan Province, China, the multi-silique trait was continuously stable for years; however, when they were grown in Ma'erkang, Sichuan Province, where the annual average temperature is consistently 7.6 °C lower, the multi-silique trait disappeared and all plants displayed normal siliques <ns0:ref type='bibr' target='#b3'>(Chai et al., 2019)</ns0:ref>. We previously investigated the association of chromosomal regions with this trait, at the genomic and transcriptomic levels, selecting potential candidates from the differentially expressed genes (DEGs) between the multi-and single-silique plants. However, the involvement of posttranscriptional modifications and the mechanisms by which temperature regulates this multisilique trait remain unclear. AS is often responsive to cold stress in plants <ns0:ref type='bibr' target='#b21'>(Iida, 2004;</ns0:ref><ns0:ref type='bibr' target='#b39'>Palusa et al., 2007)</ns0:ref> and is a mechanism by which plants perceive temperature fluctuations and modulate the activity of their transcription factors <ns0:ref type='bibr' target='#b48'>(Seo et al., 2013)</ns0:ref>. In view of the above insights, we analyzed AS using transcriptome sequencing (RNA-seq) in this study. High-throughput RNAseq technology is a widely used, highly efficient, and economical strategy for transcriptomic profiling <ns0:ref type='bibr' target='#b54'>(Tong et al., 2013;</ns0:ref><ns0:ref type='bibr' target='#b59'>Wang et al., 2009)</ns0:ref>. It has become increasingly popular because of the following qualities <ns0:ref type='bibr' target='#b34'>(Mortazavi et al., 2008;</ns0:ref><ns0:ref type='bibr' target='#b38'>Ozsolak & Milos, 2011;</ns0:ref><ns0:ref type='bibr' target='#b33'>Marioni et al., 2008;</ns0:ref><ns0:ref type='bibr' target='#b54'>Tong et al., 2013;</ns0:ref><ns0:ref type='bibr' target='#b51'>Sultan et al., 2008)</ns0:ref>: (1) It can be used to detect and quantify the expression of genes, including those expressed at low levels; (2) it can facilitate the annotation of genes and lead to the discovery of novel genes or transcripts; (3) the results are highly reproducible between both technical and biological replicates; and (4) it can detect AS events.</ns0:p><ns0:p>We performed transcriptome sequencing (RNA-seq) on the flower buds of zws-ms and its near-isogenic line (NIL), zws-217, which produces normal single siliques. This facilitated the identification of the AS events in both lines and the analysis of the differentially expressed AS genes and those with line-specific AS events. Combining these data with Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) annotations, we identified likely candidate genes switching on or off the multi-silique trait by altering AS events or transcriptional levels in varied environments.. To the best of our knowledge, this is the first time that the regulation of flower/fruit morphology by AS has been investigated in rapeseed, and our results provide insights into this field more generally.</ns0:p></ns0:div>
<ns0:div><ns0:head>Materials & Methods</ns0:head></ns0:div>
<ns0:div><ns0:head>Plant Materials and Growth Conditions</ns0:head><ns0:p>The rapeseed line zws-ms and its NIL, zws-217 <ns0:ref type='bibr' target='#b3'>(Chai et al., 2019)</ns0:ref>, were kept in the Crop Research Institute, Sichuan Academy of Agricultural Sciences, China. Both zws-ms and zws-217 were homozygous for almost all genes, differing from each other only in the multi-silique trait of zws-ms (Figure <ns0:ref type='figure'>1</ns0:ref>). The NILs zws-217 and zws-ms were both grown in an experimental field in the Xindu District of Chengdu in the Sichuan Basin, China, under normal environmental conditions. Additionally, the both lines were also grown in Ma'erkang, a mountainous area in western Sichuan, with a much lower annual average temperature. The annual average temperature in Xindu and Ma'erkang is 16.2 °C and 8.6 °C, respectively <ns0:ref type='bibr' target='#b3'>(Chai et al., 2019)</ns0:ref>.</ns0:p></ns0:div>
<ns0:div><ns0:head>Total RNA Extraction and Sequencing Library Construction</ns0:head><ns0:p>Three zws-ms plants (samples T01, T02, and T03) and three zws-217 plants (T04, T05, and T06) were selected for RNA isolation, as described previously <ns0:ref type='bibr' target='#b3'>(Chai et al., 2019)</ns0:ref>. Flower buds were detached from each plant at the budding stage (BBCH 57), and their total RNA was extracted using an RNA Isolation Kit (Tiangen, Beijing, China). The quality and concentration of the RNA were determined using a NanoDrop 2000 (Thermo Fisher Scientific, Waltham, MA, USA), and the sequencing libraries were generated using an RNA Library Prep Kit for Illumina (New England Biolabs, Ipswich, MA, USA).</ns0:p></ns0:div>
<ns0:div><ns0:head>Sequencing and Expression Analysis</ns0:head><ns0:p>The samples were sequenced on a HiSeq X Ten platform (Illumina, San Diego, CA, USA) and paired-end reads were generated. Low-quality reads and adaptor sequences were removed, and clean reads were used for the following analysis. TopHat2 <ns0:ref type='bibr' target='#b24'>(Kim et al., 2013)</ns0:ref> was used to map the clean reads onto the Brassica napus reference genome <ns0:ref type='bibr' target='#b4'>(Chalhoub et al., 2014)</ns0:ref> with default parameters '--read-mismatches 2 --read-edit-dist 2 --library-type fr---max-intron-length 5000000'. The number of fragments per kilobase of transcripts per million fragments mapped (FPKM) was calculated to represent the gene expression level, and the DESeq R package <ns0:ref type='bibr' target='#b0'>(Anders & Huber, 2010)</ns0:ref> was used to analyze the differential expression. The P-value was adjusted using Benjamini and Hochberg's approach to control the false discovery rate (FDR).</ns0:p><ns0:p>The relative expression levels of each transcript calculated using DESeq were used to define the DEGs, which were defined as having a fold change > 4 and an FDR < 0.01. Pearson's correlation coefficients were determined for the three biological replicates of each line to determine the reliability of the DEGs. Moreover, real-time quantitative PCR (qPCR) was performed to validate the transcriptome sequencing. Since the validation for transcriptome sequencing data from plants in Xindu had been confirmed previously <ns0:ref type='bibr' target='#b3'>(Chai et al., 2019)</ns0:ref>, we only validated the data from plants in colder Ma'erkang herein. Amplification reactions were performed on iQ™5 Real-Time PCR System (Biorad) as follows: an initial denaturation step at 95 °C for 3 min, 39 cycles at 95 °C for 10 s, 60 °C for 30 s, and 72 °C for 30 s. After each run, a melt curve was acquired to check for amplification specificity by heating the samples from 60 °C to 95 °C. Three biological replicates were applied.</ns0:p></ns0:div>
<ns0:div><ns0:head>AS Event Analysis</ns0:head><ns0:p>The cleaned sequence data were aligned to the reference genome using TopHat2 <ns0:ref type='bibr' target='#b24'>(Kim et al., 2013)</ns0:ref> with default settings mentioned above. The resultant gapped alignment data in a binary alignment format were then used as an input for Cufflinks and Cuffcompare, which were run using the default settings to assemble the transcripts and identify splicing junctions from the alignment data. For the AS detection and annotation, the AS events were annotated with ASprofile <ns0:ref type='bibr' target='#b11'>(Florea et al., 2013)</ns0:ref>, which uses Cufflinks and Cuffcompare outputs as input data.</ns0:p><ns0:p>Default parameters of the software were used.</ns0:p></ns0:div>
<ns0:div><ns0:head>Annotation of Genes</ns0:head><ns0:p>Gene function was annotated based on the following databases: Nr (NCBI nonredundant protein sequences), Nt (NCBI nonredundant nucleotide sequences), Pfam (Protein family), KOG/COG (Clusters of Orthologous Groups of proteins), Swiss-Prot (a manually annotated and reviewed protein sequence database), KO (KEGG Ortholog database), and GO (Gene Ontology).</ns0:p><ns0:p>The GO enrichment analysis of the DEGs was performed using the GOseq R packages based on a Wallenius noncentral hypergeometric distribution <ns0:ref type='bibr' target='#b66'>(Young et al., 2010)</ns0:ref>, which can adjust for gene length bias in the DEGs.</ns0:p><ns0:p>The KEGG database <ns0:ref type='bibr' target='#b23'>(Kanehisa et al., 2007)</ns0:ref> is a resource used to explore the high-level functions and utilities of the biological system, such as the cell, organism, and ecosystem, from molecular-level information, especially using large-scale molecular datasets generated from genome sequencing and other high-throughput experimental technologies (http://www.genome.jp/kegg/). KOBAS <ns0:ref type='bibr' target='#b32'>(Mao et al., 2005)</ns0:ref> software was used to test the statistical enrichment of the DEGs in the various KEGG pathways. Default parameters were used.</ns0:p></ns0:div>
<ns0:div><ns0:head>Results</ns0:head></ns0:div>
<ns0:div><ns0:head>Transcriptome Sequencing</ns0:head><ns0:p>Flower buds from three plants of both the multi-silique line zws-ms and the single-silique NIL zws-217 (Figure <ns0:ref type='figure'>1</ns0:ref>) were sampled for RNA extraction. The sequencing saturation and cluster analysis of the samples were determined to ensure the validity of the data. In total, 65.6 Gb of clean data were generated, with an average Q30 value of 90.54%. Each sample generated about 36.65 M clean reads with an average GC content of 47.23% (Table <ns0:ref type='table' target='#tab_2'>S1</ns0:ref>). The average proportion of total reads mapped to the reference genome for each sample was 73.72% (Table <ns0:ref type='table'>S2</ns0:ref>).</ns0:p><ns0:p>PeerJ reviewing PDF | (2019:12:44287:4:0:NEW 6 Sep 2020)</ns0:p><ns0:p>Manuscript to be reviewed</ns0:p><ns0:p>Validation of this transcriptome sequencing data was previously confirmed by qPCR <ns0:ref type='bibr' target='#b3'>(Chai et al., 2019)</ns0:ref>. Similarly, samples from colder area Ma'erkang also generated abundant data, which was validated by comparing the relative transcript levels of eight DEGs in zws-ms and zws-217 by qPCR. The qPCR analysis (Figure <ns0:ref type='figure'>S1</ns0:ref>) showed that all genes had similar trends in expression as those observed by transcriptome sequencing (described below). Each sample generated about 22.92M clean reads with an average GC content of 46.27%, and Q30 value of 92.95% (Table <ns0:ref type='table'>S3</ns0:ref>); average proportion of total reads mapped to the reference genome for each sample was 88.87% (Table <ns0:ref type='table' target='#tab_0'>S4</ns0:ref>).</ns0:p></ns0:div>
<ns0:div><ns0:head>AS Event Identification and Analysis</ns0:head><ns0:p>According to description by <ns0:ref type='bibr' target='#b43'>Reddy (2007)</ns0:ref>, alternative splicing events were sorted into 5 classes:</ns0:p><ns0:p>Alternative 3' splice site, Alternative 5' splice site, Exon Skipping, Intron Retention and Mutually Exclusive Exons. The six samples grown in Xindu under normal conditions displayed an average of 205,496 AS events (Table <ns0:ref type='table' target='#tab_2'>1</ns0:ref>; Table <ns0:ref type='table'>S5</ns0:ref>). The proportions of each AS type were analyzed in both zws-ms and zws-217. The two lines shared highly similar ratios of each AS type, with the alternative 5' splice site and alternative 3' splice site types being the most commonly observed, at 43.48% and 42.77% of AS events for both lines, respectively. The mutually exclusive exons type was the next most common (6.61%), followed by the Intron Retention type (5.92%), and the least common types was Exon Skipping, which represented just 1.22% of the AS events (Figure <ns0:ref type='figure'>2a</ns0:ref>; Table <ns0:ref type='table' target='#tab_2'>1</ns0:ref>).</ns0:p><ns0:p>As to the plants grown in colder area Ma'erkang, the two lines also shared highly similar ratios of each AS type: the alternative 5' splice site and alternative 3' splice site types represented the greatest proportion, at 42.13% and 41.29%, respectively; while the exon skipping accounted least proportion 1.51% (Figure <ns0:ref type='figure'>2b</ns0:ref>; Table <ns0:ref type='table' target='#tab_2'>1</ns0:ref>). The number of each type of AS events were significantly reduced in colder area Ma'erkang, except the Intron Retention.</ns0:p></ns0:div>
<ns0:div><ns0:head>Annotation of the Alternatively Spliced Genes</ns0:head><ns0:p>To study the biological functions of the genes with AS events, GO and KEGG pathway enrichment analyses were performed. The GO annotations for AS genes from plants in Xindu included 17 terms involved in biological processes (BP; Figure <ns0:ref type='figure' target='#fig_8'>3a</ns0:ref>), 17 terms associated with cellular components (CC), and 20 terms involved in molecular functions (MF). The most highly enriched BP terms observed in the alternatively spliced genes included 'cellular process', 'single-organism process' and 'metabolic process'. The most common CC categories were 'cell', 'cell part' and 'organelle'. In the MF category, the most enriched terms were 'binding', 'catalytic activity' and 'nucleic acid binding transcription factor activity'. Plants grown in Ma'erkang showed highly similar GO data to that in Xindu: 22 terms involved in BP (Figure <ns0:ref type='figure' target='#fig_8'>3b</ns0:ref>), 15 terms associated with CC, and 15 terms involved in MF. Moreover, data from Ma'erkang and Xindu showed the same top-3 most enriched terms in each category.</ns0:p><ns0:p>These KEGG pathways were classified into five major groups: metabolism, genetic information processing, cellular processes, environmental information processing, and organismal systems. Of these, the subgroups 'biosynthesis of amino acids', 'carbon metabolism', 'ribosome', and 'RNA transport' contained the highest number of annotated genes (Figure <ns0:ref type='figure' target='#fig_7'>4a</ns0:ref>). Data from Ma'erkang showed similar subgroups containing the most of AS genes: 'ribosome', 'carbon metabolism', 'biosynthesis of amino acids' and 'plant hormones signal transduction' contained the highest number of annotated genes (Figure <ns0:ref type='figure' target='#fig_7'>4b</ns0:ref>).</ns0:p></ns0:div>
<ns0:div><ns0:head>DEGs with AS and Their Arabidopsis Orthologs</ns0:head><ns0:p>DESeq software was used to identify the different expression levels of the AS genes in zws-ms and zws-217. From Xindu, eleven differentially expressed AS genes were identified, of which four were upregulated and seven were downregulated in zws-ms (Table <ns0:ref type='table'>2</ns0:ref>).</ns0:p><ns0:p>The Arabidopsis orthologs of these differentially expressed AS genes were identified using</ns0:p><ns0:p>The Arabidopsis Information Resource (TAIR; https://www.arabidopsis.org; Table <ns0:ref type='table'>3</ns0:ref>). The AT1G10760 encodes an α-glucan, water dikinase (GWD) required for starch degradation.</ns0:p><ns0:p>When grown in Ma'erkang, the two lines generated increased number of differentially expressed AS genes significantly to 130 (Table <ns0:ref type='table'>S6</ns0:ref>), including 52 unregulated and 78 down regulated AS genes. Four AS genes were annotated to 'response to cold (GO:0009409)':</ns0:p><ns0:p>BnaAnng17190D, BnaC01g27600D, BnaC08g39130D and BnaC09g53990D; three AS genes, BnaA07g19340D, BnaC01g27600D, BnaC08g39130D, were annotated to 'response to heat (GO:0009408)'; three were related to 'response to freezing (GO:0050826)', including BnaA05g28590D, BnaC06g15710D and BnaC08g36010D; BnaCnng24040D was found relevant to 'temperature stimulus (GO:0009266)'. Moreover, BnaC08g36010D, BnaC08g39130D and BnaC08g39360D were annotated to 'regulation of flower development (GO:0009909)', 'plant ovule development (GO:0048481)' and 'fruit development (GO:0010154)', respectively (Table <ns0:ref type='table'>S6</ns0:ref>). Compared with the 11 differentially expressed AS genes from normal conditions in Xindu, three of them (BnaA07g04500D, BnaAnng30260D and BnaC06g16950D) maintained the same expression models. In other words, these three genes were upregulated under both normal and colder conditions. While the other 8 genes (BnaA02g02630D, BnaA02g03080D, BnaA04g16220D, BnaA09g45000D, BnaA09g45260D, BnaC02g06440D, BnaC07g33980D and BnaC08g49610D) did not show significant difference between zms-ms and zws-217 in expression level in Ma'erkang.</ns0:p></ns0:div>
<ns0:div><ns0:head>Genes with Line-specific AS Events</ns0:head><ns0:p>Genes with line-specific AS events, defined as those genes with a particular AS event(s) that occurred only in zws-ms or in zws-217, were also identified and analyzed. Unlike the abovementioned general classifications, we sorted AS events into 12 finer subclasses, in order to identify them more specifically: transcription start site (TSS) and transcription terminal site AS genes were detected, of which 187 could be annotated (Table <ns0:ref type='table'>S7</ns0:ref>). Eight genes related to 'ovule development', 'flower morphogenesis' and other similar processes were highlighted and Manuscript to be reviewed considered important for further study in the coming future ( <ns0:ref type='formula'>7</ns0:ref>) BnaC03g32190D were all found line-specificaly in zws-217 from Xindu, and none of them were identified in either liens from Ma'erkang. BnaC04g26180D was annotated with 'development (GO:0048481)'; BnaC07g25280D was annotated as 'flower morphogenesis; organ morphogenesis (GO:0009887)' and 'vegetative to reproductive phase transition of meristem (GO:0010228)'; BnaC03g32190D was annotated as 'double fertilization forming a zygote and endosperm (GO:0009567)'; (8) BnaCnng68400D was associated with 'carpel development (GO:0048440).' An IR and two AE events of this gene were detected specifically in multi-silique line under normal conditions, while in colder area, they were not identified in neither zws-ms nor zws-217.</ns0:p></ns0:div>
<ns0:div><ns0:head>Discussion</ns0:head><ns0:p>As an important post-transcriptional metabolic event, AS is involved in many plant growth and developmental processes, such as flowering induction <ns0:ref type='bibr' target='#b8'>(Eckardt, 2002;</ns0:ref><ns0:ref type='bibr' target='#b50'>Slotte et al., 2009)</ns0:ref> and the responses to environmental fluctuations and pathogen attacks <ns0:ref type='bibr' target='#b1'>(Barbazuk et al., 2008)</ns0:ref>. To the best of our knowledge, AS events have seldom been reported to regulate the development of flower/fruit morphology in higher plants. This study is the first to analyze the role of AS events in rapeseed flower/fruit development as a whole, let alone those related to the multi-silique trait.</ns0:p><ns0:p>We previously described the morphology and inheritance of the multi-silique trait in B.</ns0:p><ns0:p>napus <ns0:ref type='bibr' target='#b22'>(Jiang et al., 1998)</ns0:ref>, investigating the associated regions of chromosomes at the genomic level and transcriptomically exploring the DEGs in multi-silique and single-silique plants <ns0:ref type='bibr' target='#b3'>(Chai et al., 2019)</ns0:ref>. The multi-silique trait was found to be controlled by three recessive alleles and was significantly affected by environment; however, the mechanisms by which environmental factors affect this trait remained unknown, even if we knew that temperature could switch on/off the multi-silique trait <ns0:ref type='bibr' target='#b3'>(Chai et al., 2019)</ns0:ref>. As mentioned above, AS is a pathway by which the environment could regulate plant physiology, therefore in this study, we analyzed AS events in order to investigate the mechanism by which plants perceive temperature fluctuations.</ns0:p><ns0:p>In this study, we sampled the buds of three individual plants from zws-ms and zws-217 lines in both Xindu and colder area Ma'erkang, and then subjected them to RNA-seq. All of the four groups generated sufficient data, which was validated by qPCR in earlier and present study successively. The samples in Ma'erkang group generated less data, mainly due to the less sequencing depth; nevertheless, this still provided enough data of high quality and assured the accuracy of the subsequent analysis.</ns0:p><ns0:p>We identified all of the genes with AS events in the zws-ms and zws-217 plants. In Xindu group, 11 AS genes were significantly differentially expressed between the multi-silique zws-ms line and its NIL, zws-217, which produces normal siliques. We analyzed their annotations and orthologs in Arabidopsis. One such ortholog, AT5G15470 (also known as Galacturonosyltransferase 14, GAUT14), is involved in cell wall pectin biosynthesis <ns0:ref type='bibr' target='#b2'>(Caffall et al., 2009)</ns0:ref>, and the gaut13 gaut14 double mutant was previously shown to be defective in pollen tube growth <ns0:ref type='bibr'>(Wang et al., 2013)</ns0:ref>. AT3G15420 (the ortholog of BnaA02g03080D) and AT3G10070 (the ortholog of BnaA09g45000D) encode subunits of the transcription factor complexes TFIIIC and TAF12, respectively. The former does not appear to be substantially involved in plant development; however, some members of the TAF family are involved in the regulation of morphology. The transgenic expression of TAF10 from clustered yellowtops (Flaveria trinervia) in Arabidopsis limited the development of the indeterminate inflorescence and resulted in the production of deformed leaves <ns0:ref type='bibr' target='#b12'>(Furumoto et al., 2005)</ns0:ref>. By contrast, the taf mutant in Arabidopsis has abnormal phyllotaxis and lacks proper vegetative meristem activity <ns0:ref type='bibr' target='#b53'>(Tamada et al., 2007)</ns0:ref>, indicating the important roles played by the TAFs in plant morphological development. Another DEG AS gene, BnaA04g16220D, is not annotated, and its Arabidopsis ortholog AT1G14800 is simply listed as an uncategorized nucleic acid-binding, OB-fold-like protein. The AS gene orthologs AT2G04900 and AT1G15060 encode an unknown protein and an uncategorized alpha/beta hydrolase family protein, respectively, so their roles in the regulation of the multi-silique trait are also currently unclear.</ns0:p><ns0:p>Another ortholog for differentially expressed AS gene, AT3G54620, is reported to encode a bZIP transcription factor-like protein. Members of this protein family are typically reported to regulate plant tolerance of environmental stresses. The transgenic expression of the maize (Zea mays) gene ZmbZIP72 in Arabidopsis enhanced its drought and salt tolerance <ns0:ref type='bibr' target='#b65'>(Ying et al., 2012)</ns0:ref>, while BnbZIP3, a ramie (Boehmeria nivea) bZIP transcription factor, also increased the drought, salinity, and heavy metal tolerances of transgenic Arabidopsis <ns0:ref type='bibr' target='#b19'>(Huang et al., 2016)</ns0:ref>. These genes are also involved in the regulation of other processes; for example, the repression of a bZIP transcription factor gene OsABI5 expression in rice resulted in low fertility <ns0:ref type='bibr' target='#b71'>(Zou et al., 2008)</ns0:ref>, while the transgenic expression of tomato (Solanum lycopersicum) SlbZIP2 in tobacco (Nicotiana benthamiana) increased leaf thickness <ns0:ref type='bibr' target='#b49'>(Seong et al., 2016)</ns0:ref>. To date, however, there are no reports of bZIP genes playing a significant role in flower/fruit morphology.</ns0:p><ns0:p>Other AS gene orthologs included AT5G16210, encoding a member of the HEAT repeatcontaining protein family, which are considered to be involved in intracellular transport <ns0:ref type='bibr' target='#b17'>(Hernández-Torres et al., 2014;</ns0:ref><ns0:ref type='bibr' target='#b37'>Oeffinger et al., 2004)</ns0:ref>. Although BnaC06g16950D is not annotated, its ortholog, AT3G59000, was identified as encoding an F-box/RNI-like superfamily protein in Arabidopsis, which typically function in the plant hormone signaling pathways <ns0:ref type='bibr' target='#b13'>(Gao et al., 2009)</ns0:ref>. Similarly, the ortholog AT4G16900 encodes a TIR-NBS-LRR class protein, which are known to be involved in disease resistance <ns0:ref type='bibr' target='#b63'>(Xun et al., 2019)</ns0:ref> and hormonal responses <ns0:ref type='bibr' target='#b47'>(Sarazin et al., 2015)</ns0:ref>. Moreover, <ns0:ref type='bibr' target='#b18'>Hewezi et al. (2006)</ns0:ref> unexpectedly found that these proteins are associated with developmental abnormalities; transgenic sunflowers (Helianthus annuus) expressing the antisense sequence complementing PLFOR48, which encodes a TIR-NBS-LRR-type protein,</ns0:p><ns0:p>showed stunted growth and a reduction in apical dominance; whereas the pods of transgenic tobacco (N. tabacum) lacking PLFOR48 expression were smaller and showed severe Manuscript to be reviewed deformations. This indicates that TIR-NBS-LRR-type proteins can regulate the morphology of plants, including fruit morphology, to some extent. Finally, AT1G10760, the ortholog of AS gene BnaC08g49610D, which encodes a GWD protein required for starch degradation, is involved in carbohydrate metabolism <ns0:ref type='bibr' target='#b35'>(Nadolska-Orczyk et al., 2017)</ns0:ref>. This gene was also reported to regulate seed size; <ns0:ref type='bibr' target='#b42'>Pirone et al. (2017)</ns0:ref> found that the length and width of the mature seeds were reduced in the gwd1 Arabidopsis mutant, while their density was increased.</ns0:p><ns0:p>To summarize, AT5G15470, AT3G10070, AT3G54620, AT4G16900, and AT1G10760 are all known to be involved in plant development; therefore, their corresponding rapeseed orthologs, BnaA02g02630D, BnaA09g45000D, BnaAnng30260D, BnaC07g33980D, and BnaC08g49610D, the expression levels of which differed significantly between zws-ms and zws-217, are considered to be potential candidate genes regulating the multi-silique trait.</ns0:p><ns0:p>After that, we continued to investigate data from Ma'erkang, where it is colder and the multi-silique trait in zws-ms line disappeared. Due to the importance of the above-mention 11 differentially expressed AS genes, we first paid attention to them and found that three of them had the same expression models as in Xindu. In other words, they were independent of temperature; in addition, combined with their annotations, they were excluded from the potential candidate genes responding to environmental factors. On the other hand, other 8 AS genes stopped being differentially expressed between zms-ms and zws-217 in Ma'erkang. That is to say, the expression level of these 8 AS genes (BnaA02g02630D, BnaA02g03080D, BnaA04g16220D, BnaA09g45000D, BnaA09g45260D, BnaC02g06440D, BnaC07g33980D and BnaC08g49610D) were environment-specifc. Besides, BnaC08g36010D, BnaC08g39130D and BnaC08g39360D, which were annotated to flower/ovule/fruit-related terms, were differentially expressed between zws-ms and zws-217 specifically in Ma'erkang. Moreover, we also found 9 temperature-responding AS genes differentially expressed in colder area: BnaAnng17190D, BnaC01g27600D, BnaC08g39130D, BnaC09g53990D, BnaA07g19340D, BnaA05g28590D, BnaC06g15710D, BnaC08g36010D and BnaCnng24040D. The lower temperature motivated more responding genes and this also explained the reason for increased number of AS genes clustered in each KEGG pathway.</ns0:p><ns0:p>We also explored the line-specific AS genes, which were similarly expressed between zwsms and zws-217, but contained stable and particular AS event(s) that differed between these two lines. These genes are likely to qualitatively regulate the multi-silique trait. In this case, we could Manuscript to be reviewed obtain better results by fine-classify the AS types into 12 subclasses, rather than 5 classes mentioned above. Because fine classifications could better identify differences between AS types more precisely and subtly. Thus, we found 205 genes of this type, of which 187 could be annotated. Due to the rarity of the multi-silique trait, we did not obtain much useful information from the KEGG pathway analysis. This meant that we were unable to relate this metabolic pathway information to the multi-silique trait directly; however, the GO analysis provided more potential clues. Among these, eight genes were considered to be associated with flower/carpel/ovule development. BnaC06g32640D is annotated as being involved in the regulation of the vegetative-to-reproductive phase transition in the meristem (GO:0010228) and in ovule development (GO:0048481). The IR of it in zws-217 seemed to block the multi-silique, while the TSS occurred in zws-ms as if it was positively related to mulsi-silique trait. Its Arabidopsis ortholog, AT1G71692, is annotated as AGAMOUS-LIKE12 (AGL12). <ns0:ref type='bibr' target='#b41'>Peng et al. (2015)</ns0:ref> isolated the BnFUL gene in rapeseed, which is homologous to AGL8 in Arabidopsis.</ns0:p><ns0:p>Although BnFUL was hypothesized to be involved in enhancing pod-shattering resistance, when introduced into Arabidopsis, two of the five transgenic plants expressing BnFUL unexpectedly had a multi-silique phenotype. However, the mechanisms by which BnFUL generates this multisilique phenotype remain elusive thus far, making the AGL12 gene identified in this study a potentially important candidate gene. The ortholog gene of BnaCnng68400D, AT5G15020, encodes an SIN3-LIKE 2 protein (SNL2) known to be important for seed germination or dormancy <ns0:ref type='bibr' target='#b58'>(Wang et al., 2016;</ns0:ref><ns0:ref type='bibr'>Wang et al., 2013)</ns0:ref>. The zws-ms line-specific AS events of it were detected in Xindu, where the two NILs were distinct from each other in the multi-silique trait, while in Ma'erkang, where the distinguishing trait was switched off, these AS event disappeared in zws-ms. This showed a positive correlation to the trait.</ns0:p><ns0:p>The other six selected line-specific AS genes showed particular AS events in zws-217 in Xindu, implying potential inhibition of the multi-silique morphology. However, these AS events did not occur in zws-ms or zws-217 in Ma'erkang, representing unknown complexity of the mechanisms, which were strongly influenced by environment. Fortunately, their orthologs in Arabidopsis provided many useful clues. AT2G20180 and AT4G00050, orthologs of Manuscript to be reviewed flowering phenotype of the pif mutants <ns0:ref type='bibr' target='#b26'>(Leivar & Monte, 2014)</ns0:ref>. AT5G17270 encodes a prenylyltransferase superfamily protein; however, to the best of our knowledge, there have been no reports about its development-related functions. The Arabidopsis ortholog of BnaC05g34570D is AT3G18600, which encodes a P-loop-containing nucleoside triphosphate hydrolase. While few studies have reported the functions of these proteins, <ns0:ref type='bibr' target='#b29'>Liu et al. (2016)</ns0:ref> reported that, in sesame (Sesamum indicum), one gene encoding a P-loop-containing nucleoside triphosphate hydrolase showed a reduced expression level in sterile buds, indicating that they may play a role in specifying/determining tapetal fate and development. Another line-specific AS ortholog, AT3G54660, encodes a glutathione reductase (GR), which was found to increase the fineness (mass per unit length) and bundle strength of cotton (Gossypium hirsutum) fiber when transgenically expressed <ns0:ref type='bibr' target='#b55'>(Tuttle et al., 2015)</ns0:ref>. Since cotton fibers are single cells initiating from the epidermis of the outer integument of the ovules <ns0:ref type='bibr' target='#b45'>(Ruan et al., 2004)</ns0:ref>, it can be inferred that GR regulates ovule development to some extent. AT3G28730, also known as structurespecific recognition protein SSRP1, was also found to regulate floral development, as the ssrp1-2 mutant Arabidopsis produced small and deformed petals with shorter stamens <ns0:ref type='bibr' target='#b31'>(Lolas et al., 2010)</ns0:ref>.</ns0:p><ns0:p>To date, there is some evidence to show that the line-specific AS orthologs AT2G20180, AT4G00050 and AT3G54660 are related to the regulation of flower/fruit morphology, with clear roles reported for AT1G71692 and AT3G28730. Consequently, their orthologs in rapeseed, BnaC06g32640D and BnaC07g25280D, respectively, are considered to be important candidate genes regulating the multi-silique trait by conferring or removing some specific line-specific AS events in varied environments. In addition, BnaCnng68400D, of which AS events represented a positive correlation to morphology with or without multi-siliques, was also noteworthy.</ns0:p><ns0:p>Some of the genes/loci controlling silique development in Brassica plants have previously been reported. In addition to those regulating traits such as the seed weight and silique length <ns0:ref type='bibr' target='#b30'>(Liu et al., 2015)</ns0:ref> and the number of seeds per silique in B. napus <ns0:ref type='bibr'>(Li et al., 2015)</ns0:ref>, some genes related to silique morphology have been cloned and functionally analyzed. <ns0:ref type='bibr' target='#b62'>Xiao et al. (2013)</ns0:ref> fine-mapped a multi-locular silique gene, Bjln1, to a 208-kb region on chromosome A7 in Brassica juncea and then revealed that it was the mutations in the CDS and promoter of BjuA07.CLV1 gene (equivalent to Bjln1) to cause the multi-locular trait <ns0:ref type='bibr' target='#b61'>(Xiao et al., 2018)</ns0:ref> Manuscript to be reviewed CLAVATA3 in Arabidopsis, caused the production of multi-locular siliques in B. rapa. However, the multi-silique (or multi-pistil) phenotype of zws-ms is different from the above-motioned multi-locular trait; zws-ms produces three pods on each carpopodium, rather than multiple loculi per pod.</ns0:p><ns0:p>Few studies have investigated this multi-silique trait in rapeseed; however, there have been similar reports of multi-pistil traits in other crops, particularly in wheat (Triticum aestivum): <ns0:ref type='bibr' target='#b7'>Duan et al. (2015)</ns0:ref> discovered a male-sterile wheat mutant, dms, with a dwarf status and multipistils, a pleiotropic phenotype found to be controlled by a single recessive gene, which was not identified. <ns0:ref type='bibr'>Guo et al. (2019)</ns0:ref> reported another multi-ovary trait in the wheat line DUOII, which was controlled by a dominant gene, and used a proteomics approach to propose some candidate proteins. <ns0:ref type='bibr' target='#b64'>Yang et al. (2017)</ns0:ref> mapped a gene promoting the formation of three pistils (Pis1) to chromosome 2D and identified some candidate genes according to their annotations, while <ns0:ref type='bibr' target='#b70'>Zhu et al. (2019)</ns0:ref> discovered a wheat multi-pistil mutant, 12TP, which was found to contain a semidominant mutation located on chromosome arm 2DL. Although several studies have explored the multi-pistil trait in wheat, no one has identified any of the specific genes responsible yet.</ns0:p><ns0:p>To sum up, the seven candidate genes mentioned above, including the five differentially expressed AS genes of interest (BnaA02g02630D, BnaA09g45000D, BnaAnng30260D, BnaC07g33980D, and BnaC08g49610D) and the two genes with line-specific AS events (BnaC06g32640D and BnaC07g25280D), are therefore hypothesized to regulate the multisilique trait in rapeseed zws-ms, based on their AS expression levels or line-specific AS events altered by environment. These findings lay a foundation for further functional analyses in future.</ns0:p></ns0:div>
<ns0:div><ns0:head>Conclusions</ns0:head><ns0:p>The utilization of heterosis is a way to increase the yield or improve the quality of crops.</ns0:p><ns0:p>Exploring new germplasm resources and genes, as well as clarifying their inheritance, is the foundation of obtaining of excellent hybrid. This study provides a novel inspection into the multi-silique trait in rapeseed from the transcriptional perspective by AS responding to environment, deepening the understanding of its molecular mechanism. Further function verifications are now undergoing.</ns0:p><ns0:p>PeerJ reviewing PDF | (2019:12:44287:4:0:NEW 6 Sep 2020)</ns0:p><ns0:p>Manuscript to be reviewed</ns0:p><ns0:note type='other'>Figure 1</ns0:note><ns0:p>Multi-silique trait in zws-ms, compared with the single siliques of its near-isogenic line zws-217. Manuscript to be reviewed Numbers of alternative splicing events in six samples.</ns0:p><ns0:p>Note: T01, T02, and T03: Buds of three independent zws-ms plants at the budding stage; T04, T05, and T06: Buds of three independent zws-217 plants at the budding stage. Manuscript to be reviewed</ns0:p></ns0:div><ns0:figure xml:id='fig_0'><ns0:head /><ns0:label /><ns0:figDesc>following orthologs were identified: (1) AT5G15470, the ortholog of BnaA02g02630D, encodes galacturonosyltransferase 14 (GAUT14); (2) AT3G15420 encodes the transcription factor TFIIIC (tau55-related protein); (3) AT1G14800 encodes a nucleic acid-binding, OB-fold-like protein; (4) AT2G04900 encodes an unknown protein; (5) AT3G10070 encodes one of two Arabidopsis proteins with similarity to the TBP-associated factor, TAF12; (6) AT1G15060 encodes an alpha/beta hydrolase family protein; (7) AT3G54620 encodes a bZIP transcription factor-like protein; (8) AT5G16210 encodes a HEAT repeat-containing protein; (9) AT3G59000 encodes an F-box/RNI-like superfamily protein; (10) AT4G16900, the ortholog of BnaC07g33980D, encodes a member of the disease resistance protein (TIR-NBS-LRR class) family; and (11)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_1'><ns0:head>(</ns0:head><ns0:label /><ns0:figDesc>TTS) equaled to original Alternative 5' first exon and Alternative 3' last exon, respectively; Mutually exclusive exons was subdivided into Alternative exon ends (AE) and Approximate AE (XAE); the original Intron retention was subdivided into single Intron retention (IR), Approximate IR (XIR), Multi-IR (MIR) and Approximate MIR (XMIR); the Cassette exon was then was subdivided into single Skipped exon (SKIP), Approximate SKIP (XSKIP), Multi-exon SKIP (MSKIP) and Approximate MSKIP (XMSKIP). In total, 205 line-specifically expressed</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_2'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2019:12:44287:4:0:NEW 6 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_3'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2019:12:44287:4:0:NEW 6 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_4'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2019:12:44287:4:0:NEW 6 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_5'><ns0:head>BnaC07g00780D</ns0:head><ns0:label /><ns0:figDesc>and BnaC03g32190D respectively, both encode phytochrome interacting factors (PIFs). Several transcription factors (AP1, SVP, LFY, AG, and SEP3) involved in the regulation of flowering are known to bind to the PIFs, suggesting a direct link with the reported PeerJ reviewing PDF | (2019:12:44287:4:0:NEW 6 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_6'><ns0:head>(</ns0:head><ns0:label /><ns0:figDesc>A) Primary inflorescences. (B) Siliques.</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_7'><ns0:head>Figure 4 Classified</ns0:head><ns0:label>4</ns0:label><ns0:figDesc>Figure 4</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_8'><ns0:head>Alternative 3 '</ns0:head><ns0:label>3</ns0:label><ns0:figDesc>splice site: different-size mRNAs are produced depending on the usage of a proximal or distal 3' splice site; Alternative 5' splice site: different-size mRNAs are produced depending on the use of a proximal or distal 5' splice site; Exon Skipping: an exon is either included or excluded from the mRNA; Intron Retention: an intron is either retained or excised in the mRNA, resulting in different-size transcripts; Mutually Exclusive Exons: adjacent exons are spliced in such a way that only one of them is included at a time in the mRNA.PeerJ reviewing PDF | (2019:12:44287:4:0:NEW 6 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='25,42.52,70.87,333.62,672.95' type='bitmap' /></ns0:figure>
<ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='26,42.52,229.87,525.00,230.25' type='bitmap' /></ns0:figure>
<ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='27,42.52,280.87,525.00,214.50' type='bitmap' /></ns0:figure>
<ns0:figure type='table' xml:id='tab_0'><ns0:head>Table 4 ,</ns0:head><ns0:label>4</ns0:label><ns0:figDesc>TableS7): (1) An IR event (32749874-32749905 bp on chromosome C06) of BnaC06g32640D occurred line-specifically in zws-217 in Xindu, while this event disappeared in both lines in Ma'erkang. Furthermore, a TSS</ns0:figDesc><ns0:table><ns0:row><ns0:cell>event (32750089-32750270 bp on chromosome C06) only happening in multi-silique zws-ms in</ns0:cell></ns0:row><ns0:row><ns0:cell>Xindu, appeared in both lines when planted in Ma'erkang. This gene was annotated as</ns0:cell></ns0:row><ns0:row><ns0:cell>'vegetative to reproductive phase transition of meristem (GO:0010228)' and 'Biological</ns0:cell></ns0:row><ns0:row><ns0:cell>Process: ovule development (GO:0048481)'; (2) BnaC07g00780D was associated with</ns0:cell></ns0:row><ns0:row><ns0:cell>'reproductive structure development (GO:0048608)'. In Xindu, two specific AE events</ns0:cell></ns0:row><ns0:row><ns0:cell>(1070954-1071329 bp and 1070976-1071329 bp on chromosome C07, respectively) of it were</ns0:cell></ns0:row><ns0:row><ns0:cell>observed in zws-217, but they both disappeared in two lines when planted in Ma'erkang; (3)</ns0:cell></ns0:row><ns0:row><ns0:cell>BnaC04g31460D and (4) BnaC05g34570D were related to 'regulation of flower development</ns0:cell></ns0:row><ns0:row><ns0:cell>(GO:0009909)'. An SKIP event (33329566-33329615 bp on chromosome C04) for</ns0:cell></ns0:row><ns0:row><ns0:cell>BnaC04g31460D and an IR event (33879446-33879532 on chromosome C05) for</ns0:cell></ns0:row><ns0:row><ns0:cell>BnaC05g34570D were identified only in zws-217 from Xindu; Similarly, two AE events</ns0:cell></ns0:row><ns0:row><ns0:cell>(27558813-27558932 bp and 27558838-27558932 bp on chromosome C04) for (5)</ns0:cell></ns0:row><ns0:row><ns0:cell>BnaC04g26180D, two IR events (31448257-31448346 bp and 31448264-31448353 bp on</ns0:cell></ns0:row><ns0:row><ns0:cell>chromosome C07) for (6) BnaC07g25280D and one IR event (19819830-19820358 bp on</ns0:cell></ns0:row><ns0:row><ns0:cell>chromosome C07) for (</ns0:cell></ns0:row></ns0:table></ns0:figure>
<ns0:figure type='table' xml:id='tab_2'><ns0:head>Table 1 (on next page)</ns0:head><ns0:label>1</ns0:label><ns0:figDesc /><ns0:table /></ns0:figure>
<ns0:note place='foot'>PeerJ reviewing PDF | (2019:12:44287:4:0:NEW 6 Sep 2020)</ns0:note>
</ns0:body>
" | "Response to Editor
Section Editor comments:
The manuscript appears somewhat incomplete as I can see that there are apparent post-transcriptional processing of candidate genes in two distinct environments, but there does not appear to be any contextual sequence examples demonstrating the alternate splicing observed. I would have expected some examples of splice variants of the candidates represented in their contextual setting. As there was no follow-up in an experimental role a preview of the candidate model would go far to bring sense to the findings. Pointing to the candidates is good in one sense, and hypothesizing a role is another, but the evidence provided here still appears too conceptualized. With the culmination of numerous candidates, is there speculation of the more probable candidates and a model representation that can be followed up upon? Pointing to the reference gene ID without representation of the alternate splicing observations is not a helpful lead to the reader. I would propose some additional splicing observations in the context of the environments be supplied. I can see where the manuscript is going, but there is a need for additional information.
Response:
Dear Editor, thanks for your professional advice. According to it, we revised manuscript again: We added more information about these genes, indicating the AS examples about them (in Results and Discussion sections). Moreover, based on the information, we clarified the candidate genes. We also corrected some typos or grammar mistakes. Hope this new-updated manuscript can meet your requirement.
If further revisions are needed, please kindly let us know.
" | Here is a paper. Please give your review comments after reading it. |
9,779 | "<ns0:abstract xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:p>Under physiological conditions, retinal pigment epithelium (RPE) is a cellular monolayer composed of mitotically quiescent cells. Tight junctions and adherens junctions maintain the polarity of RPE cells, and are required for cellular functions. In proliferative vitreoretinopathy (PVR), upon retinal tear, RPE cells lose cell-cell contact, undergo epithelial-mesenchymal transition (EMT), and ultimately transform into myofibroblasts, leading to the formation of fibrocellular membranes on both surfaces of the detached retina and on the posterior hyaloids, which causes tractional retinal detachment. In PVR, RPE cells are crucial contributors, and multiple signaling pathways, including SMADdependent pathway, Rho pathway, MAPK pathways, Jagged/Notch pathway, and Wnt/βcatenin pathway, are activated. These pathways mediate the EMT of RPE cells, which play a key role in the pathogenesis of PVR. This review summarizes the current body of knowledge on the polarized phenotype of RPE, the role of cell-cell contact, and the molecular mechanisms underlying the RPE EMT in PVR, emphasizing key insights into potential approaches to prevent PVR.</ns0:p></ns0:div>
</ns0:abstract>
<ns0:body xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:head>Introduction</ns0:head><ns0:p>Proliferative vitreoretinopathy (PVR) is a complex blinding disease that occurs after rhegmatogenous retinal detachment (RRD), surgical interventions, or ocular trauma. As a prolonged and exaggerated scarring process, PVR is characterized by the formation of contractile fibrocellular membranes in the vitreous cavity and on the inner and outer surfaces of the retina <ns0:ref type='bibr' target='#b30'>(Committee 1983;</ns0:ref><ns0:ref type='bibr' target='#b105'>Mudhar 2020;</ns0:ref><ns0:ref type='bibr' target='#b141'>Tosi et al. 2014)</ns0:ref>. At present, surgical interventions, including vitrectomy, membrane peeling, pneumatic retinopexy, and scleral buckle, remain the mainstay of treatment in PVR. Although work in recent decades has led to advancements in surgical techniques and management, PVR cannot be effectively treated and is still the most common cause of failure to reattach the retina <ns0:ref type='bibr' target='#b29'>(Coffee et al. 2014;</ns0:ref><ns0:ref type='bibr' target='#b73'>Khan et al. 2015;</ns0:ref><ns0:ref type='bibr' target='#b103'>Mitry et al. 2012;</ns0:ref><ns0:ref type='bibr' target='#b152'>Wickham et al. 2011</ns0:ref>). In addition, in spite of successful anatomic reattachment, the visual function of such cases cannot be improved, due to the retinal damage resulting from the mechanical contraction of fibrous membranes. Therefore, in order to improve postoperative visual function and reduce the incidence of this serious complication, it is particularly important to explore new prophylactic and therapeutic approaches based on a deeper understanding of the pathogenesis of PVR.</ns0:p><ns0:p>A growing body of evidence indicates that the mechanisms of PVR are orchestrated by multiple elements <ns0:ref type='bibr' target='#b63'>(Idrees et al. 2019;</ns0:ref><ns0:ref type='bibr' target='#b69'>Jin et al. 2017;</ns0:ref><ns0:ref type='bibr' target='#b115'>Pastor et al. 2016)</ns0:ref>, such as growth factors <ns0:ref type='bibr' target='#b20'>(Charteris 1998;</ns0:ref><ns0:ref type='bibr' target='#b109'>Ni et al. 2020;</ns0:ref><ns0:ref type='bibr' target='#b117'>Pennock et al. 2014;</ns0:ref><ns0:ref type='bibr' target='#b153'>Wubben et al. 2016)</ns0:ref>, cytokines <ns0:ref type='bibr' target='#b7'>(Bastiaans et al. 2018;</ns0:ref><ns0:ref type='bibr' target='#b54'>Harada et al. 2006;</ns0:ref><ns0:ref type='bibr' target='#b88'>Limb et al. 1991)</ns0:ref>, extracellular matrix proteins <ns0:ref type='bibr' target='#b39'>(Feist et al. 2014;</ns0:ref><ns0:ref type='bibr' target='#b100'>Miller et al. 2017</ns0:ref>) and various cells <ns0:ref type='bibr' target='#b38'>(Eastlake et al. 2016;</ns0:ref><ns0:ref type='bibr' target='#b118'>Pennock et al. 2011;</ns0:ref><ns0:ref type='bibr' target='#b125'>Shu & Lovicu 2017)</ns0:ref>. According to the histopathology of PVR, the fibrocellular membrane of PVR is composed of excessive extracellular matrix (ECM) and multiple types of cells, and retinal pigment epithelial (RPE) cells have been indicated as the most consistently present and the most abundant <ns0:ref type='bibr' target='#b2'>(Amarnani et al. 2017;</ns0:ref><ns0:ref type='bibr' target='#b35'>Ding et al. 2017;</ns0:ref><ns0:ref type='bibr' target='#b61'>Hiscott et al. 1989;</ns0:ref><ns0:ref type='bibr' target='#b95'>Machemer & Laqua 1975)</ns0:ref>, proving that the RPE cell plays a crucial role in PVR. Under physiological condition, the polarized RPE cell is non-proliferative by cell-cell contact. However, when the eye suffers from a retinal break or trauma, RPE cells are exposed to various growth factors and cytokines that are produced by activated immune cells, leading to the disruption of junctional complexes in RPE cells. Subsequently, activated RPE cells detach from Bruch's membrane, migrate through the defect of the retina, proliferate, and transform into myofibroblasts, forming fibrotic membranes <ns0:ref type='bibr'>(Chen et al. 2015;</ns0:ref><ns0:ref type='bibr' target='#b104'>Morescalchi et al. 2013;</ns0:ref><ns0:ref type='bibr' target='#b113'>Palma-Nicolás & López-Colomé 2013)</ns0:ref>. In an analogous process to exaggerated wound healing response, these membranes can attach to the retina and contract, resulting in further retinal detachment and poor vision <ns0:ref type='bibr' target='#b28'>(Chiba 2014;</ns0:ref><ns0:ref type='bibr' target='#b46'>Garweg et al. 2013)</ns0:ref>. It is noteworthy that due to the loss of cell-cell contact, RPE cells undergo epithelialmesenchymal transition (EMT), which is pivotal in the development of PVR. During EMT, RPE cells transdifferentiate into mesenchymal cells that are characterized by increased motility, and enhanced ability to proliferate, resist apoptosis and produce extracellular matrix proteins, thus participating in PVR <ns0:ref type='bibr' target='#b136'>(Tamiya & Kaplan 2016;</ns0:ref><ns0:ref type='bibr' target='#b166'>Zhang et al. 2018c)</ns0:ref>. These indicate that in-depth knowledge of EMT may provide insight into potential approaches to prevent PVR. Therefore, this review focuses on the polarized phenotype of RPE and molecular mechanisms of RPE cell EMT, discussing the role of RPE cells in PVR.</ns0:p></ns0:div>
<ns0:div><ns0:head>Survey methodology</ns0:head><ns0:p>We used the PubMed database to search available literature based on keywords including 'proliferative vitreoretinopathy(PVR)' and 'retinal pigment epithelial cell'. To include more information on the polarity of RPE, we also searched articles about the structure and function of cell-cell junctions in RPE cells that explored the role of cell-cell contact in EMT.</ns0:p></ns0:div>
<ns0:div><ns0:head n='1.'>The Polarized Retinal Pigment Epithelial Cell</ns0:head><ns0:p>The human RPE cell achieves terminal differentiation at four to six weeks of gestation and subsequently remains mitotically quiescent <ns0:ref type='bibr' target='#b93'>(Lutty & McLeod 2018;</ns0:ref><ns0:ref type='bibr' target='#b130'>Stern & Temple 2015)</ns0:ref>. The RPE, which is situated between the photoreceptors and the choroid, plays many complex roles indispensable to the health of the neural retina and the choroid. These roles include recycling of components of the visual cycle, absorption of light to protect from photo-oxidative stress, production of essential growth factors, immunological regulation of the eye, phagocytosis of photoreceptor outer segments generated during daily photoreceptor renewal, and transportation across the blood retina barrier (BRB) <ns0:ref type='bibr' target='#b41'>(Ferrington et al. 2016;</ns0:ref><ns0:ref type='bibr' target='#b42'>Fields et al. 2019;</ns0:ref><ns0:ref type='bibr' target='#b97'>Mateos et al. 2014;</ns0:ref><ns0:ref type='bibr' target='#b107'>Naylor et al. 2019;</ns0:ref><ns0:ref type='bibr' target='#b132'>Strauss 2005;</ns0:ref><ns0:ref type='bibr' target='#b145'>Vigneswara et al. 2015)</ns0:ref>. In order to maintain these multiple functions, RPE cells display a highly specialized structural and functional polarity.</ns0:p><ns0:p>Similar to other epithelia, the RPE displays three characteristics of the epithelial phenotype: apical plasma membrane, junctional complexes, and basolateral domain. RPE cells display structural polarity, with apical microvilli and melanosomes, and basal microinfolds. The abundant melanin granules in RPE cells absorb stray light, a process that is essential for visual function <ns0:ref type='bibr' target='#b132'>(Strauss 2005)</ns0:ref>. In a polarized cell, the distributions of surface proteins on the apical and basal plasma membranes are different, contributing to the performance of cellular functions <ns0:ref type='bibr' target='#b75'>(Khristov et al. 2018</ns0:ref>). However, a highly polarized distribution of ion channels, transporters and receptors in RPE is different from that observed in conventional extraocular epithelia <ns0:ref type='bibr' target='#b82'>(Lehmann et al. 2014)</ns0:ref>. For example, Na, K-ATPase <ns0:ref type='bibr' target='#b129'>(Sonoda et al. 2009</ns0:ref>) and monocarboxylate transporters (MCT) 1 <ns0:ref type='bibr' target='#b34'>(Deora et al. 2005</ns0:ref>) are localized to the apical aspect of RPE cells, while chloride transporter CFTR <ns0:ref type='bibr' target='#b96'>(Maminishkis et al. 2006</ns0:ref>) is basally located. On the apical plasma membrane, RPE cells phagocytize the photoreceptor outer segments, which are regulated by polarized receptors. <ns0:ref type='bibr' target='#b14'>Bulloj et al. (2018)</ns0:ref> found that binding of Semaphorin 4D (sema4D) to RPE apical receptor Plexin-B1 suppresses outer segment internalization, contributing to the maintenance of photoreceptor function and longevity. The RPE also transports fluid out of the subretinal space, and regulates bidirectional nutrient transport between the outer retina and the choroid, in a manner dependent on the polarized distribution of membrane channels and transporters <ns0:ref type='bibr' target='#b132'>(Strauss 2005)</ns0:ref>. The RPE basolaterally secretes extracellular matrix components and factors, which participate in ECM remodeling and maintain the outer BRB (oBRB) function <ns0:ref type='bibr' target='#b17'>(Caceres & Rodriguez-Boulan 2020)</ns0:ref>. Therefore, the polarized phenotype of the RPE is vital to both the oBRB and is the basis of the homeostasis of the outer retina <ns0:ref type='bibr' target='#b17'>(Caceres & Rodriguez-Boulan 2020;</ns0:ref><ns0:ref type='bibr' target='#b82'>Lehmann et al. 2014)</ns0:ref>. The disruption of RPE polarity contributes to the development of several retinal diseases, such as PVR and age-related macular degeneration (AMD). A comprehensive understanding of the way in which this polarity is achieved may provide insights into the pathogenesis of PVR.</ns0:p><ns0:p>However, most available data on RPE polarity is contributed by studies performed on RPEimmortalized cell lines that show partial preservation of the RPE phenotype, and were extrapolated from data obtained from the prototype Madin-Darby Canine Kidney (MDCK) cell line <ns0:ref type='bibr' target='#b82'>(Lehmann et al. 2014)</ns0:ref>. The detailed mechanisms that determine RPE polarization remain unclear. Some scholars believe that junctional complexes, including adherens junctions (AJs) and tight junctions (TJs), are essential for building epithelial cell polarity and maintaining the integrity of epithelial layers such as RPE <ns0:ref type='bibr' target='#b110'>(Niessen 2007;</ns0:ref><ns0:ref type='bibr' target='#b116'>Pei et al. 2019;</ns0:ref><ns0:ref type='bibr' target='#b136'>Tamiya & Kaplan 2016)</ns0:ref>.</ns0:p><ns0:p>Tight junctions are complex cell-cell junctions formed by transmembrane proteins interactions with peripheral cytoplasmic proteins (Fig <ns0:ref type='figure'>1</ns0:ref>). Transmembrane proteins include occludin, members of the claudin family, and junctional adhesion molecules (JAMs). Peripheral cytoplasmic proteins, such as zonula occludens (ZOs), form bridges between transmembrane proteins and the actin filament cytoskeleton and play a key role in the assembly and organization of TJs <ns0:ref type='bibr' target='#b8'>(Bazzoni & Dejana 2004;</ns0:ref><ns0:ref type='bibr' target='#b9'>Bazzoni et al. 2000;</ns0:ref><ns0:ref type='bibr' target='#b107'>Naylor et al. 2019)</ns0:ref>.</ns0:p><ns0:p>The RPE tight junctions regulate the paracellular movement of solutes via size and charge selectivity <ns0:ref type='bibr' target='#b10'>(Benedicto et al. 2017;</ns0:ref><ns0:ref type='bibr' target='#b16'>Caceres et al. 2017;</ns0:ref><ns0:ref type='bibr' target='#b107'>Naylor et al. 2019)</ns0:ref>.Occludin and claudins determine the permeability and semi-selectivity of the TJs, and as such play critical roles in the oBRB <ns0:ref type='bibr' target='#b3'>(Balda et al. 2000;</ns0:ref><ns0:ref type='bibr' target='#b42'>Fields et al. 2019;</ns0:ref><ns0:ref type='bibr' target='#b44'>Furuse et al. 1998;</ns0:ref><ns0:ref type='bibr' target='#b53'>Günzel & Yu 2013;</ns0:ref><ns0:ref type='bibr' target='#b119'>Rosenthal et al. 2017)</ns0:ref>. JAMs regulate TJ assembly and function by recruiting other proteins to the TJ and play an important role in the barrier property of TJs <ns0:ref type='bibr' target='#b5'>(Balda & Matter 2016;</ns0:ref><ns0:ref type='bibr' target='#b111'>Orlova et al. 2006;</ns0:ref><ns0:ref type='bibr' target='#b123'>Shin et al. 2006</ns0:ref>). In patients with RRD, damage to TJs elicits the breakdown of oBRB and promotes the penetration of growth factors and cytokines, aggravating PVR. As well as having a barrier function, TJs define the physical separation between apical and basal domains of the plasma membrane, to maintain RPE cell polarity <ns0:ref type='bibr' target='#b19'>(Campbell et al. 2017;</ns0:ref><ns0:ref type='bibr' target='#b48'>González-Mariscal et al. 2014;</ns0:ref><ns0:ref type='bibr' target='#b128'>Sluysmans et al. 2017)</ns0:ref>. The two extracellular loops of occludin mediate adhesion of adjacent cells and block the movement of plasma components. The C-terminal domain combines directly with ZOs, subsequently interacting with the actin cytoskeleton, which is essential to organizing and maintaining cell polarization <ns0:ref type='bibr' target='#b5'>(Balda & Matter 2016;</ns0:ref><ns0:ref type='bibr' target='#b43'>Furuse et al. 1994;</ns0:ref><ns0:ref type='bibr' target='#b123'>Shin et al. 2006;</ns0:ref><ns0:ref type='bibr' target='#b139'>Tarau et al. 2019)</ns0:ref>. <ns0:ref type='bibr' target='#b40'>Feng et al. (2019)</ns0:ref> demonstrated that during EMT, the breakdown of TJs resulting from loss of claudin-1 causes ARPE-19 cells to lose their epithelial phenotype and transform into fibroblasts, promoting the development of PVR. TJs are involved in the regulation of signaling pathways that govern various cellular functions such as proliferation, migration, and differentiation <ns0:ref type='bibr' target='#b11'>(Bhat et al. 2018;</ns0:ref><ns0:ref type='bibr' target='#b122'>Shi et al. 2018;</ns0:ref><ns0:ref type='bibr' target='#b128'>Sluysmans et al. 2017)</ns0:ref>. <ns0:ref type='bibr' target='#b144'>Vietor et al. (2001)</ns0:ref> found that decreased amounts of occludin can cause up-regulation and translocation of the adhesion junction protein β-catenin, which interacts with the transcription factor lymphoid enhancer-binding factor (LEF)/T cell factor (TCF) in the nucleus, leading to a loss of the polarized epithelial phenotype in EpH4 cells. ZOs, adaptor proteins within the TJ complex, exhibit dual localization at TJs and in the nucleus. Under injury or stress, the disruption of TJs increases ZO-2 nuclear accumulation, driving its interaction with transcription factors, and inducing MDCK epithelial cell proliferation <ns0:ref type='bibr' target='#b65'>(Islas et al. 2002;</ns0:ref><ns0:ref type='bibr' target='#b122'>Shi et al. 2018;</ns0:ref><ns0:ref type='bibr' target='#b142'>Traweger et al. 2003)</ns0:ref>. In differentiated RPE cells, the interaction between ZO-1 with ZO-1-associated nucleic acid-binding protein (ZONAB) maintains cell-cell contact by sequestering ZONAB at the TJ or in the cytoplasm, maintaining cells dormancy. However, when damage to TJs decreases ZO-1 levels, ZONAB is translocated into the nucleus, leading to the up-regulation of cyclin D1 (CD1) and subsequent cell proliferation <ns0:ref type='bibr' target='#b4'>(Balda et al. 2003;</ns0:ref><ns0:ref type='bibr' target='#b48'>González-Mariscal et al. 2014)</ns0:ref>. Therefore, TJs provide a structural foundation for the maintenance of cell-cell contact. <ns0:ref type='bibr' target='#b47'>Georgiadis et al. (2010)</ns0:ref> demonstrated that the overexpression of ZONAB or knockdown of ZO-1 could result in increased RPE proliferation and the development of EMT. Recent research has confirmed that during EMT, ZO-1 is decreased in ARPE-19 cells, and the knockdown of either ZO-1 or AJ protein E-cadherin leads to the downregulation of the other protein, indicating the existence of an interaction between the two junctional complexes <ns0:ref type='bibr' target='#b6'>(Bao et al. 2019)</ns0:ref>. Due to the importance of TJs in the maintenance of integrity and functionality of epithelial cells, several researchers have focused on novel factors that stimulate the formation of TJs, such as nicotinamide <ns0:ref type='bibr' target='#b56'>(Hazim et al. 2019</ns0:ref>) and lysophosphatidic acid <ns0:ref type='bibr' target='#b86'>(Lidgerwood et al. 2018)</ns0:ref>. Studies into these factors may produce well-differentiated RPE cell lines and a platform to enable the rapid expansion of our understanding of many RPE functions and retinal pathologies. This approach could be conducive to finding novel therapeutic interventions for PVR.</ns0:p><ns0:p>Besides the TJ complex described above, another type of junctional complex called AJs plays a key role in the maintenance of the integrity of epithelial cells and cell-cell contact (Fig <ns0:ref type='figure'>1</ns0:ref>). Cadherins, the major proteins of AJs, belong to the glycoprotein superfamily, of which there are more than 20 members. The cytoplasmic domain of cadherins regulates interactions between cadherins and catenins, including β-catenin, α-catenin, and p120-catenin, and other scaffolding proteins such as ZO-1, to maintain cell shape and modulate cell proliferation <ns0:ref type='bibr' target='#b0'>(Aberle et al. 1994;</ns0:ref><ns0:ref type='bibr' target='#b108'>Nelson & Nusse 2004;</ns0:ref><ns0:ref type='bibr' target='#b151'>Wheelock & Johnson 2003)</ns0:ref>. In quiescent adult RPE cells, epithelial cadherins (E-and/or P-cadherin) sequester β-catenin at the AJs to maintain cell-cell contact. Reduction of cadherin levels or dissociation of AJs allows β-catenin to translocate into the nucleus, where it interacts with the transcription factor LEF, and activates the transcription of various genes, including Snail and cyclin D1, which participate in RPE cell EMT via the canonical Wnt/β-catenin signaling pathway <ns0:ref type='bibr' target='#b50'>(Gonzalez & Medici 2014;</ns0:ref><ns0:ref type='bibr' target='#b77'>Lamouille et al. 2014;</ns0:ref><ns0:ref type='bibr' target='#b108'>Nelson & Nusse 2004;</ns0:ref><ns0:ref type='bibr' target='#b157'>Yang et al. 2018)</ns0:ref> . <ns0:ref type='bibr' target='#b137'>Tamiya et al. (2010)</ns0:ref> suggested that the loss of Pcadherin causes the loss of cell-cell contact and initiates RPE cell migration and EMT. These events coincide with a switch in cadherin isoform expression from P-to N-cadherin. In addition, hepatocyte growth factor (HGF) and its receptor c-Met can destabilize cell-cell adhesion and elicit nuclear translocation of β-catenin, resulting in RPE cell migration <ns0:ref type='bibr' target='#b87'>(Lilien & Balsamo 2005;</ns0:ref><ns0:ref type='bibr' target='#b90'>Liou et al. 2002)</ns0:ref>. Jin et al found that HGF induces loss or redistribution of junctional proteins ZO-1, occludin, and β-catenin in RPE explants, potentially damaging barrier function and increasing the migration of RPE cells, resulting in retinal detachment(RD) and PVR <ns0:ref type='bibr' target='#b67'>(Jin et al. 2002;</ns0:ref><ns0:ref type='bibr' target='#b68'>Jin et al. 2004</ns0:ref>). Given the importance of HGF in the interruption of RPE junction, HGF may be a potential target for the prevention and treatment of PVR. However, this possibility needs further study.</ns0:p><ns0:p>Under physiological conditions in the eye, TJs and AJs maintain the specialized structural and functional polarity of RPE cells and play a pivotal role in the maintenance of cell-cell contact; they sequester EMT signaling effectors ZONAB and β-catenin at the junction or cytoplasm to prevent cells from responding to mitotic factors, causing cells to leave the cellcycle (Fig <ns0:ref type='figure'>1</ns0:ref>). Thus, normally, RPE cells form a cobblestone-like monolayer of immotile, polarized, and mitotically quiescent cells. However, once junctional complexes break down, RPE cells undergo EMT, which is an important contributor to proliferative vitreoretinopathy. In this pathological process, RPE cells lose their structural and functional polarity and transdifferentiate into mesenchymal cells, which proliferate, resist apoptosis, possess migratory ability, and produce abundant ECM, leading to the formation of an aberrant scar-like fibrocellular membrane.</ns0:p></ns0:div>
<ns0:div><ns0:head n='2.'>De-differentiated RPE and Fibrocellular Membrane</ns0:head><ns0:p>Proliferative vitreoretinopathy is characterized by the formation of fibrocellular membranes composed of proliferative and migratory cells and excessive, aberrant ECM. Histopathological analysis of PVR has demonstrated that PVR membranes have contractile activity and strain the retina, leading to tractional retinal detachment (TRD), which is responsible for blurring vision.</ns0:p><ns0:p>Several studies <ns0:ref type='bibr' target='#b39'>(Feist et al. 2014;</ns0:ref><ns0:ref type='bibr' target='#b135'>Takahashi et al. 2010</ns0:ref>) have found that the cellular components of PVR membranes include RPE cells, myofibroblasts, fibroblasts, glial cells and macrophages, and that myofibroblasts are critical for the formation and contractile activity of fibrocellular membranes. Based on the indirect immunofluorescence evaluation of human PVR membranes, <ns0:ref type='bibr' target='#b39'>Feist et al. (2014)</ns0:ref> showed that myofibroblasts originate principally from RPE cells through EMT. Myofibroblasts are characterized by increased expression of alpha-smooth muscle actin (α-SMA) and incorporation of α-SMA into newly formed actin stress fibers, which enhances their contractile properties. Myofibroblasts also secrete excessive matrix and profibrogenic factors, promoting the contraction of PVR membranes that ultimately cause irreversible loss of vision <ns0:ref type='bibr' target='#b45'>(Gamulescu et al. 2006;</ns0:ref><ns0:ref type='bibr' target='#b60'>Hinz et al. 2001;</ns0:ref><ns0:ref type='bibr' target='#b125'>Shu & Lovicu 2017;</ns0:ref><ns0:ref type='bibr' target='#b136'>Tamiya & Kaplan 2016;</ns0:ref><ns0:ref type='bibr' target='#b140'>Tomasek et al. 2002)</ns0:ref>.</ns0:p><ns0:p>In addition to myofibroblasts, abnormally increased ECM reinforces the continuous contractile tension of PVR membranes, and this mechanical tension, together with specialized ECM proteins, regulates myofibroblast differentiation and its function, contributing to PVR. In PVR membranes, the primary components of ECM are collagen and fibronectin. The majority of collagen fibrils are type Ⅰ collagen, which is synthesized by RPE cells and Müller cells.</ns0:p><ns0:p>Collagen fibrils provide tensile strength to the ECM, and activate Rho, resulting in the translocation of myocardin-related transcription factor (MRTF) into the nucleus and promoting RPE cell EMT <ns0:ref type='bibr' target='#b52'>(Guettler et al. 2008;</ns0:ref><ns0:ref type='bibr' target='#b102'>Miralles et al. 2003)</ns0:ref>. Fibronectin may also play a significant role in PVR. During pathological ECM remodeling, fibronectin is one of the earliest ECM components recruited, serving as a scaffold for other ECM proteins <ns0:ref type='bibr' target='#b70'>(Kadler et al. 2008;</ns0:ref><ns0:ref type='bibr'>Miller et al. 2017;</ns0:ref><ns0:ref type='bibr' target='#b101'>Miller et al. 2014)</ns0:ref>. Extra domain (ED)-A fibronectin, a splice variant of fibronectin, is increased in TGF-β2-induced RPE cells and induces myofibroblast differentiation, participating in PVR <ns0:ref type='bibr' target='#b74'>(Khankan et al. 2011)</ns0:ref>.</ns0:p><ns0:p>Under normal conditions, ECM breakdown by proteases such as matrix-metalloproteases (MMPs) plays a crucial role in ECM remodeling and the release of growth factors, maintaining tissue homeostasis in cooperation with ECM synthesis, reassembly, and chemical modification <ns0:ref type='bibr' target='#b13'>(Bonnans et al. 2014;</ns0:ref><ns0:ref type='bibr' target='#b31'>Craig et al. 2015;</ns0:ref><ns0:ref type='bibr' target='#b89'>Lindsey et al. 2016</ns0:ref>). As mentioned above, the polarized RPE is able to basolaterally secrete the extracellular matrix components fibronectin and collagens, MMP and tissue inhibitors of MMPs (TIMPs), which participate in ECM remodeling. However, under pathological conditions such as inflammation and retinal injury, RPE cells lose their apical-basal polarity, undergo EMT and abnormally secrete MMPs, TIMPs and ECM proteins, leading to dysregulated ECM remodeling <ns0:ref type='bibr' target='#b51'>(Greene et al. 2017)</ns0:ref>. Such ECM has aberrant composition and organization and mechanical properties, and enhances matrix stiffness and strain, which disrupts the normal structure and function of the retina, exacerbating the progression of PVR.</ns0:p></ns0:div>
<ns0:div><ns0:head n='3.'>RPE and Epithelial-mesenchymal Transition</ns0:head></ns0:div>
<ns0:div><ns0:head n='3.1'>EMT of RPE Cell</ns0:head><ns0:p>Epithelial-mesenchymal transition is an important biological process, in which epithelial cells transdifferentiate into mesenchymal cells. Although EMT can occur in normal embryonic development and wound healing, it also participates in pathological processes such as fibrosis, cancer progression, and PVR. There are three distinct subtypes of EMT: type 1 occurs during tissue and embryo development, type 2 is involved in wound healing and organ fibrosis, and type 3 is associated with cancer progression and metastasis <ns0:ref type='bibr' target='#b36'>(Dongre & Weinberg 2019;</ns0:ref><ns0:ref type='bibr' target='#b71'>Kalluri & Weinberg 2009)</ns0:ref>. This review focuses on type 2 EMT, which is crucial to PVR. During EMT, due to junctional complexes damage, RPE cells relinquish their apical-basal polarity, reorganize their cytoskeletal architecture, and convert into spindle-shaped cells (Fig <ns0:ref type='figure'>1</ns0:ref>). These cells downregulate the expression of epithelial proteins such as E-cadherin and ZO-1, and increase expression of mesenchymal drivers including N-cadherin, vimentin, α-SMA and fibronectin <ns0:ref type='bibr' target='#b85'>(Li et al. 2020)</ns0:ref>. This mesenchymal transdifferentiation of RPE cells can increase the directional Manuscript to be reviewed motility of individual cells, confer resistance to apoptosis, and facilitate cell proliferation and dysregulated ECM remodeling, eventually leading to the formation of PVR membranes.</ns0:p></ns0:div>
<ns0:div><ns0:head n='3.2'>Transcription Factors of EMT</ns0:head><ns0:p>The details of the molecular mechanisms that drive RPE cell EMT and lead to PVR remain to be clarified. Emerging evidence suggests that diverse extracellular inductive signals, including soluble cytokines and growth factors, and ECM components, can modulate the expression and activity of EMTassociated transcription factors and act together to control the initiation and progression of EMT in responding epithelial cells <ns0:ref type='bibr' target='#b155'>(Yang et al. 2020</ns0:ref>). Among the various transcription factors involved in the induction of EMT, core transcription factors including Snail 1, Snail 2(also known as Slug), Twist 1 and zinc-finger E-box-binding (Zeb) 1 have been identified as important regulators of RPE cell EMT. These factors impact the expression of genes that control repression of the epithelial phenotype and activation of the mesenchymal phenotype <ns0:ref type='bibr' target='#b12'>(Boles et al. 2020;</ns0:ref><ns0:ref type='bibr' target='#b40'>Feng et al. 2019;</ns0:ref><ns0:ref type='bibr' target='#b83'>Li et al. 2019;</ns0:ref><ns0:ref type='bibr' target='#b84'>Li et al. 2014;</ns0:ref><ns0:ref type='bibr' target='#b92'>Liu et al. 2009;</ns0:ref><ns0:ref type='bibr' target='#b113'>Palma-Nicolás & López-Colomé 2013)</ns0:ref>. For example, thrombin can repress the expression of E-cadherin by stimulating Snail 2 expression and promote the expression of N-cadherin by phosphoinositide 3kinase (PI3K)/PKC-ζ/mTOR signaling in Rat RPE cells (Palma-Nicolás & López-Colomé 2013). During RPE dedifferentiation in primary culture, Zeb1 is overexpressed and binds to the MITF A promoter to repress the cyclin dependent kinase inhibitor, p21CDKN1a, resulting in RPE cell proliferation and EMT <ns0:ref type='bibr' target='#b92'>(Liu et al. 2009</ns0:ref>). These EMT transcription factors often act in concert, functionally cooperating at target genes by the convergence of signaling pathways. However, the molecular details of how these transcription factors contribute to EMT are still elusive <ns0:ref type='bibr' target='#b77'>(Lamouille et al. 2014;</ns0:ref><ns0:ref type='bibr' target='#b131'>Stone et al. 2016)</ns0:ref>.</ns0:p></ns0:div>
<ns0:div><ns0:head n='3.3'>Epigenetic Factors of EMT</ns0:head><ns0:p>Due to the importance of epigenetic regulation of EMT, epigenetic modifiers have attracted increasing attention. Unlike transcription factors, epigenetic modifications are more stable/longterm. However, evidence has shown that epigenetic modifiers work in concert with transcription factors at different molecular layers to regulate the EMT process <ns0:ref type='bibr' target='#b127'>(Skrypek et al. 2017)</ns0:ref>. Several epigenetic factors have been described including DNA methylation, histone modification and non-coding RNA. Because of the specific machinery utilized for EMT activation, these modifications are characterized by cell type specificity. In RPE cells, Methyl-CpG-binding protein 2 (MeCP2), a DNA methylation reader, plays a crucial role in the induction of EMT, and DNA methylation may participate in the pathogenesis of PVR <ns0:ref type='bibr' target='#b58'>(He et al. 2015;</ns0:ref><ns0:ref type='bibr' target='#b85'>Li et al. 2020)</ns0:ref>. <ns0:ref type='bibr' target='#b58'>He et al. (2015)</ns0:ref> found high levels of expression of MeCP2 in all human PVR membranes, and concluded that MeCP2 mediates α-SMA expression through Ras GTPase activating protein (RASAL1). Furthermore, DNA methylation inhibitor 5-Aza-2' deoxycytidine (5-AZA-dC) reportedly inhibits the expression of TGF-β-induced α-SMA and FN in human fetal RPE cells. It appears that 5-AZA-dC may have therapeutic value in the treatment of PVR. However, the mechanisms underlying the blockade of α-SMA and FN expression are complex, and further investigation is warranted.</ns0:p><ns0:p>Recently, the role of histone modifications associated with EMT has been assessed in RPE cells. However, there has been little research into the regulation of RPE cell EMT by histone modification. <ns0:ref type='bibr' target='#b12'>Boles et al. (2020)</ns0:ref> reported that TGF-β1 and TNF-a co-treatment (TNT) induces an EMT program in adult human RPE stem cell (RPESC)-RPE cells, involving an apparent reorganization of H3K27ac and H3K4me1 patterns at distal enhancers. The regions that gain H3K27ac tend to have a high H3K4me1/H3K4me3 ratio, indicating that they have enhancer activity and are associated with upregulated genes. <ns0:ref type='bibr' target='#b154'>Xiao et al. (2014)</ns0:ref> found that the expression of histone deacetylases (HDACs) in TGF-β-induced EMT of RPE cells was increased, and that Trichostatin A (TSA), a class I and II HDAC inhibitor, attenuated TGF-β2-induced EMT by inhibiting the canonical SMAD pathway and the non-canonical signaling pathways, including Akt, p38MAPK, ERK1/2 pathways and Notch pathway. Therefore, histone modifications may participate in the regulation of RPE cell EMT, and HDAC inhibitors may have potential as drugs for the prevention and treatment of PVR.</ns0:p><ns0:p>The study of EMT mechanisms at the RNA level has provided new perspectives on the treatment of PVR <ns0:ref type='bibr' target='#b72'>(Kaneko & Terasaki 2017;</ns0:ref><ns0:ref type='bibr' target='#b147'>Wang et al. 2016)</ns0:ref>. MicroRNAs (miRNAs) are small noncoding RNAs that contribute to cellular processes by regulating gene expression. In differentiated RPE cells, microRNA-204 is highly expressed, and represses the expression of type Ⅱ TGF-β receptors and Snail 2, maintaining epithelial structure and function. In contrast, low expression levels of miR-204 and anti-miR-204 promote RPE cells proliferation, participating in EMT <ns0:ref type='bibr' target='#b146'>(Wang et al. 2010)</ns0:ref>. MicroRNA-194 overexpression can also suppress RPE cell EMT by attenuating the expression of Zeb1 <ns0:ref type='bibr' target='#b32'>(Cui et al. 2019</ns0:ref>). In addition to miRNAs, long non-coding RNAs (lncRNAs) contribute to the regulation of RPE EMT <ns0:ref type='bibr' target='#b163'>(Zhang et al. 2019)</ns0:ref>. In RPE cells treated with PVR vitreous or TGF-β1, MALAT1 expression is increased, and knockdown of MALAT1 attenuates the phosphorylation of SMAD2/3 and the expression of Snail, Slug, and Zeb1, preventing cell migration and proliferation <ns0:ref type='bibr' target='#b156'>(Yang et al. 2016</ns0:ref>). In patients with PVR, MALAT1 is increased in the blood, and is reduced after surgery. Thus, MALAT1 may be a potential prognostic and diagnostic indicator for PVR <ns0:ref type='bibr' target='#b168'>(Zhou et al. 2015)</ns0:ref>.</ns0:p></ns0:div>
<ns0:div><ns0:head n='3.4'>Signaling Pathways of EMT</ns0:head><ns0:p>During RPE cell EMT, extracellular signals change the expression of genes encoding epithelial and mesenchymal proteins and mediate cellular behavior such as cell migration, proliferation, and apoptosis through a network of interacting signaling pathways that contribute to the development of PVR <ns0:ref type='bibr' target='#b23'>(Chen et al. 2014a;</ns0:ref><ns0:ref type='bibr' target='#b24'>Chen et al. 2014b;</ns0:ref><ns0:ref type='bibr' target='#b78'>Lee-Rivera et al. 2015)</ns0:ref>. Among these, transforming growth factor-β (TGF-β) and its intracellular cascades play a key role in the EMT of RPE cells.</ns0:p><ns0:p>TGF-β induces EMT of RPE cells via two pathways: the classical SMAD-dependent pathway and the SMAD-independent pathway (Fig <ns0:ref type='figure' target='#fig_2'>2</ns0:ref>) <ns0:ref type='bibr' target='#b18'>(Cai et al. 2018;</ns0:ref><ns0:ref type='bibr' target='#b57'>He et al. 2017;</ns0:ref><ns0:ref type='bibr' target='#b59'>Heffer et al. 2019;</ns0:ref><ns0:ref type='bibr' target='#b64'>Ishikawa et al. 2015;</ns0:ref><ns0:ref type='bibr' target='#b133'>Takahashi et al. 2015;</ns0:ref><ns0:ref type='bibr' target='#b158'>Yao et al. 2019;</ns0:ref><ns0:ref type='bibr' target='#b162'>Zhang et al. 2017;</ns0:ref><ns0:ref type='bibr' target='#b164'>Zhang et al. 2018b;</ns0:ref><ns0:ref type='bibr' target='#b169'>Zhou et al. 2017</ns0:ref>). In the SMAD dependent pathway, TGF-β binds to cell surface receptor complexes, and activates type Ⅰ TGF-β receptors, which phosphorylate SMAD2 and SMAD3. The activated SMADs combine with SMAD4 to form a SMAD complex, which then enters the nucleus and combines with regulatory elements to regulate the expression of key genes associated with EMT. In addition to SMAD-dependent signaling, TGFβ induces EMT through SMAD independent signaling pathways including Rho GTPase-dependent pathways <ns0:ref type='bibr' target='#b80'>(Lee et al. 2008)</ns0:ref>, PI3K/Akt pathway <ns0:ref type='bibr' target='#b62'>(Huang et al. 2017;</ns0:ref><ns0:ref type='bibr' target='#b159'>Yokoyama et al. 2012)</ns0:ref>, mitogen-activated kinase (MAPK) pathways <ns0:ref type='bibr' target='#b26'>(Chen et al. 2017;</ns0:ref><ns0:ref type='bibr' target='#b79'>Lee et al. 2020;</ns0:ref><ns0:ref type='bibr' target='#b99'>Matoba et al. 2017;</ns0:ref><ns0:ref type='bibr' target='#b121'>Schiff et al. 2019)</ns0:ref> and Jagged/Notch signaling pathway <ns0:ref type='bibr' target='#b162'>(Zhang et al. 2017)</ns0:ref>. The MAPK signaling pathways include extracellular signal-regulated kinase(ERK) MAPK pathway, p38 MAPK pathway, and JUN Nterminal kinase (JNK) pathway <ns0:ref type='bibr' target='#b114'>(Parrales et al. 2013;</ns0:ref><ns0:ref type='bibr' target='#b121'>Schiff et al. 2019;</ns0:ref><ns0:ref type='bibr' target='#b154'>Xiao et al. 2014;</ns0:ref><ns0:ref type='bibr' target='#b161'>Zhang et al. 2018a</ns0:ref>).</ns0:p><ns0:p>The Rho pathway has been reported to regulate the assembly and organization of the actin cytoskeleton and associated gene expression, and may be essential for the fibrotic response of RPE cells in PVR. In TGF-β1-treated ARPE-19 cells, activated RhoA or its downstream effector Rho kinase (ROCK) increase the kinase activity of LIM kinase (LIMK) which then phosphorylates cofilin. This phosphorylation attenuates the activity of cofilin, which promotes actin polymerization and reorganizes the actin cytoskeleton, leading to stress fiber formation <ns0:ref type='bibr' target='#b80'>(Lee et al. 2008)</ns0:ref>. TGF-β-induced RhoA activation also facilitates cell migration and increases α-SMA expression in primary RPE cells <ns0:ref type='bibr' target='#b143'>(Tsapara et al. 2010)</ns0:ref>. <ns0:ref type='bibr' target='#b66'>Itoh et al. (2007)</ns0:ref> demonstrated that ROCK inhibitor Y27632 and RhoA inhibitor, simvastatin, suppress TGF-β2-induced type Ⅰ collagen expression in ARPE-19 cells, and confirmed the existence of crosstalk between the SMAD pathway and the Rho pathway. Some studies have suggested that activated SMAD3 induces NET1 gene expression to regulate RhoA activation in RPE cells <ns0:ref type='bibr' target='#b81'>(Lee et al. 2010)</ns0:ref>. Moreover, thrombin can activate Rho and ROCK, leading to myosin light chain (MLC) phosphorylation and actin stress fiber formation in EMT of RPE cells (Fig <ns0:ref type='figure'>3</ns0:ref>) <ns0:ref type='bibr' target='#b120'>(Ruiz-Loredo et al. 2011)</ns0:ref>. Therefore, ROCK inhibitor and RhoA inhibitor may be new potential therapeutic target drugs for PVR.</ns0:p><ns0:p>The PI3K/Akt pathway mediates a broad range of cellular functions, such as cell transformation, migration, proliferation, apoptosis, and gene expression <ns0:ref type='bibr' target='#b1'>(Aguilar-Solis et al. 2017;</ns0:ref><ns0:ref type='bibr' target='#b91'>Liu et al. 2019)</ns0:ref>. During PVR, binding of TGF-β to its receptor activates PI3K, resulting in the phosphorylation of Akt; activated Akt inhibits glycogen synthase kinase 3β (GSK-3β), promoting EMT in RPE cells <ns0:ref type='bibr' target='#b126'>(Shukal et al. 2020;</ns0:ref><ns0:ref type='bibr' target='#b161'>Zhang et al. 2018a</ns0:ref>). Researchers have found that inhibition or knockdown of GSK-3β promotes cell migration and collagen contraction in ARPE-19 cells, while GSK-3β overexpression and PI3K/Akt inhibitor reverse these cellular responses <ns0:ref type='bibr' target='#b62'>(Huang et al. 2017)</ns0:ref>. Some studies have shown that thrombin can activate PI3K, resulting in increased cyclin D1 expression and RPE cell proliferation, processes that are involved in the development of PVR through PDK1/Akt and PKCζ/mTORC signaling (Fig <ns0:ref type='figure'>3</ns0:ref>) <ns0:ref type='bibr' target='#b78'>(Lee-Rivera et al. 2015;</ns0:ref><ns0:ref type='bibr' target='#b113'>Palma-Nicolás & López-Colomé 2013;</ns0:ref><ns0:ref type='bibr' target='#b114'>Parrales et al. 2013</ns0:ref>).</ns0:p><ns0:p>In addition to the PI3K-AKT pathway, other kinase pathways contribute to EMT in cooperation with the SMAD-dependent signaling pathways. In human RPE cells, TGF-β activates TGF-β-activated kinase 1 (TAK1), which subsequently transduces signals to several downstream effectors, including p38 <ns0:ref type='bibr' target='#b59'>(Heffer et al. 2019)</ns0:ref>, JNK <ns0:ref type='bibr' target='#b76'>(Kimura et al. 2015)</ns0:ref> and nuclear factor-κB (NF-κB) <ns0:ref type='bibr' target='#b25'>(Chen et al. 2016b)</ns0:ref>, which participate in EMT. <ns0:ref type='bibr' target='#b37'>Dvashi et al. (2015)</ns0:ref> found that TAK1 inhibitor caused a reduction in both p38 and SMAD2/3 activity, attenuating cell migration, cell contractility and α-SMA expression in TGF-β1-induced RPE cells. Moreover, the ERK MAPK pathway plays a role in TGF-β-induced EMT and cooperates with other signaling pathways in the regulation of EMT in RPE cells. Recent studies <ns0:ref type='bibr' target='#b24'>(Chen et al. 2014b;</ns0:ref><ns0:ref type='bibr' target='#b138'>Tan et al. 2017;</ns0:ref><ns0:ref type='bibr' target='#b154'>Xiao et al. 2014)</ns0:ref> have shown that blocking the ERK1/2 pathway inhibits the phosphorylation of SMAD2 and the Jagged/Notch pathway. Inhibition of the Jagged/Notch signaling pathway can alleviate TGF-β2-induced EMT by regulating the expression of Snail, <ns0:ref type='bibr'>Slug and Zeb1 (Fig 3)</ns0:ref>; this also suppresses the ERK1/2 signaling <ns0:ref type='bibr' target='#b24'>(Chen et al. 2014b</ns0:ref>).</ns0:p><ns0:p>The contribution of growth factors other than TGF-β, such as HGF, fibroblast growth factor (FGF), epidermal growth factor (EGF) and platelet derived growth factor (PDGF) should also be factored in with regard to the induction of RPE EMT. These factors bind to and stimulate the autophosphorylation of transmembrane receptors on Tyr, subsequently participating in RPE cell EMT via PI3K/Akt pathway, ERK MAPK pathway, p38 MAPK pathway (Fig <ns0:ref type='figure'>3</ns0:ref>) <ns0:ref type='bibr' target='#b22'>(Chen et al. 2016a;</ns0:ref><ns0:ref type='bibr' target='#b112'>Ozal et al. 2020)</ns0:ref>. <ns0:ref type='bibr' target='#b21'>Chen et al. (2012)</ns0:ref> explored the role of Wnt/β-catenin signaling in PVR, and found that when EGTA disrupted contact inhibition in RPE cells, EGF+FGF2 could activate Wnt signaling and increase nuclear levels of β-catenin, which interacts with TCF and/or LEF, leading to cell proliferation (Fig 3 <ns0:ref type='figure'>)</ns0:ref>; and EGF+FGF2 cooperated with TGF-β1 to induce EMT through SMAD/Zeb1/2 signaling. Acting together, various inductive signals received by RPE cells from their niche can trigger the activation of EMT programs by individual intracellular cascades or the crosstalk of multiple intracellular signaling pathways.</ns0:p></ns0:div>
<ns0:div><ns0:head n='3.5'>Interventions of RPE EMT</ns0:head><ns0:p>Therapeutic interventions against RPE EMT have largely been explored in mechanistic experiments using in vitro cell culture and in vivo animal models. To date, some promising drug candidates have been trialed in preclinical studies of PVR, including TGF-β receptor inhibitors, peroxisome proliferator-activated receptor (PPAR)-γ agonists, retinoic acid receptor-γ (RAR-γ) agonists and methotrexate <ns0:ref type='bibr' target='#b124'>(Shu et al. 2020;</ns0:ref><ns0:ref type='bibr'>Zhou et al. 2020)</ns0:ref>. <ns0:ref type='bibr' target='#b106'>Nassar et al. (2014)</ns0:ref> found that TGF-β receptor 1 inhibitor LY-364947 (LY) attenuates RPE cell transdifferentiation in vitro, and that intravitreal injection of LY completely prevents PVR and TRD in vivo. Evidence is emerging to show that the up-regulation of PPAR-γ expression may be beneficial for the treatment of fibrosis in several organs <ns0:ref type='bibr' target='#b149'>(Wang et al. 2019)</ns0:ref>. <ns0:ref type='bibr' target='#b55'>Hatanaka et al. (2012)</ns0:ref> reported that PPAR-γ agonist pioglitazone could prevent TGF-β-induced morphological changes and the upregulation of EMT-related markers in primary monkey RPE cells, through inhibition of the SMAD pathway. Some drugs, including dichloroacetate (DCA) <ns0:ref type='bibr' target='#b126'>(Shukal et al. 2020)</ns0:ref>, salinomycin (SNC) <ns0:ref type='bibr' target='#b59'>(Heffer et al. 2019)</ns0:ref>, resveratrol <ns0:ref type='bibr' target='#b64'>(Ishikawa et al. 2015)</ns0:ref>, protein kinase A inhibitor H89 <ns0:ref type='bibr' target='#b94'>(Lyu et al. 2020</ns0:ref>) and heavy chain-hyaluronan/pentraxin3 <ns0:ref type='bibr' target='#b57'>(He et al. 2017)</ns0:ref>, reportedly inhibit EMT in an in vitro EMT cell model and prevent PVR development by blocking the activation of theTGF-β pathway. Thus, inhibition of EMT by pharmacological agents may be an effective strategy to prevent PVR development.</ns0:p></ns0:div>
<ns0:div><ns0:head>Conclusion</ns0:head><ns0:p>Clinical and experimental studies have shown that RPE cells play an important role in PVR. Junctional complexes are crucial for the maintenance of RPE polarity. Under the influence of growth factors and cytokines, RPE cells lose cell-cell contact and apical-basal polarity, and undergo EMT via multiple signaling pathways, which promote cell proliferation, migration, and ECM production. RPE cells further transform into myofibroblasts and form fibrocellular membranes that have contractile activity and strain the retina, leading to tractional retinal detachment in PVR. As a complex refractory blinding disorder, PVR involves multiple signaling pathways and factors. In addition, the specialized polarity of RPE cells is fundamental for retinal homeostasis, and RPE EMT plays a key role in the development of PVR. Nevertheless, further research into the mechanisms underlying RPE polarity and EMT is needed to prevent this devastating complication. A deeper understanding of RPE polarization is fundamental for elucidating the mechanism of EMT initiation and progression, and is essential to exploring the potential pharmacologic prophylactic and therapeutic approaches to PVR. Various factors, such as microenvironmental signals, transcription factors, and epigenetic factors, participate in the regulation of EMT at different molecular levels. Further studies about the detailed molecular mechanisms of EMT are needed to facilitate the development of therapeutic strategies for PVR. </ns0:p></ns0:div><ns0:figure xml:id='fig_0'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:06:49749:1:0:NEW 19 Aug 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_1'><ns0:head /><ns0:label /><ns0:figDesc>Figure 1</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_2'><ns0:head>Figure 2</ns0:head><ns0:label>2</ns0:label><ns0:figDesc>Figure 2</ns0:figDesc></ns0:figure>
<ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='27,42.52,70.87,525.00,386.25' type='bitmap' /></ns0:figure>
<ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='30,42.52,70.87,525.00,369.75' type='bitmap' /></ns0:figure>
<ns0:note place='foot'>PeerJ reviewing PDF | (2020:06:49749:1:0:NEW 19 Aug 2020)</ns0:note>
</ns0:body>
" | "Dear Prof. Pfeffer,
Thank you for your letter and for the referees’ comments regarding our manuscript entitled “Polarity and Epithelial-mesenchymal Transition of Retinal Pigment Epithelial Cells in Proliferative Vitreoretinopathy” (49749). Those comments were highly insightful and enabled us to greatly improve the quality of our manuscript. We studied comments carefully and have edited the manuscript to address their concerns. Based on your comment and request, we have made extensive modification in the original manuscript.
Revisions in the tracked changes manuscript are shown using red text for additions and the strikethrough feature for deletions. In the following pages are our point-by-point responses to each of the comments of the reviewers. All the lines indicated below are in the revised manuscript. We hope that the revisions in the manuscript and our accompanying responses will be sufficient to render our manuscript suitable for publication in PeerJ.
Thank you and all the reviewers for the kind advice.
Sincerely,
Jinsong Zhao
Institute: Eye Center of the second hospital, Jilin University, Changchun city, Jilin Province, China.
Address: Eye Center of the second hospital,
Jilin University,
4026 Yatai Street,
Changchun, Jilin, 130021,
China;
E-mail：jinsongzhao2003@163.com.
Tel：86-0431-81136523
Responses to the comments of Reviewer 1
Basic reporting
This manuscript by Zou and colleagues covers the topic of epithelial mesenchymal transition (EMT) in relation to an ocular fibrotic complication termed proliferative vitreoretinopathy. The main targeted audience is, therefore, people interested in ocular fibrosis. However, as EMT plays a key role in other fibrotic complications as well as in cancer metastasis, the manuscript may have some cross-disciplinary interest.
While there are a fair amount of overlap with past reviews on the topic, the manuscript also describes recent studies that have been published in the last three years.
The Introduction does clearly summarize the subject area, current problems, and the motivation.
Response: Thank you very much for your positive evaluation of our work. We have added more recent studies in the revised manuscript.
Experimental design
The review is organized into subsections but could be better organized into coherent paragraphs in order to avoid jumping around and remove unnecessary repetitions.
Response: We apologize for the confusion generated by the previous version of the manuscript. According to the reviewer’s comment, we have reorganized our review. We sincerely hope that our logic is now easier to follow with this new version.
Validity of the findings
The authors claims that they have reviewed “emphasizing key insights into potential approaches to prevent PVR” (from the abstract). However, I was not sure which part actually emphasized key insights.
Response: We agree with the comment. We have added comments on potential prophylactic and therapeutic approaches for PVR throughout the manuscript. For example:
Line 182-188: Due to the importance of TJs in the maintenance of integrity and functionality of epithelial cells, several researchers have focused on novel factors that stimulate the formation of TJs, such as nicotinamide (Hazim et al. 2019) and lysophosphatidic acid (Lidgerwood et al. 2018). Studies into these factors may produce well-differentiated RPE cell lines and a platform to enable the rapid expansion of our understanding of many RPE functions and retinal pathologies. This approach could be conducive to finding novel therapeutic interventions for PVR.
Line 209-211: Given the importance of HGF in the interruption of RPE junction, HGF may be a potential target for the prevention and treatment of PVR. However, this possibility needs further study.
Line 331-332: HDAC inhibitors may be potential drugs in the prevention and treatment of PVR.
We have also added a section “Interventions of RPE EMT” at line 425-443 as follows: Therapeutic interventions against RPE EMT have largely been explored in mechanistic experiments using in vitro cell culture and in vivo animal models. To date, some promising drug candidates have been trialed in preclinical studies of PVR, including TGF- β receptor inhibitors, peroxisome proliferator-activated receptor (PPAR)-γ agonists, retinoic acid receptor-γ (RAR-γ) agonists and methotrexate (Shu et al. 2020; Zhou et al. 2020).Nassar et al. (2014) found that TGF-β receptor 1 inhibitor LY-364947 (LY) attenuates RPE cell transdifferentiation in vitro, and that intravitreal injection of LY completely prevents PVR and TRD in vivo. Evidence is emerging to show that the up-regulation of PPAR-γ expression may be beneficial for the treatment of fibrosis in several organs(Wang et al. 2019). Hatanaka et al. (2012) reported that PPAR-γ agonist pioglitazone could prevent TGF-β-induced morphological changes and the up-regulation of EMT-related markers in primary monkey RPE cells, through inhibition of the SMAD pathway. Some drugs, including dichloroacetate (DCA)(Shukal et al. 2020), salinomycin (SNC)(Heffer et al. 2019), resveratrol(Ishikawa et al. 2015), protein kinase A inhibitor H89 (Lyu et al. 2020) and heavy chain-hyaluronan/pentraxin3(He et al. 2017), reportedly inhibit EMT in an in vitro EMT cell model and prevent PVR development by blocking the activation of theTGF-β pathway. Thus, inhibition of EMT by pharmacological agents may be an effective strategy to prevent PVR development.
The manuscript conclude with the need for more research on mechanisms involved in PVR, and does not necessarily identify gaps or future directions.
Response: We agree with the suggestion and have added a comment about unresolved questions and future directions in the conclusion of the revised manuscript (Line 451-461) as follows: As a complex refractory blinding disorder, PVR involves multiple signaling pathways and factors. In addition, the specialized polarity of RPE cells is fundamental for retinal homeostasis, and RPE EMT plays a key role in the development of PVR. Nevertheless, further research into the mechanisms underlying RPE polarity and EMT is needed to prevent this devastating complication. A deeper understanding of RPE polarization is fundamental for elucidating the mechanism of EMT initiation and progression, and is essential to exploring the potential pharmacologic prophylactic and therapeutic approaches to PVR. Various factors, such as microenvironmental signals, transcription factors, and epigenetic factors, participate in the regulation of EMT at different molecular levels. Further studies about the detailed molecular mechanisms of EMT are needed to facilitate the development of therapeutic strategies for PVR.
Comments for the Author
The vast number of cited references clearly demonstrate that the authors covered many articles. While the authors have covered many relevant studies, it becomes a little hard to follow due to the following reasons.
(1) The written English in certain places make it harder to follow. Improvement of the English language will significantly enhance the readability of the manuscript and ensures that the audience clearly understand your text. Improvement in the use of commas, in particular, will be beneficial.
Repeated typos: “RPE cells EMT” should be “RPE cell EMT”
Other typos: extracelluar (line 128), Samd (should be Smad, Fig2)
Suggested change to use of “release (line 121)”, “therefore (line 124)”, “transmit (line 128)”, “through (line 274)”
Consider rewriting sentence: Line 226 (;they change ~)
Response: We apologize for the mistakes in the manuscript. We carefully checked the entire manuscript for typographic, grammatical and formatting errors and have corrected the errors in the revised manuscript by us as well as the editing company. Necessary changes have been made in the revised manuscript.
(2) Description of studies using RPE and other cell types are often intermingled without clarification. Furthermore, studies that has been conducted on the human RPE cell line ARPE-19, which never becomes truly polarized, is often treated as RPE cells within the manuscript. It will be very helpful if the authors distinguish between non-RPE, RPE and ARPE-19.
Response: We apologize for the ambiguous expression in the original manuscript. We completely agree with the reviewer that the human RPE cell line ARPE-19 shows partial preservation of the RPE phenotype. In the revised version of the manuscript, we have identified the cells for each study to distinguish between non-RPE, RPE and ARPE-19. For example,
Line 159-161: Feng et al. (2019) demonstrated that during EMT, the breakdown of TJs resulting from loss of claudin-1 causes ARPE-19 cells to lose their epithelial phenotype and transform into fibroblasts, promoting the development of PVR.
Line 163-167: Vietor et al. (2001) found that decreased amounts of occludin can cause up-regulation and translocation of the adhesion junction protein β-catenin, which interacts with the transcription factor lymphoid enhancer-binding factor (LEF)/T cell factor (TCF) in the nucleus, leading to a loss of the polarized epithelial phenotype in EpH4 cells.
Line 168-170: Under injury or stress, the disruption of TJs increases ZO-2 nuclear accumulation, driving its interaction with transcription factors, and inducing MDCK epithelial cell proliferation (Islas et al. 2002; Shi et al. 2018; Traweger et al. 2003).
Line 178-182: Recent research has confirmed that during EMT, ZO-1 is decreased in ARPE-19 cells, and the knockdown of either ZO-1 or AJ protein E-cadherin leads to the downregulation of the other protein, indicating the existence of an interaction between the two junctional complexes (Bao et al. 2019).
(3)Section 3. RPE and Fibrocellular Membrane – please clarify between “differentiated RPE” and “de-differentiated RPE (EMT-RPE)” as the “RPE” that contributes to PVR membrane formation is the EMT-RPE.
Response: Thank you for pointing this out. We have corrected this to De-differentiated RPE and Fibrocellular Membrane.
Minor comments
1. Not sure if Line155-165 is necessary as it does not seem to have relevance to EMT or PVR.
Response: We accept the reviewer’s comment. We have rewritten the text in the revised version of the manuscript.
2. Line 261-262 – explain the function of cofilin and how it is regulated by phosphorylation.
Response: As suggested by the reviewer, we have rewritten the sentence in the revised manuscript (line 372-376) as follows: In TGF-β1-treated ARPE-19 cells, activated RhoA or its downstream effector Rho kinase (ROCK) increase the kinase activity of LIM kinase (LIMK) which then phosphorylates cofilin. This phosphorylation attenuates the activity of cofilin, which promotes actin polymerization and reorganizes the actin cytoskeleton, leading to stress fiber formation (Lee et al. 2008).
3. Line 295-303 reads like an add-on after writing the manuscript. It will be great if it can be incorporated into the section above.
Response: Accordingly, we have modified the expression to address the reviewer’s point in the revised manuscript (line 412-424) as follows: The contribution of growth factors other than TGF-β, such as HGF, fibroblast growth factor (FGF), epidermal growth factor (EGF) and platelet derived growth factor (PDGF) should also be factored in with regard to the induction of RPE EMT. These factors bind to and stimulate the autophosphorylation of transmembrane receptors on Tyr, subsequently participating in RPE cell EMT via PI3K/Akt pathway, ERK MAPK pathway, p38 MAPK pathway (Fig 3) (Chen et al. 2016a; Ozal et al. 2020).Chen et al. (2012) explored the role of Wnt/β-catenin signaling in PVR, and found that when EGTA disrupted contact inhibition in RPE cells, EGF+FGF2 could activate Wnt signaling and increase nuclear levels of β-catenin, which interacts with TCF and/or LEF, leading to cell proliferation (Fig 3); and EGF+FGF2 cooperated with TGF-β1 to induce EMT through SMAD/Zeb1/2 signaling. Acting together, various inductive signals received by RPE cells from their niche can trigger the activation of EMT programs by individual intracellular cascades or the crosstalk of multiple intracellular signaling pathways.
Responses to the comments of Reviewer 2 (Shikun He)
Basic reporting
The review is within the scope of the journal.
there are numbers of reviews that are similar to the current review, however, previous review did not make a clear connection of the RPE Polarity to EMT.
The Introduction of the review is adequate.
Response: We appreciate the reviewer’s positive evaluation of our work.
Experimental design
It is a comprehensive review, it covered most of the information in the field.
The uncovered information, not cited reference and the suggestion for the review organization are listed in the following section of General comments for the author.
Response: Thank you very much for your positive evaluation of our work. Following your comments, we have made necessary changes in the revised manuscript.
Validity of the findings
General speaking, the review is well developed and supported argument that meets the goals set out in the Introduction.
The conclusion did not include the contents of unresolved questions / gaps.
Response: We accept the reviewer’s comment. In the revised version of the manuscript, we have rewritten the conclusion (line 445-461) as follows: Clinical and experimental studies have shown that RPE cells play an important role in PVR. Junctional complexes are crucial for the maintenance of RPE polarity. Under the influence of growth factors and cytokines, RPE cells lose cell-cell contact and apical-basal polarity, and undergo EMT via multiple signaling pathways, which promote cell proliferation, migration, and ECM production. RPE cells further transform into myofibroblasts and form fibrocellular membranes that have contractile activity and strain the retina, leading to tractional retinal detachment in PVR. As a complex refractory blinding disorder, PVR involves multiple signaling pathways and factors. In addition, the specialized polarity of RPE cells is fundamental for retinal homeostasis, and RPE EMT plays a key role in the development of PVR. Nevertheless, further research into the mechanisms underlying RPE polarity and EMT is needed to prevent this devastating complication. A deeper understanding of RPE polarization is fundamental for elucidating the mechanism of EMT initiation and progression, and is essential to exploring the potential pharmacologic prophylactic and therapeutic approaches to PVR. Various factors, such as microenvironmental signals, transcription factors, and epigenetic factors, participate in the regulation of EMT at different molecular levels. Further studies about the detailed molecular mechanisms of EMT are needed to facilitate the development of therapeutic strategies for PVR.
Comments for the Author
This is an interesting review paper; The authors summarized some of previous publications in the study of the role of RPE EMT in the pathogenesis of PVR. the diagrams look nice.
Following questions should be addressed before the consideration of the paper to be published in the journal.
Major points:
1. Lack of the information of Clinical evidence of RPR EMT in the pathogenesis of PVR
Response: We thank the reviewer for pointing out this issue. We indeed should add the information of clinical evidence of RPR EMT in the pathogenesis of PVR. However, clinical evidence of RPE EMT in PVR is based on histopathological studies. Several studies have shown that cytokeratin, an RPE cell marker, co-localized with vimentin in PVR membranes. Feist et al. (2014) showed that the majority of myofibroblasts in human PVR membranes originated from RPE cells, suggesting that RPE cells are capable of undergoing EMT and migrating into neuroretina, and that this process plays a major role in the pathogenesis of PVR. We have included this study in the manuscript (line 232-234) as follows: Based on the indirect immunofluorescence evaluation of human PVR membranes, Feist et al. (2014) showed that myofibroblasts originate principally from RPE cells through EMT.
A comment to the histopathology of PVR has been also included at line 64-68: According to the histopathology of PVR, the fibrocellular membrane of PVR is composed of excessive extracellular matrix (ECM) and multiple types of cells, and retinal pigment epithelial (RPE) cells have been indicated as the most consistently present and the most abundant (Amarnani et al. 2017; Ding et al. 2017; Hiscott et al. 1989; Machemer & Laqua 1975), proving that the RPE cell plays a crucial role in PVR.
2. the authors mentioned the Polarity of RPE, what are the characteristics of RPE Polarity physiologically and what is the phenotypes changes in the condition of EMT? In factor, the authors put the RPE Polarity in the remarkable position, however, there is not much description of the relevance of RPE Polarity to EMT, in order to make its title consistence with contents, the review should be reconstructed and extended.
Response: Based on helpful comments from the reviewers, we have edited the text further. We have added a comment to the RPE polarity at line 103-136 and a comment to the phenotypes changes in the condition of EMT at line 274-281.
We have also added this statement at line 212-222 which addresses the relevance of RPE Polarity to EMT: Under physiological conditions in the eye, TJs and AJs maintain the specialized structural and functional polarity of RPE cells and play a pivotal role in the maintenance of cell-cell contact; they sequester EMT signaling effectors ZONAB and β-catenin at the junction or cytoplasm to prevent cells from responding to mitotic factors, causing cells to leave the cell-cycle (Fig 1). Thus, normally, RPE cells form a cobblestone-like monolayer of immotile, polarized, and mitotically quiescent cells. However, once junctional complexes break down, RPE cells undergo EMT, which is an important contributor to proliferative vitreoretinopathy. In this pathological process, RPE cells lose their structural and functional polarity and transdifferentiate into mesenchymal cells, which proliferate, resist apoptosis, possess migratory ability, and produce abundant ECM, leading to the formation of an aberrant scar-like fibrocellular membrane.
3. the role of Epigenetic factors in the contribution of RPE EMT has not been incorporated into the review paper.
Response: As suggested by the reviewer, a comment on the contribution of epigenetic factors in RPE EMT has been included at line 303-347.
4. Review is not only a pile up of previous publications but also your view and comments should be included. In the section of conclusion, the authors should incorporate what are the unsolved question in the field and what is the speculation.
Response：We sincerely appreciate the reviewer’s valuable suggestion. We have added our comments throughout the text and rewritten this following conclusion at line 445-461: Clinical and experimental studies have shown that RPE cells play an important role in PVR. Junctional complexes are crucial for the maintenance of RPE polarity. Under the influence of growth factors and cytokines, RPE cells lose cell-cell contact and apical-basal polarity, and undergo EMT via multiple signaling pathways, which promote cell proliferation, migration, and ECM production. RPE cells further transform into myofibroblasts and form fibrocellular membranes that have contractile activity and strain the retina, leading to tractional retinal detachment in PVR. As a complex refractory blinding disorder, PVR involves multiple signaling pathways and factors. In addition, the specialized polarity of RPE cells is fundamental for retinal homeostasis, and RPE EMT plays a key role in the development of PVR. Nevertheless, further research into the mechanisms underlying RPE polarity and EMT is needed to prevent this devastating complication. A deeper understanding of RPE polarization is fundamental for elucidating the mechanism of EMT initiation and progression, and is essential to exploring the potential pharmacologic prophylactic and therapeutic approaches to PVR. Various factors, such as microenvironmental signals, transcription factors, and epigenetic factors, participate in the regulation of EMT at different molecular levels. Further studies about the detailed molecular mechanisms of EMT are needed to facilitate the development of therapeutic strategies for PVR.
Minor:
1. There are 3 types of EMT, in the current review which type EMT refers to?
Response: This review focuses on type 2 EMT. We have rewritten the sentence in the revised manuscript (line 270-274) as follows: There are three distinct subtypes of EMT: type 1 occurs during tissue and embryo development, type 2 is involved in wound healing and organ fibrosis, and type 3 is associated with cancer progression and metastasis (Dongre & Weinberg 2019; Kalluri & Weinberg 2009). This review focuses on type 2 EMT, which is crucial to PVR.
2. What are the markers in RPE cells when EMT occurs; in the condition of EMT, which molecular markers are obtained and which lost?
Response: We have added a clarifying comment in the revised manuscript (line 274-281) as follows: During EMT, due to junctional complexes damage, RPE cells relinquish their apical-basal polarity, reorganize their cytoskeletal architecture, and convert into spindle-shaped cells (Fig 1). These cells downregulate the expression of epithelial proteins such as E-cadherin and ZO-1, and increase expression of mesenchymal drivers including N-cadherin, vimentin, α-SMA and fibronectin (Li et al. 2020). This mesenchymal transdifferentiation of RPE cells can increase the directional motility of individual cells, confer resistance to apoptosis, and facilitate cell proliferation and dysregulated ECM remodeling, eventually leading to the formation of PVR membranes.
3. Interruption of junction of RPE is a critical step in the induction of EMT, there are some key publications about this; among the previous study, HGF has been shown to play an important role in the interruption of RPE junction, treatment of RPE explants with HGF results in rapid disassembly of tight and adherens junctions associated with loss or redistribution of junctional proteins, decreased TER, and increased migration of RPE cells from the monolayer (Jin et al. IOVS, August 2002, Vol. 43, No. 8,2782-2790; Invest Ophthalmol Vis Sci. 2004;45:323–329). Those informative message has been neglected.
Response: We apologize for the neglect of those informative messages. We have added these key studies about the interruption of RPE junction induced by HGF in the revised manuscript (line 206-209) as follows: Jin et al found that HGF induces loss or redistribution of junctional proteins ZO-1, occludin, and β-catenin in RPE explants, potentially damaging barrier function and increasing the migration of RPE cells, resulting in retinal detachment(RD) and PVR (Jin et al. 2002; Jin et al. 2004).
4. there are much more transcription factors involved in the induction of EMT, the authors need to look up previous publication closely.
Response: We completely agree with the reviewer that a growing number of transcription factors have been described to induce EMT. We searched the recent literature, and have added more factors. However, we largely focus on core transcription factors of EMT, so we did not included all transcription factors involved in the induction of EMT. Necessary changes to the statements have been made in the revised manuscript (line 283-292): The details of the molecular mechanisms that drive RPE cell EMT and lead to PVR remain to be clarified. Emerging evidence suggests that diverse extracellular inductive signals, including soluble cytokines and growth factors, and ECM components, can modulate the expression and activity of EMTassociated transcription factors and act together to control the initiation and progression of EMT in responding epithelial cells (Yang et al. 2020). Among the various transcription factors involved in the induction of EMT, core transcription factors including Snail 1, Snail 2(also known as Slug), Twist 1 and zinc-finger E-box-binding (Zeb) 1 have been identified as important regulators of RPE cell EMT. These factors impact the expression of genes that control repression of the epithelial phenotype and activation of the mesenchymal phenotype.
5. line 304, RNA and EMT, it should be miRNA not RNA
Response: Considering the reviewer’s suggestion, we have changed “RNA and EMT” into “Epigenetic Factors of EMT”.
6. It may be better to put section 3 “ RPE and Fibrocellular Membrane” in the second place.
Response: According to the reviewer’s comment, we have reconstructed the paper.
7. no information about the intervention of RPE EMT.
Response: Thank you very much for the suggestion. We have added a section “3.5 Interventions of RPE EMT” in the revised manuscript (line 425-443) as follows: Therapeutic interventions against RPE EMT have largely been explored in mechanistic experiments using in vitro cell culture and in vivo animal models. To date, some promising drug candidates have been trialed in preclinical studies of PVR, including TGF- β receptor inhibitors, peroxisome proliferator-activated receptor (PPAR)-γ agonists, retinoic acid receptor-γ (RAR-γ) agonists and methotrexate (Shu et al. 2020; Zhou et al. 2020).Nassar et al. (2014) found that TGF-β receptor 1 inhibitor LY-364947 (LY) attenuates RPE cell transdifferentiation in vitro, and that intravitreal injection of LY completely prevents PVR and TRD in vivo. Evidence is emerging to show that the up-regulation of PPAR-γ expression may be beneficial for the treatment of fibrosis in several organs(Wang et al. 2019). Hatanaka et al. (2012) reported that PPAR-γ agonist pioglitazone could prevent TGF-β-induced morphological changes and the up-regulation of EMT-related markers in primary monkey RPE cells, through inhibition of the SMAD pathway. Some drugs, including dichloroacetate (DCA)(Shukal et al. 2020), salinomycin (SNC)(Heffer et al. 2019), resveratrol(Ishikawa et al. 2015), protein kinase A inhibitor H89 (Lyu et al. 2020) and heavy chain-hyaluronan/pentraxin3(He et al. 2017), reportedly inhibit EMT in an in vitro EMT cell model and prevent PVR development by blocking the activation of theTGF-β pathway. Thus, inhibition of EMT by pharmacological agents may be an effective strategy to prevent PVR development.
Responses to the comments of Reviewer 3
Basic reporting
There are many reviews about RPE cell EMT and PVR, but previous reviews did not pay much attention to the relationship between RPE polarity and EMT. So I think it’s reasonable for this review to be published in the journal after revision, the review is within the scope of the journal. The introduction does adequately introduce the subject.
Response: Thank you for your recognition of our work.
Experimental design
This is a review manuscript which content is within the Aims and Scope of the journal. The survey methodology is consistent with a comprehensive, unbiased coverage of the subject. But I think there is logical problem in this review manuscript. And not all sources were adequately cited.
Response: We apologize for the logical mistakes in the previous version of the manuscript. This comment is valuable to our review. According to the reviewer’s comment, we have reorganized the contents of this review.
We have also added more recent studies in the revised manuscript.
Validity of the findings
The review is a well developed and supported argument that meets the goals set out in the Introduction. But the conclusion doesn’t include the contents of unresolved questions / gaps/ future directions.
Response: We have rewritten the conclusion in the revised manuscript.
Comments for the Author
To be frank, this is an interesting review manuscript. Although there are many reviews about RPE cell EMT and PVR, but previous reviews did not pay much attention to the relationship between RPE polarity and EMT. But there still has some questions, the following questions should be addressed before the manuscript to be fully considered to be published in the journal.
Major points:
1. The title include the polarity of RPE cell, it’s the main difference from previous review about EMT and PVR, but what is polarity of RPE, and what’s the change once EMT occurs, what’s the real relationship between the polarity of RPE cell and EMT in PVR, all this questions were not answered clearly. The author should adding a new part about RPE polarity and EMT.
Response: We sincerely appreciate the reviewer’s suggestion. According to the reviewer’s comments, we have provided more details to describe the polarity of RPE at line 103-136, the changes during EMT at line 274-281 and the relationship between the polarity of RPE cell and EMT in PVR at line 212-222.
2.The manuscript mentioned the relationship between lncRNA and EMT, is it necessary in this manuscript, if it’s necessary, why not adding other epigenetic factors in this manuscript?
Response: As suggested by the reviewer, we have added more information regarding the epigenetic regulation of EMT on line 303-347.
3. In the conclusion, the authors should mention what are the unsolved questions in the field and what is the future direction.
Response: We completely agree. We have added this comment on the unsolved questions and the future direction in the conclusion of the revised manuscript (line 451-461). As a complex refractory blinding disorder, PVR involves multiple signaling pathways and factors. In addition, the specialized polarity of RPE cells is fundamental for retinal homeostasis, and RPE EMT plays a key role in the development of PVR. Nevertheless, further research into the mechanisms underlying RPE polarity and EMT is needed to prevent this devastating complication. A deeper understanding of RPE polarization is fundamental for elucidating the mechanism of EMT initiation and progression, and is essential to exploring the potential pharmacologic prophylactic and therapeutic approaches to PVR. Various factors, such as microenvironmental signals, transcription factors, and epigenetic factors, participate in the regulation of EMT at different molecular levels. Further studies about the detailed molecular mechanisms of EMT are needed to facilitate the development of therapeutic strategies for PVR.
Minor points:
1. There are some new papers about EMT and RPE cells, which also summarized the EMT of RPE cells and transcription factors of EMT, the authors should find new reference from recent publication.
Response: Considering the reviewer’s suggestion, we have added more recent studies in the revised manuscript.
2. The logical structure should be adjusted, I think it would be better to change the place of part2 (RPE and EMT) and part 3 (RPE and Fibrocellular Membrane).
Response: We accept the reviewer’s comment. In the revised version of the manuscript, we have made corrections according to the reviewer’s comments.
3. As we all known that there are 3 types of EMT, the authors should mention that which kind of EMT happens in PVR in this manuscript.
Response: This review focuses on type 2 EMT. Necessary changes to the statements have been made in the revised manuscript (line 270-274).
4. The figures looks good, but there are some spelling mistakes, such as in Figure 2, Samd2/3 should be replaced with Smad2/3, Samd4 should be replaced with Smad4.
Response: We apologize for these mistakes in the figure. We have corrected these errors in the revised Figure 2.
" | Here is a paper. Please give your review comments after reading it. |
9,780 | "<ns0:abstract xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:p>Under physiological conditions, retinal pigment epithelium (RPE) is a cellular monolayer composed of mitotically quiescent cells. Tight junctions and adherens junctions maintain the polarity of RPE cells, and are required for cellular functions. In proliferative vitreoretinopathy (PVR), upon retinal tear, RPE cells lose cell-cell contact, undergo epithelial-mesenchymal transition (EMT), and ultimately transform into myofibroblasts, leading to the formation of fibrocellular membranes on both surfaces of the detached retina and on the posterior hyaloids, which causes tractional retinal detachment. In PVR, RPE cells are crucial contributors, and multiple signaling pathways, including SMADdependent pathway, Rho pathway, MAPK pathways, Jagged/Notch pathway, and Wnt/βcatenin pathway, are activated. These pathways mediate the EMT of RPE cells, which play a key role in the pathogenesis of PVR. This review summarizes the current body of knowledge on the polarized phenotype of RPE, the role of cell-cell contact, and the molecular mechanisms underlying the RPE EMT in PVR, emphasizing key insights into potential approaches to prevent PVR.</ns0:p></ns0:div>
</ns0:abstract>
<ns0:body xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:head>Introduction</ns0:head><ns0:p>Proliferative vitreoretinopathy (PVR) is a complex blinding disease that occurs after rhegmatogenous retinal detachment (RRD), surgical interventions, or ocular trauma. As a prolonged and exaggerated scarring process, PVR is characterized by the formation of contractile fibrocellular membranes in the vitreous cavity and on the inner and outer surfaces of the retina <ns0:ref type='bibr' target='#b30'>(Committee 1983;</ns0:ref><ns0:ref type='bibr' target='#b105'>Mudhar 2020;</ns0:ref><ns0:ref type='bibr' target='#b141'>Tosi et al. 2014)</ns0:ref>. At present, surgical interventions, including vitrectomy, membrane peeling, pneumatic retinopexy, and scleral buckle, remain the mainstay of treatment in PVR. Although work in recent decades has led to advancements in surgical techniques and management, PVR cannot be effectively treated and is still the most common cause of failure to reattach the retina <ns0:ref type='bibr' target='#b29'>(Coffee et al. 2014;</ns0:ref><ns0:ref type='bibr' target='#b73'>Khan et al. 2015;</ns0:ref><ns0:ref type='bibr' target='#b103'>Mitry et al. 2012;</ns0:ref><ns0:ref type='bibr' target='#b152'>Wickham et al. 2011</ns0:ref>). In addition, in spite of successful anatomic reattachment, the visual function of such cases cannot be improved, due to the retinal damage resulting from the mechanical contraction of fibrous membranes. Therefore, in order to improve postoperative visual function and reduce the incidence of this serious complication, it is particularly important to explore new prophylactic and therapeutic approaches based on a deeper understanding of the pathogenesis of PVR.</ns0:p><ns0:p>A growing body of evidence indicates that the mechanisms of PVR are orchestrated by multiple elements <ns0:ref type='bibr' target='#b63'>(Idrees et al. 2019;</ns0:ref><ns0:ref type='bibr' target='#b69'>Jin et al. 2017;</ns0:ref><ns0:ref type='bibr' target='#b115'>Pastor et al. 2016)</ns0:ref>, such as growth factors <ns0:ref type='bibr' target='#b20'>(Charteris 1998;</ns0:ref><ns0:ref type='bibr' target='#b109'>Ni et al. 2020;</ns0:ref><ns0:ref type='bibr' target='#b118'>Pennock et al. 2014;</ns0:ref><ns0:ref type='bibr' target='#b154'>Wubben et al. 2016)</ns0:ref>, cytokines <ns0:ref type='bibr' target='#b7'>(Bastiaans et al. 2018;</ns0:ref><ns0:ref type='bibr' target='#b54'>Harada et al. 2006;</ns0:ref><ns0:ref type='bibr' target='#b88'>Limb et al. 1991)</ns0:ref>, extracellular matrix proteins <ns0:ref type='bibr' target='#b39'>(Feist et al. 2014;</ns0:ref><ns0:ref type='bibr' target='#b100'>Miller et al. 2017</ns0:ref>) and various cells <ns0:ref type='bibr' target='#b38'>(Eastlake et al. 2016;</ns0:ref><ns0:ref type='bibr' target='#b119'>Pennock et al. 2011;</ns0:ref><ns0:ref type='bibr' target='#b126'>Shu & Lovicu 2017)</ns0:ref>. According to the histopathology of PVR, the fibrocellular membrane of PVR is composed of excessive extracellular matrix (ECM) and multiple types of cells, and retinal pigment epithelial (RPE) cells have been indicated as the most consistently present and the most abundant <ns0:ref type='bibr' target='#b2'>(Amarnani et al. 2017;</ns0:ref><ns0:ref type='bibr' target='#b35'>Ding et al. 2017;</ns0:ref><ns0:ref type='bibr' target='#b61'>Hiscott et al. 1989;</ns0:ref><ns0:ref type='bibr' target='#b95'>Machemer & Laqua 1975)</ns0:ref>, proving that the RPE cell plays a crucial role in PVR. Under physiological condition, the polarized RPE cell is non-proliferative by cell-cell contact. However, when the eye suffers from a retinal break or trauma, RPE cells are exposed to various growth factors and cytokines that are produced by activated immune cells, leading to the disruption of junctional complexes in RPE cells. Subsequently, activated RPE cells detach from Bruch's membrane, migrate through the defect of the retina, proliferate, and transform into myofibroblasts, forming fibrotic membranes <ns0:ref type='bibr'>(Chen et al. 2015;</ns0:ref><ns0:ref type='bibr' target='#b104'>Morescalchi et al. 2013;</ns0:ref><ns0:ref type='bibr' target='#b113'>Palma-Nicolás & López-Colomé 2013)</ns0:ref>. In an analogous process to exaggerated wound healing response, these membranes can attach to the retina and contract, resulting in further retinal detachment and poor vision <ns0:ref type='bibr' target='#b28'>(Chiba 2014;</ns0:ref><ns0:ref type='bibr' target='#b46'>Garweg et al. 2013)</ns0:ref>. It is noteworthy that due to the loss of cell-cell contact, RPE cells undergo epithelialmesenchymal transition (EMT), which is pivotal in the development of PVR. During EMT, RPE cells transdifferentiate into mesenchymal cells that are characterized by increased motility, and enhanced ability to proliferate, resist apoptosis and produce extracellular matrix proteins, thus participating in PVR <ns0:ref type='bibr' target='#b136'>(Tamiya & Kaplan 2016;</ns0:ref><ns0:ref type='bibr' target='#b167'>Zhang et al. 2018c)</ns0:ref>. These indicate that in-depth knowledge of EMT may provide insight into potential approaches to prevent PVR. Therefore, this review focuses on the polarized phenotype of RPE and molecular mechanisms of RPE cell EMT, discussing the role of RPE cells in PVR.</ns0:p></ns0:div>
<ns0:div><ns0:head>Survey methodology</ns0:head><ns0:p>We used the PubMed database to search available literature based on keywords including 'proliferative vitreoretinopathy(PVR)' and 'retinal pigment epithelial cell'. To include more information on the polarity of RPE, we also searched articles about the structure and function of cell-cell junctions in RPE cells that explored the role of cell-cell contact in EMT.</ns0:p></ns0:div>
<ns0:div><ns0:head n='1.'>The Polarized Retinal Pigment Epithelial Cell</ns0:head><ns0:p>The human RPE cell achieves terminal differentiation at four to six weeks of gestation and subsequently remains mitotically quiescent <ns0:ref type='bibr' target='#b93'>(Lutty & McLeod 2018;</ns0:ref><ns0:ref type='bibr' target='#b131'>Stern & Temple 2015)</ns0:ref>. The RPE, which is situated between the photoreceptors and the choroid, plays many complex roles indispensable to the health of the neural retina and the choroid. These roles include recycling of components of the visual cycle, absorption of light to protect from photo-oxidative stress, production of essential growth factors, immunological regulation of the eye, phagocytosis of photoreceptor outer segments generated during daily photoreceptor renewal, and transportation across the blood retina barrier (BRB) <ns0:ref type='bibr' target='#b41'>(Ferrington et al. 2016;</ns0:ref><ns0:ref type='bibr' target='#b42'>Fields et al. 2019;</ns0:ref><ns0:ref type='bibr' target='#b97'>Mateos et al. 2014;</ns0:ref><ns0:ref type='bibr' target='#b107'>Naylor et al. 2019;</ns0:ref><ns0:ref type='bibr' target='#b133'>Strauss 2005;</ns0:ref><ns0:ref type='bibr' target='#b145'>Vigneswara et al. 2015)</ns0:ref>. In order to maintain these multiple functions, RPE cells display a highly specialized structural and functional polarity.</ns0:p><ns0:p>Similar to other epithelia, the RPE displays three characteristics of the epithelial phenotype: apical plasma membrane, junctional complexes, and basolateral domain. RPE cells display structural polarity, with apical microvilli and melanosomes, and basal microinfolds. The abundant melanin granules in RPE cells absorb stray light, a process that is essential for visual function <ns0:ref type='bibr' target='#b133'>(Strauss 2005)</ns0:ref>. In a polarized cell, the distributions of surface proteins on the apical and basal plasma membranes are different, contributing to the performance of cellular functions <ns0:ref type='bibr' target='#b75'>(Khristov et al. 2018</ns0:ref>). However, a highly polarized distribution of ion channels, transporters and receptors in RPE is different from that observed in conventional extraocular epithelia <ns0:ref type='bibr' target='#b82'>(Lehmann et al. 2014)</ns0:ref>. For example, Na, K-ATPase <ns0:ref type='bibr' target='#b130'>(Sonoda et al. 2009</ns0:ref>) and monocarboxylate transporters (MCT) 1 <ns0:ref type='bibr' target='#b33'>(Deora et al. 2005</ns0:ref>) are localized to the apical aspect of RPE cells, while chloride transporter CFTR <ns0:ref type='bibr' target='#b96'>(Maminishkis et al. 2006</ns0:ref>) is basally located. On the apical plasma membrane, RPE cells phagocytize the photoreceptor outer segments, which are regulated by polarized receptors. <ns0:ref type='bibr' target='#b14'>Bulloj et al. (2018)</ns0:ref> found that binding of Semaphorin 4D (sema4D) to RPE apical receptor Plexin-B1 suppresses outer segment internalization, contributing to the maintenance of photoreceptor function and longevity. The RPE also transports fluid out of the subretinal space, and regulates bidirectional nutrient transport between the outer retina and the choroid, in a manner dependent on the polarized distribution of membrane channels and transporters <ns0:ref type='bibr' target='#b133'>(Strauss 2005)</ns0:ref>. The RPE basolaterally secretes extracellular matrix components and factors, which participate in ECM remodeling and maintain the outer BRB (oBRB) function <ns0:ref type='bibr' target='#b17'>(Caceres & Rodriguez-Boulan 2020)</ns0:ref>. Therefore, the polarized phenotype of the RPE is vital to both the oBRB and is the basis of the homeostasis of the outer retina <ns0:ref type='bibr' target='#b17'>(Caceres & Rodriguez-Boulan 2020;</ns0:ref><ns0:ref type='bibr' target='#b82'>Lehmann et al. 2014)</ns0:ref>. The disruption of RPE polarity contributes to the development of several retinal diseases, such as PVR and age-related macular degeneration (AMD). A comprehensive understanding of the way in which this polarity is achieved may provide insights into the pathogenesis of PVR.</ns0:p><ns0:p>However, most available data on RPE polarity is contributed by studies performed on RPEimmortalized cell lines that show partial preservation of the RPE phenotype, and were extrapolated from data obtained from the prototype Madin-Darby Canine Kidney (MDCK) cell line <ns0:ref type='bibr' target='#b82'>(Lehmann et al. 2014)</ns0:ref>. The detailed mechanisms that determine RPE polarization remain unclear. Some scholars believe that junctional complexes, including adherens junctions (AJs) and tight junctions (TJs), are essential for building epithelial cell polarity and maintaining the integrity of epithelial layers such as RPE <ns0:ref type='bibr' target='#b110'>(Niessen 2007;</ns0:ref><ns0:ref type='bibr' target='#b117'>Pei et al. 2019;</ns0:ref><ns0:ref type='bibr' target='#b136'>Tamiya & Kaplan 2016)</ns0:ref>.</ns0:p><ns0:p>Tight junctions are complex cell-cell junctions formed by transmembrane proteins interactions with peripheral cytoplasmic proteins (Fig <ns0:ref type='figure' target='#fig_1'>1</ns0:ref>). Transmembrane proteins include occludin, members of the claudin family, and junctional adhesion molecules (JAMs). Peripheral cytoplasmic proteins, such as zonula occludens (ZOs), form bridges between transmembrane proteins and the actin filament cytoskeleton and play a key role in the assembly and organization of TJs <ns0:ref type='bibr' target='#b8'>(Bazzoni & Dejana 2004;</ns0:ref><ns0:ref type='bibr' target='#b9'>Bazzoni et al. 2000;</ns0:ref><ns0:ref type='bibr' target='#b107'>Naylor et al. 2019)</ns0:ref>.</ns0:p><ns0:p>The RPE tight junctions regulate the paracellular movement of solutes via size and charge selectivity <ns0:ref type='bibr' target='#b10'>(Benedicto et al. 2017;</ns0:ref><ns0:ref type='bibr' target='#b15'>Caceres et al. 2017;</ns0:ref><ns0:ref type='bibr' target='#b107'>Naylor et al. 2019)</ns0:ref>.Occludin and claudins determine the permeability and semi-selectivity of the TJs, and as such play critical roles in the oBRB <ns0:ref type='bibr' target='#b3'>(Balda et al. 2000;</ns0:ref><ns0:ref type='bibr' target='#b42'>Fields et al. 2019;</ns0:ref><ns0:ref type='bibr' target='#b44'>Furuse et al. 1998;</ns0:ref><ns0:ref type='bibr' target='#b53'>Günzel & Yu 2013;</ns0:ref><ns0:ref type='bibr' target='#b120'>Rosenthal et al. 2017)</ns0:ref>. JAMs regulate TJ assembly and function by recruiting other proteins to the TJ and play an important role in the barrier property of TJs <ns0:ref type='bibr' target='#b5'>(Balda & Matter 2016;</ns0:ref><ns0:ref type='bibr' target='#b111'>Orlova et al. 2006;</ns0:ref><ns0:ref type='bibr' target='#b124'>Shin et al. 2006</ns0:ref>). In patients with RRD, damage to TJs elicits the breakdown of oBRB and promotes the penetration of growth factors and cytokines, aggravating PVR. As well as having a barrier function, TJs define the physical separation between apical and basal domains of the plasma membrane, to maintain RPE cell polarity <ns0:ref type='bibr' target='#b19'>(Campbell et al. 2017;</ns0:ref><ns0:ref type='bibr' target='#b48'>González-Mariscal et al. 2014;</ns0:ref><ns0:ref type='bibr' target='#b129'>Sluysmans et al. 2017)</ns0:ref>. The two extracellular loops of occludin mediate adhesion of adjacent cells and block the movement of plasma components. The C-terminal domain combines directly with ZOs, subsequently interacting with the actin cytoskeleton, which is essential to organizing and maintaining cell polarization <ns0:ref type='bibr' target='#b5'>(Balda & Matter 2016;</ns0:ref><ns0:ref type='bibr' target='#b43'>Furuse et al. 1994;</ns0:ref><ns0:ref type='bibr' target='#b124'>Shin et al. 2006;</ns0:ref><ns0:ref type='bibr' target='#b139'>Tarau et al. 2019)</ns0:ref>. <ns0:ref type='bibr' target='#b40'>Feng et al. (2019)</ns0:ref> demonstrated that during EMT, the breakdown of TJs resulting from loss of claudin-1 causes ARPE-19 cells to lose their epithelial phenotype and transform into fibroblasts, promoting the development of PVR. TJs are involved in the regulation of signaling pathways that govern various cellular functions such as proliferation, migration, and differentiation <ns0:ref type='bibr' target='#b11'>(Bhat et al. 2018;</ns0:ref><ns0:ref type='bibr' target='#b123'>Shi et al. 2018;</ns0:ref><ns0:ref type='bibr' target='#b129'>Sluysmans et al. 2017)</ns0:ref>. <ns0:ref type='bibr' target='#b144'>Vietor et al. (2001)</ns0:ref> found that decreased amounts of occludin can cause up-regulation and translocation of the adhesion junction protein β-catenin, which interacts with the transcription factor lymphoid enhancer-binding factor (LEF)/T cell factor (TCF) in the nucleus, leading to a loss of the polarized epithelial phenotype in EpH4 cells. ZOs, adaptor proteins within the TJ complex, exhibit dual localization at TJs and in the nucleus. Under injury or stress, the disruption of TJs increases ZO-2 nuclear accumulation, driving its interaction with transcription factors, and inducing MDCK epithelial cell proliferation <ns0:ref type='bibr' target='#b65'>(Islas et al. 2002;</ns0:ref><ns0:ref type='bibr' target='#b123'>Shi et al. 2018;</ns0:ref><ns0:ref type='bibr' target='#b142'>Traweger et al. 2003)</ns0:ref>. In differentiated RPE cells, the interaction between ZO-1 with ZO-1-associated nucleic acid-binding protein (ZONAB) maintains cell-cell contact by sequestering ZONAB at the TJ or in the cytoplasm, maintaining cells dormancy. However, when damage to TJs decreases ZO-1 levels, ZONAB is translocated into the nucleus, leading to the up-regulation of cyclin D1 (CD1) and subsequent cell proliferation <ns0:ref type='bibr' target='#b4'>(Balda et al. 2003;</ns0:ref><ns0:ref type='bibr' target='#b48'>González-Mariscal et al. 2014)</ns0:ref>. Therefore, TJs provide a structural foundation for the maintenance of cell-cell contact. <ns0:ref type='bibr' target='#b47'>Georgiadis et al. (2010)</ns0:ref> demonstrated that the overexpression of ZONAB or knockdown of ZO-1 could result in increased RPE proliferation and the development of EMT. Recent research has confirmed that during EMT, ZO-1 is decreased in ARPE-19 cells, and the knockdown of either ZO-1 or AJ protein E-cadherin leads to the downregulation of the other protein, indicating the existence of an interaction between the two junctional complexes <ns0:ref type='bibr' target='#b6'>(Bao et al. 2019)</ns0:ref>. Due to the importance of TJs in the maintenance of integrity and functionality of epithelial cells, several researchers have focused on novel factors that stimulate the formation of TJs, such as nicotinamide <ns0:ref type='bibr' target='#b56'>(Hazim et al. 2019</ns0:ref>) and lysophosphatidic acid <ns0:ref type='bibr' target='#b86'>(Lidgerwood et al. 2018)</ns0:ref>. Studies into these factors may produce well-differentiated RPE cell lines and a platform to enable the rapid expansion of our understanding of many RPE functions and retinal pathologies. This approach could be conducive to finding novel therapeutic interventions for PVR.</ns0:p><ns0:p>Besides the TJ complex described above, another type of junctional complex called AJs plays a key role in the maintenance of the integrity of epithelial cells and cell-cell contact (Fig <ns0:ref type='figure' target='#fig_1'>1</ns0:ref>). Cadherins, the major proteins of AJs, belong to the glycoprotein superfamily, of which there are more than 20 members. The cytoplasmic domain of cadherins regulates interactions between cadherins and catenins, including β-catenin, α-catenin, and p120-catenin, and other scaffolding proteins such as ZO-1, to maintain cell shape and modulate cell proliferation <ns0:ref type='bibr' target='#b0'>(Aberle et al. 1994;</ns0:ref><ns0:ref type='bibr' target='#b108'>Nelson & Nusse 2004;</ns0:ref><ns0:ref type='bibr' target='#b151'>Wheelock & Johnson 2003)</ns0:ref>. In quiescent adult RPE cells, epithelial cadherins (E-and/or P-cadherin) sequester β-catenin at the AJs to maintain cell-cell contact. Reduction of cadherin levels or dissociation of AJs allows β-catenin to translocate into the nucleus, where it interacts with the transcription factor LEF, and activates the transcription of various genes, including Snail and cyclin D1, which participate in RPE cell EMT via the canonical Wnt/β-catenin signaling pathway <ns0:ref type='bibr' target='#b49'>(Gonzalez & Medici 2014;</ns0:ref><ns0:ref type='bibr' target='#b77'>Lamouille et al. 2014;</ns0:ref><ns0:ref type='bibr' target='#b108'>Nelson & Nusse 2004;</ns0:ref><ns0:ref type='bibr' target='#b158'>Yang et al. 2018)</ns0:ref> . <ns0:ref type='bibr' target='#b137'>Tamiya et al. (2010)</ns0:ref> suggested that the loss of Pcadherin causes the loss of cell-cell contact and initiates RPE cell migration and EMT. These events coincide with a switch in cadherin isoform expression from P-to N-cadherin. In addition, hepatocyte growth factor (HGF) and its receptor c-Met can destabilize cell-cell adhesion and elicit nuclear translocation of β-catenin, resulting in RPE cell migration <ns0:ref type='bibr' target='#b87'>(Lilien & Balsamo 2005;</ns0:ref><ns0:ref type='bibr' target='#b90'>Liou et al. 2002)</ns0:ref>. Jin et al found that HGF induces loss or redistribution of junctional proteins ZO-1, occludin, and β-catenin in RPE explants, potentially damaging barrier function and increasing the migration of RPE cells, resulting in retinal detachment(RD) and PVR <ns0:ref type='bibr' target='#b67'>(Jin et al. 2002;</ns0:ref><ns0:ref type='bibr' target='#b68'>Jin et al. 2004</ns0:ref>). Given the importance of HGF in the interruption of RPE junction, HGF may be a potential target for the prevention and treatment of PVR. However, this possibility needs further study.</ns0:p><ns0:p>Under physiological conditions in the eye, TJs and AJs maintain the specialized structural and functional polarity of RPE cells and play a pivotal role in the maintenance of cell-cell contact; they sequester EMT signaling effectors ZONAB and β-catenin at the junction or cytoplasm to prevent cells from responding to mitotic factors, causing cells to leave the cellcycle (Fig <ns0:ref type='figure' target='#fig_1'>1</ns0:ref>). Thus, normally, RPE cells form a cobblestone-like monolayer of immotile, polarized, and mitotically quiescent cells. However, once junctional complexes break down, RPE cells undergo EMT, which is an important contributor to proliferative vitreoretinopathy. In this pathological process, RPE cells lose their structural and functional polarity and transdifferentiate into mesenchymal cells, which proliferate, resist apoptosis, possess migratory ability, and produce abundant ECM, leading to the formation of an aberrant scar-like fibrocellular membrane.</ns0:p></ns0:div>
<ns0:div><ns0:head n='2.'>De-differentiated RPE and Fibrocellular Membrane</ns0:head><ns0:p>Proliferative vitreoretinopathy is characterized by the formation of fibrocellular membranes composed of proliferative and migratory cells and excessive, aberrant ECM. Histopathological analysis of PVR has demonstrated that PVR membranes have contractile activity and strain the retina, leading to tractional retinal detachment (TRD), which is responsible for blurring vision.</ns0:p><ns0:p>Several studies <ns0:ref type='bibr' target='#b39'>(Feist et al. 2014;</ns0:ref><ns0:ref type='bibr' target='#b135'>Takahashi et al. 2010</ns0:ref>) have found that the cellular components of PVR membranes include RPE cells, myofibroblasts, fibroblasts, glial cells and macrophages, and that myofibroblasts are critical for the formation and contractile activity of fibrocellular membranes. Based on the indirect immunofluorescence evaluation of human PVR membranes, <ns0:ref type='bibr' target='#b39'>Feist et al. (2014)</ns0:ref> showed that myofibroblasts originate principally from RPE cells through EMT. Myofibroblasts are characterized by increased expression of alpha-smooth muscle actin (α-SMA) and incorporation of α-SMA into newly formed actin stress fibers, which enhances their contractile properties. Myofibroblasts also secrete excessive matrix and profibrogenic factors, promoting the contraction of PVR membranes that ultimately cause irreversible loss of vision <ns0:ref type='bibr' target='#b45'>(Gamulescu et al. 2006;</ns0:ref><ns0:ref type='bibr' target='#b60'>Hinz et al. 2001;</ns0:ref><ns0:ref type='bibr' target='#b126'>Shu & Lovicu 2017;</ns0:ref><ns0:ref type='bibr' target='#b136'>Tamiya & Kaplan 2016;</ns0:ref><ns0:ref type='bibr' target='#b140'>Tomasek et al. 2002)</ns0:ref>.</ns0:p><ns0:p>In addition to myofibroblasts, abnormally increased ECM reinforces the continuous contractile tension of PVR membranes, and this mechanical tension, together with specialized ECM proteins, regulates myofibroblast differentiation and its function, contributing to PVR. In PVR membranes, the primary components of ECM are collagen and fibronectin. The majority of collagen fibrils are type Ⅰ collagen, which is synthesized by RPE cells and Müller cells.</ns0:p><ns0:p>Collagen fibrils provide tensile strength to the ECM, and activate Rho, resulting in the translocation of myocardin-related transcription factor (MRTF) into the nucleus and promoting RPE cell EMT <ns0:ref type='bibr' target='#b52'>(Guettler et al. 2008;</ns0:ref><ns0:ref type='bibr' target='#b102'>Miralles et al. 2003)</ns0:ref>. Fibronectin may also play a significant role in PVR. During pathological ECM remodeling, fibronectin is one of the earliest ECM components recruited, serving as a scaffold for other ECM proteins <ns0:ref type='bibr' target='#b70'>(Kadler et al. 2008;</ns0:ref><ns0:ref type='bibr'>Miller et al. 2017;</ns0:ref><ns0:ref type='bibr' target='#b101'>Miller et al. 2014)</ns0:ref>. Extra domain (ED)-A fibronectin, a splice variant of fibronectin, is increased in TGF-β2-induced RPE cells and induces myofibroblast differentiation, participating in PVR <ns0:ref type='bibr' target='#b74'>(Khankan et al. 2011)</ns0:ref>.</ns0:p><ns0:p>Under normal conditions, ECM breakdown by proteases such as matrix-metalloproteases (MMPs) plays a crucial role in ECM remodeling and the release of growth factors, maintaining tissue homeostasis in cooperation with ECM synthesis, reassembly, and chemical modification <ns0:ref type='bibr' target='#b13'>(Bonnans et al. 2014;</ns0:ref><ns0:ref type='bibr' target='#b31'>Craig et al. 2015;</ns0:ref><ns0:ref type='bibr' target='#b89'>Lindsey et al. 2016</ns0:ref>). As mentioned above, the polarized RPE is able to basolaterally secrete the extracellular matrix components fibronectin and collagens, MMP and tissue inhibitors of MMPs (TIMPs), which participate in ECM remodeling. However, under pathological conditions such as inflammation and retinal injury, RPE cells lose their apical-basal polarity, undergo EMT and abnormally secrete MMPs, TIMPs and ECM proteins, leading to dysregulated ECM remodeling <ns0:ref type='bibr' target='#b51'>(Greene et al. 2017)</ns0:ref>. Such ECM has aberrant composition and organization and mechanical properties, and enhances matrix stiffness and strain, which disrupts the normal structure and function of the retina, exacerbating the progression of PVR.</ns0:p></ns0:div>
<ns0:div><ns0:head n='3.'>RPE and Epithelial-mesenchymal Transition</ns0:head></ns0:div>
<ns0:div><ns0:head n='3.1'>EMT of RPE Cell</ns0:head><ns0:p>Epithelial-mesenchymal transition is an important biological process, in which epithelial cells transdifferentiate into mesenchymal cells. Although EMT can occur in normal embryonic development and wound healing, it also participates in pathological processes such as fibrosis, cancer progression, and PVR. There are three distinct subtypes of EMT: type 1 occurs during tissue and embryo development, type 2 is involved in wound healing and organ fibrosis, and type 3 is associated with cancer progression and metastasis <ns0:ref type='bibr' target='#b36'>(Dongre & Weinberg 2019;</ns0:ref><ns0:ref type='bibr' target='#b71'>Kalluri & Weinberg 2009)</ns0:ref>. This review focuses on type 2 EMT, which is crucial to PVR. During EMT, due to junctional complexes damage, RPE cells relinquish their apical-basal polarity, reorganize their cytoskeletal architecture, and convert into spindle-shaped cells (Fig <ns0:ref type='figure' target='#fig_1'>1</ns0:ref>). These cells downregulate the expression of epithelial proteins such as E-cadherin and ZO-1, and increase expression of mesenchymal drivers including N-cadherin, vimentin, α-SMA and fibronectin <ns0:ref type='bibr'>(Li et al. 2020)</ns0:ref>. This mesenchymal transdifferentiation of RPE cells can increase the directional Manuscript to be reviewed motility of individual cells, confer resistance to apoptosis, and facilitate cell proliferation and dysregulated ECM remodeling, eventually leading to the formation of PVR membranes.</ns0:p></ns0:div>
<ns0:div><ns0:head n='3.2'>Transcription Factors of EMT</ns0:head><ns0:p>The details of the molecular mechanisms that drive RPE cell EMT and lead to PVR remain to be clarified. Emerging evidence suggests that diverse extracellular inductive signals, including soluble cytokines and growth factors, and ECM components, can modulate the expression and activity of EMTassociated transcription factors and act together to control the initiation and progression of EMT in responding epithelial cells <ns0:ref type='bibr' target='#b156'>(Yang et al. 2020</ns0:ref>). Among the various transcription factors involved in the induction of EMT, core transcription factors including Snail 1, Snail 2(also known as Slug), Twist 1 and zinc-finger E-box-binding (Zeb) 1 have been identified as important regulators of RPE cell EMT. These factors impact the expression of genes that control repression of the epithelial phenotype and activation of the mesenchymal phenotype <ns0:ref type='bibr' target='#b12'>(Boles et al. 2020;</ns0:ref><ns0:ref type='bibr' target='#b40'>Feng et al. 2019;</ns0:ref><ns0:ref type='bibr' target='#b83'>Li et al. 2019;</ns0:ref><ns0:ref type='bibr' target='#b84'>Li et al. 2014;</ns0:ref><ns0:ref type='bibr' target='#b92'>Liu et al. 2009;</ns0:ref><ns0:ref type='bibr' target='#b113'>Palma-Nicolás & López-Colomé 2013)</ns0:ref>. For example, thrombin can repress the expression of E-cadherin by stimulating Snail 2 expression and promote the expression of N-cadherin by phosphoinositide 3kinase (PI3K)/PKC-ζ/mTOR signaling in Rat RPE cells (Palma-Nicolás & López-Colomé 2013). During RPE dedifferentiation in primary culture, Zeb1 is overexpressed and binds to the MITF A promoter to repress the cyclin dependent kinase inhibitor, p21CDKN1a, resulting in RPE cell proliferation and EMT <ns0:ref type='bibr' target='#b92'>(Liu et al. 2009</ns0:ref>). These EMT transcription factors often act in concert, functionally cooperating at target genes by the convergence of signaling pathways. However, the molecular details of how these transcription factors contribute to EMT are still elusive <ns0:ref type='bibr' target='#b77'>(Lamouille et al. 2014;</ns0:ref><ns0:ref type='bibr' target='#b132'>Stone et al. 2016)</ns0:ref>.</ns0:p></ns0:div>
<ns0:div><ns0:head n='3.3'>Epigenetic Factors of EMT</ns0:head><ns0:p>Due to the importance of epigenetic regulation of EMT, epigenetic modifiers have attracted increasing attention. Evidence has shown that epigenetic modifiers work in concert with transcription factors at different molecular layers to regulate the EMT process <ns0:ref type='bibr' target='#b128'>(Skrypek et al. 2017)</ns0:ref>. Several epigenetic factors have been described including DNA methylation, histone modification and non-coding RNA. Because of the specific machinery utilized for EMT activation, these modifications are characterized by cell type specificity. In RPE cells, Methyl-CpG-binding protein 2 (MeCP2), a DNA methylation reader, plays a crucial role in the induction of EMT, and DNA methylation may participate in the pathogenesis of PVR <ns0:ref type='bibr' target='#b58'>(He et al. 2015;</ns0:ref><ns0:ref type='bibr'>Li et al. 2020)</ns0:ref>. <ns0:ref type='bibr' target='#b58'>He et al. (2015)</ns0:ref> found high levels of expression of MeCP2 in all human PVR membranes, and concluded that MeCP2 mediates α-SMA expression through Ras GTPase activating protein (RASAL1). Furthermore, DNA methylation inhibitor 5-Aza-2' deoxycytidine (5-AZA-dC) reportedly inhibits the expression of TGF-β-induced α-SMA and FN in human fetal RPE cells. It appears that 5-AZA-dC may have therapeutic value in the treatment of PVR. However, the mechanisms underlying the blockade of α-SMA and FN expression are complex, and further investigation is warranted.</ns0:p><ns0:p>Recently, the role of histone modifications associated with EMT has been assessed in RPE cells. However, there has been little research into the regulation of RPE cell EMT by histone modification. <ns0:ref type='bibr' target='#b12'>Boles et al. (2020)</ns0:ref> reported that TGF-β1 and TNF-a co-treatment (TNT) induces an EMT program in adult human RPE stem cell (RPESC)-RPE cells, involving an apparent reorganization of H3K27ac and H3K4me1 patterns at distal enhancers. The regions that gain H3K27ac tend to have a high H3K4me1/H3K4me3 ratio, indicating that they have enhancer activity and are associated with upregulated genes. <ns0:ref type='bibr' target='#b155'>Xiao et al. (2014)</ns0:ref> found that the expression of histone deacetylases (HDACs) in TGF-β-induced EMT of RPE cells was increased, and that Trichostatin A (TSA), a class I and II HDAC inhibitor, attenuated TGF-β2-induced EMT by inhibiting the canonical SMAD pathway and the non-canonical signaling pathways, including Akt, p38MAPK, ERK1/2 pathways and Notch pathway. Therefore, histone modifications may participate in the regulation of RPE cell EMT, and HDAC inhibitors may have potential as drugs for the prevention and treatment of PVR.</ns0:p><ns0:p>The study of EMT mechanisms at the RNA level has provided new perspectives on the treatment of PVR <ns0:ref type='bibr' target='#b72'>(Kaneko & Terasaki 2017;</ns0:ref><ns0:ref type='bibr' target='#b147'>Wang et al. 2016)</ns0:ref>. MicroRNAs (miRNAs) are small noncoding RNAs that contribute to cellular processes by regulating gene expression. In differentiated RPE cells, microRNA-204 is highly expressed, and represses the expression of type Ⅱ TGF-β receptors and Snail 2, maintaining epithelial structure and function. In contrast, low expression levels of miR-204 and anti-miR-204 promote RPE cells proliferation, participating in EMT <ns0:ref type='bibr' target='#b146'>(Wang et al. 2010)</ns0:ref>. MicroRNA-194 overexpression can also suppress RPE cell EMT by attenuating the expression of Zeb1 <ns0:ref type='bibr' target='#b32'>(Cui et al. 2019</ns0:ref>). In addition to miRNAs, long non-coding RNAs (lncRNAs) contribute to the regulation of RPE EMT <ns0:ref type='bibr' target='#b164'>(Zhang et al. 2019)</ns0:ref>. In RPE cells treated with PVR vitreous or TGF-β1, MALAT1 expression is increased, and knockdown of MALAT1 attenuates the phosphorylation of SMAD2/3 and the expression of Snail, Slug, and Zeb1, preventing cell migration and proliferation <ns0:ref type='bibr' target='#b157'>(Yang et al. 2016</ns0:ref>). In patients with PVR, MALAT1 is increased in the blood, and is reduced after surgery. Thus, MALAT1 may be a potential prognostic and diagnostic indicator for PVR <ns0:ref type='bibr' target='#b169'>(Zhou et al. 2015)</ns0:ref>.</ns0:p></ns0:div>
<ns0:div><ns0:head n='3.4'>Signaling Pathways of EMT</ns0:head><ns0:p>During RPE cell EMT, extracellular signals change the expression of genes encoding epithelial and mesenchymal proteins and mediate cellular behavior such as cell migration, proliferation, and apoptosis through a network of interacting signaling pathways that contribute to the development of PVR <ns0:ref type='bibr' target='#b23'>(Chen et al. 2014a;</ns0:ref><ns0:ref type='bibr' target='#b24'>Chen et al. 2014b;</ns0:ref><ns0:ref type='bibr' target='#b78'>Lee-Rivera et al. 2015)</ns0:ref>. Among these, transforming growth factor-β (TGF-β) and its intracellular cascades play a key role in the EMT of RPE cells.</ns0:p><ns0:p>TGF-β induces EMT of RPE cells via two pathways: the classical SMAD-dependent pathway and the SMAD-independent pathway (Fig <ns0:ref type='figure' target='#fig_2'>2</ns0:ref>) <ns0:ref type='bibr' target='#b18'>(Cai et al. 2018;</ns0:ref><ns0:ref type='bibr' target='#b57'>He et al. 2017;</ns0:ref><ns0:ref type='bibr' target='#b59'>Heffer et al. 2019;</ns0:ref><ns0:ref type='bibr' target='#b64'>Ishikawa et al. 2015;</ns0:ref><ns0:ref type='bibr' target='#b134'>Takahashi et al. 2015;</ns0:ref><ns0:ref type='bibr' target='#b159'>Yao et al. 2019;</ns0:ref><ns0:ref type='bibr' target='#b163'>Zhang et al. 2017;</ns0:ref><ns0:ref type='bibr' target='#b165'>Zhang et al. 2018b;</ns0:ref><ns0:ref type='bibr' target='#b171'>Zhou et al. 2017</ns0:ref>). In the SMAD dependent pathway, TGF-β binds to cell surface receptor complexes, and activates type Ⅰ TGF-β receptors, which phosphorylate SMAD2 and SMAD3. The activated SMADs combine with SMAD4 to form a SMAD complex, which then enters the nucleus and combines with regulatory elements to regulate the expression of key genes associated with EMT. In addition to SMAD-dependent signaling, TGFβ induces EMT through SMAD independent signaling pathways including Rho GTPase-dependent pathways <ns0:ref type='bibr' target='#b80'>(Lee et al. 2008)</ns0:ref>, PI3K/Akt pathway <ns0:ref type='bibr' target='#b62'>(Huang et al. 2017;</ns0:ref><ns0:ref type='bibr' target='#b160'>Yokoyama et al. 2012)</ns0:ref>, mitogen-activated kinase (MAPK) pathways <ns0:ref type='bibr' target='#b26'>(Chen et al. 2017;</ns0:ref><ns0:ref type='bibr' target='#b79'>Lee et al. 2020;</ns0:ref><ns0:ref type='bibr' target='#b98'>Matoba et al. 2017;</ns0:ref><ns0:ref type='bibr' target='#b122'>Schiff et al. 2019)</ns0:ref> and Jagged/Notch signaling pathway <ns0:ref type='bibr' target='#b163'>(Zhang et al. 2017)</ns0:ref>. The MAPK signaling pathways include extracellular signal-regulated kinase(ERK) MAPK pathway, p38 MAPK pathway, and JUN Nterminal kinase (JNK) pathway <ns0:ref type='bibr' target='#b114'>(Parrales et al. 2013;</ns0:ref><ns0:ref type='bibr' target='#b122'>Schiff et al. 2019;</ns0:ref><ns0:ref type='bibr' target='#b155'>Xiao et al. 2014;</ns0:ref><ns0:ref type='bibr' target='#b162'>Zhang et al. 2018a</ns0:ref>).</ns0:p><ns0:p>The Rho pathway has been reported to regulate the assembly and organization of the actin cytoskeleton and associated gene expression, and may be essential for the fibrotic response of RPE cells in PVR. In TGF-β1-treated ARPE-19 cells, activated RhoA or its downstream effector Rho kinase (ROCK) increase the kinase activity of LIM kinase (LIMK) which then phosphorylates cofilin. This phosphorylation attenuates the activity of cofilin, which promotes actin polymerization and reorganizes the actin cytoskeleton, leading to stress fiber formation <ns0:ref type='bibr' target='#b80'>(Lee et al. 2008)</ns0:ref>. TGF-β-induced RhoA activation also facilitates cell migration and increases α-SMA expression in primary RPE cells <ns0:ref type='bibr' target='#b143'>(Tsapara et al. 2010)</ns0:ref>. <ns0:ref type='bibr' target='#b66'>Itoh et al. (2007)</ns0:ref> demonstrated that ROCK inhibitor Y27632 and RhoA inhibitor, simvastatin, suppress TGF-β2-induced type Ⅰ collagen expression in ARPE-19 cells, and confirmed the existence of crosstalk between the SMAD pathway and the Rho pathway. Some studies have suggested that activated SMAD3 induces NET1 gene expression to regulate RhoA activation in RPE cells <ns0:ref type='bibr' target='#b81'>(Lee et al. 2010)</ns0:ref>. Moreover, thrombin can activate Rho and ROCK, leading to myosin light chain (MLC) phosphorylation and actin stress fiber formation in EMT of RPE cells (Fig <ns0:ref type='figure'>3</ns0:ref>) <ns0:ref type='bibr' target='#b121'>(Ruiz-Loredo et al. 2011)</ns0:ref>. Therefore, ROCK inhibitor and RhoA inhibitor may be new potential therapeutic target drugs for PVR.</ns0:p><ns0:p>The PI3K/Akt pathway mediates a broad range of cellular functions, such as cell transformation, migration, proliferation, apoptosis, and gene expression <ns0:ref type='bibr' target='#b1'>(Aguilar-Solis et al. 2017;</ns0:ref><ns0:ref type='bibr' target='#b91'>Liu et al. 2019)</ns0:ref>. During PVR, binding of TGF-β to its receptor activates PI3K, resulting in the phosphorylation of Akt; activated Akt inhibits glycogen synthase kinase 3β (GSK-3β), promoting EMT in RPE cells <ns0:ref type='bibr' target='#b127'>(Shukal et al. 2020;</ns0:ref><ns0:ref type='bibr' target='#b162'>Zhang et al. 2018a)</ns0:ref>. Researchers have found that inhibition or knockdown of GSK-3β promotes cell migration and collagen contraction in ARPE-19 cells, while GSK-3β overexpression and PI3K/Akt inhibitor reverse these cellular responses <ns0:ref type='bibr' target='#b62'>(Huang et al. 2017)</ns0:ref>. Some studies have shown that thrombin can activate PI3K, resulting in increased cyclin D1 expression and RPE cell proliferation, processes that are involved in the development of PVR through PDK1/Akt and PKCζ/mTORC signaling (Fig <ns0:ref type='figure'>3</ns0:ref>) <ns0:ref type='bibr' target='#b78'>(Lee-Rivera et al. 2015;</ns0:ref><ns0:ref type='bibr' target='#b113'>Palma-Nicolás & López-Colomé 2013;</ns0:ref><ns0:ref type='bibr' target='#b114'>Parrales et al. 2013</ns0:ref>).</ns0:p><ns0:p>In addition to the PI3K-AKT pathway, other kinase pathways contribute to EMT in cooperation with the SMAD-dependent signaling pathways. In human RPE cells, TGF-β activates TGF-β-activated kinase 1 (TAK1), which subsequently transduces signals to several downstream effectors, including p38 <ns0:ref type='bibr' target='#b59'>(Heffer et al. 2019)</ns0:ref>, JNK <ns0:ref type='bibr' target='#b76'>(Kimura et al. 2015)</ns0:ref> and nuclear factor-κB (NF-κB) <ns0:ref type='bibr' target='#b25'>(Chen et al. 2016b)</ns0:ref>, which participate in EMT. <ns0:ref type='bibr' target='#b37'>Dvashi et al. (2015)</ns0:ref> found that TAK1 inhibitor caused a reduction in both p38 and SMAD2/3 activity, attenuating cell migration, cell contractility and α-SMA expression in TGF-β1-induced RPE cells. Moreover, the ERK MAPK pathway plays a role in TGF-β-induced EMT and cooperates with other signaling pathways in the regulation of EMT in RPE cells. Recent studies <ns0:ref type='bibr' target='#b24'>(Chen et al. 2014b;</ns0:ref><ns0:ref type='bibr' target='#b138'>Tan et al. 2017;</ns0:ref><ns0:ref type='bibr' target='#b155'>Xiao et al. 2014)</ns0:ref> have shown that blocking the ERK1/2 pathway inhibits the phosphorylation of SMAD2 and the Jagged/Notch pathway. Inhibition of the Jagged/Notch signaling pathway can alleviate TGF-β2-induced EMT by regulating the expression of Snail, <ns0:ref type='bibr'>Slug and Zeb1 (Fig 3)</ns0:ref>; this also suppresses the ERK1/2 signaling <ns0:ref type='bibr' target='#b24'>(Chen et al. 2014b</ns0:ref>).</ns0:p><ns0:p>The contribution of growth factors other than TGF-β, such as HGF, fibroblast growth factor (FGF), epidermal growth factor (EGF) and platelet derived growth factor (PDGF) should also be factored in with regard to the induction of RPE EMT. These factors bind to and stimulate the autophosphorylation of transmembrane receptors on Tyr, subsequently participating in RPE cell EMT via PI3K/Akt pathway, ERK MAPK pathway, p38 MAPK pathway (Fig <ns0:ref type='figure'>3</ns0:ref>) <ns0:ref type='bibr' target='#b22'>(Chen et al. 2016a;</ns0:ref><ns0:ref type='bibr' target='#b112'>Ozal et al. 2020)</ns0:ref>. <ns0:ref type='bibr' target='#b21'>Chen et al. (2012)</ns0:ref> explored the role of Wnt/β-catenin signaling in PVR, and found that when EGTA disrupted contact inhibition in RPE cells, EGF+FGF2 could activate Wnt signaling and increase nuclear levels of β-catenin, which interacts with TCF and/or LEF, leading to cell proliferation (Fig 3 <ns0:ref type='figure'>)</ns0:ref>; and EGF+FGF2 cooperated with TGF-β1 to induce EMT through SMAD/Zeb1/2 signaling. Acting together, various inductive signals received by RPE cells from their niche can trigger the activation of EMT programs by individual intracellular cascades or the crosstalk of multiple intracellular signaling pathways.</ns0:p></ns0:div>
<ns0:div><ns0:head n='3.5'>Interventions of RPE EMT</ns0:head><ns0:p>Therapeutic interventions against RPE EMT have largely been explored in mechanistic experiments using in vitro cell culture and in vivo animal models. To date, some promising drug candidates have been trialed in preclinical studies of PVR, including TGF-β receptor inhibitors, peroxisome proliferator-activated receptor (PPAR)-γ agonists, retinoic acid receptor-γ (RAR-γ) agonists and methotrexate <ns0:ref type='bibr' target='#b125'>(Shu et al. 2020;</ns0:ref><ns0:ref type='bibr' target='#b168'>Zhou et al. 2020)</ns0:ref>. <ns0:ref type='bibr' target='#b106'>Nassar et al. (2014)</ns0:ref> found that TGF-β receptor 1 inhibitor LY-364947 (LY) attenuates RPE cell transdifferentiation in vitro, and that intravitreal injection of LY completely prevents PVR and TRD in vivo. Evidence is emerging to show that the up-regulation of PPAR-γ expression may be beneficial for the treatment of fibrosis in several organs <ns0:ref type='bibr' target='#b149'>(Wang et al. 2019)</ns0:ref>. <ns0:ref type='bibr' target='#b55'>Hatanaka et al. (2012)</ns0:ref> reported that PPAR-γ agonist pioglitazone could prevent TGF-β-induced morphological changes and the upregulation of EMT-related markers in primary monkey RPE cells, through inhibition of the SMAD pathway. Some drugs, including dichloroacetate (DCA) <ns0:ref type='bibr' target='#b127'>(Shukal et al. 2020)</ns0:ref>, salinomycin (SNC) <ns0:ref type='bibr' target='#b59'>(Heffer et al. 2019)</ns0:ref>, resveratrol <ns0:ref type='bibr' target='#b64'>(Ishikawa et al. 2015)</ns0:ref>, protein kinase A inhibitor H89 <ns0:ref type='bibr' target='#b94'>(Lyu et al. 2020</ns0:ref>) and heavy chain-hyaluronan/pentraxin3 <ns0:ref type='bibr' target='#b57'>(He et al. 2017)</ns0:ref>, reportedly inhibit EMT in an in vitro EMT cell model and prevent PVR development by blocking the activation of theTGF-β pathway. Thus, inhibition of EMT by pharmacological agents may be an effective strategy to prevent PVR development.</ns0:p></ns0:div>
<ns0:div><ns0:head>Conclusion</ns0:head><ns0:p>Clinical and experimental studies have shown that RPE cells play an important role in PVR. Junctional complexes are crucial for the maintenance of RPE polarity. Under the influence of growth factors and cytokines, RPE cells lose cell-cell contact and apical-basal polarity, and undergo EMT via multiple signaling pathways, which promote cell proliferation, migration, and ECM production. RPE cells further transform into myofibroblasts and form fibrocellular membranes that have contractile activity and strain the retina, leading to tractional retinal detachment in PVR. As a complex refractory blinding disorder, PVR involves multiple signaling pathways and factors. In addition, the specialized polarity of RPE cells is fundamental for retinal homeostasis, and RPE EMT plays a key role in the development of PVR. Nevertheless, further research into the mechanisms underlying RPE polarity and EMT is needed to prevent this devastating complication. A deeper understanding of RPE polarization is fundamental for elucidating the mechanism of EMT initiation and progression, and is essential to exploring the potential pharmacologic prophylactic and therapeutic approaches to PVR. Various factors, such as microenvironmental signals, transcription factors, and epigenetic factors, participate in the regulation of EMT at different molecular levels. Further studies about the detailed molecular mechanisms of EMT are needed to facilitate the development of therapeutic strategies for PVR. </ns0:p></ns0:div><ns0:figure xml:id='fig_0'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:06:49749:2:0:NEW 15 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_1'><ns0:head>Figure 1</ns0:head><ns0:label>1</ns0:label><ns0:figDesc>Figure 1</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_2'><ns0:head>Figure 2</ns0:head><ns0:label>2</ns0:label><ns0:figDesc>Figure 2</ns0:figDesc></ns0:figure>
<ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='27,42.52,70.87,525.00,386.25' type='bitmap' /></ns0:figure>
<ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='30,42.52,70.87,525.00,369.75' type='bitmap' /></ns0:figure>
</ns0:body>
" | "Dear Prof. Pfeffer,
Please find attached a revised version of our manuscript “Polarity and Epithelial-mesenchymal Transition of Retinal Pigment Epithelial Cells in Proliferative Vitreoretinopathy” (49749), which we would like to resubmit for publication as a literature review in PeerJ.
I am very grateful for your comments regarding the manuscript. Your comments and those of the reviewers were highly insightful and enabled us to greatly improve the quality of our manuscript. In the following pages are our point-by-point responses to each of the comments of the reviewers.
Revisions in the manuscript are shown using red text for additions and the strikethrough feature for deletions. According to your advice, we amended the relevant parts in the manuscript. We hope that the revisions in the manuscript and our accompanying responses will be sufficient to render our manuscript suitable for publication in PeerJ.
Thank you and all the reviewers for the kind advice.
Sincerely,
Jinsong Zhao
Institute: Eye Center of the second hospital, Jilin University, Changchun city, Jilin Province, China.
Address: Eye Center of the second hospital,
Jilin University,
4026 Yatai Street,
Changchun, Jilin, 130021,
China;
E-mail：jinsongzhao2003@163.com.
Tel：86-0431-81136523
Responses to the comments of Reviewer 2 (Shikun He)
Comments for the author
The authors answered all questions raised by reviewers, the manuscript was improved a lot.
However, there is one concept which need to be corrected, in line 296, the authors mentioned “Unlike transcription factors, epigenetic modifications are more stable/long-term”. the statement is not right；in fact, Epigenetic modifications are highly dynamic process in developmental biology and pathogenesis of diseases.
Response: We agree with the reviewer that epigenetic modifications are highly dynamic process in developmental biology and pathogenesis of diseases. Accordingly, we have modified the expression to address the reviewer’s point in the revised manuscript (line 304-308) as follows: Due to the importance of epigenetic regulation of EMT, epigenetic modifiers have attracted increasing attention. Evidence has shown that epigenetic modifiers work in concert with transcription factors at different molecular layers to regulate the EMT process (Skrypek et al. 2017). Several epigenetic factors have been described including DNA methylation, histone modification and non-coding RNA.
" | Here is a paper. Please give your review comments after reading it. |
9,781 | "<ns0:abstract xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:p>Due to varietal differences, diminutive size, and similar morphological characters, it is difficult to classify and identify Liriomyza spp., a genus comprised of economicallyimportant, highly-polyphagous insect pests. In this study, we reconfirmed the morphological characteristics of three closely-related invasive leafminers, L. trifolii, L. sativae, and L. huidobrensis. Morphological results showed that characteristics imparted by the male genitalia were the most reliable morphological features for identification. The colors exhibited by vertical setae were variable among species, and the ratio of the length of the ultimate section of vein CuA 1 divided by penultimate section also varied within species. Although the patterns of abdominal tergites were diverse among Liriomyza spp., L. trifolii exhibited a unique pattern with a yellow patch at the 5 th black visible tergite; this pattern can be profiled as a prominent characteristic for morphological identification. In order to identify the three Liriomyza spp. quickly and accurately, we developed an improved molecular identification method using multiplex PCR based on the gene encoding mitochondrial cytochrome oxidase I (COI); this method enabled direct identification based on the size of amplified products. The results of this study provide a valuable reference for the identification of Liriomyza spp., which will ultimately improve our ability to control individual species.</ns0:p></ns0:div>
</ns0:abstract>
<ns0:body xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:head>INTRODUCTION</ns0:head><ns0:p>Leafminer flies (Diptera: Agromyzidae), especially Liriomyza trifolii, L. sativae and L. huidobrensis, are invasive insect pests in many countries. They are polyphagous, economicallysignificant pests that cause severe damage to many ornamental and vegetable crops worldwide <ns0:ref type='bibr' target='#b31'>(Spencer, 1973</ns0:ref><ns0:ref type='bibr' target='#b32'>(Spencer, , 1990;;</ns0:ref><ns0:ref type='bibr' target='#b24'>Reitz et al., 1999)</ns0:ref>. Both larvae and adults cause serious damage to crops <ns0:ref type='bibr' target='#b21'>(Musgrave et al., 1975;</ns0:ref><ns0:ref type='bibr' target='#b16'>Minkenberg & Van Lenteren, 1986)</ns0:ref>. The damage caused by larval feeding on leaves can reduce photosynthetic capacity, and leaf mining activity can cause premature leaf drop resulting in reduced yields <ns0:ref type='bibr' target='#b10'>(Johnson et al., 1983;</ns0:ref><ns0:ref type='bibr' target='#b2'>Chandler & Gilstrap, 1987)</ns0:ref>. Moreover, indirect damage occurs when adults pierce leaves for feeding and oviposition, thus increasing plant susceptibility to disease <ns0:ref type='bibr' target='#b36'>(Zitter & Tsai, 1977;</ns0:ref><ns0:ref type='bibr' target='#b19'>Motteoni & Broadbent, 1988)</ns0:ref>. The rapid life cycle and high growth rate of Liriomyza spp. can lead to serious crop losses. Accurate identification of Liriomyza is important for implementing effective control strategies, because insecticide resistance and tolerance to environmental stress varies among species <ns0:ref type='bibr' target='#b3'>(Chang et al., 2017;</ns0:ref><ns0:ref type='bibr' target='#b8'>Gao et al., 2017)</ns0:ref>.</ns0:p><ns0:p>Closely-related Liriomyza spp. are similar in morphology at the adult stage <ns0:ref type='bibr' target='#b23'>(Oudman et al., 1995;</ns0:ref><ns0:ref type='bibr' target='#b14'>Lei et al., 1996;</ns0:ref><ns0:ref type='bibr' target='#b5'>Chen, 1999;</ns0:ref><ns0:ref type='bibr'>Scheffer et al., 2001)</ns0:ref>, and adult males can only be identified with certainty according to genitalia, which is both time-consuming and difficult. Identification at the early developmental stages of Liriomyza infestation is necessary for effective control; however, the absence of morphological characters makes identification difficult and larvae cannot be collected directly due to their mining behavior <ns0:ref type='bibr' target='#b23'>(Oudman et al., 1995;</ns0:ref><ns0:ref type='bibr' target='#b6'>Chiu et al., 2000;</ns0:ref><ns0:ref type='bibr' target='#b18'>Morgan et al., 2000;</ns0:ref><ns0:ref type='bibr'>Scheffer et al., 2001)</ns0:ref>.</ns0:p><ns0:p>Since morphological identification of female adults, larvae and pupae of Liriomyza species is PeerJ reviewing PDF | (2020:06:50114:1:1:NEW 9 Sep 2020)</ns0:p><ns0:p>Manuscript to be reviewed complex and difficult, molecular methods of identification are required. Immature developmental stages are the most common forms intercepted at ports of entry, therefore, it is important to identify these interceptions accurately and rapidly. With the development of mitochondrial and other molecular markers <ns0:ref type='bibr' target='#b1'>(Carapelli et al., 2018;</ns0:ref><ns0:ref type='bibr' target='#b4'>Chen et al., 2019)</ns0:ref>, several molecular methods have been developed to identify Liriomyza species <ns0:ref type='bibr' target='#b15'>(Menken & Ulenberg, 1983;</ns0:ref><ns0:ref type='bibr' target='#b35'>Zehnder et al., 1983;</ns0:ref><ns0:ref type='bibr' target='#b23'>Oudman et al., 1995;</ns0:ref><ns0:ref type='bibr' target='#b6'>Chiu et al., 2000;</ns0:ref><ns0:ref type='bibr' target='#b18'>Morgan et al., 2000)</ns0:ref>. Multiplex PCR is a cost-effective, rapid, accurate method where identification can be determined by PCR product size with speciesspecific primers <ns0:ref type='bibr' target='#b22'>(Nakamura et al., 2013)</ns0:ref>.</ns0:p><ns0:p>In this study, we re-verified morphological characteristics of three leafminers, L. trifolii, L. sativae and L. huidobrensis. A new morphological characteristic for detection of L. trifolii was investigated, and an improved molecular method for identification was developed based on multiplex PCR. This study provides approaches that can be deployed for identification of Liriomyza species, which will ultimately help future control efforts.</ns0:p></ns0:div>
<ns0:div><ns0:head>MATERIALS AND METHODS</ns0:head></ns0:div>
<ns0:div><ns0:head>Insects</ns0:head><ns0:p>The three species of Liriomyza spp. were collected from areas where leafminers occur in China.</ns0:p><ns0:p>In this study, 263 individuals of three species were selected for further data analysis (Table <ns0:ref type='table' target='#tab_0'>S1</ns0:ref>).</ns0:p><ns0:p>These were collected at the larval stage, tagged with relevant information and transported to the laboratory for pupation and emergence as adults. After preliminary morphological identification, adults were labeled, immersed in 70% ethanol and stored at -20 o C. After dissecting and photographing the samples, the remaining tissues were stored in 100% ethanol for DNA extraction and molecular analysis.</ns0:p></ns0:div>
<ns0:div><ns0:head>Morphological identification</ns0:head><ns0:p>Samples were examined with a stereomicroscope (Zeiss Stemi 2000c) and photographed with a wide depth of field (Zeiss Smartzoom 5). Male genitalia and wings were dissected, and slides were prepared and photographed with the Axio imager A2 (Zeiss, Germany).</ns0:p><ns0:p>Differences in the ratios of ultimate section lengths of vein CuA 1 among different Liriomyza species were determined by one-way analysis of variance (ANOVA), followed by Tukey's multiple comparisons. All statistical analyses were performed using SPSS v. 16.0 (SPSS, Chicago, IL, USA), and statistical significance was determined when P < 0.05.</ns0:p></ns0:div>
<ns0:div><ns0:head>Molecular identification and primer selection for multiplex PCR</ns0:head><ns0:p>Genomic DNA of Liriomyza species was extracted using the AxyPrep TM Multisource Genomic DNA Kit (Axygen, USA). A partial sequence of the mitochondrial cytochrome oxidase I (COI) gene was amplified with common primers F, 5′-CAACATTTATTTTGATTTTTTGG-3′ and R, 5′-TCCAATGCACTAATCTGCCATATTA-3′ <ns0:ref type='bibr' target='#b29'>(Simon et al., 1994;</ns0:ref><ns0:ref type='bibr' target='#b33'>Yang et al., 2010)</ns0:ref> using protocols described by <ns0:ref type='bibr' target='#b4'>Chen et al. (2019)</ns0:ref>, to molecular cross-checking and verification all of Liriomyza species in this study using sequencing, accession number can be found in Table <ns0:ref type='table' target='#tab_0'>S1</ns0:ref>.</ns0:p><ns0:p>For multiplex PCR, full-length COI genes of three Liriomyza species were downloaded from NCBI (https://www.ncbi.nlm.nih.gov/) and aligned using Clustal X. To develop a rapid PeerJ reviewing PDF | (2020:06:50114:1:1:NEW 9 Sep 2020)</ns0:p><ns0:p>Manuscript to be reviewed identification method, three species-specific primers and a common reverse primer were mixed to amplify DNA from different Liriomyza species. The PCR conditions were as follows: denaturation at 94 °C for 3 min; 35 cycles at 94 °C for 1 min, 58 °C for 1 min and 72 °C for 1 min; followed by extension at 72 °C for 10 min. PCR was conducted in a 25 μL reaction volume containing 2 μL (100 ng) of DNA template, 1 μL (10 μM) of each primer, 12.5 μL of 2× Taq Master mix (Vazyme Biotech Co., Ltd) and 6.5 μL ddH 2 O. PCR products were separated in 1.0% agarose gels, and primers that amplified only one specific band for each species are shown in Table <ns0:ref type='table' target='#tab_0'>1</ns0:ref>.</ns0:p></ns0:div>
<ns0:div><ns0:head>RESULTS</ns0:head></ns0:div>
<ns0:div><ns0:head>Morphological identification</ns0:head><ns0:p>The distiphallus, which is part of the male genitalia, is a very small, fragile structure enclosed by membranes located at the terminus of the aedeagus. For L. trifolii, the morphological characteristics of the distiphallus include one distal bulb with marked constriction between lower and upper halves in dorsoventral view; the bulb is lightly sclerotized with a long basal stem (Fig. <ns0:ref type='figure'>1A</ns0:ref>). For L. sativae, the distiphallus is characterized by one distal bulb with a slight constriction between upper and lower halves in the dorsoventral view; the bulb is more intensely sclerotized with a shorter basal stem (Fig. <ns0:ref type='figure'>1B</ns0:ref>). For L. huidobrensis, the distiphallus contains two distal bulbs; these meet at rims that extend in an anteroventral orientation (Fig. <ns0:ref type='figure'>1C</ns0:ref>).</ns0:p><ns0:p>With respect to vertical setae, L. trifolii exhibits inner and outer vertical setae on a yellow background; whereas vertical setae are present on a black background for L. huidobrensis. In L. <ns0:ref type='table' target='#tab_2'>PDF | (2020:06:50114:1:1:NEW 9 Sep 2020)</ns0:ref> Manuscript to be reviewed sativae, outer and inner vertical setae are presented on black and yellow backgrounds, respectively <ns0:ref type='bibr' target='#b31'>(Spencer, 1973)</ns0:ref>. In this study, only 86.1% (192/223) of L. trifolii had yellow inner and outer vertical setae; 9.9% (22/223) had yellow inner vertical setae and undetermined color for outer setae, and 4.0% (9/223) had yellow inner and black outer vertical setae (Table <ns0:ref type='table' target='#tab_1'>2</ns0:ref>; Figs. <ns0:ref type='figure' target='#fig_2'>2A-C</ns0:ref>). For L. sativae, 17.6% (6/34) had black inner and outer vertical setae, 58.8% (20/34) had yellow inner and black outer vertical setae, and 23.5% (8/34) had outer black setae with an undetermined color for inner vertical setae (Table <ns0:ref type='table' target='#tab_1'>2</ns0:ref>; Fig. <ns0:ref type='figure' target='#fig_2'>2D-F</ns0:ref>). For L. huidobrensis, 100% (6/6) exhibited black inner and outer vertical setae (Table <ns0:ref type='table' target='#tab_1'>2</ns0:ref>; Fig. <ns0:ref type='figure' target='#fig_2'>2G-I</ns0:ref>). These results show that characteristics of vertical setae are not reliable for identifying Liriomyza species.</ns0:p></ns0:div>
<ns0:div><ns0:head>PeerJ reviewing</ns0:head><ns0:p>Wing pattern ratios were calculated as the length of the ultimate section of vein CuA 1 divided by the penultimate section ('a' and 'b', see Fig. <ns0:ref type='figure' target='#fig_11'>3A-C</ns0:ref>). In this study, 'a' was 2.70 ± 0.31 times the length of 'b' in L. trifolii, and 'a' was 2.72 ± 0.37 times the length of 'b' in L. sativae.</ns0:p><ns0:p>For L. huidobrensis, 'a' was 2.20 ± 0.24 times the length of 'b' (F 2,237 = 7.345, P < 0.05) (Fig. <ns0:ref type='figure' target='#fig_4'>4</ns0:ref>).</ns0:p><ns0:p>Although the ratio of L. huidobrensis was significantly different from the other two species (P < 0.05), there was no significant difference between L. trifolii and L. sativae (P = 0.907). Many L. trifolii individuals exhibited truncated or missing dm-cu cross veins. Furthermore, we noted inconsistency between left and right forewing patterns within individual samples (Fig. <ns0:ref type='figure' target='#fig_11'>3A</ns0:ref>, with dashed lines).</ns0:p><ns0:p>In L. trifolii, the 2 nd -5 th visible tergites were generally divided by a yellow medial furrow in male adults; furthermore, there was a yellow patch at the 5 th black visible tergite that can distinguish L. trifolii from other Liriomyza species (Figure <ns0:ref type='figure' target='#fig_5'>5A-C</ns0:ref>). In L. sativae and L.</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:06:50114:1:1:NEW 9 Sep 2020)</ns0:p><ns0:p>Manuscript to be reviewed huidobrensis, only the second visible tergite is divided by a yellow medial furrow and no yellow patch is evident on the 5 th tergite (Fig. <ns0:ref type='figure' target='#fig_5'>5D-I</ns0:ref>).</ns0:p></ns0:div>
<ns0:div><ns0:head>Molecular detection of Liriomyza spp.</ns0:head><ns0:p>Candidate primers for species-specific detection of Liriomyza were based on the alignment of 262 (L. sativae), 612 (L. trifolii), and 959 (L. huidobrensis) COI sequences. We designed one reverse primer, 1181 R, that was common to all three Liriomyza species. The position of forward primers was selected to produce <1000 bp amplicons when paired with the reverse primer with at least 300 bp nucleotides between species. In addition, sites were selected where the number of differential nucleotides was >2 bp to increase the specificity of the primers (Fig. <ns0:ref type='figure' target='#fig_6'>6</ns0:ref>).</ns0:p><ns0:p>The three Liriomyza species could be differentiated by specific PCR products in 1.0% agarose gels, and the resulting PCR products were 569, 919, and 222 bp for L. trifolii, L. sativae and L. huidobrensis, respectively (Fig. <ns0:ref type='figure' target='#fig_12'>7A</ns0:ref>). The validity of multiplex PCR for identification was further confirmed by using the system with different developmental stages; the approach worked equally well for larvae, pupae and adults of the three Liriomyza species (Fig. <ns0:ref type='figure' target='#fig_12'>7B</ns0:ref>). Populations from different geographical regions were also obtained to evaluate the reliability of speciesspecific primers. The results obtained by multiplex PCR (Fig. <ns0:ref type='figure' target='#fig_8'>S1</ns0:ref>) and subsequent sequence analysis of COI (Fig. <ns0:ref type='figure' target='#fig_9'>S2</ns0:ref>) showed that geography did not impact the reliability of primers.</ns0:p></ns0:div>
<ns0:div><ns0:head>DISCUSSION</ns0:head><ns0:p>The morphological characteristics used for Liriomyza identification have primarily followed Manuscript to be reviewed <ns0:ref type='bibr' target='#b31'>Spencer's (1973)</ns0:ref> criteria. However, variability in life stages, emergence times and sample preservation result in large differences in body color and markings, which can make current morphological criteria unreliable for identification <ns0:ref type='bibr' target='#b31'>(Spencer, 1973;</ns0:ref><ns0:ref type='bibr' target='#b12'>Kang et al., 1996;</ns0:ref><ns0:ref type='bibr'>Shao, 2004)</ns0:ref>.</ns0:p><ns0:p>Currently, the identification of Liriomyza spp. based on morphology is restricted to male adults because there are no reliable features for species-level identification of female adults or immature developmental stages <ns0:ref type='bibr' target='#b7'>(EPPO, 2005)</ns0:ref>. The identification of adults requires the examination of the male adult genitalia. In general, the distiphallus provides reliable detection of the three Liriomyza species and has considerable diagnostic value <ns0:ref type='bibr' target='#b31'>(Spencer, 1973;</ns0:ref><ns0:ref type='bibr' target='#b28'>Shiao, 2004)</ns0:ref>. However, differences in distiphalluses between species are subtle and dissection is difficult for nonprofessionals. Consequently, features of distiphallic structure should be cross-checked with other external morphological characteristics to ensure that identification is valid.</ns0:p><ns0:p>According to <ns0:ref type='bibr' target='#b31'>Spencer (1973)</ns0:ref>, coloration of the vertical setae is an important external feature that can distinguish L. trifolii and L. sativae without dissection; however, this feature is unstable and lacks clear interspecific boundaries. Results of the current study show that reliance on coloration of vertical setae can result in misidentification of L. trifolii and L. sativae; thus, this feature should only be used as a supplement for identification. The ratio of the length of the ultimate section of vein CuA 1 is unreliable since most ratio values overlapped among Liriomyza species. In this study, we also evaluated the patterns of abdominal tergites and discovered that the yellow patch at the 5 th black visible tergite of L. trifolii is a new, reliable morphological characteristic for identification. Similar findings were reported for abdominal color patterns for PeerJ reviewing PDF | (2020:06:50114:1:1:NEW 9 Sep 2020)</ns0:p><ns0:p>Manuscript to be reviewed six Liriomyza species <ns0:ref type='bibr' target='#b28'>(Shiao, 2004)</ns0:ref>.</ns0:p><ns0:p>Molecular methods for insect identification can be used with different developmental stages, including immature stages where morphological features may be lacking. Furthermore, molecular assays may facilitate identification of atypical or damaged samples. However, the specificity of molecular assays may be limited because they were developed for a particular purpose and evaluated against a restricted number of species <ns0:ref type='bibr' target='#b22'>(Nakamura et al., 2013)</ns0:ref>. Multiplex PCR assays were recently developed for identification of Liriomyza species <ns0:ref type='bibr' target='#b17'>(Miura et al., 2004;</ns0:ref><ns0:ref type='bibr' target='#b9'>Guan et al., 2006;</ns0:ref><ns0:ref type='bibr' target='#b22'>Nakamura et al., 2013)</ns0:ref> and are based on amplification of a target gene region using species-specific primer combinations. Multiplex PCR assays are easier and faster than other molecular methods, such as RAPD-PCR, PCR-RFLP, DNA barcoding and real-time PCR <ns0:ref type='bibr' target='#b6'>(Chiu et al., 2000;</ns0:ref><ns0:ref type='bibr' target='#b18'>Morgan et al., 2000;</ns0:ref><ns0:ref type='bibr'>Scheffer et al., 2001;</ns0:ref><ns0:ref type='bibr' target='#b13'>Kox et al., 2005;</ns0:ref><ns0:ref type='bibr' target='#b25'>Scheffer et al., 2006;</ns0:ref><ns0:ref type='bibr' target='#b0'>Blacket et al., 2015;</ns0:ref><ns0:ref type='bibr' target='#b30'>Sooda et al., 2017)</ns0:ref>; furthermore, multiplex PCR assays are more sensitive than enzyme electrophoresis methods <ns0:ref type='bibr' target='#b35'>(Zehnder et al. 1983;</ns0:ref><ns0:ref type='bibr' target='#b15'>Menken & Ulenberg, 1983</ns0:ref><ns0:ref type='bibr' target='#b16'>, 1986;</ns0:ref><ns0:ref type='bibr' target='#b23'>Oudman et al., 1995)</ns0:ref>. In general, the reliability and sensitivity of multiplex PCR represents a great improvement in molecular identification protocols and will enable us to manage invasive pests more effectively.</ns0:p></ns0:div>
<ns0:div><ns0:head>CONCLUSIONS</ns0:head><ns0:p>Invasive Liriomyza spp. comprise a group of insect pests that cause considerable economic loss and serious quarantine problems. In this study, morphological features were re-evaluated for L. trifolii, L. sativae, and L. huidobrensis, and the discriminative ability of traditional PeerJ reviewing PDF | (2020:06:50114:1:1:NEW 9 Sep 2020)</ns0:p><ns0:p>Manuscript to be reviewed morphological characteristics, such as male genitalia, abdominal color patterns, length of CuA 1 and abdominal tergite patterns were reevaluated. Furthermore, we developed an improved molecular identification method using multiplex PCR based on COI to identify the three Liriomyza species quickly and accurately. This study provides valuable tools for the identification of Liriomyza spp. using both morphological and molecular criteria. Manuscript to be reviewed Manuscript to be reviewed Manuscript to be reviewed </ns0:p></ns0:div><ns0:figure xml:id='fig_0'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:06:50114:1:1:NEW 9 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_1'><ns0:head>FIGURE LEGENDS Figure 1 .</ns0:head><ns0:label>LEGENDS1</ns0:label><ns0:figDesc>FIGURE LEGENDS</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_2'><ns0:head>Figure 2 .</ns0:head><ns0:label>2</ns0:label><ns0:figDesc>Figure 2. Color characteristics of outer and inner vertical setae in three Liriomyza species. A-C, L. trifolii; D-F,</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_3'><ns0:head>Figure 3 .</ns0:head><ns0:label>3</ns0:label><ns0:figDesc>Figure 3. Comparison of wing pattern ratios for three Liriomyza species. Ratios were calculated as the length of</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_4'><ns0:head>Figure 4 .</ns0:head><ns0:label>4</ns0:label><ns0:figDesc>Figure 4. Ratios showing lengths of the ultimate section of vein CuA 1 divided by the penultimate section.</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_5'><ns0:head>Figure 5 .</ns0:head><ns0:label>5</ns0:label><ns0:figDesc>Figure 5. Abdominal color patterns of three Liriomyza species. A-C, L. trifolii; D-F, L. sativae; G-I, L.</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_6'><ns0:head>Figure 6 .</ns0:head><ns0:label>6</ns0:label><ns0:figDesc>Figure 6. Alignment of COI sequences in Liriomyza spp. Sequences surrounded by rectangular boxes represent</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_7'><ns0:head>Figure 7 .</ns0:head><ns0:label>7</ns0:label><ns0:figDesc>Figure 7. Agarose gel electrophoresis of multiplex PCR products. (A) PCR products in three different</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_8'><ns0:head>Figure S1 .</ns0:head><ns0:label>S1</ns0:label><ns0:figDesc>Figure S1. Agarose gel electrophoresis of multiplex PCR products from different geographical populations of</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_9'><ns0:head>Figure S2 .</ns0:head><ns0:label>S2</ns0:label><ns0:figDesc>Figure S2. Alignment of COI sequences from different geographical population using species-specific primers.</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_10'><ns0:head /><ns0:label /><ns0:figDesc>Figure 1</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_11'><ns0:head>Figure 3</ns0:head><ns0:label>3</ns0:label><ns0:figDesc>Figure 3</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_12'><ns0:head>Figure 7</ns0:head><ns0:label>7</ns0:label><ns0:figDesc>Figure 7</ns0:figDesc></ns0:figure>
<ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='24,42.52,229.87,525.00,210.75' type='bitmap' /></ns0:figure>
<ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='25,42.52,275.62,525.00,331.50' type='bitmap' /></ns0:figure>
<ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='30,42.52,229.87,525.00,177.00' type='bitmap' /></ns0:figure>
<ns0:figure type='table' xml:id='tab_0'><ns0:head>Table 1 .</ns0:head><ns0:label>1</ns0:label><ns0:figDesc>Primers used in this study.</ns0:figDesc><ns0:table><ns0:row><ns0:cell>Primer name</ns0:cell><ns0:cell>Nucleotide sequence (5′-3′)</ns0:cell><ns0:cell>Ta (Tm) ºC</ns0:cell><ns0:cell>Product size (bp)</ns0:cell><ns0:cell>GenBank number</ns0:cell></ns0:row><ns0:row><ns0:cell cols='2'>Lt612 CAATTACAATACTATTAACAGACCG</ns0:cell><ns0:cell>58 (48.5)</ns0:cell><ns0:cell>569</ns0:cell><ns0:cell>MT919718</ns0:cell></ns0:row><ns0:row><ns0:cell cols='2'>Ls262 AGCTCCAGACATAGCATTTCCTCG</ns0:cell><ns0:cell>58 (58.9)</ns0:cell><ns0:cell>919</ns0:cell><ns0:cell>MT919719</ns0:cell></ns0:row><ns0:row><ns0:cell cols='2'>Lh959 TTCAGATGGCTTGCCACATTACACG</ns0:cell><ns0:cell>58 (59.9)</ns0:cell><ns0:cell>222</ns0:cell><ns0:cell>MT919720</ns0:cell></ns0:row><ns0:row><ns0:cell cols='3'>LR1181 GAATAAATCCKGCTATAATTGCAAATAC 58 (50.9)</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell></ns0:row></ns0:table><ns0:note>PeerJ reviewing PDF | (2020:06:50114:1:1:NEW 9 Sep 2020)</ns0:note></ns0:figure>
<ns0:figure type='table' xml:id='tab_1'><ns0:head>Table 2 (on next page)</ns0:head><ns0:label>2</ns0:label><ns0:figDesc /><ns0:table /></ns0:figure>
<ns0:figure type='table' xml:id='tab_2'><ns0:head>Table 2 .</ns0:head><ns0:label>2</ns0:label><ns0:figDesc>The data of color characteristic of outer and inner vertical setae position in three Liriomyza species.</ns0:figDesc><ns0:table /><ns0:note>PeerJ reviewing PDF | (2020:06:50114:1:1:NEW 9 Sep 2020)</ns0:note></ns0:figure>
</ns0:body>
" | "Reviewer 1
Basic reporting
1. The one and only issue that, in my view, may prevent it from being acceptable even after revision, if not dealt with convincingly, is intraspecific variability. This is possibly the key reason why such tests are not valid/usable and may mislead other authors into using an invalid test in their research.
Response: Thanks for your valuable comment. We have selected different geographical populations of those three Liriomyza species to verify the reliability of our primers. The results showed that different geographical populations did not affect the reliability of primers. Primers work in different geographic populations, based on gel electrophoresis and sequencing. I have submitted the relevant results as supplementary materials (Figure S1 and Figure S2).
Figure S1. Agarose gel electrophoresis of multiplex PCR products from different geographical populations of Liriomyza. Lt Lanes 1-10 indicate L. trifolii populations from: (1) Hengshui, (2) Hangzhou, (3) Dongguan, (4) Zhangzhou, (5) Qionghai, (6), Nanning, (7) Changzhou, (8) Nanchang, (9), Huizhou, and (10) Huzhou. Ls lanes 1-2 indicate L. sativae populations from Shangqiu and Luoyang, respectively. Lh lanes 1-2 represent L. huidobrensis populations from Kunming and the laboratory, respectively.
Figure S2. Alignment of COI sequences from different geographical population using species-specific primers. The shaded nucleotides represent divergent sites. Nucleotides bounded by red and blue dashed rectangles represent forward and reverse primers, respectively. Lt sequences represent the following L. trifolii populations: Lt 1, Hengshui; Lt 2, Hangzhou; Lt 3, Dongguan; Lt 4, Zhangzhou; Lt 5, Qionghai; Lt 6, Nanning; Lt 7, Changzhou; Lt 8, Nanchang; Lt 9, Huizhou; and Lt 10, Huzhou. Ls sequences represent L. sativae populations from Shangqiu (Ls 1) and Luoyang (Ls 2). Lh sequences represent L. huidobrensis populations from Kunming (Lh 1) and the laboratory (Lh 2).
2. Language is not always clear and should be improved before publication. As an example see lines 27-19, 45-49, 57-60, 145-149, 154-157, 192-194.
Response: Thanks for your suggestions and modifications. Our manuscript has been polished by English native speaker, some errors have been corrected.
3. Background is appropriate, but some recent studies (eg: Chen et al, 2019 Sci.Rep; Carapelli et al, 2018 Genes) may be included.
Response: Relevant studies has been added, see Line 67.
References:
Carapelli A, Soltani A, Leo C, Vitale M, Amri M, Mediouni-Ben Jemâa J. 2018. Cryptic diversity hidden within the leafminer genus Liriomyza (Diptera: Agromyzidae). Genes 9: 554.
Chen J Y, Chang Y W, Tang X T, Zheng S Z, Du Y Z. 2019. Population genetics of Liriomyza trifolii (Diptera: Agromyzidae) and comparison with four Liriomyza species in China based on COI, EF-1a and microsatellites loci. Scientific Reports 9: 17856.
4. Furthermore it is not always clear form the introduction and discussion sections:
(a) which morphological characters are considered as inequivocous and which are considered dubious and re-evaluated. A clear distinction between male genital structures, being inequivocous, and the other three, being equivocous, stems from the results of this study. Does this represent the state of the art before this study? Did other authors comment on the validity of these characters in the past?
Response: Male adult genitalia probably comprise the most important taxonomic character in the Agromyzidae since Nowakowski (1962), and they subsequently were generally adopted as a major character by most leading specialists, such as Spencer (1990). Nevertheless, most of these characters are presented by means of drawings, and only a few photos (e.g., Shiao, 2004) can be found in the literature. Genitalia drawings are somewhat difficult to interpret, as different authors use different ways to express the genital characters, and the correct species identification always relies strongly on the quality, precision, and details of the drawings. In addition, those genitalia drawings are more or less personalized and somewhat difficult to compare with actual dissected objects, especially for a beginner. For example, some authors use outline drawings to represent overall shapes; some prefer dotted drawings to emphasize the different areas or various degrees of sclerotization of the hard parts of the genitalia.
Yes, the previous description genital structures were a stable feature, but there is no clear picture. We believe that using photography with a larger focal depth can overcome these defects of using traditional photos and also avoid misjudgement by reading descriptions or drawings.
References:
Nowakowski J T. 1962. Introduction to a systematic revision of the family Agromyzidae (Diptera) with some remarks on host plant selection by these flies. Polska Akademia Nauk Instytut Zoologiczny Annales Zoologici 20: 67-183.
Spencer K A. 1990. Host Specialization in the World Agromyzidae (Diptera) 45, Series Entomologica, Kluwer, Dordrecht, The Netherlands.
Shiao S F. 2004. Morphological diagnosis of six Liriomyza species (Diptera, Agromyzidae) of quarantine importance in Taiwan. Applied Entomology and Zoology 39: 27-39.
(b) which are the strength and weaknesses of the molecular tests developed in the past and which is the knowledge gap that the authors are willing to cover. Curiously, the authors suggest that 'using samples from different geographic regions' (line 197) may be a limitation, while I would argue that using samples from one single location is the key limitation of this study (see below). A paper (Zehnder et al 1986) is referenced to in the text but it is not present in the bibliography. Please cross check text and bibliography throughout.
Response: Yes, the reliability of different geographic populations needs to be checked in this study. We have selected different geographical populations of those three Liriomyza species to verify the reliability of our primers (see above).
A paper (Zehnder et al 1986) has been checked and deleted in the text.
5. Figures and Tables：
(a) In Figure 4 please specify what interval the central box represents and add marks (usually asterisks are used) to convey the results of the post-hoc tests. In Figure 6 please indicate which positions are variable within each species. The image (as is) suggests that sequences are invariable within species, that is certainly not the case. In Figure 7 it is not clear which species panel B refers to. Please prepare a montage figure where all developmental stages of all species are shown and emend the caption accordingly.
Response: These concerns and errors have been corrected. Interval in the central box means median in Figure 4; For Figure 6, no mutation affecting the specificity of primers was found based on gel and sequencing data of different geographical populations and it's not easy to add all the invalid mutations in figure 6; we have supplemented samples of different developmental stages of L. sativae, and L. huidobrensis in Figure 7.
(b) In Table 1 the accession number of the complete mitochondrial genome of the three species may be removed. Table 2, see below. No mention is made to data deposition, while I deem that new cox1 sequences should be deposited in GenBank.
Response: Accession number in Table 1 has been removed and changed to new accession numbers of sequences by species-specific primer designed in this study. Table 2 and accession number of new cox1 sequences (see below).
Experimental design
1. The research question is well defined (i.e. developing an improved molecular test for species discrimination) but the knowledge gap this is filling is not well defined or justified. In fact, the limitations of previous molecular methods similarly based on PCR are not presented and it is not clear how this new test improves over available tests.
Response: The characteristics of our multiplex PCR assay: (1) the position of forward primers was selected to produce <1000bp amplicons in combination with the reverse primer which can easy for sequencing (generally, the length of one commercial sequencing is ~ 1000 bp). (2) It also to leave at least 300bp nucleotides between each species which can easy to distinguish on gel. (3) In addition, the selected primer sites where the number of differential nucleotides >2 bp for the specificity of the primers.
2. Some aspects of the methodology are not clearly presented and it cannot be evaluated if the investigation has been conducted under rigorous standards. More specifically, my main concern regards intraspecific variability, an issue that appears to have been overlooked by the authors but, in my view, needs some consideration.
(a) how many individuals, and how many from which species, were sequenced (lines 98-102)? Did the authors sequence all individuals listed in Supplementary Table 1?
Response: We have selected different geographical populations of those three Liriomyza species to verify the reliability of our primers. The results showed that different geographical populations did not affect the reliability of primers (see above).
Yes, we sequenced all the samples in Table S1, and access numbers can be found in Table S1.
(b) how many sequences were taken from GenBank (lines 103-104)? How many from each species? Which is the geographical representativity of these sequences? Which is the level of intra and interspecific variability?
Response: Only three COI genes were selected from the mitochondrial genome sequence of each species in GenBank for species-specific primers design (access numbers were JN570506.1, NC_015926.1, and JQ862474.1 for L. trifolii, L. sativae, and L. huidobrensis, respectively). Yes, we ignored interspecific variation in previous manuscript and only focus on the differences among species. We have selected different geographical populations of those three Liriomyza species to verify the reliability of our primers.
(c) how many variable sites were observed between species (lines 105-107)? How many different sites were fixed and invariable within species?
Response: Sorry, this sentence was not clear. “These were aligned with the COI sequences of those three leafminer flies,” which just means three full length of Liriomyza species COI genes (download from NCBI) were aligned and used for primer design.
(d) how many specimens were tested to validate the test? Which developmental stages? Three bands per species (adults) and three band from three different developmental stages (stages and species are not indicated) are shown in figure 7. Is this the full panel of tests conducted to validate the essay? I think that a minimum of few tens of individuals per species (including different developmental stages) may be used here.
Response: In fact, we have verified the reliability of these primers in daily identification experiments more than hundreds of times. In addition, we have supplemented samples of different developmental stages of L. sativae, and L. huidobrensis and different geographical populations of those three species. We believe that the species-specific primers designed in this study is reliable.
(e) line 153 says that the reverse primer is 'common only to those three Liriomyza species'. I would say that it 'is common' to the three species, while the authors cannot know if it is common 'only' to these. This is relevant as the primer may amplify in other species, something that has not been tested.
Response: Yes, the description of these statements needs to be precise and clear. We have changed the description to “common to those three Liriomyza species”.
3. Methods (meaning how experiments were performed) are described with sufficient detail. How many samples, and which samples were included is not always clear (see above).
Response: I have explained those concerns in above.
Validity of the findings
1. As from my comments above, I think that natural intraspecies genetic variability is a crucial aspect here. Considering that (a) sequence data used for test development may or may not include a sufficient number of reference sequences from outside of China (see my abovestated request for clarification), (b) samples used to validate the test are from China (see also my abovestated comment on their number), and (c) the fact that we are talking about three invesive species whose native range is not China and may display a sizeable level of genetic diversity in their native range that is not present in China, I would argue that, strictly speaking, the test is valid for China only.
Response: We have changed these descriptions and the title of the manuscript has been changed to “Revalidation of morphological characters and multiplex PCR assay for the identification of three congener invasive Liriomyza species (Diptera: Agromyzidae) in China”.
2. Underlying data has generally been provided, but a few pieces of information should be addedd: (a) Gen Bank accessions for new sequences (see my comment above).
Response: New sequences have been submitted and changed in Table S1 and Table 1.
(b) patterns of setae color in Table 2 are not clearly reported. I suggest the authors list, for each species, the visible phenotypes (including uncertain phenotypes that are descibed in the text) and report their absolute and relative frequencies. Here the 'unit of observation' is the individual insect, not the seta. This has been correctly presented in the text but not, in my view, in the table.
Response: I'm sorry to confuse you. Yes, Table 2 should be showed that the number or ratio of samples with different characteristics of the vertical setae position among the three Liriomyza species. We have re-created Table 2 by the “unit of observation individual”.
(c) statistical analysis on wings is correctly conducted (an overall test followed by multiple post hoc tests), but the results are not clearly presented. The p value of the overall test should be reported as well as p values for each of the three post hoc comparisons (or at least the significant ones). Furthermore, in the methods section it is declared that p=0.05 is used as a significance threshold, while in the result section a p=0.001 is considered not significant.
Response: Thanks for your comments, those errors have been corrected see Line 142-145.
Reviewer 2
Basic reporting
1. The English is not always clear and many sentences are vague and imprecise (see attachment for further details). In my opinion, it should be improved, before publication.
Response: Thanks for your suggestions and modifications. Our manuscript has been polished by English native speaker, some errors have been corrected.
2. The literature references should be improved in some points (e.g., lines 168-170). The background and the aims are sufficiently described and clear throughout the text.
Response: These concerns and errors have been corrected. Reference has been added in Line 176.
3. The structure of the manuscript is clear, as well as the figures and tables are relevant to the manuscript aim. However, I would suggest to improve the Figure 1 by both increasing the resolution and by adding more annotation to help the reader understanding the morphological description. I think the Figure references in the text should be improved (see attachment for further details).
Response: These concerns and errors have been corrected. In Figure 1, we added a red line to indicate the length of basal stem. We have re-cited the figure1 according to the description of each species as figure 1A, figure 1B and figure 1C. Similar figure citation insufficient in the manuscript have been checked and revised.
4. The raw data are not shared (e.g., the sequences obtained from their first PCR are not shared and no accession number is given).
Response: New sequences have been submitted. We sequenced all the samples in Table S1, and access numbers can be found in Table S1. In addition, Accession number in Table 1 has been removed and changed to new accession numbers of sequences by species-specific primer designed in this study.
Experimental design
1. The way in which the methods and the results are described opens to some important critical issues (also highlighted in the attached pdf). The methods are not sufficiently described to be reproducible. For examples, it is not clear how many sequences they used to design the primers. If the sequences were only one per species, how did the Authors account for intra-specific variability of the cox1 gene?
Response: Only one set of data was selected from the mitochondrial genome sequence of each species in GenBank for species-specific primers design (access numbers were JN570506.1, NC_015926.1, and JQ862474.1 for L. trifolii, L. sativae, and L. huidobrensis, respectively). Yes, we ignored interspecific variation in previous manuscript and only focus on the differences among species. We have selected different geographical populations of those three Liriomyza species to verify the reliability of our primers. The results showed that different geographical populations did not affect the reliability of primers. Primers work in different geographic populations, based on gel electrophoresis and sequencing. I have submitted the relevant results as supplementary materials (Figure S1 and Figure S2).
Figure S1. Agarose gel electrophoresis of multiplex PCR products from different geographical populations of Liriomyza. Lt Lanes 1-10 indicate L. trifolii populations from: (1) Hengshui, (2) Hangzhou, (3) Dongguan, (4) Zhangzhou, (5) Qionghai, (6), Nanning, (7) Changzhou, (8) Nanchang, (9), Huizhou, and (10) Huzhou. Ls lanes 1-2 indicate L. sativae populations from Shangqiu and Luoyang, respectively. Lh lanes 1-2 represent L. huidobrensis populations from Kunming and the laboratory, respectively.
Figure S2. Alignment of COI sequences from different geographical population using species-specific primers. The shaded nucleotides represent divergent sites. Nucleotides bounded by red and blue dashed rectangles represent forward and reverse primers, respectively. Lt sequences represent the following L. trifolii populations: Lt 1, Hengshui; Lt 2, Hangzhou; Lt 3, Dongguan; Lt 4, Zhangzhou; Lt 5, Qionghai; Lt 6, Nanning; Lt 7, Changzhou; Lt 8, Nanchang; Lt 9, Huizhou; and Lt 10, Huzhou. Ls sequences represent L. sativae populations from Shangqiu (Ls 1) and Luoyang (Ls 2). Lh sequences represent L. huidobrensis populations from Kunming (Lh 1) and the laboratory (Lh 2).
2. Moreover, no other Liriomyza species was included in the alignment as a control that the primers were actually specific for the three Liriomyza species under study. In the analysis, specimens and sequences of Liriomyza langei were not included and it is known that this latter species cannot be morphologically distinguished from L. huidobrensis (Scheffer et al 2014). Therefore, in my opinion, the Authors should have at least included in the cox1 alignment the sequences of L. langei (available on public databases). In this way, the Authors could confirm that the species-specific selection through multiplex PCR was potentially effective also in other areas where L. huidobrensis is present (e.g., America).
Response: Thanks for your valuable comment. However, we could not confirm the validity of the result if we design primer for L. langei, because there are no L. langei in China.
Furthermore, it is not clear from the methods (and the results), whether the Authors sequenced or not the multiplex PCR product to confirm that they actually selected the cox1 of the wanted species
Response: It was sequenced and the sequencing results matched the species. Accession number in Table 1 has been removed and changed to new accession numbers of sequences by species-specific primer designed in this study.
Validity of the findings
1. In my opinion, the novelty and impact of the present work strongly depends on few aspects. The first one is the inclusion of specimens or at least cox1 sequences of L. langei in the alignment to check the efficiency of the primers designed.
Response: Thanks for your valuable comment. However, we could not confirm the validity of the result if we design primer for L. langei, because there are no L. langei in China.
2. The second aspect is the sequencing of the multiplex PCR products to confirm that the primer selection actually works.
Response: It was sequenced and the sequencing results matched the species. Accession number in Table 1 has been removed and changed to new accession numbers of sequences by species-specific primer designed in this study.
3. A third aspect that should be better and clearly described is the application of this method to all the developmental stages of L. sativae and L. huidobrensis. The authors reported that this method worked only for all the developmental stages of L. trifolii. If we cannot distinguish larvae from these three species, then this method is not effective to develop useful control strategies for these pest insects.
Response: We have supplemented samples of different developmental stages of L. sativae, and L. huidobrensis in Figure 7.
Figure 7. Agarose gel electrophoresis of multiplex PCR products. (A) PCR products in three different Liriomyza species. (B) PCR products from different developmental stages of the three Liriomyza species. Each experiment contained three biological samples. Abbreviations: Lt, L. trifolii; Ls, L. sativae; Lh, L. huidobrensis.
4. Finally, the submission of the cox1 raw data on public databases is fundamental for reproducibility.
Response: We sequenced all the samples in Table S1, and access numbers can be found in Table S1.
Some comments in PDF file.
1. In Line 55, Should the references be organized in chronological order?
Response: The order of references has been rearranged.
2. In Line 79, Which ones? I see that there are some information in table S1, but these latter do not seem to be referred to all the samples included in the present study...
Response: The morphological identification section used all of samples (263 individuals) collected from the areas where Liriomyza occurred in China.
In Line 81, At line 58 you wrote that 'but larvae cannot be collected owing to their mining behavior'... Can you further and clearly explain your method of collection? Then you collected information, which ones? Ecological, geographical, on the hosts?
Response: The collected information can be found in Table S1 including collection location and host plant. We collected at larval stage, and then back to the laboratory for pupation and emergence. The larvae of Liriomyza cause damage by feeding inside foliar tunnels then the larvae exit the leaves for pupation, so we just collected the leaves with larvae inside. Therefore, it is impossible to collect larvae directly.
In Line 83, Based on which characters did you preliminary identified the specimens, especially considering that generally it's the genitalia dissection that actually can help in the identification?
Response: Based on the position of vertical setae, the specimens can be preliminarily identified. It's very time-consuming to identify with genital dissection characteristics.
In Line 89, Did you prepared slides also for wings? If not, why dissect them?
Response: The dissection of the wings is for the convenience of photographing the characteristics of the wings.
In Line 91, Can you show in a figure or explain which part of the body is it? Not all of us are expert in Liriomyza morphology or insect morphology in general, so it would be helpful understanding which characters you investigated.
Response: The wing patterns of Liriomyza species see blow.
In line 92, How did you calculate the length of the ultimate section of the vein?
Response: The length can be calculated automatically by microscope and corresponding software.
In Line 94, Which figure? Average of what?
Response: Figure 4. Average of the ratio of the length of ultimate section of vein CuA1 divided by penultimate section.
In line 102, Which PCR conditions did you use? How did you prepared samples for sequencing and which sequencing method did you use? More importantly did you submitted these cox1 sequences on public databases? If yes, what is their accession number? Without these information, it is pretty difficult to replicate your analysis...
Response: The PCR conditions were as follows using protocols described by Chen et al. (2019): denaturation at 94 °C for 3 min; 35 cycles at 94 °C for 1 min, 51 °C for 1 min and 72 °C for 1 min; followed by extension at 72 °C for 10 min. 25μL PCR mixture consisted of 2μL (100ng) of DNA template,1μL (10μM) each of primers (F and R), 12.5μL of 2×Taq Master mix (Vazyme Biotech Co., Ltd) and 8.5 μL ddH2O. PCR products were then checked in 1.0% agarose gel. Amplified products were purified using a gel extraction kit (Axygen, USA) and sequenced. We sequenced all the samples in Table S1, and access numbers can be found in Table S1.
In line 103, How many sequences did you downloaded from GenBank? Just one for L. sativae, L. trifolii and L. huidobrensis? Can you list the accession numbers here or in a table?
Response: Only three COI sequences were selected from the mitochondrial genome sequence of each species in GenBank for species-specific primers design (access numbers were JN570506.1, NC_015926.1, and JQ862474.1 for L. trifolii, L. sativae, and L. huidobrensis, respectively).
In Line 104, Not clear at all... Which sequence ('these') did you align with 'those three'?
Response: Three Liriomyza species COI genes.
In Line 109, you miss 0.5ul of something to have 25ul of PCR mix... Is the DNA volume of 2.5ul?
Response: We have corrected this error. ddH2O should be 6.5μL in Line 112.
In line 118, I understand the reference for this description is Figure 1. I think you should add at the end of each description not only that we can visualize the distiphallus in Fig.1 but also which between trifolii, sativae and huidobrensis is Fig.1A, B or C... Moreover, the pictures are not clear to me... For example I cannot see any clear difference between A and B; both seems to me that have a distal bulb with a long basal stem. I think the pictures should be in higher resolution that these ones, as well as better annotated. For example, even if there is a scale it is pretty hard to see whether the basal stem is longer in L. trifolii than in the other species
Response: These concerns and errors have been corrected. In Figure 1, we added a red line to indicate the length of basal stem. We have re-cited the figure1 according to the description of each species as figure 1A, figure 1B and figure 1C. Similar figure citation insufficient in the manuscript have been checked and revised.
In Line 125, References?
Response: Reference has been added in Line 130.
Reference:
Spencer K A. 1973. Agromyzidae (Diptera) of economic importance 9, Series Entomologica. Bath, The Hague Publishers.
In Line 128, This sentence is not clear to me...Which character could you observe that was not as expected (i.e., inner and outer vertical setae on yellow ground)? What does it mean: 'outer inner vertical setae' at line 129?
Response: I'm sorry to confuse you. Yes, Table 2 should be showed that the number or ratio of samples with different characteristics of the vertical setae position among the three Liriomyza species. We have re-created Table 2 by the “unit of observation individual”. It should be “outer or inner vertical setae” at line 129.
The classical types of inner and outer vertical setae for three Liriomyza species (see below).
In Line 134, Again can you refer to the figure before in the text, every time you describe the characters? And in particular, can you explain which of the Liriomyza species you analyze is represented by Fig.2A, B, C, D, E, till Fig. 2?
Response: These concerns and errors have been corrected. We have re-cited the figure 2 according to the description of each species as figure 2A-C, figure 2D-F and figure 2G-I. Similar figure citation insufficient in the manuscript have been checked and revised.
In line 136, Rephrase the sentence... What does it mean? Is the ratio the wing pattern itself? Was this difference significant? If yes, can you tell the f and p value? If it was not significant, you cannot use this character even to distinguish L huidobrensis....
Response: The ratio of the length of the ultimate section of vein CuA1 divided by the penultimate section (a and b section). In this study, a is 2.70 ± 0.31 times length of b in L. trifolii, and a is 2.72 ± 0.37 times length of b in L. sativae. For L. huidobrensis, a is 2.20 ± 0.24 times the length of b (F2,237 = 7.345, P < 0.05). Although the ratio of L. huidobrensis was different from the other two species (P < 0.05), there is no significant difference between L. trifolii and L. sativae (P = 0.907).
In Line 142, In many individual that miss the dm-cu what? Maybe you should delete 'which'... Furthermore, what the dm-cu is? Maybe you should explain the morphological abbreviation before describing the specimens as in all morphological description/re-description....
Response: We have corrected this error., “which” has been deleted. The wing patterns of Liriomyza species see above.
In line 145, Same as before... Refer to figure while describing please. Even if you define the figure annotation in the caption, it would be easier to know already from the main text which part of the Figure look at
Response: These concerns and errors have been corrected. We have re-cited the Figure 5.
In Line 153, Are these number the single sites? So did you design the primers based on one nucleotide difference? How can you say that? From this manuscript, it seems that you did not use any other Liriomyza cox1 sequence apart from the three in which you are interested in. So how can you be sure that the reverse primer did not amplify other species? Moreover, I can see from the fig 6 that there is also a A/C mutation...Did you try to design the reverse primer 10 nucleotides before the 1181 position? It seems a more conserved region to me, for designing a common primer.
Response: Yes, we ignored interspecific variation in previous manuscript and only focus on the differences among species. We have selected different geographical populations of those three Liriomyza species to verify the reliability of our primers. The results showed that different geographical populations did not affect the reliability of primers. Primers work in different geographic populations, based on gel electrophoresis and sequencing (see above). In fact, we have verified the reliability of these primers in daily identification experiments more than hundreds of times. In addition, we have supplemented samples of different developmental stages of L. sativae, and L. huidobrensis and different geographical populations of those three species. We believe that the species-specific primers designed in this study is reliable.
In Line 155, What does it mean?
Response: The characteristics of our multiplex PCR assay: (1) the position of forward primers was selected to produce <1000bp amplicons in combination with the reverse primer which can easy for sequencing (generally, the length of one commercial sequencing is ~ 1000 bp). (2) It also to leave at least 300bp nucleotides between each species which can easy to distinguish on gel. (3) In addition, the selected primer sites where the number of differential nucleotides >2 bp for the specificity of the primers.
In Line 158, Did you sequenced these multiplex PCR product to be sure that the primers worked well in selecting the species? Because considering the difficulty in the morphological identification, it would be better to have a double check by sequencing the multiplex PCR products before saying that this is THE method to identify these three Liriomyza species... Finally, are you saying also that this method did not work for all the developmental stages of L. sativae and L. huidobrensis? If yes, where is the utility of this method if you cannot identify larvae or adults of the other two pest species?
Response: It was sequenced and the sequencing results matched the species. Accession number in Table 1 has been removed and changed to new accession numbers of sequences by species-specific primer designed in this study. We have supplemented samples of different developmental stages of L. sativae, and L. huidobrensis in Figure 7.
In Line 167, Maybe you should add some reference to this statement.
Response: These concerns and errors have been corrected. Reference has been added in Line 176.
In line 187, Is this true for all the Liriomyza species or only for the three under study?
Response: Only for the three under study, the error has been corrected.
In line 193, What does it mean? Not having enough morphological differentiation does not mean a lack of deep research...
Response: The error has been corrected.
In Line 207, ...if it only worked for all developmental stages of all the three species. And this seems to not be the case. If I am wrong, you should definitely improve the results section.
Response: Yes, this multiplex PCR method worked for all developmental stages of all the three species.
In Line 219, I think this is a bit too much of an overstatement... In my opinion, it cannot be a reference study if you confirm that the morphological characters alone cannot be resolutive to identify the three Liriomyza species. And, again in my opinion, it cannot be a reference study if you cannot apply the multiplex PCR to all the developmental stages of all the three species, as well as if you do not specify whether you sequenced the multiplex PCR product or not...
Response: The sentence has been changed to “This study provides valuable tools for the identification of Liriomyza spp. using both morphological and molecular criteria.”. In this manuscript, morphological characters of three Liriomyza species are examined and evaluated, we think this study can provide some reference for the identification of Liriomyza species. For molecular level, this multiplex PCR method worked for all developmental stages of all the three species.
" | Here is a paper. Please give your review comments after reading it. |
9,782 | "<ns0:abstract xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:p>After the first confirmed case of the novel coronavirus disease (COVID-19) was found, it is of considerable significance to divide the risk levels of various provinces or provincial municipalities in Mainland China and predict the spatial distribution characteristics of infectious diseases. In this paper, we predict the epidemic risk of each province based on geographical proximity information, spatial inverse distance information, economic distance and Baidu migration index. Simulation study revealed that the information based on geographical economy matrix and migration index could well predict the spatial spread of the epidemic. The results reveal that the accuracy rate of the prediction is over 87.10% with a rank difference of 3.1. The results based on prior information will guide government agencies and medical and health institutions to implement responses to major public health emergencies when facing the epidemic situation.</ns0:p></ns0:div>
</ns0:abstract>
<ns0:body xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:head n='1.'>Introduction</ns0:head><ns0:p>The ongoing outbreak of coronavirus disease in 2019 (COVID- 19), has caused 4653 deaths, along with 86,045 confirmed cases and four suspected cases in China, as of June 18, 2020 (24:00 GMT+8), according to the National Health Commission of the People's Republic of China (NHCPRC, 2020). The recorded deaths associated with COVID-19 notably exceeds the other two coronaviruses (severe acute respiratory syndrome coronavirus, SARS-CoV, and Middle East respiratory syndrome coronavirus, MERS-CoV). As the epidemic continues to spread, it poses a high threat to global public health and economy (Bogoch et al., 2020; J.T. Wu et al., 2020). While the epidemic situation in Mainland China has been controlled, the outbreak outside Mainland China is reported to begin on a large scale. According to the World Health Organization (WHO), as of June 18, (2020), a total of 200 countries (or regions) have found confirmed cases, with a total of 14406440 confirmed cases and 601846 deaths. The United States, India, Brazil, South Africa and Colombia are the Top-5 countries with a more severe epidemic, among which 3833271 cases have been confirmed in The United States, with a total death of 142877 people. The emergence of COVID-19 coincides with the largest population migration season in China, that is, the spring festival tourism season. The virus spreads rapidly throughout the country and around the world. At the early stage of the outbreak, most cases were scattered, and some were linked to the Hunan Seafood Wholesale Market (J.T. Wu et al., 2020). The Chinese government has implemented control measures, including setting up special hospitals and travel restrictions to mitigate the spread of the virus. Besides, the 31 provinces, districts and cities in Mainland China have also launched the first-level response to public health emergencies. On January After the first case is confirmed, scientific precautionary measures need to be taken. Due to the vast land of Mainland China and natural resources, the natural and economic conditions of each province and city are different. Thus, at the initial stage of the epidemic development, we need to estimate the risk of the epidemic in each province and provincial municipality, which could guide the authorities concern in carrying out preventive measures accordingly. Therefore, it is essential to forecast the development of the epidemic situation in each province and city, which will be related to the allocation of medical resources and ensure the supply of food, such as rice and water, to cater people's need. The review of the novel coronavirus has made many outstanding achievements which relate to lots of aspects from the biological characteristics of the virus, the clinical characteristics of patients and the prediction of the number of people infected, future preventive measures as well as risk assessment. However, there are no much works in the literature on the division of regional risks in Mainland China. The existing literature only evaluated the epidemic situation and the related factors affecting the epidemic situation, and did not analyze the specific situation of regional epidemic situation, nor did it analyze how the regional differences of epidemic situation formed. These modeling methods are not specific to the region, which is not enough to provide accurate and useful control suggestions. Based on the researches before the event and little factual information available, the local governments can acquire some practical virus control advice to guide strategies for situational awareness and intervention, reducing the probability of regional infection, and suppressing the epidemic as soon Manuscript to be reviewed corresponding geographical distance, economic distance and Baidu migration index.</ns0:p><ns0:p>The main contribution of this paper is to use fewer epidemic data or even zero epidemic data to simulate the spread of an epidemic in Mainland China for a period of time. It is because we cannot gets the data of the epidemic before it comes. Additionally, using the data of the epidemic situation itself to predict the epidemic situation, its practicality and generalization are not very strong. For better control the epidemic situation, it is necessary to divide the risk of the epidemic situation in each region, which is neither wholly sealed nor free control. However according to the risk level of the epidemic situation in different regions, the corresponding epidemic control measures should be taken. In the classification of the epidemic situation in 31 provinces or provincial municipalities in Mainland China, our method only predicted the epidemic level of 4 provinces or provincial municipalities incorrectly, which is an excellent academic achievement in the actual epidemic prevention and control.</ns0:p></ns0:div>
<ns0:div><ns0:head n='2.'>Inspiration and Novelty</ns0:head><ns0:p>The first case of novel coronavirus confirmed in Wuhan was reported in December 2019. The study covers the epidemic situation in 31 provinces or provincial municipalities in Mainland China. Firstly, the most apparent feature of the epidemic development is that the number of confirmed cases in neighbouring provinces or provincial municipalities of Wuhan is significantly higher than that in other parts of Mainland China. Thus, the authors verified whether the spread of the epidemic is related to the proximity of geographical space; secondly, the more developed cities are, the more infected cases recorded compared with economically backward provinces or provincial municipalities. Because the highly developed cities have closer economic exchanges with other cities, which resulted in a large number of population flows and hence the large-scale spread of the epidemic; thirdly, the migration data of the population will effectively assist the prediction of infectious diseases, especially the population emigration rate from the severely affected areas of the epidemic, which has immeasurable value for the prediction of infectious cases of the epidemic. Based on these three characteristics of the outbreaks, inspired by the actual epidemic data, this paper studies the regional characteristics of the epidemic development, and predicts the provinces or provincial municipalities risk level of the epidemic.</ns0:p><ns0:p>As for the novelty and importance of this work, this research work is rarely seen in previous studies. First of all, as far as novelty is concerned, we try to use as little or no real data as possible to predict the real development of the epidemic situation, which is the starting point of our research work. Because using the data of the epidemic situation itself to predict the development of the epidemic in the future might be inappropriate.</ns0:p><ns0:p>Because of the coming of the next large-scale epidemic, we do not have data on the epidemic itself, but we still need to carry out prevention and control of the epidemic. Zero epidemic data is the value of our research work, but can produce immeasurable value in the next outbreak.</ns0:p><ns0:p>The importance of this work is self-evident as the correctness of epidemic prevention and control not only involves the safety of life and property but also involves the stable operation of society. The prevention and control of the epidemic situation do not mean that it is completely sealed off. Still, some put and lose, since total closure means economic stagnation, which is not an optimal strategy. We need to carry out corresponding control strategies according to the severity of the epidemic, which is the optimal trade-off between economic stagnation and epidemic prevention and control. This work is the foothold of our whole research, helping each region to classify the epidemic situation according to its severity.</ns0:p></ns0:div>
<ns0:div><ns0:head n='3.'>Materials and Methods</ns0:head></ns0:div>
<ns0:div><ns0:head n='3.1'>Description of the latest epidemic data</ns0:head><ns0:p>China had 86045 COVID-2019 cases by 24:00 July 18, 2020, of which 2007 cases were imported from abroad. This paper only studies the epidemic situation in China, so the imported cases are removed from the samples to obtain the domestic epidemic data. The specific epidemic data and ranking are shown in Table <ns0:ref type='table'>1</ns0:ref>.</ns0:p><ns0:p>At present, the epidemic situation in Mainland China is stable, with 68135 confirmed cases. Hubei Province, the centre of the epidemic situation, account for 79.19% of the total confirmed cases. It shows that there is a strong relationship between the spread of the epidemic and the regional spread speed of the epidemic is very fast. In the six provinces adjacent to Hubei, namely Anhui, Hunan, Shaanxi, Jiangxi, Henan and Chongqing, the number of confirmed cases reached 5123, accounting for 5.95% of the total confirmed cases. It shows that in the early stage of the epidemic, as long as the seven provinces are well controlled, 85.14% of the cases in the whole country can be stabilized, which is an essential measure for epidemic prevention and control. With the global spread of the epidemic, China's imported cases began to increase, especially in economically developed provinces, such as Beijing, Shanghai, Guangdong; and some import and export and tourism provinces, such as</ns0:p><ns0:p>Heilongjiang and Inner Mongolia.</ns0:p></ns0:div>
<ns0:div><ns0:head n='3.2'>Epidemic prediction based on spatial geographic adjacency information</ns0:head><ns0:p>To study the spatial clustering characteristics of the novel coronavirus disease, we first examine the geographical spatial correlation between regions. The spatial weight W adopts a simple geographical weight, that is, for 31 provinces or provincial municipalities across the country, the weight of 1 is assigned if there is a common boundary between them, 0 and if they are not adjacent. We simply call this kind of spatial matrix as a 0-1 matrix, and its basic form is ( <ns0:ref type='formula'>1</ns0:ref>) Manuscript to be reviewed China.</ns0:p><ns0:formula xml:id='formula_0'>) ( j i j and i j and i w ij      adjacent not are 0 adjacent are 1</ns0:formula><ns0:p>According to the response of major public health emergencies in China, there are four levels(I, II, III and IV). Therefore, it is necessary to divide the epidemic situation into four levels based on the risk and severity, to facilitate the corresponding provinces or provincial municipalities to initiate the corresponding health emergencies response. Based on the above geographical adjacency information (GAI), we partitioned the 31 provinces or provincial municipalities into four risk levels with Hubei as the centre of infectious diseases. Table <ns0:ref type='table'>3</ns0:ref> presents the specific hazard classification. According to the geographical proximity between the province and the epidemic centre, the highest risk level is level 1; if the province is adjacent to the epidemic centre, it is level 2; if the province is adjacent to a level 2 Province, it is level 3, and so on. To show the transmission process of the epidemic more clearly, we have drawn Figure <ns0:ref type='figure'>1</ns0:ref>. The case first spread in the centre of the epidemic 17</ns0:p><ns0:p>(Hubei), then to its neighbouring provinces (1-2 level), then to its neighbouring provinces too (2-3 level), and finally to 31 provinces or provincial municipalities in China(3-4 level).</ns0:p><ns0:p>From Table <ns0:ref type='table'>3</ns0:ref> and Figure <ns0:ref type='figure'>1</ns0:ref>, we concluded that with Hubei as the centre, all other provinces or provincial municipalities in Mainland China are reachable through at most two provinces or provincial municipalities. As the capital of Hubei Province, Wuhan is known as the 'thoroughfare of nine provinces'. It has the largest water, land and air transportation hub in endoland region of China and the shipping centre in the middle reaches of the Yangtze River. Its high-speed rail network connects most part of the country, and it is the only city in the central part of China that can directly navigate five of the World continents. Therefore, the geographical location of Wuhan, Hubei Province, led to the national outbreak of the novel coronavirus disease. However, if only based on the 0-1 adjacency of geographic information, the interaction infection between provinces or provincial municipalities is not considered. For example, the epidemic situation in Hubei may affect Anhui, Jiangxi, Henan, Hunan, Chongqing and Shaanxi. In turn, the epidemic situation in Anhui may also affect Hubei, Jiangsu, Jiangxi, Shandong, Henan and Zhejiang. In consideration of this mutual influence, we still take Hubei as the centre of the epidemic situation, assigning Hubei 'epidemic index 1' according to the principle of outward diffusion and one-step diffusion (Martini NF,2014), and spread its epidemic index outward. Six neighbouring provinces or provincial municipalities obtained 1/6 of the epidemic index, respectively. By analogy, after Anhui obtained the epidemic index, it was spread to six neighbouring provinces or provincial municipalities, each of which obtained 1 / 6 of the epidemic index of Anhui. After simple programming and</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:05:49417:1:3:NEW 14 Aug 2020)</ns0:p><ns0:p>Manuscript to be reviewed simulation 1 , the corresponding epidemic index and ranking are obtained as presented in Table <ns0:ref type='table' target='#tab_4'>4</ns0:ref>.</ns0:p><ns0:p>Through simple geographical adjacency and epidemic transmission principle, the epidemic index can reflect the risk situation of provinces or provincial municipalities. Although there is a particular gap in the real situation, it has specific reference value for epidemic prevention. The risk level of provinces or provincial municipalities according to the epidemics index is given in Table <ns0:ref type='table'>5</ns0:ref>, keeping the consistency of the number of provinces at each level reference to the risk level provides in Table <ns0:ref type='table'>3</ns0:ref>.</ns0:p><ns0:p>In comparison with the real data 2 , we found that there is a difference between the predicted and the real level of epidemic in the provinces or provincial municipalities and only depend on the geographical adjacency.</ns0:p><ns0:p>From the epidemic index and risk level (Table <ns0:ref type='table' target='#tab_4'>4</ns0:ref> and Table <ns0:ref type='table'>5</ns0:ref>), 11 (Zhejiang) and 19 (Guangdong) of the third level risk provinces or provincial municipalities shifted down to the second level; 22 (Chongqing) and 27</ns0:p><ns0:p>(Shaanxi) of the second level risk provinces or provincial municipalities moved up to the third level risk provinces or provincial municipalities. This gap indicates that the spread of the epidemic is not only related to the geographical proximity, but also the economic level. The higher the economic level of a province is, the closer its interaction with other provinces is, which is more likely to lead to the spread of novel coronavirus disease. We can see the difference between the estimated results of the epidemic index of the third risk level provinces or provincial municipalities and the real risk level (Table <ns0:ref type='table'>5</ns0:ref>-Real division results), which further verifies the previous hypothesis. 1 (Beijing) and 9 (Shanghai) entered the third level of the real situation (Table <ns0:ref type='table'>5</ns0:ref>) from the fourth level of the epidemic index; 5 (Inner Mongolia), 24 (Guizhou), 26 (Tibet) and 28 (Gansu) entered the fourth level of the real situation from the third level of the epidemic index. To sum up, we need to find a matrix that can measure the economic distance between every two provinces and then apply it to explain the transmission of novel coronavirus disease.</ns0:p></ns0:div>
<ns0:div><ns0:head n='3.3'>Epidemic prediction based on spatial Euclidean inverse distance</ns0:head><ns0:p>Before we carry out the impact of economic distance on the spread of the epidemic, we need to measure the impact of Euclidean distance on the spread of the epidemic. It is because, the simple spatial geographic information is adjacent to each other, whether it can comprehensively and accurately measure the spread of the epidemic or not remains a discussion issue. The point is that, can we verify whether the spread of epidemic will <ns0:ref type='bibr' target='#b0'>1</ns0:ref> The principle of programming is that according to the theory that the epidemic situation spreads from the center to the outside, the epidemic index is equally distributed to all the neighbors connected with it. The provinces or provincial municipalities that get the epidemic index are then transmitted to all their neighbors. After a round of transmission, we calculate the epidemic transmission index of 31 provinces and provincial municipalities. (2)</ns0:p><ns0:formula xml:id='formula_1'>      j i j i d W ij ij 0 / 1</ns0:formula><ns0:p>Space anti-Euclidean distance matrix can realize simple multi-level propagation. That is to say, novel coronavirus disease cases in one province or municipality may spread to 30 other provinces and municipalities throughout the country. The closer the distance, the higher the impact coefficient of the epidemic. Based on the theory of distance inverse weighting (Zubaedah R et al., 2020), we still take Hubei as the centre of the epidemic situation, according to the principle of outward diffusion and multi-step diffusion, we assign Hubei 'epidemic index 1' and spread its epidemic index outward. The remaining 30 provinces or provincial municipalities in the country obtained the epidemic index , respectively. By analogy, after Anhui obtained the epidemic index,</ns0:p><ns0:formula xml:id='formula_2'>ij d / 1</ns0:formula><ns0:p>then each of the remaining 30 provinces or the provincial municipalities in Anhui, obtained the epidemic index of Anhui. In this way, the transmission of epidemic index in all provinces or provincial municipalities is ij d / 1 completed, and then calculated the final epidemic index obtained in each province and city. After programming and simulation <ns0:ref type='bibr' target='#b3'>3</ns0:ref> , Table <ns0:ref type='table'>6</ns0:ref> displays the corresponding epidemic index and ranking.</ns0:p><ns0:p>After the inverse weighting of Euclidean distance, the ranking of epidemic index becomes more reasonable.</ns0:p><ns0:p>Therefore, according to the hazard classification in Table <ns0:ref type='table'>3</ns0:ref>, the hazard levels of provinces or provincial municipalities based on the Euclidean inverse distance are shown in Table <ns0:ref type='table'>7</ns0:ref> below.</ns0:p><ns0:p>From Table <ns0:ref type='table'>7</ns0:ref>, we can observe that the simulation results are more reasonable than that of Table <ns0:ref type='table'>5</ns0:ref>. For instance，Guangdong (19) is not adjacent to Hubei Province, but it leaps from the third level dangerous city in Table <ns0:ref type='table'>5，to</ns0:ref> the second level dangerous city under the inverse distance matrix, which is more consistent with the real situation. The number of errors in the three dangerous provinces or provincial municipalities of the predicted results and the real epidemic situation level has changed from 7 in Table <ns0:ref type='table'>5</ns0:ref> to 4 in Table <ns0:ref type='table'>7</ns0:ref>, which is more accurate and reasonable. However, there are still some shortcomings. 11 (Zhejiang), 1 (Beijing) and 3 (Hebei) are far away from Hubei, but their predicted risk levels based on the inverse distance matrix all show a backward shift.</ns0:p><ns0:p>Similarly, 26 (Tibet), 28 (Gansu) and 31 (Xinjiang) are economically backward provinces, but the predicted risk levels have moved forward. Based on the above analysis, it is necessary to introduce economic distance to simulate the spread of the epidemic.</ns0:p></ns0:div>
<ns0:div><ns0:head n='3.4'>Epidemic prediction based on geographical distance and economic distance</ns0:head><ns0:p>Predicting the spread of the epidemic considering only the geographical distance of the provinces or provincial municipalities can reflect limited information. Therefore, we further take into account the economic influence among regions, to better reflect the reality of epidemic transmission. For example, Tianjin connects with Beijing and Hebei, but the two regions have different economic ties with Tianjin. Based on China's economic reality, we assume that the closer the economic exchanges between two cities with the same economic level, the higher the spread of the epidemic, and therefore this study uses the difference of per capita GDP between regions as an indicator to measure the economic distance between regions referring to Lin Guangping (2005). Its basic form is</ns0:p><ns0:formula xml:id='formula_3'>       j i j i y y E j i ij 0<ns0:label>(3)</ns0:label></ns0:formula><ns0:p>Where, is the average per capita GDP of the i th province and city in 2014-2018. We construct the i y geographical-economic matrix (combining the geographical and economic information) based on the spatial inverse distance matrix and economic distance matrix (Standardization). The basic form of the geographicaleconomic matrix is</ns0:p><ns0:formula xml:id='formula_4'> W (4) ij ij ij E W W    </ns0:formula><ns0:p>Where, is the weighting coefficient of geographical distance and economic distance. A simulated result <ns0:ref type='bibr' target='#b5'>4</ns0:ref>  for the epidemic index and its ranking, taken ＝0.5, is presented in Table <ns0:ref type='table' target='#tab_5'>8</ns0:ref>.</ns0:p></ns0:div>
<ns0:div><ns0:head></ns0:head><ns0:p>According to the hazard classification in Table <ns0:ref type='table'>3</ns0:ref>, we present the hazard levels of provinces or provincial <ns0:ref type='bibr' target='#b5'>4</ns0:ref> The principle of programming is based on the theory of epidemic spreading from the center to the outside. However, this simulation is different from the simulation implementation in Section 3.2 and 3.3, because the spread from each region to another region is not completely equal weight spread, but the reciprocal weighted spread based on geographical and economic distance. In other words, the closer the two regions are in geographical and economic distance, the greater the epidemic index will be. See attachment code for details.</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:05:49417:1:3:NEW 14 Aug 2020)</ns0:p><ns0:p>Manuscript to be reviewed municipalities based on the geographical economic distance (GED) in Table <ns0:ref type='table'>9</ns0:ref>.</ns0:p><ns0:p>Form Table <ns0:ref type='table' target='#tab_5'>8</ns0:ref>, the result of the matrix of geographical-economic, indicates that it may be more reasonable to predict the spread of the epidemic, taking into account both the geographical and economic distances. The ranking of 1 (Beijing) increased from 28 to 23, and that of 2 (Tianjin) increased from 31 to 28, but the predicted risk level is still not well adjusted. 13 (Fujian) prediction results still move forward, because Fujian itself is relatively close to Hubei Province in geographical distance, and the per capita GDP is relatively high. The epidemic index of this kind of provinces or provincial municipalities is challenging to measure simply by geographical distance and economic distance, so the next part of this paper will introduce Baidu's migration index.</ns0:p></ns0:div>
<ns0:div><ns0:head n='3.5'>Epidemic prediction based on migration index weighting of Baidu</ns0:head><ns0:p>Usually, when an epidemic occurs, the more people moving out of a central city, the higher the spread of the epidemic. The spread of the epidemic can thus be closely related to the number of people moving out of the central city (Wuhan). Baidu 5 provides daily real-time migration rate; therefore, we quote January 28 and 29 <ns0:ref type='bibr' target='#b8'>6</ns0:ref> and No. 30, the emigration rate from Wuhan to 31 provinces or provincial municipalities in Mainland China, and a simple average, as the migration index from Wuhan. In this paper, the epidemic index generated by geographicaleconomic distance and the Baidu migration index (BMI) is used to weigh the impact of the Hubei epidemic on the whole country. Simulation study 7 reveals that when the epidemic index and the Baidu migration index are 1:2 weighted, it well reflects the spread of the epidemic. Table <ns0:ref type='table'>10</ns0:ref> gives the weighted epidemic index.</ns0:p><ns0:p>After the weighting of the Baidu migration index, the predicted epidemic index and ranking are more closely in line with the reality of epidemic transmission. In order to understand the difference between the ranking of the epidemic index and the real situation, we took the rank of the ranked difference by using the real epidemic data on March 4, 2020, as shown in Table <ns0:ref type='table' target='#tab_1'>11</ns0:ref> below.</ns0:p><ns0:p>According to the risk level in Table <ns0:ref type='table'>3</ns0:ref>, the risk level of provinces or provincial municipalities weighted by Baidu migration index is shown in Table <ns0:ref type='table' target='#tab_3'>12</ns0:ref>.</ns0:p><ns0:p>From the information reflected in Table <ns0:ref type='table'>10</ns0:ref>-12, the epidemic index weighted by Baidu migration index can accurately reflect the actual epidemic information. The difference between the predicted epidemic ranks and the real one is relatively small. In one hand, 1 (Beijing) and 3 (Hebei) have also risen from the third level of risk cities predicted by geographical-economic distance to the second level of risk cities. In the other hand, 26 (Tibet), 28 (Gansu) and 31 (Xinjiang) are economically backward provinces. Based on the geographical-economic distance prediction, they are third-level dangerous provinces, and a well-corrected migration index weighting based on Baidu. In the prediction of provinces or provincial municipalities with a difficult epidemic situation, we note a reversed prediction ranking for Zhejiang and Jiangsu. Because Jiangsu's economic strength is higher than Zhejiang's, and after the outbreak, there were 18800 people from Wuhan, Hubei Province to Wenzhou, Zhejiang Province from January 23 to 27 <ns0:ref type='bibr' target='#b11'>8</ns0:ref> , which led to the deviation of our model prediction.</ns0:p><ns0:p>The prediction ranking deviation of Heilongjiang Province is significant, because Harbin, as a famous ice and snow tourism city, is the first choice of many tourists, especially in the epidemic area, which provides objective conditions for epidemic input. Before the outbreak, there was a massive flow of people during the ice and snow tourism season. Data shows that from December 1, 2019, to January 31, 2020, despite the impact of the epidemic, Harbin still receives about 70000 Hubei tourists, including 43899 registered accommodation and 10450 Wuhan tourists <ns0:ref type='bibr' target='#b12'>9</ns0:ref> .</ns0:p></ns0:div>
<ns0:div><ns0:head n='4.'>Results</ns0:head><ns0:p>From section 3.2 to section 3.5, this study discusses the epidemic situation based on geographic adjacency information (GAI), Euclidean inverse distance (EID), geographical-economic distance (GED) and Baidu migration index (BMI). Tables 13-A to 13-D provides the predicted and real results. Confusion matrix (Tables <ns0:ref type='table' target='#tab_7'>13</ns0:ref>) shows that the accuracy of our prediction results has increased from 54.84% to 87.10%, and the prediction results are stable. The dynamic spread of the epidemic does not reduce the robustness of our model. Our paper was written in March 2020 and revised in July 2020, but the difference between the predicted level and the actual epidemic development remains unchanged.</ns0:p></ns0:div>
<ns0:div><ns0:head n='5.'>Discussion</ns0:head><ns0:p>8 Data link: https://baijiahao.baidu.com/s?id=1657417825137559938&wfr=spider&for=pc</ns0:p><ns0:p>The research results of this paper are of considerable practical significance, but there are still some areas to be improved in the future. First, as the global epidemic continues to spread, imported cases should be considered when predicting the spread of the epidemic in Mainland China. One can add the import and export volume of each province to simulate the spread of the epidemic; Second, the transmission speed of the novel coronavirus to various provinces and municipalities in Mainland China is also related to the traffic level of various provinces and municipalities and can be simulated by the number of railways and highways in various provinces and municipalities. Third, the epidemic prevention and control measures of a province have a tangible impact on the spread of the epidemic. We can position the quality of epidemic prevention and control measures through the data of public opinion.</ns0:p><ns0:p>Although the research approach of this manuscript needs to be improved to achieve more comprehensive and systematic results. But it does achieve a good prediction accuracy, and the accuracy of this prediction does not appear any deviation with the development of the epidemic, so the robustness of the model is very good.</ns0:p><ns0:p>Therefore, for researchers and policy makers, the policies and measures in next outbreak can be based on the results of our model. When the first case is found, we can quickly classify the epidemic level in China in the future. According to the level results, the corresponding level of medical and health response was carried out. This is the optimal strategy between total closure and free control, which can minimize the economic loss and case infection rate.</ns0:p></ns0:div>
<ns0:div><ns0:head n='6.'>Conclusions</ns0:head><ns0:p>In this paper, we simulate the spread of novel coronavirus disease and used geographical adjacency information, Euclidean spatial distance, geographical-economic distance, and Baidu migration index to predict the spread of the epidemic index and risk level of each province. The conclusions of this paper are as follows.</ns0:p><ns0:p>First, the accuracy of forecasting the risk level of provinces or provincial municipalities based on merely geographical adjacency information is about 54.84%. There are some differences between the simulation results and the actual epidemic situation, and the prediction results have specific reference value. This is because, in the early stage of epidemic development, we do not know much about the virus and the spatial information of the whole region. In a short period, we can start the corresponding level response of major public health emergencies according to simple geographic information.</ns0:p><ns0:p>Secondly, the accuracy of forecasting the risk level of provinces or provincial municipalities based on the spatial Euclidean inverse distance and the geographical-economic distance matrix is about 70%. However, it indicates a better ranking of the epidemic index based on the geographical-economic matrix is more reasonable</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:05:49417:1:3:NEW 14 Aug 2020)</ns0:p><ns0:p>Manuscript to be reviewed than that of the spatial Euclidean inverse distance. The simulated results reveal that the spread of the epidemic is related to the economic level of each province. That is, the higher the economic level of each province, the closer the economic exchanges with other provinces, which provides objective conditions for the large-scale spread of the epidemic. Therefore, when provinces or provincial municipalities start to respond to the level of major public health emergencies, the provinces or provincial municipalities with the highest economic level can appropriately improve the level based on the neighbouring relationship. Thirdly, based on the geographical-economic distance matrix and Baidu migration index, the accuracy rate of the epidemic risk level prediction is 87.10%, which can reflect the real epidemic situation. Through the simulation of these non-epidemic data, we can get the ranking of the epidemic index and the risk level of each province, which has an outstanding practical value.</ns0:p><ns0:p>According to the predicted results, the provincial and municipal governments and medical and health institutions can start the corresponding response of major public health emergencies, and make preparation for medical and health care, and issue a control measures to prevent the spread of the epidemic. These can minimize economic loss and the number of infected people from the epidemic.</ns0:p><ns0:p>The research results of this paper can not only be used as a reference for national prevention and control measures, but also as a reference sample for the prevention and control measures of each prefecture-level city within a province, and even as a vital basis source for the prevention and control measures of other countries in the world. Government officials, experts and scholars of other countries can simulate the spread of epidemic according to their country's situation, predict the spread of epidemic index and the corresponding risk level of provinces or provincial municipalities. Going by this can reduce the risk of the epidemic spread and national economic loss.</ns0:p></ns0:div>
<ns0:div><ns0:head>Figure 1 Epidemic transmission map</ns0:head><ns0:p>Epidemic transmission map Manuscript to be reviewed Manuscript to be reviewed Manuscript to be reviewed Manuscript to be reviewed Manuscript to be reviewed</ns0:p></ns0:div><ns0:figure xml:id='fig_0'><ns0:head>23 ,</ns0:head><ns0:label>23</ns0:label><ns0:figDesc>2020, the local government of Wuhan suspended all public transport and closed all entry-exit traffic. Other cities in Hubei Province announced similar traffic control measures shortly after Wuhan's instructions. Two months later, Wuhan, Hubei province, the thoroughfare of nine provinces, was reopened and 'reconnected' with the outside world on April 8, 2020. Since January 2020, many scholars have studied different aspects of the novel coronavirus, including biological characterization of the virus (R.J. Lu et al., 2020), medical diagnosis (R. Liu et al., 2020; C.C. Lai, et al., 2020), clinical characteristics of patients (W.J. Guan et al., 2020;W.J. Yang et al., 2020;H.J. Chen et al.,2020;H.S. Shi et al.,2020;J. Fuk-Woo Chan et al.,2020;N.S. Chen et al.,2020), comparison with SARS (Wilder-Smith et al., 2020), estimation of the reproductive number (S. Zhang et al., 2020;J. M. Read et al., 2020), future trends and the reporting ratio (Q.Y. Liu et al., 2020;D. Benvenuto et al., 2020;Y. Chen et al., 2020). Unlike other diseases, the prevention and suppression of transmission of infectious diseases have become particularly momentous.</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_1'><ns0:head /><ns0:label /><ns0:figDesc>Samrat K. Dey et al., (2020) use a visual exploratory data analysis approach to analyze the epidemiological outbreak of COVID-19. The result shows that it is highly momentous to readily provide information to begin the evaluation necessary to understand the risks and begin containment activities. Sibylle Bernard Stoecklin et al., (2020) define a contact and follow-up procedure by the level of risk of infection and suggest that effective collaboration between all parties involved in the surveillance and response to emerging threats is required to detect imported cases early and to implement adequate control measures. Maged N. Kamel Boulos1 and Estella M. Geraghty (2020) offers pointers to, and describes, a range of practical online/mobile GIS and mapping dashboards and applications for tracking the 2019/2020 coronavirus epidemic and associated events as they unfold around the world. Péter Boldog et.al (2020) developed a computing tool to assess the risk of outbreaks of COVID-2019 outside China, and consider key parameters, such as: (I) the evolution of cumulative number of cases outside mainland China ; (II) connectivity between destination countries and China, including baseline travel frequency, travel restrictions and the effectiveness of entry inspection; (III) effectiveness of control measures in the country of destination. Many scholars assess the risks of other diseases or infectious diseases, such as cardiovascular (Amoolya Vusirikala et al., 2019), CHIKV (Marina Mariconti et al., 2019), H7N9(Ping Zhang, 2019), swine fever (K. Mintiens et al. 2003), severe dengue in Thailand (Zhiwei Xu, 2019). Mohsen Ahmadi et al., (2020) thought that the effect of climatic factors on spreading of COVID-2019 could play a vital role in the next coronavirus outbreak and the result of sensitivity analysis shows a direct relationship between the population density, intra-provincial movement and the infection outbreak.</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_2'><ns0:head /><ns0:label /><ns0:figDesc>as possible. The development of spatial information systems and spatial econometrics has made it more efficient and convenient for us to measure the distance between regions (entwiste B et al., 1997；Arslan O et.al ,2013). Boyda DC et.al (2019) using GIS and spatial analysis methods, this paper reviewed and summarized the transmission of HIV in Africa. Due to the limitations of simple geographic information, Kurata H (2015) began to use the Euclidean distance matrix to measure the distance between two regions and generally used the Euclidean inverse distance to measure the proximity of two regions (Zubaedah R et al., 2020; Lumijarvi J, 2004). However, European inverse distance still cannot accurately reflect the strength of the relationship between the two regions, and hence, some economists have introduced economic distance into the spatial information system (Bastawrous A et al., 2020; Najafi Allamdarlo H, 2018). In the real epidemic prevention and control, population migration and motion can spread the epidemic rapidly, so it is an essential factor in the epidemic situation (Wong IA, 2020; Yue JC, 2011). In this paper, we introduced these factors into the spatial risk map theory to measure the distance between the regions (Pourghasemi HR,2020;Ijumulana J et.al,2020;Fusade-Boyer,2020) and dynamically measure the spread of the epidemic in China. Pourghasemi HR (2020) used a spatial model and risk map to analyze the epidemic situation in Iran from January 19, 2020, to July 14, 2020. The results show that the spatial model and risk map are significant for the analysis and simulation of epidemic spread. In this paper, we predict the spread trend of the epidemic, divide the risk level of provinces or provincial municipalities, to guide the governments of provinces or provincial municipalities to start the corresponding public health event response. It is achieved based on the initial location of the epidemic and according to the PeerJ reviewing PDF | (2020:05:49417:1:3:NEW 14 Aug 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_3'><ns0:head /><ns0:label /><ns0:figDesc>decline with farther distance, based on the spatial Euclidean inverse distance (EID)? Instead of geographical adjacency information, between all cities given the same weight index. Therefore, from the longitude and latitude data of the provincial capitals of each province in Mainland China, we determine the Euclidean distance for ij d each pair of provinces (i and j), and then construct the space inverse distance matrix based on the Euclidean distance matrix, which is the reciprocal of the distance between space elements. Following the expression of the Euclidean distance matrix.</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_4'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:05:49417:1:3:NEW 14 Aug 2020)</ns0:figDesc></ns0:figure>
<ns0:figure type='table' xml:id='tab_0'><ns0:head>Table 2</ns0:head><ns0:label>2</ns0:label><ns0:figDesc>displays the neighboring information of 31 provinces or provincial municipalities in Mainland</ns0:figDesc><ns0:table /><ns0:note>PeerJ reviewing PDF | (2020:05:49417:1:3:NEW 14 Aug 2020)</ns0:note></ns0:figure>
<ns0:figure type='table' xml:id='tab_1'><ns0:head>1 Table 1</ns0:head><ns0:label>11</ns0:label><ns0:figDesc>Domestic epidemic situation in China (as of 18 June 2020, 24:00 ) 2</ns0:figDesc><ns0:table><ns0:row><ns0:cell /><ns0:cell>*Data</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>3 4 5 6</ns0:cell><ns0:cell>source: News</ns0:cell><ns0:cell>Region</ns0:cell><ns0:cell>Cumulative confirmed</ns0:cell><ns0:cell>Overseas import</ns0:cell><ns0:cell>Domestic confirmed</ns0:cell><ns0:cell>Epidemic level</ns0:cell><ns0:cell>Rank</ns0:cell><ns0:cell>Sina real-</ns0:cell></ns0:row><ns0:row><ns0:cell>7</ns0:cell><ns0:cell>time</ns0:cell><ns0:cell>Hubei</ns0:cell><ns0:cell>68135</ns0:cell><ns0:cell>1</ns0:cell><ns0:cell>68134</ns0:cell><ns0:cell>1</ns0:cell><ns0:cell>1</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell>Guangdong</ns0:cell><ns0:cell>1659</ns0:cell><ns0:cell>264</ns0:cell><ns0:cell>1395</ns0:cell><ns0:cell>2</ns0:cell><ns0:cell>2</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell>Henan</ns0:cell><ns0:cell>1276</ns0:cell><ns0:cell>3</ns0:cell><ns0:cell>1273</ns0:cell><ns0:cell>2</ns0:cell><ns0:cell>3</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell>Zhejiang</ns0:cell><ns0:cell>1270</ns0:cell><ns0:cell>51</ns0:cell><ns0:cell>1219</ns0:cell><ns0:cell>2</ns0:cell><ns0:cell>4</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell>Hunan</ns0:cell><ns0:cell>1019</ns0:cell><ns0:cell>1</ns0:cell><ns0:cell>1018</ns0:cell><ns0:cell>2</ns0:cell><ns0:cell>5</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell>Anhui</ns0:cell><ns0:cell>991</ns0:cell><ns0:cell>1</ns0:cell><ns0:cell>990</ns0:cell><ns0:cell>2</ns0:cell><ns0:cell>6</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell>Jiangxi</ns0:cell><ns0:cell>932</ns0:cell><ns0:cell>3</ns0:cell><ns0:cell>929</ns0:cell><ns0:cell>2</ns0:cell><ns0:cell>7</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell>Shandong</ns0:cell><ns0:cell>797</ns0:cell><ns0:cell>34</ns0:cell><ns0:cell>763</ns0:cell><ns0:cell>3</ns0:cell><ns0:cell>8</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell>Beijing</ns0:cell><ns0:cell>929</ns0:cell><ns0:cell>174</ns0:cell><ns0:cell>755</ns0:cell><ns0:cell>3</ns0:cell><ns0:cell>9</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell>Jiangsu</ns0:cell><ns0:cell>655</ns0:cell><ns0:cell>24</ns0:cell><ns0:cell>631</ns0:cell><ns0:cell>3</ns0:cell><ns0:cell>10</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell>Sichuan</ns0:cell><ns0:cell>599</ns0:cell><ns0:cell>4</ns0:cell><ns0:cell>595</ns0:cell><ns0:cell>3</ns0:cell><ns0:cell>11</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell>Chongqing</ns0:cell><ns0:cell>583</ns0:cell><ns0:cell>7</ns0:cell><ns0:cell>576</ns0:cell><ns0:cell>3</ns0:cell><ns0:cell>12</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell>Heilongjiang</ns0:cell><ns0:cell>947</ns0:cell><ns0:cell>386</ns0:cell><ns0:cell>561</ns0:cell><ns0:cell>3</ns0:cell><ns0:cell>13</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell>Shanghai</ns0:cell><ns0:cell>733</ns0:cell><ns0:cell>391</ns0:cell><ns0:cell>342</ns0:cell><ns0:cell>3</ns0:cell><ns0:cell>14</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell>Hebei</ns0:cell><ns0:cell>349</ns0:cell><ns0:cell>10</ns0:cell><ns0:cell>339</ns0:cell><ns0:cell>3</ns0:cell><ns0:cell>15</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell>Fujian</ns0:cell><ns0:cell>364</ns0:cell><ns0:cell>68</ns0:cell><ns0:cell>296</ns0:cell><ns0:cell>3</ns0:cell><ns0:cell>16</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell>Guangxi</ns0:cell><ns0:cell>255</ns0:cell><ns0:cell>3</ns0:cell><ns0:cell>252</ns0:cell><ns0:cell>3</ns0:cell><ns0:cell>17</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell>Shaanxi</ns0:cell><ns0:cell>322</ns0:cell><ns0:cell>79</ns0:cell><ns0:cell>243</ns0:cell><ns0:cell>3</ns0:cell><ns0:cell>18</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell>Yunnan</ns0:cell><ns0:cell>188</ns0:cell><ns0:cell>14</ns0:cell><ns0:cell>174</ns0:cell><ns0:cell>3</ns0:cell><ns0:cell>19</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell>Hainan</ns0:cell><ns0:cell>171</ns0:cell><ns0:cell>2</ns0:cell><ns0:cell>169</ns0:cell><ns0:cell>3</ns0:cell><ns0:cell>20</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell>Jilin</ns0:cell><ns0:cell>155</ns0:cell><ns0:cell>5</ns0:cell><ns0:cell>150</ns0:cell><ns0:cell>4</ns0:cell><ns0:cell>21</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell>Guizhou</ns0:cell><ns0:cell>147</ns0:cell><ns0:cell>1</ns0:cell><ns0:cell>146</ns0:cell><ns0:cell>4</ns0:cell><ns0:cell>22</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell>Tianjin</ns0:cell><ns0:cell>203</ns0:cell><ns0:cell>66</ns0:cell><ns0:cell>137</ns0:cell><ns0:cell>4</ns0:cell><ns0:cell>23</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell>Shanxi</ns0:cell><ns0:cell>201</ns0:cell><ns0:cell>67</ns0:cell><ns0:cell>134</ns0:cell><ns0:cell>4</ns0:cell><ns0:cell>24</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell>Liaoning</ns0:cell><ns0:cell>164</ns0:cell><ns0:cell>33</ns0:cell><ns0:cell>131</ns0:cell><ns0:cell>4</ns0:cell><ns0:cell>25</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell>Gansu</ns0:cell><ns0:cell>167</ns0:cell><ns0:cell>75</ns0:cell><ns0:cell>92</ns0:cell><ns0:cell>4</ns0:cell><ns0:cell>26</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell>Xinjiang</ns0:cell><ns0:cell>106</ns0:cell><ns0:cell>14</ns0:cell><ns0:cell>92</ns0:cell><ns0:cell>4</ns0:cell><ns0:cell>27</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell>Inner Mongolia</ns0:cell><ns0:cell>249</ns0:cell><ns0:cell>172</ns0:cell><ns0:cell>77</ns0:cell><ns0:cell>4</ns0:cell><ns0:cell>28</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell>Ningxia</ns0:cell><ns0:cell>75</ns0:cell><ns0:cell>2</ns0:cell><ns0:cell>73</ns0:cell><ns0:cell>4</ns0:cell><ns0:cell>29</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell>Qinghai</ns0:cell><ns0:cell>18</ns0:cell><ns0:cell>1</ns0:cell><ns0:cell>17</ns0:cell><ns0:cell>4</ns0:cell><ns0:cell>30</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell>Tibet</ns0:cell><ns0:cell>1</ns0:cell><ns0:cell>0</ns0:cell><ns0:cell>1</ns0:cell><ns0:cell>4</ns0:cell><ns0:cell>31</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>8</ns0:cell><ns0:cell cols='7'>dynamic tracking of novel coronavirus disease (all data sources in this article are from Sina News, if not specified)</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>9</ns0:cell><ns0:cell cols='7'>+ Data link:https://news.sina.cn/zt_d/yiqing0121?ua=iPhone10%2C2__weibo__10.1.1__iphone__os12.4.1&from</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>10</ns0:cell><ns0:cell cols='2'>=10A1193010&wm=3049_0135</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row></ns0:table><ns0:note>PeerJ reviewing PDF | (2020:05:49417:1:3:NEW 14 Aug 2020)</ns0:note></ns0:figure>
<ns0:figure type='table' xml:id='tab_2'><ns0:head>Table 2 (on next page)</ns0:head><ns0:label>2</ns0:label><ns0:figDesc /><ns0:table><ns0:row><ns0:cell>Hazard classification of provinces or provincial municipalities based on geographical proximity information</ns0:cell></ns0:row><ns0:row><ns0:cell>Hazard classification of provinces or provincial municipalities based on geographical</ns0:cell></ns0:row><ns0:row><ns0:cell>proximity information</ns0:cell></ns0:row></ns0:table><ns0:note>PeerJ reviewing PDF | (2020:05:49417:1:3:NEW 14 Aug 2020)</ns0:note></ns0:figure>
<ns0:figure type='table' xml:id='tab_3'><ns0:head>1 Table 2 .</ns0:head><ns0:label>12</ns0:label><ns0:figDesc>The adjacent information of 31 provinces or provincial municipalities</ns0:figDesc><ns0:table><ns0:row><ns0:cell>S/N</ns0:cell><ns0:cell>Region</ns0:cell><ns0:cell>Adjacent information</ns0:cell><ns0:cell>S/N</ns0:cell><ns0:cell>Region</ns0:cell><ns0:cell>Adjacent information</ns0:cell></ns0:row><ns0:row><ns0:cell>1</ns0:cell><ns0:cell>Beijing</ns0:cell><ns0:cell>2、3</ns0:cell><ns0:cell>17</ns0:cell><ns0:cell>Hubei</ns0:cell><ns0:cell>12、14、16、18、22、27</ns0:cell></ns0:row><ns0:row><ns0:cell>2</ns0:cell><ns0:cell>Tianjin</ns0:cell><ns0:cell>1、3</ns0:cell><ns0:cell>18</ns0:cell><ns0:cell>Hunan</ns0:cell><ns0:cell>14、17、19、20、22、24</ns0:cell></ns0:row><ns0:row><ns0:cell>3</ns0:cell><ns0:cell>Hebei</ns0:cell><ns0:cell>1、2、4、5、6、15、16</ns0:cell><ns0:cell>19</ns0:cell><ns0:cell>Guangdong</ns0:cell><ns0:cell>13、14、18、20、21</ns0:cell></ns0:row><ns0:row><ns0:cell>4</ns0:cell><ns0:cell>Shanxi</ns0:cell><ns0:cell>3、5、16、27</ns0:cell><ns0:cell>20</ns0:cell><ns0:cell>Guangxi</ns0:cell><ns0:cell>18、19、24、25</ns0:cell></ns0:row><ns0:row><ns0:cell>5</ns0:cell><ns0:cell>Inner Mongolia</ns0:cell><ns0:cell>3、4、6、7、8、27、 28、30</ns0:cell><ns0:cell>21</ns0:cell><ns0:cell>Hainan</ns0:cell><ns0:cell>19</ns0:cell></ns0:row><ns0:row><ns0:cell>6</ns0:cell><ns0:cell>Liaoning</ns0:cell><ns0:cell>3、5、7</ns0:cell><ns0:cell>22</ns0:cell><ns0:cell>Chongqing</ns0:cell><ns0:cell>17、18、23、24、27</ns0:cell></ns0:row><ns0:row><ns0:cell>7</ns0:cell><ns0:cell>Jilin</ns0:cell><ns0:cell>5、6、8</ns0:cell><ns0:cell>23</ns0:cell><ns0:cell>Sichuan</ns0:cell><ns0:cell>22、24、25、26、27、28、29</ns0:cell></ns0:row><ns0:row><ns0:cell>8</ns0:cell><ns0:cell>Heilongjiang</ns0:cell><ns0:cell>5、7</ns0:cell><ns0:cell>24</ns0:cell><ns0:cell>Guizhou</ns0:cell><ns0:cell>18、20、22、23、25</ns0:cell></ns0:row><ns0:row><ns0:cell>9</ns0:cell><ns0:cell>Shanghai</ns0:cell><ns0:cell>10、11</ns0:cell><ns0:cell>25</ns0:cell><ns0:cell>Yunnan</ns0:cell><ns0:cell>20、23、24、26</ns0:cell></ns0:row><ns0:row><ns0:cell>10</ns0:cell><ns0:cell>Jiangsu</ns0:cell><ns0:cell>9、11、12、15</ns0:cell><ns0:cell>26</ns0:cell><ns0:cell>Tibet</ns0:cell><ns0:cell>23、25、29、31</ns0:cell></ns0:row><ns0:row><ns0:cell>11</ns0:cell><ns0:cell>Zhejiang</ns0:cell><ns0:cell>9、10、12、13、14</ns0:cell><ns0:cell>27</ns0:cell><ns0:cell>Shaanxi</ns0:cell><ns0:cell>4、5、16、17、22、23、28、 30</ns0:cell></ns0:row><ns0:row><ns0:cell>12</ns0:cell><ns0:cell>Anhui</ns0:cell><ns0:cell>10、11、14、15、16、17</ns0:cell><ns0:cell>28</ns0:cell><ns0:cell>Gansu</ns0:cell><ns0:cell>5、23、27、29、30、31</ns0:cell></ns0:row><ns0:row><ns0:cell>13</ns0:cell><ns0:cell>Fujian</ns0:cell><ns0:cell>11、14、19</ns0:cell><ns0:cell>29</ns0:cell><ns0:cell>Qinghai</ns0:cell><ns0:cell>23、26、28、31</ns0:cell></ns0:row><ns0:row><ns0:cell>14</ns0:cell><ns0:cell>Jiangxi</ns0:cell><ns0:cell>11、12、13、17、18、19</ns0:cell><ns0:cell>30</ns0:cell><ns0:cell>Ningxia</ns0:cell><ns0:cell>5、27、28</ns0:cell></ns0:row><ns0:row><ns0:cell>15</ns0:cell><ns0:cell>Shandong</ns0:cell><ns0:cell>3、10、12、16</ns0:cell><ns0:cell>31</ns0:cell><ns0:cell>Xinjiang</ns0:cell><ns0:cell>26、28、29</ns0:cell></ns0:row><ns0:row><ns0:cell>16</ns0:cell><ns0:cell>Henan</ns0:cell><ns0:cell>3、4、12、15、17、27</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>2</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row></ns0:table><ns0:note>* the numbers of provinces or provincial municipalities in this paper are in the order of Table2.3 1</ns0:note></ns0:figure>
<ns0:figure type='table' xml:id='tab_4'><ns0:head>Table 4 .</ns0:head><ns0:label>4</ns0:label><ns0:figDesc>Epidemic index and ranking of 31 provinces or provincial municipalities in China</ns0:figDesc><ns0:table><ns0:row><ns0:cell>Region</ns0:cell><ns0:cell>Epidemic index</ns0:cell><ns0:cell>Rank</ns0:cell><ns0:cell>Region</ns0:cell><ns0:cell>Epidemic index</ns0:cell><ns0:cell>Rank</ns0:cell></ns0:row><ns0:row><ns0:cell>Hubei</ns0:cell><ns0:cell>1.272</ns0:cell><ns0:cell>1</ns0:cell><ns0:cell>Shanxi</ns0:cell><ns0:cell>0.121</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>Anhui</ns0:cell><ns0:cell>0.330</ns0:cell><ns0:cell>2</ns0:cell><ns0:cell>Guangxi</ns0:cell><ns0:cell>0.118</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>Shaanxi</ns0:cell><ns0:cell>0.328</ns0:cell><ns0:cell>3</ns0:cell><ns0:cell>Fujian</ns0:cell><ns0:cell>0.113</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>Jiangxi</ns0:cell><ns0:cell>0.326</ns0:cell><ns0:cell>4</ns0:cell><ns0:cell>Tibet</ns0:cell><ns0:cell>0.090</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>Hunan</ns0:cell><ns0:cell>0.315</ns0:cell><ns0:cell>5</ns0:cell><ns0:cell>Ningxia</ns0:cell><ns0:cell>0.077</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>Henan</ns0:cell><ns0:cell>0.298</ns0:cell><ns0:cell>6</ns0:cell><ns0:cell>Qinghai</ns0:cell><ns0:cell>0.075</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>Chongqing</ns0:cell><ns0:cell>0.271</ns0:cell><ns0:cell>7</ns0:cell><ns0:cell>Xinjiang</ns0:cell><ns0:cell>0.071</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>Cichuan</ns0:cell><ns0:cell>0.205</ns0:cell><ns0:cell>8</ns0:cell><ns0:cell>Jilin</ns0:cell><ns0:cell>0.070</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>Zhejiang</ns0:cell><ns0:cell>0.191</ns0:cell><ns0:cell>9</ns0:cell><ns0:cell>Liaoning</ns0:cell><ns0:cell>0.064</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>Inner Mongolia</ns0:cell><ns0:cell>0.183</ns0:cell><ns0:cell>10</ns0:cell><ns0:cell>Yunnan</ns0:cell><ns0:cell>0.059</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>Hebei</ns0:cell><ns0:cell>0.177</ns0:cell><ns0:cell>11</ns0:cell><ns0:cell>Shanghai</ns0:cell><ns0:cell>0.049</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>Jiangsu</ns0:cell><ns0:cell>0.156</ns0:cell><ns0:cell>12</ns0:cell><ns0:cell>Heilongjiang</ns0:cell><ns0:cell>0.049</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>Guangdong</ns0:cell><ns0:cell>0.142</ns0:cell><ns0:cell>13</ns0:cell><ns0:cell>Beijing</ns0:cell><ns0:cell>0.047</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>Guizhou</ns0:cell><ns0:cell>0.136</ns0:cell><ns0:cell>14</ns0:cell><ns0:cell>Tianjin</ns0:cell><ns0:cell>0.040</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>Gansu</ns0:cell><ns0:cell>0.133</ns0:cell><ns0:cell>15</ns0:cell><ns0:cell>Hainan</ns0:cell><ns0:cell>0.031</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>Shandong</ns0:cell><ns0:cell>0.126</ns0:cell><ns0:cell>16</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row></ns0:table><ns0:note>*After the epidemic index is spread, it is standardized to make the total epidemic index of all provinces or provincial 3 municipalities be 100.1</ns0:note></ns0:figure>
<ns0:figure type='table' xml:id='tab_5'><ns0:head>Table 8 .</ns0:head><ns0:label>8</ns0:label><ns0:figDesc>Epidemic index and ranking of 31 provinces or provincial municipalities in Mainland China 2 (Geographic economic matrix)</ns0:figDesc><ns0:table><ns0:row><ns0:cell>Region</ns0:cell><ns0:cell>Epidemic Index</ns0:cell><ns0:cell>Rank</ns0:cell><ns0:cell>Region</ns0:cell><ns0:cell>Epidemic Index</ns0:cell><ns0:cell>Rank</ns0:cell></ns0:row><ns0:row><ns0:cell>Hebei</ns0:cell><ns0:cell>27.488</ns0:cell><ns0:cell>1</ns0:cell><ns0:cell>Hainan</ns0:cell><ns0:cell>2.307</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>Hunan</ns0:cell><ns0:cell>3.464</ns0:cell><ns0:cell>2</ns0:cell><ns0:cell>Guizhou</ns0:cell><ns0:cell>2.306</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>Jiangxi</ns0:cell><ns0:cell>3.186</ns0:cell><ns0:cell>3</ns0:cell><ns0:cell>Shandong</ns0:cell><ns0:cell>2.291</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>Anhui</ns0:cell><ns0:cell>2.901</ns0:cell><ns0:cell>4</ns0:cell><ns0:cell>Sichuan</ns0:cell><ns0:cell>2.285</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>Henan</ns0:cell><ns0:cell>2.655</ns0:cell><ns0:cell>5</ns0:cell><ns0:cell>Chongqing</ns0:cell><ns0:cell>2.266</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>Fujian</ns0:cell><ns0:cell>2.627</ns0:cell><ns0:cell>6</ns0:cell><ns0:cell>Shanxi</ns0:cell><ns0:cell>2.265</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>Guangdong</ns0:cell><ns0:cell>2.627</ns0:cell><ns0:cell>7</ns0:cell><ns0:cell>Beijing</ns0:cell><ns0:cell>2.252</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>Jiangsu</ns0:cell><ns0:cell>2.625</ns0:cell><ns0:cell>8</ns0:cell><ns0:cell>Heilongjiang</ns0:cell><ns0:cell>2.245</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>Zhejiang</ns0:cell><ns0:cell>2.524</ns0:cell><ns0:cell>9</ns0:cell><ns0:cell>Hebei</ns0:cell><ns0:cell>2.234</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>Guangxi</ns0:cell><ns0:cell>2.500</ns0:cell><ns0:cell>10</ns0:cell><ns0:cell>Ningxia</ns0:cell><ns0:cell>2.204</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>Shanghai</ns0:cell><ns0:cell>2.439</ns0:cell><ns0:cell>11</ns0:cell><ns0:cell>Inner Mongolia</ns0:cell><ns0:cell>2.191</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>Yunan</ns0:cell><ns0:cell>2.390</ns0:cell><ns0:cell>12</ns0:cell><ns0:cell>Tianjin</ns0:cell><ns0:cell>2.183</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>Shaanxi</ns0:cell><ns0:cell>2.383</ns0:cell><ns0:cell>13</ns0:cell><ns0:cell>Qinghai</ns0:cell><ns0:cell>2.122</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>Gansu</ns0:cell><ns0:cell>2.337</ns0:cell><ns0:cell>14</ns0:cell><ns0:cell>Liaoning</ns0:cell><ns0:cell>2.061</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>Tibet</ns0:cell><ns0:cell>2.330</ns0:cell><ns0:cell>15</ns0:cell><ns0:cell>Jilin</ns0:cell><ns0:cell>1.995</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>Xinjiang</ns0:cell><ns0:cell>2.317</ns0:cell><ns0:cell>16</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row></ns0:table></ns0:figure>
<ns0:figure type='table' xml:id='tab_6'><ns0:head>Table 11 .</ns0:head><ns0:label>11</ns0:label><ns0:figDesc>Rank difference between the epidemic index ranking and real epidemic ranking</ns0:figDesc><ns0:table><ns0:row><ns0:cell>Region</ns0:cell><ns0:cell>Prediction Rank</ns0:cell><ns0:cell>Real Rank</ns0:cell><ns0:cell>Rank Difference</ns0:cell><ns0:cell>Region</ns0:cell><ns0:cell>Prediction Rank</ns0:cell><ns0:cell>Real Rank</ns0:cell><ns0:cell>Rank Difference</ns0:cell></ns0:row><ns0:row><ns0:cell>Hubei</ns0:cell><ns0:cell>1</ns0:cell><ns0:cell>1</ns0:cell><ns0:cell>0</ns0:cell><ns0:cell>Hebei</ns0:cell><ns0:cell>17</ns0:cell><ns0:cell>15</ns0:cell><ns0:cell>2</ns0:cell></ns0:row><ns0:row><ns0:cell>Hunan</ns0:cell><ns0:cell>2</ns0:cell><ns0:cell>5</ns0:cell><ns0:cell>3</ns0:cell><ns0:cell>Beijing</ns0:cell><ns0:cell>18</ns0:cell><ns0:cell>13</ns0:cell><ns0:cell>5</ns0:cell></ns0:row><ns0:row><ns0:cell>Henan</ns0:cell><ns0:cell>3</ns0:cell><ns0:cell>3</ns0:cell><ns0:cell>0</ns0:cell><ns0:cell>Hainan</ns0:cell><ns0:cell>19</ns0:cell><ns0:cell>20</ns0:cell><ns0:cell>1</ns0:cell></ns0:row><ns0:row><ns0:cell>Guangdong</ns0:cell><ns0:cell>4</ns0:cell><ns0:cell>2</ns0:cell><ns0:cell>2</ns0:cell><ns0:cell>Guizhou</ns0:cell><ns0:cell>20</ns0:cell><ns0:cell>21</ns0:cell><ns0:cell>1</ns0:cell></ns0:row><ns0:row><ns0:cell>Jiangxi</ns0:cell><ns0:cell>5</ns0:cell><ns0:cell>7</ns0:cell><ns0:cell>2</ns0:cell><ns0:cell>Gansu</ns0:cell><ns0:cell>21</ns0:cell><ns0:cell>26</ns0:cell><ns0:cell>5</ns0:cell></ns0:row><ns0:row><ns0:cell>Anhui</ns0:cell><ns0:cell>6</ns0:cell><ns0:cell>6</ns0:cell><ns0:cell>0</ns0:cell><ns0:cell>Shanxi</ns0:cell><ns0:cell>22</ns0:cell><ns0:cell>23</ns0:cell><ns0:cell>1</ns0:cell></ns0:row><ns0:row><ns0:cell>Jiangsu</ns0:cell><ns0:cell>7</ns0:cell><ns0:cell>9</ns0:cell><ns0:cell>2</ns0:cell><ns0:cell>Xinjiang</ns0:cell><ns0:cell>23</ns0:cell><ns0:cell>27</ns0:cell><ns0:cell>4</ns0:cell></ns0:row><ns0:row><ns0:cell>Zhejiang</ns0:cell><ns0:cell>8</ns0:cell><ns0:cell>4</ns0:cell><ns0:cell>4</ns0:cell><ns0:cell>Heilongjiang</ns0:cell><ns0:cell>24</ns0:cell><ns0:cell>12</ns0:cell><ns0:cell>12</ns0:cell></ns0:row><ns0:row><ns0:cell>Fujian</ns0:cell><ns0:cell>9</ns0:cell><ns0:cell>16</ns0:cell><ns0:cell>7</ns0:cell><ns0:cell>Tibet</ns0:cell><ns0:cell>25</ns0:cell><ns0:cell>31</ns0:cell><ns0:cell>6</ns0:cell></ns0:row><ns0:row><ns0:cell>Shanghai</ns0:cell><ns0:cell>10</ns0:cell><ns0:cell>14</ns0:cell><ns0:cell>4</ns0:cell><ns0:cell>Inner Mongolia</ns0:cell><ns0:cell>26</ns0:cell><ns0:cell>28</ns0:cell><ns0:cell>2</ns0:cell></ns0:row><ns0:row><ns0:cell>Chongqing</ns0:cell><ns0:cell>11</ns0:cell><ns0:cell>10</ns0:cell><ns0:cell>1</ns0:cell><ns0:cell>Liaoning</ns0:cell><ns0:cell>27</ns0:cell><ns0:cell>24</ns0:cell><ns0:cell>3</ns0:cell></ns0:row><ns0:row><ns0:cell>Sichuan</ns0:cell><ns0:cell>12</ns0:cell><ns0:cell>11</ns0:cell><ns0:cell>1</ns0:cell><ns0:cell>Tianjin</ns0:cell><ns0:cell>28</ns0:cell><ns0:cell>22</ns0:cell><ns0:cell>6</ns0:cell></ns0:row><ns0:row><ns0:cell>Shaanxi</ns0:cell><ns0:cell>13</ns0:cell><ns0:cell>18</ns0:cell><ns0:cell>5</ns0:cell><ns0:cell>Ningxia</ns0:cell><ns0:cell>29</ns0:cell><ns0:cell>29</ns0:cell><ns0:cell>0</ns0:cell></ns0:row><ns0:row><ns0:cell>Shandong</ns0:cell><ns0:cell>14</ns0:cell><ns0:cell>8</ns0:cell><ns0:cell>6</ns0:cell><ns0:cell>Qinghai</ns0:cell><ns0:cell>30</ns0:cell><ns0:cell>30</ns0:cell><ns0:cell>0</ns0:cell></ns0:row><ns0:row><ns0:cell>Yunnan</ns0:cell><ns0:cell>15</ns0:cell><ns0:cell>19</ns0:cell><ns0:cell>4</ns0:cell><ns0:cell>Jilin</ns0:cell><ns0:cell>31</ns0:cell><ns0:cell>25</ns0:cell><ns0:cell>6</ns0:cell></ns0:row><ns0:row><ns0:cell>Guangxi</ns0:cell><ns0:cell>16</ns0:cell><ns0:cell>17</ns0:cell><ns0:cell>1</ns0:cell><ns0:cell>Average Rank</ns0:cell><ns0:cell>--</ns0:cell><ns0:cell>--</ns0:cell><ns0:cell>3.1</ns0:cell></ns0:row><ns0:row><ns0:cell>2</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row></ns0:table><ns0:note>PeerJ reviewing PDF | (2020:05:49417:1:3:NEW 14 Aug 2020)</ns0:note></ns0:figure>
<ns0:figure type='table' xml:id='tab_7'><ns0:head>Table 13 (on next page)</ns0:head><ns0:label>13</ns0:label><ns0:figDesc /><ns0:table><ns0:row><ns0:cell>Confusion matrix</ns0:cell></ns0:row><ns0:row><ns0:cell>Confusion matrix</ns0:cell></ns0:row></ns0:table><ns0:note>PeerJ reviewing PDF | (2020:05:49417:1:3:NEW 14 Aug 2020)</ns0:note></ns0:figure>
<ns0:figure type='table' xml:id='tab_8'><ns0:head>Table 13 -</ns0:head><ns0:label>13</ns0:label><ns0:figDesc>A. Confusion matrix based on GAITable 13-B. Confusion matrix based on EID Table 13-C. Confusion matrix based on GED Table 13-D. Confusion matrix</ns0:figDesc><ns0:table /></ns0:figure>
<ns0:note place='foot' n='2'>The Real division results are based on the ranking in Table1and Table5.PeerJ reviewing PDF | (2020:05:49417:1:3:NEW 14 Aug 2020)</ns0:note>
<ns0:note place='foot' n='3'>The principle of programming is based on the theory of epidemic spreading from the center to the outside. However, this simulation is different from the simulation implementation in Section 3.2, because the spread from each region to another region is not completely equal weight spread, but the reciprocal weighted spread based on distance. In other words, the closer the two regions are, the greater the epidemic index will be. See attachment code for details.</ns0:note>
<ns0:note place='foot' n='5'>Baidu migration index website: https://qianxi.baidu.com/? From = Shoubai × city = 0<ns0:ref type='bibr' target='#b8'>6</ns0:ref> On January 29, Wuhan issued the order to seal the city, so this paper selects the day before and the day after as the outward migration index of Wuhan.<ns0:ref type='bibr' target='#b10'>7</ns0:ref> The principle of programming is based on the theory of epidemic spreading from the center to the outside. However, this simulation is different from the simulation implementation in Section 3.2 to 3.4, because the spread from each region to another region is not completely equal weight spread, but the reciprocal weighted spread based on geographical, economic distance and Baidu Index. See attachment code for details.</ns0:note>
<ns0:note place='foot' n='9'>Data link: http://www.hlj.chinanews.com/hljnews/2020/0210/55385.html PeerJ reviewing PDF | (2020:05:49417:1:3:NEW 14 Aug 2020)</ns0:note>
</ns0:body>
" | "PeerJ No.: #2020:05:49417:0:2NEW2 Jun 2020
Title: Regional infectious risk prediction of COVID-19 based on geo-spatial data
Response to Editor and Reviewers
30th July, 2020
Dear Editor,
We would like to express our deepest gratitude and appreciation to you and to the reviewers for those constructive comments and suggestions related to our manuscript. We have completed the suggested revision of the manuscript, as per the editor and reviewers’ comments. We have subsequently checked all of the general and specific comments and errors provided by the editor and reviewers, and we have made necessary changes according to their suggestions. We look forward to your positive response to our collective efforts and we believe that the revised manuscript is now suitable for the publication in PeerJ.
Sincerely,
Cheng Xuewei, Ph.D.
Han Zhaozhou, Ph.D.
Abba Badamasi, Ph.D.
Wang Hong, Ph.D.
Point-by-point responses to editor and reviewer comments
Comments from Editor:
Editor comments (Theerapong Krajaejun)
MAJOR REVISIONS
Dear authors,
Three experts in the field have reviewed your manuscript. Based on their comments, the manuscript has been returned for major revision, giving that the study is new and interesting, but it requires substantial work to improve its data presentation and value, such as a clear rationale of the study, data update, figure(s) for better understanding the results, and extensive literature review and appropriate citations. If you decide to revise the manuscript, please address the reviewers' concerns (listed below) point-by-point.
With kind regards,
Theerapong Krajaejun
Academic Editor, PeerJ
[# PeerJ Staff Note: Please ensure that all review comments are addressed in a rebuttal letter and any edits or clarifications mentioned in the letter are also inserted into the revised manuscript where appropriate. It is a common mistake to address reviewer questions in the rebuttal letter but not in the revised manuscript. If a reviewer raised a question then your readers will probably have the same question so you should ensure that the manuscript can stand alone without the rebuttal letter. Directions on how to prepare a rebuttal letter can be found at:
https://peerj.com/benefits/academic-rebuttal-letters/ #]
[# PeerJ Staff Note: The review process has identified that the English language must be improved. PeerJ can provide language editing services - please contact us at editorial.support@peerj.com for pricing (be sure to provide your manuscript number and title) #]
Responses: Responses: Thank you for your comments. Following the comments of reviewers and editors, we have seriously revised the paper to make the expression of the paper clearer and make it easier for readers to understand our research.
Reviewer 1 (Mohsen Ahmadi):
Comment 1:
Basic reporting
The manuscript subject is new and interesting but it should be corrected as follow comments;
Responses: Thank you for your appreciation for our paper. We revised research manuscript in detail according to your comments.
Comment 1:
The introduction is well-written however it should concentrate on the novelty and importance of the work.
Responses: Thank you for your appreciation. We have concentrated on the novelty and importance of the work, adding to section 2 Inspiration and Novelty. The details are as follows.
As for the novelty and importance of this work, this research work is rarely seen in previous studies. First of all, as far as novelty is concerned, we try to use as little or no real data as possible to predict the real development of the epidemic situation, which is the starting point of our research work. Because, using the data of the epidemic situation itself to predict the development of the epidemic in the future, might be considered inappropriate. In addition, because of the coming of the next large-scale epidemic, we do not have data on the epidemic itself, but we still need to carry out prevention and control of the epidemic. Zero epidemic data is the value of our research work because it will produce immeasurable value in the next outbreak.
The importance of this work is self-evident. Because the correctness of epidemic prevention and control not only involves the safety of life and property but also involves the stable operation of society. The prevention and control of the epidemic situation does not mean that it is completely sealed off, but some put and lose. Because the total closure means economic stagnation, which is not an optimal strategy. We need to carry out corresponding control strategies according to the severity of the epidemic, which is the optimal trade-off between economic stagnation and epidemic prevention and control. This work is the foothold of our whole research, helping each region to classify the epidemic situation according to its severity.
Comment 2:
Equation should be enumerated and cited in the manuscript.
Responses: Thank you for your comments. Equation should be enumerated and cited in the manuscript according to your comments.
Comment 3:
The format of commas is not correct.
Responses: Thank you for your comments. We have already revised the format of commas and check all punctuation of this paper earnestly.
Comment 4:
The manuscript is the lock of figures. it needs some figures or plot for reader to be helpful.
Responses: Thank you for your comments. Originally, there were two maps about the spread of the epidemic in our paper, but after consulting with the editorial department of PeerJ for many times, they said that our map was not in line with the internationally recognized boundary (which is the standard recognized by the Chinese government), so we had to delete it. This is one of the reasons why reviewers see the lag of our paper data. We put these two maps in the other figures folder so that reviewers can view them without causing political conflict. However, in order to give reviewers a better understanding of our epidemic transmission model, we have drawn figure 1 to explain how the epidemic spreads within China.
Comment 5:
The manuscript should consider the sensitivity analysis to find the effective parameters. it should also add to the literature review, Investigation of effective climatology parameters on COVID-19 outbreak in Iran(Science of the Total Environment), Environmental Science and Pollution Research (Modeling and sensitivity analysis of NOx emissions and mechanical efficiency for diesel engine), Application of gene expression programming and sensitivity analyses in analyzing effective parameters in gastric cancer tumor size and location(Soft Computing)
Responses: Sensitivity analysis is very important, but the research method in this paper does not involve parameters, so it is impossible to do sensitivity analysis.Thank you for providing so many excellent papers for my reference, we have done a more detailed literature review.
Comment 6:
You should also use some artificial intelligence methods for analysis. you should also use prediction methods of Presentation of new thermal conductivity expression for-water and CuO-water nanofluids using gene expression programming (Journal of Thermal Analysis and Calorimetry), Presentation of a new hybrid approach for forecasting economic growth using artificial intelligence approaches(Neural Computing and Applications0, A gene expression programming model for economic growth using knowledge-based economy indicators(Journal of Modelling in Management), Statistical and Econometrical Analysis of Knowledge-Based Economy Indicators Affecting Economic Growth in Iran: The new evidence of Principal Component Analysis-Tukey and ARDL(Journal of Policy Modeling).
Responses: Thank you for your comments. It is our next research work to predict the development of epidemic situation by using machine learning method. We are well aware of the importance of machine learning in epidemic prediction, and we will publish our research results in future papers.
Experimental design
the manuscript has an Experimental data and should explain the data description statistic
Responses: Thank you for your comments, we have made a detailed statistical description and interpretation of the epidemic data. For details, please refer to Section 2.1 Description of the latest epidemic data.
Validity of the findings
validation the manuscript is incorrect and should be improved.
Responses: Thank you for your comments, we have revised and improved validation of findings. For details, please refer to 4 Results.
Reviewer 2:
Basic reporting
The English is mostly clear but it could do with some editing.
The major issue with this research, and it is for this reason that I recommend that it is rejected, is that there is an absence of references to previous works. There need to be references cited backing up the method used by the researchers. I realize that this is a new pandemic but previous epidemiological studies can be referenced as can the motivation for the research methodology used. This is one of the basic principles of scientific research and as such it should not be published in its current form.
It is very strange that this is a paper on geospatial methods and yet no maps or Figures of any description are provided.
The paper refers to a submission date of March and yet I only received this paper more than three months later suggesting that this is not the first journal to receive this paper. Especially given that this is a fastly moving pandemic, the data should be as up to date as possible.
Responses: Thank you for your comments. Our method has a certain degree of innovation, can refer to previous works may be relatively less. But we used our own method to simulate the experiment, and the results predicted the development of the epidemic situation in China very well. As a new infectious disease, COVID-2019 has presented a large number of academic achievements in a short period of time, but the types of its achievements are basically clinical research. On the basis of the reviewers' opinions, this paper further enriches its own literature research, so as to achieve a higher level of research results.
Originally, there were two maps about the spread of the epidemic in our paper, but after consulting with the editorial department of PeerJ for many times, they said that our map was not in line with the internationally recognized boundary (which is the standard recognized by the Chinese government), so we had to delete it. This is one of the reasons why reviewers see the lag of our paper data. We put these two maps in the other figures folder for the reviewers to see. However, in order to give reviewers a better understanding of our epidemic transmission model, we have drawn figure 1 to explain how the epidemic spreads within China.
We selected the data of March 2020 to start the research. It will take a long time for the whole paper to be completed. It is already may when we finish the paper. We asked Abba badamasi, the author of this paper, to conduct several rounds of language check. After the review, we submitted this paper to PeerJ. The editor of journal pointed out that our paper was not standardized, and we made several rounds of revision. In particular, the cognitive inconsistency of the map boundary in the paper may delay the submission to your reviewers. I'm sorry to be here. We updated the data in the revised version (as of July 18, 2020), and found that the difference between the predicted data of our model and the real updated data did not change over time, which is enough to illustrate the robustness of our model.
Comments for the author
It may well be that the paper has a contribution to add to the understanding of the pandemic.
Unfortunately, it reads more like a report than a scientific paper and therefore I have to recommend that it is rejected. If the authors wish to try and publish elsewhere, I recommend that they make sure the pandemic information is as up to date as possible and they follow the generally accepted approach of making sure the research is placed within the established body of literature.
Responses: Thank you for your comments，we have updated the data of the paper and improved the literature review. I hope our paper is more like an academic paper and more suitable to be published in PeerJ.
Reviewer 3:
Basic reporting
An extended background on Covid-19 is covered in Introduction. Introduction should be written clearly and concisely on the problem the authors would like to address and why it is an important problem to tackle. The Introduction section should also contains the description of the approach along with the brief discussion on the justification to support the approach taken. The findings of the approach as well as the contributions of the paper are not highlighted. The approach employed by the authors has practical value that merits sharing with the wider academic community, but more work needs to be done to emphasize their technical aspects as well as to more convincingly demonstrate the original contribution that this work makes to the scholarly knowledge base. Both findings and contributions should be clearly stated in the introduction part of the paper.
A vague and inconsistent expression such as “more than two months ago”, “geographic information adjacency”instead of“geographic adjacency information”throughout the paper should be avoided. In terms of literature review, a list of COVID related studies and a brief literature review on risk in coronavirus studies are described in the Introduction section. An important first step that will need to be taken to address this shortcoming is the inclusion of a more comprehensive literature review. In its present form, the brief discussion of existing work in the Introduction is inadequate in breadth and depth. While cursory mention is made in this section of other studies and the fact that they do not support risk mapping inside Mainland China, there is no attempt to present a comparison of their features, strengths, and weaknesses, and more importantly, to show how the present work builds upon and extends what has already been done. Also conspicuously missing from the literature review is coverage of the relevant scientific literature on the use of spatial indexes, economic indexes, and mobility indexes and the intersection between these areas, especially as they relate to surveillance and prediction of risk within both infectious and non-infectious diseases.
Furthermore, additional visualizations such as the map of Mainland China with provincial information as well as the relevant index numbers from (Table 1) will increase the readability for the readers who are not familiar with the topology of the provinces. Some tables need clarification particularly on the use of different colors (red vs. blue) of fonts. Instead of varying the colors, varying the font style such as Bold vs. Italics is recommended.
Responses: Thank you for your comments, we have explained in more detail in Section 2 inspiration and novelty about the problems we want to solve and their importance. For the description of the approach, we have added a new paragraph in section Introduction.
The main contribution of this paper is to use less epidemic data or even zero epidemic data to simulate the spread of an epidemic in China for a period of time. Because we can't get the data of the epidemic before it comes. If we use the data of the epidemic situation itself to predict the epidemic situation, its practicality and generalization are not very strong. How to better control the epidemic situation, it is necessary to divide the risk of the epidemic situation in each region, which is neither completely sealed nor free control. But according to the risk level of the epidemic situation in different regions, the corresponding epidemic control measures should be taken. In the classification of epidemic situation in 31 provinces or provincial municipalities in China, our method only predicted the epidemic level of 4 provinces or provincial municipalities incorrectly, which is a great academic achievement in the actual epidemic prevention and control. These contents are added to Section Introduction of the original manuscript.
Originally, there were two maps about the spread of the epidemic in our paper to make additional visualization and increase the readability for the readers, but after consulting with the editorial department of PeerJ for many times, they said that our map was not in line with the internationally recognized boundary (which is the standard recognized by the Chinese government), so we had to delete it. This is one of the reasons why reviewers see the lack of figures. We put these two maps in the other figures folder so that reviewers can view them without causing political conflict. However, in order to give reviewers a better understanding of our epidemic transmission model, we have drawn figure 1 to explain how the epidemic spreads within China.About changing the color identification in the form to Bold and Italics, it has been modified according to your requirements.
About the relevant scientific literature on the use of spatial indexes, economic indexes, and mobility indexes and the intersection between these areas, especially as they relate to surveillance and prediction of risk within both infectious and non-infectious diseases. We have revised this content in the manuscript. The details are as follows.
In fact, the development of spatial information systems and spatial econometrics has made it more efficient and convenient for us to measure the distance between regions (entwiste B et.al ,1997；Alslan O et.al ,2013). Boyda DC et.al (2019) using GIS and spatial analysis methods, this paper reviewed and summarized the transmission of HIV in Africa. Due to the limitations of simple geographic information, Kurata H (2015) began to use the Euclidean distance matrix to measure the distance between two regions, and generally used the Euclidean inverse distance to measure the proximity of two regions (zubaedah R et.al ,2020;Lumijarvi J,2004). However, European inverse distance still can not accurately reflect the strength of the relationship between the two regions, so some economists have introduced economic distance into the spatial information system (bastawrus a et.al ,2020;Najafi Allamdarlo H,2018). In the real epidemic prevention and control, population migration and motion can spread the epidemic rapidly, so it is an important factor in the epidemic situation (Wong IA, 2020; Yue JC, 2011). In this paper, these factors to measure the distance between the two regions are introduced into the spatial risk map theory (Pourghasemi HR,2020;Ijumulana J et.al,2020;Fusade-Boyer,2020)to dynamically measure the spread of the epidemic in China. Pourghasemi HR (2020) used spatial model and risk map to analyze the epidemic situation in Iran from January 19, 2020 to July 14, 2020. The results show that spatial model and risk map are very important for the analysis and Simulation of epidemic spread.
Experimental design
The 'Methods and Results' section gives a step by step description of the method. This part of the manuscript could be enhanced through the addition of more detailed technical on“programming and simulation”in every step of the method as well as justification of the decisions and choices made, with reference to extant research findings and (spatial risk mapping) theory that underpinned or informed the various aspects of the study approach. It would also be worthwhile outlining alternative techniques and approaches that may have been considered, including the reasons they were deemed unsuitable.
Responses: Thank you for your comments. We have explained our methods and results in more detail and systematically in the manuscript, including our programming and simulation process. Let us choose the method and theory more suitable for our research problems.
Validity of the findings
As part of the evaluation, the results are evaluated against the actual risk division results from a news website on daily reporting information about the disease. Coronavirus is an ongoing pandemic and this study compare the results with the information from one specific point in time, it is important to report and discuss the source, and when the snapshot was taken to further validate the results.
The 'Discussion' section is needed to add. If the manuscript is to move beyond mere 'show and tell' to make a real contribution to the field, the Discussion will need to be substantially developed to (a) situate the outcomes, findings, and lessons learned from the study in the wider context of the literature surveyed in the (expanded) literature review; (b) offer evidence-based practical recommendations for researchers, policy planners and system developers that are grounded in the researchers’ findings and experiences/observations; and (c) consider how the present work might apply to or otherwise have value that transcends the discipline of risk mapping in epidemiology.
Responses: Thank you for your comments. This paper has compared and analyzed the predicted results with the real epidemic results, and reported the data sources of the real epidemic results and the corresponding data reporting time. And based on different information, we do some column verification work. See section 4 Results.
This paper further enriches the discussion chapters of the paper and clarifies the further research direction. It provides more practical suggestions for researchers, policy makers and system developers, and promotes the discipline of risk mapping to a higher level. The details are as follows.
Although the research approach of this manuscript need to be improved to achieve more comprehensive and systematic results. But it does achieve a good prediction accuracy, and the accuracy of this prediction does not appear any deviation with the development of the epidemic, so the robustness of the model is very good. Therefore, for researchers and policy makers, the policies and measures in next outbreak can be based on the results of our model. When the first case is found, we can quickly classify the epidemic level in China in the future. According to the level results, the corresponding level of medical and health response was carried out. This is the optimal strategy between total closure and free control, which can minimize the economic loss and case infection rate.
Comments for the Author
Comment 1:
Line (52 - 67), existing work is described. More technical insights are needed and how and why the existing work is not suitable for the problem the authors are tackling?
Responses: Thank you for your comments. We have revised Introduction of manuscript. The existing literature only evaluated the epidemic situation and the related factors affecting the epidemic situation, and did not analyze the specific situation of regional epidemic situation, nor did it analyze how the regional differences of epidemic situation formed. These modeling methods are not specific to the region, which is not enough to improve the accurate and effective control suggestions.
Comment 2:
Line (75 - 79) In the last paragraph of the introduction, the approach should be summarized and the findings should be briefly delivered? Importantly, the main contributions of the paper should be highlighted.
Responses: Thank you for your comments. We have revised Introduction about the main contributions.The details are as follows.
The main contribution of this paper is to use less epidemic data or even zero epidemic data to simulate the spread of an epidemic in China for a period of time. Because we can't get the data of the epidemic before it comes. If we use the data of the epidemic situation itself to predict the epidemic situation, its practicality and generalization are not very strong. How to better control the epidemic situation, it is necessary to divide the risk of the epidemic situation in each region, which is neither completely sealed nor free control. But according to the risk level of the epidemic situation in different regions, the corresponding epidemic control measures should be taken. In the classification of epidemic situation in 31 provinces or provincial municipalities in China, our method only predicted the epidemic level of 4 provinces or provincial municipalities incorrectly, which is a great academic achievement in the actual epidemic prevention and control.
Comment 3:
Line 82 -”more than two months ago”, needs clarification Introduction and Inspiration sections can be combined together and add a related work section covering the discussion on how spatial proximity information, economic indexes, and mobility indexes have been used disease risk mapping in general and in particular coronavirus.
Responses: Thank you for your comments. I have revised the “more than two months ago” into accurate and specific time.
Comment 4:
Line 97, 259, 263 - information adjacency -> adjacency information
Responses:Thank you for your comments. I have revised the inconsistent expressions, and a detailed check of the full manuscript has been carried out.
Comment 5:
Line 107 - “* the numbers of cities in this paper are in the order of Table 1.”,
Table 1 is the list of provinces or provincial municipalities instead of cities.
Responses:We have corrected the writing errors here. We are sorry for the misunderstanding.
Comment 6:
Line 109-“Based on the above geographical adjacency information, we partitioned the 31 provinces or provincial municipalities into 4 risk levels with Hubei as the center of infectious diseases.” Justification is needed for four risk levels and explanation on how the provinces are partitioned explicitly.
Responses: Thank you for your comments. According to the response of major public health emergencies in China, there are four levels (I, II, III and IV). Therefore, it is necessary to divide the epidemic situation into four levels based on the risk and severity, so as to facilitate the corresponding provinces or provincial municipalities to initiate the corresponding health emergencies response. Based on the above geographical adjacency information (GAI), we partitioned the 31 provinces or provincial municipalities into 4 risk levels with Hubei as the center of infectious diseases. Table 3 presents the specific hazard classification.
According to the geographical proximity between the province and the epidemic center, the highest risk level is level 1; if the province is adjacent to the epidemic center, it is level 2; if the province is adjacent to a level 2 Province, it is level 3, and so on.
In order to show the transmission process of the epidemic more clearly, we have drawn Figure 1. The case first spread in the center of the epidemic 17 (Hubei), then to its neighboring provinces (1-2 level), then to its neighboring provinces too (2-3 level), and finally to 31 provinces or provincial municipalities in China (3-4 level).
Comment 7:
Line 109 - Table 2 presents instead of Table 2 present
Responses:We have corrected the writing errors here. We are sorry for the grammatical errors.
Comment 8:
Line 112 - table 2 -> Table 2
Responses:We have corrected the writing errors here. We are sorry for the inferior errors.
Comment 9:
Line 122 - Hubei was stated twice.
Responses: We have corrected the writing errors here. We are sorry for the inferior errors.
Comment 10:
Line 124 - add scholarly references to such Principle
Responses: Thank you for your comments. We have added scholarly references in there.
Comment 11:
Line 127 - clarify “simple programming and simulation”.
Responses: Thank you for your comments. The principle of programming is that according to the theory that the epidemic situation spreads from the center to the outside, the epidemic index is equally distributed to all the neighbors connected with it. The provinces or provincial municipalities that get the epidemic index are then transmitted to all their neighbors. After a round of transmission, we calculate the epidemic transmission index of 31 provinces and provincial municipalities. In the manuscript, we use footnotes to make further explanation.
Comment 12:
Line 132 - full stop (not a comma) after Table 2.
Responses: We have corrected the writing errors here. We are sorry for the inferior errors.
Comment 13:
Line 138 - The sentence begins with In comparison with the real data, …..State the location of the real data in the paper? If it is Table 4, state it clearly.
Responses: Thank you for your comments. We are so sorry to state it unclearly, but we have stated it clearly in the manuscript.
Comment 14:
Line 139 - 142: The sentence starts with “From the epidemic index: …”, to which table the authors are referring to? The indexes the authors referred to the sentence are the indexes in Table 1 which are not related to the epidemic index.
Responses: Thank you for your comments. We are so sorry to state it unclearly, but we have stated it clearly in the manuscript.
Comment 15:
Line 146 - need a reference to“the real level division,”
Responses: Thank you for your comments. We have added a link to the real level division.
Comment 16:
Line 149 - need a reference to “the real situation”,
Responses: Thank you for your comments. We have added a link to the real situation.
Comment 17:
Line 157 - The sentence“Instead of geographical adjacency, with all of them given the same weight index.”Need to clarify“them”in the second part of the sentence.
Responses: Thank you for your comments. “them” point to “provinces and provincial municipalities” in the manuscript.
Comment 18:
Line 161 - It is express -> It is expressed
Responses: We have corrected the writing errors here. We are sorry for the inferior errors.
Comment 19:
Line 166 - need a reference to theory of distance inverse weight
Responses: Thank you for your comments. We have added scholarly references in there.
Comment 20:
Line 172 - The sentence starts with“After programming and simulation, …”. Need to elaborate on the simulation, what was simulated and how it has been done?
Responses: Thank you for your comments. The principle of programming is based on the theory of epidemic spreading from the center to the outside. However, this simulation is different from the simulation implementation in Section 3.2, because the spread from each region to another region is not completely equal weight spread, but the reciprocal weighted spread based on distance. In other words, the closer the two regions are, the greater the epidemic index will be. See attachment code for details.
Comment 21:
Line 197 - In paragraph reference to Lin Guangping (2015) is not matched with the reference #26 (which is stated as 2005)
Responses:We have corrected the writing errors here. We are sorry for the inferior errors.
Comment 22:
Line 228 - the same comment as Line 172.
Responses: Thank you for your comments. The principle of programming is based on the theory of epidemic spreading from the center to the outside. However, this simulation is different from the simulation implementation in Section 3.2 and 3.3, because the spread from each region to another region is not completely equal weight spread, but the reciprocal weighted spread based on geographical and economic distance. In other words, the closer the two regions are in geographical and economic distance, the greater the epidemic index will be. In the manuscript, we use footnotes to make further explanation. See attachment code for details.
Comment 23:
Line 234 - table 10 -> Table 10.
Responses: We have corrected the writing errors here. We are sorry for the inferior errors.
Comment 24:
Line 242 - The sentence starts with “In other hand, …”. The following sentence in Line 243 also starts with “In other hand,...”.
Responses: We have corrected the writing errors here. We are sorry for the inferior errors.
Comment 25:
Line 244 - Need clarification on the phrase“three-level dangerous provinces”.
Responses: Thank you for your comments. According to the severity of the epidemic in China, the epidemic situation is divided into four levels. the phrase“three-level dangerous provinces”means that these provinces are level 3 in epidemic division.
Comment 26:
Line 259 - The spread novel coronavirus -> The spread of novel coronavirus
Responses: We have corrected the writing errors here. We are sorry for the inferior errors.
Comment 27:
Line 259 - the same suggestion as at Line 172.
Responses: Thank you for your comments. The principle of programming is based on the theory of epidemic spreading from the center to the outside. However, this simulation is different from the simulation implementation in Section 3.2 to 3.4, because the spread from each region to another region is not completely equal weight spread, but the reciprocal weighted spread based on geographical, economic distance and Baidu Index. In the manuscript, we use footnotes to make further explanation. See attachment code for details.
Comment 28:
Line 259 - The sentence stated Euclidean spatial distance, geographical economic distance and Baidu migration index, and …….are predicted.”
“Euclidean spatial distance, geographical economic distance and Baidu migration index”are not predicted, they are used in predicting the risk of the regions in the study.
Responses: Thank you for your comments. We have revised this sentence. The spread of novel coronavirus disease is simulated and the epidemic index and risk level of each province are predicted based on geographical adjacency information, Euclidean spatial distance, geographical economic distance and Baidu migration index.
Comment 29:
Line 263-the authors stated that the accuracy of 54% for geographical information adjacency. Providing information such as (the ranks of xx/yy provinces are correctly predicted) will assist the reader.
Responses: Thank you for your comments. We added the verification of the results, specific to the prediction accuracy of each method, through the confusion matrix display, for details, see section 4 Results.
Comment 30:
Line 264 - The sentence is not clear“there are some differences between the simulation results and the actual epidemic situation, and the prediction results have specific reference values”. What are the specific reference values? With 54% of accuracy, it is obvious that there are differences but need to state them stately.
Responses: Thank you for your comments. “specific reference values” means that we can use these information about prediction results to carry out corresponding health emergency response. For the prediction results, 54% of the accuracy is that we only use the adjacency relationship, and the follow-up method will have higher accuracy.
Comment 31:
Line 269 - the same suggestion as at Line 263Responses:We have corrected the writing errors here. We are sorry for the inferior errors.
Comment 32:
Tables 4, 6, 8, 11, what is the meaning of the use of different colors (blue and red)
Responses: Thank you for your comments. The red and blue numbers in the table indicate that the predicted result is different from the real result. Specifically, the red number means that the prediction level is low, and the blue number means that the prediction level is high.
Comment 33:
Tables 4, 6, 8, when the real division results, when the information was accessed?
Responses: Thank you for your comments. We have updated the time of the real epidemic situation in the table, and the data in the revised version is July 18, 2020.
" | Here is a paper. Please give your review comments after reading it. |
9,784 | "<ns0:abstract xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:p>The SARS-CoV-2 coronavirus is wreaking havoc globally, yet as a novel pathogen knowledge of its biology is still emerging. Climate and seasonality influence the distributions of many diseases, and studies suggest at least some link between SARS-CoV-2 and weather. One such study, building species distribution models (SDMs), predicted SARS-CoV-2 risk may remain concentrated in the Northern Hemisphere, shifting northward in summer months. Others have highlighted issues with SARS-CoV-2 SDMs, notably: the primary niche of the virus is the host it infects, climate may be a weak distributional predictor, global prevalence data have issues, and the virus is not in a population equilibrium. While these issues should be considered, climate's relationship with SARS-CoV-2 is still worth exploring, as it may have some impact on the distribution of cases. To further examine if there is a link to climate, we build SDM model projections with raw SARS-CoV-2 case data and population scaled case data in the United States. The case data were from across March, 2020, before large travel restrictions and public health policies were impacting cases across the country. We show that SDMs built from population scaled case data cannot be distinguished from control models built from raw human population data, while SDMs built on raw case data fail to predict the known distribution of cases in the US from March. The population scaled analyses indicate that climate did not play a central role in early US viral distribution and that human population density was likely the primary driver. We do find slightly more population scaled viral cases in cooler areas. However, the temporal and geographic constraints on this study mean that we cannot rule out climate as a partial driver of the SARS-CoV-2 distribution.</ns0:p><ns0:p>Climate's role on SARS-CoV-2 should continue to be cautiously examined, but at this time we should assume that SARS-CoV-2 will continue to spread anywhere in the US where</ns0:p></ns0:div>
</ns0:abstract>
<ns0:body xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:head>Introduction</ns0:head></ns0:div>
<ns0:div><ns0:head>SARS-CoV-2 and climate overview</ns0:head><ns0:p>In March, 2020, the United States of America had the most reported cases of COVID-19, the disease caused by the coronavirus SARS-CoV-2 (CDC NCIRD 2020). While we are now beginning to better understand the global and US-level distributions of this virus, we are only starting to learn how abiotic variables affect its basic distribution, particularly before policy seemed to become the principal driver of the virus. With over two months of SARS-CoV-2 records across US counties by March, the role of abiotic variables was becoming easier to examine, especially as multiple sources began compiling data and making it public (The New York Times 2020; Dong, Du, and Gardner 2020; The COVID Tracking Project 2020).</ns0:p><ns0:p>Organisms are distributed within their environment based on both direct and indirect interactions with biotic and abiotic variables <ns0:ref type='bibr' target='#b26'>(Elith and Leathwick 2009</ns0:ref>) -SARS-CoV-2 is no exception. As this virus is primarily distributed by human hosts, it is constrained by human distributions and interactions (i.e., biotic variables). There are numerous documented cases on local and global scales of individual human sources for new outbreaks <ns0:ref type='bibr' target='#b37'>(Holshue et al. 2020;</ns0:ref><ns0:ref type='bibr' target='#b41'>KCDC 2020)</ns0:ref>. Furthermore, many other viruses, including coronaviruses, display marked seasonality and are affected by local climatic conditions <ns0:ref type='bibr' target='#b65'>(Price, Graham, and Ramalingam 2019;</ns0:ref><ns0:ref type='bibr' target='#b31'>Fisman 2012;</ns0:ref><ns0:ref type='bibr' target='#b48'>Lofgren et al. 2007;</ns0:ref><ns0:ref type='bibr' target='#b34'>Gaunt et al. 2010</ns0:ref>). This has prompted researchers to begin looking globally at how weather and climate (i.e., a subset of abiotic variables) may relate to the presence and abundance of the virus.</ns0:p><ns0:p>Modeling the SARS-CoV-2 viral distribution from the early part of the outbreak in relation to climate and other abiotic variables could help refine our understanding, hopefully adding to the knowledge gleaned from studies on human transmission dynamics <ns0:ref type='bibr' target='#b43'>(Kucharski et al. 2020;</ns0:ref><ns0:ref type='bibr' target='#b22'>Chinazzi et al. 2020)</ns0:ref>, which appear to explainr much of the viral distribution. Climate can also be compared directly to biotic variables, such as human population density, to disentangle their relative importances. More dense human populations can have elevated communicable disease spread, as there is more person-to-person contact <ns0:ref type='bibr' target='#b28'>(Fang et al. 2013</ns0:ref>) -the principal way SARS-CoV-2 is known to spread <ns0:ref type='bibr' target='#b70'>(Rothan and Byrareddy 2020)</ns0:ref>. Identification of abiotic variables with suitably large effect size on the distribution and spread of SARS-CoV-2 might help prevent viral spread by identifying higher risk regions and project the seasonal variation in the risk of transmission. However, it is important to make clear that the models presented here, and in other recent studies <ns0:ref type='bibr'>(Araújo and Naimi 2020;</ns0:ref><ns0:ref type='bibr' target='#b10'>Bariotakis et al. 2020)</ns0:ref>, are influenced by the variability of input parameters, the variables explored, and the restrictions inherent in modeling any complex system <ns0:ref type='bibr' target='#b27'>(Elith et al. 2011)</ns0:ref>. We present this study to serve as a baseline in our understanding, given the available data, rather than a definitive model.</ns0:p></ns0:div>
<ns0:div><ns0:head>SARS-CoV-2 and climate research</ns0:head><ns0:p>Early studies suggest that transmission of SARS-CoV-2 may have, at minimum, a loose association with climatic features. There have been numerous reports showing a correlation of case incidence with cool temperatures and low humidity <ns0:ref type='bibr' target='#b76'>(Wang et al. 2020;</ns0:ref><ns0:ref type='bibr' target='#b29'>Ficetola and Rubolini 2020;</ns0:ref><ns0:ref type='bibr' target='#b8'>Bannister-Tyrrell et al. 2020;</ns0:ref><ns0:ref type='bibr'>Araújo and Naimi 2020;</ns0:ref><ns0:ref type='bibr' target='#b66'>Chen et al. 2020;</ns0:ref><ns0:ref type='bibr' target='#b2'>Alvarez-Ramirez and Meraz 2020;</ns0:ref><ns0:ref type='bibr' target='#b57'>Notari 2020;</ns0:ref><ns0:ref type='bibr' target='#b58'>Paez et al. 2020)</ns0:ref>. Furthermore, studies that controlled for case growth rate or demographic factors found that weather was still a significant factor in the success of SARS-CoV-2 outbreaks <ns0:ref type='bibr' target='#b15'>(Bukhari and Jameel 2020;</ns0:ref><ns0:ref type='bibr' target='#b66'>Chen et al. 2020;</ns0:ref><ns0:ref type='bibr' target='#b72'>Sajadi et al. 2020)</ns0:ref>. Unfortunately, none of these studies provide a mechanism for how abiotic variables like temperature might limit or promote the person-to-person transmission of the virus. Still, lipidenveloped viruses (including coronaviruses) may be more stable outside of the host in lower humidity and cooler temperatures, conditions that are common in temperate areas in late winter and early spring <ns0:ref type='bibr' target='#b65'>(Price, Graham, and Ramalingam 2019)</ns0:ref>. Recent studies have also indicated that SARS-CoV-2 transmission rate may be negatively correlated with temperature and humidity <ns0:ref type='bibr' target='#b66'>(Qi et al. 2020;</ns0:ref><ns0:ref type='bibr' target='#b80'>Wu et al. 2020;</ns0:ref><ns0:ref type='bibr' target='#b69'>Rosario et al. 2020;</ns0:ref><ns0:ref type='bibr' target='#b0'>Ahmadi et al. 2020</ns0:ref>) <ns0:ref type='bibr'>(Sagripanti and David</ns0:ref> Lytle 2020) However, the mechanism causing these correlations could be indirect, as humans behave differently in different seasons <ns0:ref type='bibr' target='#b11'>(Bedford et al. 2015;</ns0:ref><ns0:ref type='bibr' target='#b78'>Wesolowski et al. 2017)</ns0:ref>. Recent epidemiological models have also highlighted that weather may not play a large role when most people lack immunity during the pandemic phase, but that weather has the potential to play a larger role once endemic phase <ns0:ref type='bibr'>(Baker et al. 2020)</ns0:ref> Biologists often employ species distribution modeling (SDM; alternatively called ecological niche modeling) to predict geographic ranges of species. SDMs employ environmental data (typically climate) to predict if geographic space is suitable for a given species or population <ns0:ref type='bibr' target='#b60'>(Peterson 2001</ns0:ref>). These models have proven useful in a wide variety of applications, such as invasion biology, climate change, zoonotic diseases, and speciation <ns0:ref type='bibr' target='#b26'>(Elith and Leathwick 2009;</ns0:ref><ns0:ref type='bibr' target='#b35'>Guisan and Thuiller 2005)</ns0:ref>. SDMs built to predict the spread of viral pathogens often do so by modeling the potential distribution of a known vectors or alternate hosts (R. H. <ns0:ref type='bibr' target='#b55'>Miller et al. 2012;</ns0:ref><ns0:ref type='bibr' target='#b44'>Larson et al. 2010;</ns0:ref><ns0:ref type='bibr' target='#b1'>de Almeida et al. 2019;</ns0:ref><ns0:ref type='bibr' target='#b68'>Richman et al. 2018</ns0:ref>), but some base the models directly on pathogen occurrence and correlations with environmental, climatic, or demographic covariates <ns0:ref type='bibr' target='#b49'>(Machado-Machado 2012;</ns0:ref><ns0:ref type='bibr' target='#b12'>Belkhiria, Alkhamis, and Martínez-López 2016;</ns0:ref><ns0:ref type='bibr' target='#b62'>Pigott, Bhatt, et al. 2014;</ns0:ref><ns0:ref type='bibr'>Messina et al. 2015;</ns0:ref><ns0:ref type='bibr' target='#b13'>Bhatt et al. 2013</ns0:ref>). Efforts to model global disease distribution within an ecological framework frequently draw on known or hypothesized environmental variables correlated with disease occurrence to predict suitability for transmission. Output from these models can be extrapolated into areas where a disease has not yet been reported, but suitable environmental data exist <ns0:ref type='bibr' target='#b26'>(Elith and Leathwick 2009)</ns0:ref>.</ns0:p><ns0:p>Researchers early in the pandemic created SDMs of SARS-CoV-2 using climatic variables <ns0:ref type='bibr'>(Araújo and Naimi 2020;</ns0:ref><ns0:ref type='bibr' target='#b10'>Bariotakis et al. 2020)</ns0:ref>. These studies suggested that the virus is strongly constrained by global climate patterns <ns0:ref type='bibr'>(Araújo and Naimi 2020;</ns0:ref><ns0:ref type='bibr' target='#b10'>Bariotakis et al. 2020)</ns0:ref>. Preliminary models also suggested that the virus would continue to be concentrated in the Northern Hemisphere, shifting northwards throughout the summer and then back towards its spring distribution in the fall and winter <ns0:ref type='bibr'>(Araújo and Naimi 2020)</ns0:ref>.</ns0:p><ns0:p>However, a research group recently put forth a strong rebuttal <ns0:ref type='bibr' target='#b23'>(Chipperfield et al. 2020)</ns0:ref> to some of this SDM work <ns0:ref type='bibr'>(Araújo and Naimi 2020)</ns0:ref>. Their main criticism asserts that basic assumptions underlying SDMs are violated when modeling SARS-CoV-2, due to the mode of transmission, current population disequilibrium, and failure to incorporate epidemiological data <ns0:ref type='bibr' target='#b23'>(Chipperfield et al. 2020)</ns0:ref>. Furthermore, they highlight issues pertaining to input data for global records, such as: Missing data, omitted data, and single localities representing many thousands of reported cases (hospital coordinates or political centroids) across an entire country <ns0:ref type='bibr' target='#b3'>(Araújo et al. 2019a)</ns0:ref>. This is exacerbated as many countries are reporting records based on varied criteria, such as multiple molecular tests and tests using computed tomography scans <ns0:ref type='bibr' target='#b47'>(Lippi, Simundic, and Plebani 2020)</ns0:ref>. The rebuttal additionally suggested that model evaluation and justification were insufficient. An updated draft of the original paper <ns0:ref type='bibr'>(Araújo and Naimi 2020)</ns0:ref> addressed some of these issues, and produced reasonable arguments to some others. While we agree with many arguments from the rebuttal, we believe SDMs have the potential to be useful for modeling in a number of instances, if done carefully, and have been effective when used for other diseases <ns0:ref type='bibr'>(Pigott, Golding, et al. 2014;</ns0:ref><ns0:ref type='bibr' target='#b64'>Pigott et al. 2015;</ns0:ref><ns0:ref type='bibr' target='#b19'>Carlson, Dougherty, and Getz 2016)</ns0:ref> and other instances when data are limited or incomplete <ns0:ref type='bibr' target='#b33'>(Galante et al. 2018;</ns0:ref><ns0:ref type='bibr' target='#b9'>Barbet-Massin et al. 2018;</ns0:ref><ns0:ref type='bibr' target='#b40'>Katz and Zellmer 2018;</ns0:ref><ns0:ref type='bibr' target='#b59'>Pearson et al. 2006;</ns0:ref><ns0:ref type='bibr' target='#b32'>Fois et al. 2018;</ns0:ref><ns0:ref type='bibr' target='#b36'>Hernandez et al. 2006;</ns0:ref><ns0:ref type='bibr' target='#b42'>Kiedrzyński et al. 2017)</ns0:ref>. Still, the current climate SDMs of SARS-CoV-2 may to at least some extent reflect the habitat preferences of their host. Careful consideration of host availability (human population density) and pathogen ecologies (abiotic variables related to transmission) may be necessary to frame analyses modeling the global distribution <ns0:ref type='bibr' target='#b39'>(Johnson, Escobar, and Zambrana-Torrelio 2019)</ns0:ref>, helping to better ensure that projected distributions are not simply the result of environmental variables related to human population density.</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:05:48694:1:0:NEW 14 Aug 2020)</ns0:p><ns0:p>Manuscript to be reviewed</ns0:p></ns0:div>
<ns0:div><ns0:head>SARS-CoV-2 in the US</ns0:head><ns0:p>COVID-19 was first detected in the US on January 20, 2020 <ns0:ref type='bibr' target='#b37'>(Holshue et al. 2020)</ns0:ref>, has since been detected in all 50 states, and continues to spread rapidly <ns0:ref type='bibr' target='#b22'>(Chinazzi et al. 2020)</ns0:ref>.</ns0:p><ns0:p>Quality county-level data are also publically available (The New York Times 2020; The COVID Tracking Project 2020; <ns0:ref type='bibr' target='#b25'>Dong, Du, and Gardner 2020)</ns0:ref>. As of July 22nd, 2020, SARS-CoV-2 has caused just under 145,000 deaths in the US (IHME COVID-19 team and Murray 2020).</ns0:p><ns0:p>However, like much of the world, the case distribution is rapidly changing and the case numbers continue to rise and fall. Public policy decisions appear to be the main mitigating factor against this coronavirus <ns0:ref type='bibr' target='#b45'>(Leonhardt 2020</ns0:ref>).</ns0:p><ns0:p>Here we develop a suite of data visualizations and species distribution models using both climate and human population data to determine whether the effect of climate can be appropriately disentangled from other drivers of SARS-CoV-2 transmission, given early records for the virus in the US. We feel that examining a relatively early time-point of the pandemic in the US is useful, as it in part (probably largely) predates the impact of major shifts in public health policy for the US.</ns0:p></ns0:div>
<ns0:div><ns0:head>Materials & Methods</ns0:head></ns0:div>
<ns0:div><ns0:head>Code and results deposition</ns0:head><ns0:p>The code and results developed in this study are deposited under a CC-BY-NC 4.0</ns0:p><ns0:p>License on Github (https://github.com/rsh249/cv19_enm/releases/tag/v0.0.5). All analysis code was written using R 3.6.2 (R Core Team 2019). Plots for Figures 1, 2, 3, S1, and S2 were produced using the 'ggplot2' package <ns0:ref type='bibr' target='#b79'>(Wickham 2009)</ns0:ref>.</ns0:p></ns0:div>
<ns0:div><ns0:head>Data acquisition</ns0:head><ns0:p>SARS-CoV-2 case data for US counties (one record per county, with a total sample size of 1,883 counties that had available reports) were collected from the New York Times database (The New York Times 2020) on March 31, 2020 for the March 30, 2020 data release. Countylevel data on human population densities were acquired from the 2010 United States Census through the R 'tidycensus' package <ns0:ref type='bibr' target='#b75'>(Walker, Eberwein, and Herman 2020)</ns0:ref>. Georeferencing to county centroids was performed by referencing county and state names in the GeoNames database (https://www.geonames.org/). While exact virus case geolocations or even town-level data would provide finer scale resolution to our analyses, these data are likely the best curated dataset for the US at this time.</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:05:48694:1:0:NEW 14 Aug 2020)</ns0:p><ns0:p>Manuscript to be reviewed</ns0:p></ns0:div>
<ns0:div><ns0:head>Climate data</ns0:head><ns0:p>Interpolated climate data averaged from 1970-2000 for the month of March were accessed through the WorldClim v2.1 database <ns0:ref type='bibr' target='#b30'>(Fick and Hijmans 2017)</ns0:ref>. The seven climate parameters examined for the month of March were average monthly temperature, average monthly minimum temperature, average monthly maximum temperature, average monthly precipitation, average daily solar radiation, wind speed, and water vapor pressure (a measure related to humidity). These climate parameters are consistent with possible correlates of SARS-</ns0:p><ns0:p>CoV-2 transmission in several recent studies <ns0:ref type='bibr' target='#b66'>(Qi et al. 2020;</ns0:ref><ns0:ref type='bibr' target='#b71'>Sagripanti and David Lytle 2020;</ns0:ref><ns0:ref type='bibr' target='#b0'>Ahmadi et al. 2020;</ns0:ref><ns0:ref type='bibr' target='#b69'>Rosario et al. 2020;</ns0:ref><ns0:ref type='bibr' target='#b80'>Wu et al. 2020)</ns0:ref>. Climate data were extracted from these layers for each georeferenced county record using the average from a 5km buffer around each county centroid coordinate. These local averages should represent the county generally as climate is correlated over short distances, but these averages may not be the best representation in counties that have a large area or highly clustered population centers. While county centroids may not be ideal in all circumstances, our view is that county centroids plus this buffer will average the climate relative to most cases in the area. More precise case georeferencing is not possible without high levels of individual case and movement tracking to identify the source of infection.</ns0:p></ns0:div>
<ns0:div><ns0:head>Climate Distribution Visualization</ns0:head><ns0:p>Total cumulative positive case data for SARS-CoV-2 in each reporting county on March 30, 2020 were used to extract data for the seven abiotic variables listed above. Probability density distributions for SARS-CoV-2 are produced to characterize the likelihood of case occurrence given the available range of climate values. These distributions are calculated from case occurrences and corresponding climate data using a Gaussian Kernel Density Estimator and standard bandwidth estimation. These distributions can be interpreted similarly to histograms, but are normalized to be comparable between different sample sizes. Raw case data, total cases per county, were scaled to reflect virus cases in each county unit by dividing total cases by the county population from the 2010 US Census. These population-scaled viral cases (cases / population) were used as probability density weightings, and resulting curves were standardized to an area of one. Probability densities were also calculated with raw case count data and county-level population as weightings. Probability density estimation and visualization was done with the R 'ggplot2' package <ns0:ref type='bibr' target='#b79'>(Wickham 2009)</ns0:ref>. To test whether the SARS-CoV-2 cases and human population data were correlated, we applied Spearman tests 'cor.test(method='spearman')'. resampling method. All models were built using the Maxent algorithm implemented in the 'maxnet' R package <ns0:ref type='bibr' target='#b61'>(Phillips, 2017)</ns0:ref>, including those tested within ENMeval. Operational models using the best model parameters were then built for SARS-CoV-2 using all population scaled data, raw virus data, and for human density using the county population data <ns0:ref type='bibr' target='#b61'>(Phillips 2017)</ns0:ref>. Of the seven WorldClim variables, the three temperature variables are clearly linked. Accordingly, we only used maximum temperature, as higher temperatures have been hypothesized to lower the viral distribution. Models built using all temperature variables made similar models, but are not further reported on. Reducing correlated variables is generally preferable to avoid extrapolation issues that arise from local covariance between variables <ns0:ref type='bibr' target='#b3'>(Araújo et al. 2019a</ns0:ref>).</ns0:p></ns0:div>
<ns0:div><ns0:head>Species distribution models</ns0:head><ns0:p>Given recent critiques, it has become clear that when using SDMs in relation to SARS-CoV-2 it is worth documenting how close a study may come to the gold standards set for this type of modeling <ns0:ref type='bibr'>(Araújo et al. 2019b</ns0:ref>). The mode rank for our SDMs is bronze (Table <ns0:ref type='table' target='#tab_1'>1</ns0:ref>), largely as data from an emerging disease are inherently imperfect. Niche overlap and similarity tests were conducted with the 'ecospat' library (Di <ns0:ref type='bibr' target='#b24'>Cola et al. 2017;</ns0:ref><ns0:ref type='bibr' target='#b14'>Broennimann et al. 2012;</ns0:ref><ns0:ref type='bibr' target='#b77'>Warren, Glor, and Turelli 2008)</ns0:ref> to compare the climatic niche of SARS-CoV-2 and humans. The niche overlap test considers whether there is a greater</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:05:48694:1:0:NEW 14 Aug 2020)</ns0:p><ns0:p>Manuscript to be reviewed than expected overlap between the climate space occupied by two population models (human total vs. SARS-CoV-2 positive) than would be expected given a null distribution sampled from both populations. The niche similarity test considers the degree of similarity between the density of occurrence between two population models relative to the same null distribution <ns0:ref type='bibr' target='#b77'>(Warren, Glor, and Turelli 2008)</ns0:ref>. A significant overlap between SARS-CoV-2 and total human population models indicates that the virus' distribution is not necessarily constrained by climate (i.e., that the full range of human occupied climate is accessible to the virus).</ns0:p></ns0:div>
<ns0:div><ns0:head>Results</ns0:head></ns0:div>
<ns0:div><ns0:head>Climate Distribution Visualizations</ns0:head><ns0:p>The human population climate curves are visually similar compared to the population scaled SARS-CoV-2 climate curves for all seven abiotic variables during the latest date in March (Figure <ns0:ref type='figure' target='#fig_1'>1</ns0:ref>). There do appear to be modest differences in the population scaled viral curves towards cooler temperatures and lower water vapor pressure compared to the human population curves. The raw density of SARS-CoV-2 cases (Figure <ns0:ref type='figure' target='#fig_1'>S1</ns0:ref>) is visually different from the population scaled data (Figure <ns0:ref type='figure' target='#fig_1'>1</ns0:ref>). Population scaled SARS-CoV-2 data appear to become better fitted to the human population data over time (i.e., these curves better match on March 30 than they did on March 2 or 16, 2020; Figure <ns0:ref type='figure' target='#fig_1'>1</ns0:ref>), whereas the raw SARS-CoV-2 case data become more and more fixed on a narrow range of values for all climate variables (Figure <ns0:ref type='figure' target='#fig_1'>S1</ns0:ref>).</ns0:p><ns0:p>The latter (fixed and narrow) range presumably has to do with a few virus hotspots.</ns0:p><ns0:p>There is a highly significant positive relationship (Spearman's r = 0.75, p < 2e -12 );</ns0:p><ns0:p>between the number of humans in a county and the number of viral cases; however, the population scaled viral cases do not have a strong relationship (Spearman's r = 0.05, p = 0.03) with human populations (Figure <ns0:ref type='figure' target='#fig_2'>2</ns0:ref>). Although the latter p-value is significant, this is often not all that meaningful for large datasets, whereas the weak r value better illustrates the relationship for the population scaled data.</ns0:p></ns0:div>
<ns0:div><ns0:head>Species distribution models</ns0:head><ns0:p>Model testing for SARS-CoV-2 population scaled case model identified a Maxent model with linear and quadratic feature classes and a regularization multiplier of 0.5 as the best model with a reasonably high model fit and transferability under block resampling (Avg. Test AUC = 0.82; Test AUC variance = 0.0002). Population scaled case and human population models appear to be highly similar, with areas of high suitability predicted for much of the West Coast and most of the eastern half of the US (Figure <ns0:ref type='figure'>3</ns0:ref>). There are some differences; for instance, the population scaled model outputp for SARS-CoV-2 (Figure <ns0:ref type='figure'>3A</ns0:ref>) suggests Florida is less suitable than it is in the map for human population (Figure <ns0:ref type='figure'>3B</ns0:ref>). is not simply a map of current human populations). We did this as our ultimate goal was comparing a viral model to a human model. The niche overlap and similarity tests both find that the models of SARS-CoV-2</ns0:p><ns0:p>(population scaled viral case data) and human population have significantly higher overlap (p < 0.01) than would be expected by chance (Figure <ns0:ref type='figure' target='#fig_7'>4</ns0:ref>), confirming the visual similarities between these maps. The raw data showed less overlap: The similarity test was significant, but an order of magnitude less so than for the population scaled data (p = 0.02), while the overlap test was insignificant (p = 0.20), as seen in Figure <ns0:ref type='figure'>S3</ns0:ref>.</ns0:p></ns0:div>
<ns0:div><ns0:head>Discussion</ns0:head></ns0:div>
<ns0:div><ns0:head>Main findings</ns0:head><ns0:p>Our results suggest that population scaled SARS-CoV-2 coronavirus cases from March, 2020 were highly linked with the human populations in the US, and that any influence of climate is currently hard to disentangle for SARS-CoV-2 cases in the US (Figures <ns0:ref type='figure' target='#fig_7'>1-4</ns0:ref>). Furthermore, this link was strengthening over the duration of March (Figure <ns0:ref type='figure' target='#fig_1'>1</ns0:ref>). This indicates that caution should be used when dealing with climate modeling of SARS-CoV-2, at least on a national scale. Our results, while broadly similar to what was found in the US for global scale models (Araújo and Naimi 2020; <ns0:ref type='bibr' target='#b10'>Bariotakis et al. 2020)</ns0:ref>, indicate that the pattern we observed is simply reflecting human population density. Based on this, at least some human-focused data (i.e., host data) should be compared or incorporated in any modeling exercises for SARS-CoV-2.</ns0:p><ns0:p>Accordingly, the current global results from other studies (Araújo and Naimi 2020; <ns0:ref type='bibr' target='#b10'>Bariotakis et al. 2020)</ns0:ref> should not be taken at face value without a critical comparison to human distributions.</ns0:p><ns0:p>Furthermore, this is inline with current knowledge that public policy is the main driver for</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:05:48694:1:0:NEW 14 Aug 2020)</ns0:p><ns0:p>Manuscript to be reviewed coronavirus spread in the US <ns0:ref type='bibr' target='#b45'>(Leonhardt 2020)</ns0:ref>. It is also inline with modeling efforts that have combined epidemiological variables with climate <ns0:ref type='bibr'>(Baker et al. 2020)</ns0:ref>.</ns0:p><ns0:p>It is also noteworthy that our models using the raw viral case data (Figure <ns0:ref type='figure' target='#fig_2'>S2</ns0:ref>), as opposed to population scaled data (Figure <ns0:ref type='figure'>3</ns0:ref>), performed poorly: the raw SDM did not highlight most of the known March range of SARS-CoV-2 in the US. Other SDM studies have only used raw values as inputs <ns0:ref type='bibr'>(Araújo and Naimi 2020;</ns0:ref><ns0:ref type='bibr' target='#b10'>Bariotakis et al. 2020</ns0:ref>), which our model shows to be at risk of overfitting and low model transferability to novel regions. Similarly, our analysis of raw SARS-CoV-2 case data shows that over time the probability density curves focus on a narrower range of climate as more data comes from the areas most affected by the outbreak (Figure <ns0:ref type='figure' target='#fig_1'>S1</ns0:ref>). In contrast, adjusting for population size with the population scaling results in SARS-CoV-2 climate curves more closely approaching that of the human population curves over time (Figure <ns0:ref type='figure' target='#fig_1'>1</ns0:ref>). Furthermore, county-level human population numbers only appear to be a strong correlate to raw cases, rather than population scaled cases (Figure <ns0:ref type='figure' target='#fig_2'>2</ns0:ref>). Together these observations suggest that the SARS-CoV-2 was highly successful in a few areas in March (e.g., New York City), but when scaled for the available human population in each county we can infer a much broader geographic and climatic range available. Again, our data have a few indications that climate may have had some relevance for SARS-CoV-2 distributions in March. Viral probability densities were slightly higher in cooler locales (Figure <ns0:ref type='figure' target='#fig_1'>1, S1</ns0:ref>). The population scaled data were particularly compelling, even if the climate difference is smaller, as they are less driven by the current SARS-CoV-2 hotspots.</ns0:p><ns0:p>Furthermore, our population scaled viral SDMs, while statistically identical to our human SDMs, had lower suitability for the virus in Florida and the southwestern border (Figure <ns0:ref type='figure'>3</ns0:ref>). In reality, Florida had a reasonably high number of viral cases in March.</ns0:p></ns0:div>
<ns0:div><ns0:head>Avoiding a presumably faulty prediction</ns0:head><ns0:p>Given that our SDMs for humans and the virus were statistically indistinguishable, we do not believe that a future projection of our SARS-CoV-2 SDM in the US using only climate data would be trustworthy at this point in time, and therefore do not present one. A model based on</ns0:p><ns0:p>March data would likely have suggested the SARS-CoV-2 prevalence will shift northward during the summer, which would be congruent with another early global modeling study <ns0:ref type='bibr'>(Araújo and Naimi 2020)</ns0:ref>. However, given that humans do not move northward in large numbers in the US during the summer, our predictions would be based on faulty presumptions about host resources (i.e., that there are sufficient humans in the north to harbor the bulk of cases). Most importantly, we now know that the virus has spread across much of the southern US during the PeerJ reviewing PDF | (2020:05:48694:1:0:NEW 14 Aug 2020)</ns0:p><ns0:p>Manuscript to be reviewed summer. Still, it is possible that a future model projection would turn out to be correct using updated datasets, and there could be at least some residual predictive power that is not fully encompassed by human population patterns.</ns0:p></ns0:div>
<ns0:div><ns0:head>Modeling counterarguments</ns0:head><ns0:p>It has now been argued that SDM modeling of SARS-CoV-2 can help to predict where the virus may generally be found now and in the future <ns0:ref type='bibr'>(Araújo and Naimi 2020;</ns0:ref><ns0:ref type='bibr' target='#b10'>Bariotakis et al. 2020)</ns0:ref>, as well as why it might be a fool's errand to conduct these analyses in the first place <ns0:ref type='bibr' target='#b23'>(Chipperfield et al. 2020)</ns0:ref>. Here are some of the main arguments that have been put forward against SARS-CoV-2 SDMs, followed by our opinion on the subject. 1) Issue: The virus is spreading and the population was not in equilibrium with climate or any other putative niche dimension in March, let alone today. Our thoughts: It was not in equilibrium and that is imperfect; however, SDMs have been useful for a number of non-equilibrium systems, like species invasions during their spreading phase <ns0:ref type='bibr' target='#b74'>(Václavík and Meentemeyer 2009)</ns0:ref>, and waiting for equilibrium will mean predictions are no longer as useful to conduct (i.e., would only be helpful for a next outbreak of this virus). 2) Issue: The virus may have been spreading heavily or underreported in the Global South as of March. Our thoughts: Surely this was partially true, but this may not completely explain observable climatic correlations. Even if we are working with incomplete data, it is hard to know if adding those data will significantly change conclusions based on preliminary models (i.e., will the SARS-CoV-2 model still be indistinguishable from a human population model). Still, to mitigate this problem we focused on the more consistent USlevel data. 3) Issue: Papers must strive for best practices. Our thoughts: Yes, we agree that best practices are indeed worth pursuing, even if not everyone agrees on what best practices may be. We have done our best to achieve the somewhat aspirational standards that have been put</ns0:p><ns0:p>forth <ns0:ref type='bibr' target='#b3'>(Araújo et al. 2019a)</ns0:ref>. We explicitly document this, and summarize shortcomings that are inherent with a newly spreading system (Table <ns0:ref type='table' target='#tab_1'>1</ns0:ref>). 4) Issue: Caution should be taken in claims and dissemination of research that could impact public health policy. Our thoughts: Yes, caution is advisable in terms of studies on dire subjects like SARS-CoV-2; however, cross pollination between disciplines has been important for many breakthroughs and advances in science. Further caution towards SDM methodology for modeling SARS-CoV-2 in recent review literature highlights the likely limited effect climate has on a pathogen spread via direct transmission, and thus concludes that this tool is inappropriate in this situation <ns0:ref type='bibr' target='#b16'>(Carlson et al. 2020a</ns0:ref><ns0:ref type='bibr' target='#b16'>(Carlson et al. , [b] 2020))</ns0:ref>. Important confounding variables such as human interactions, public policy, and microclimate may mask any effects climate plays on this virus <ns0:ref type='bibr' target='#b18'>(Carlson et al. 2020b;</ns0:ref><ns0:ref type='bibr' target='#b23'>Chipperfield et al. 2020</ns0:ref>). Due to this, epidemiological studies are more suited to understand patterns of transmission <ns0:ref type='bibr' target='#b18'>(Carlson et al. 2020b;</ns0:ref><ns0:ref type='bibr' target='#b23'>Chipperfield et al. 2020;</ns0:ref><ns0:ref type='bibr' target='#b16'>Carlson et al. 2020a)</ns0:ref>.</ns0:p><ns0:p>Importantly, given these problems with SDMs of SARS-CoV-2, there is concern that these studies may negatively affect public policy <ns0:ref type='bibr' target='#b23'>(Chipperfield et al. 2020;</ns0:ref><ns0:ref type='bibr' target='#b16'>Carlson et al. 2020a, [b]</ns0:ref> 2020). Others continue to highlight the possible connections between climate and SARS-CoV-2 and advocate for continued research in this area <ns0:ref type='bibr' target='#b4'>(Araújo, Mestre, and Naimi 2020)</ns0:ref>.</ns0:p></ns0:div>
<ns0:div><ns0:head>Caveats and unknowns</ns0:head><ns0:p>There are individual people who regularly go undetected for SARS-CoV-2 (e.g., those who lack symptoms or have mild symptoms) -currently estimated at up to 25% <ns0:ref type='bibr' target='#b51'>(Mandavilli 2020)</ns0:ref>. This is of course an issue for all studies of SARS-CoV-2, but was likely especially true for March when testing was less widely available in the US. Obviously it is not ideal for modeling and forecasting; however, it is inherent in any study. In fact, we have far more data available for SARS-CoV-2 than we will ever have for the vast majority of viruses and biodiversity generally.</ns0:p><ns0:p>Our results are based on data for a single, large country. While taking global data points would be ideal, we avoided these because they are known to have been more problematic and inconsistent <ns0:ref type='bibr' target='#b23'>(Chipperfield et al. 2020)</ns0:ref>. While it may somewhat hobble our ability to forecast viral distributions at other time points, as there are surely non-analogous weather systems to come (i.e., no place in the US in March was as hot as Death Valley, California in full summer heat), we feel it was a worthy trade-off. Policy, social factors, and a variety of other variables were not included in this study, despite their obvious, known, or potential importance <ns0:ref type='bibr' target='#b50'>(Maier and Brockmann 2020;</ns0:ref><ns0:ref type='bibr' target='#b46'>Leung et al. 2020;</ns0:ref><ns0:ref type='bibr' target='#b54'>A. Miller et al. 2020)</ns0:ref>; other variables such as wearing masks may be important for global patterns <ns0:ref type='bibr' target='#b50'>(Maier and Brockmann 2020;</ns0:ref><ns0:ref type='bibr' target='#b46'>Leung et al. 2020)</ns0:ref>.</ns0:p><ns0:p>Unfortunately, data on these variables are often lacking in public databases. Our goals were to focus on a rather macro scale, which may be harder to do with policy data that can vary substantially from county to county, state to state, and, for a global study, country to country.</ns0:p><ns0:p>While policies in the US probably had not been implemented for long enough to tell their impact in March data, these changes (e.g., self-isolation and mask wearing) clearly strongly influenced the trajectory of the virus after March.</ns0:p></ns0:div>
<ns0:div><ns0:head>Data accessibility and mobilization</ns0:head><ns0:p>We believe the New York Times (The New York Times 2020) providing robust countylevel data in an accessible repository with an open license for the US sets an excellent standard that should be repeated by governments, academics, and other media organizations for other parts of the world so that this type of study may be better repeated in any country or globally. In a similar vein, we have made our analytical and modeling pipeline (along with figures) available (https://github.com/rsh249/cv19_enm/releases/tag/v0.0.5). We believe it is imperative for all pipelines and scripts to be made available for any SARS-CoV-2 research to ensure that models can be improved upon and any errors can be more quickly uncovered and resolved.</ns0:p></ns0:div>
<ns0:div><ns0:head>Future directions</ns0:head><ns0:p>There are many avenues to pursue regarding SARS-CoV-2 modeling and predictions.</ns0:p><ns0:p>We are excited to see researchers from a variety of fields extending their toolkits towards understanding this virus. We hope that ecological studies like this and others can play a role without overcomplicating the research efforts put forth by epidemiologists. Still, studies should familiarize themselves with current critiques of SDMs for SARS-CoV-2 modeling and be cautious of their inputs and conclusions.</ns0:p><ns0:p>With improving data, we feel that future studies should better be able to examine the system globally while considering human populations and public policy efforts at curbing the virus. We also believe that it will at some point in the US and elsewhere be worth examining death rates in different areas, as it would be helpful to know if climate or other abiotic variables might impact this heartbreaking statistic. Coupling regularly updated data with automated online resources would also be particularly helpful in learning how this virus may spread.</ns0:p></ns0:div>
<ns0:div><ns0:head>Conclusions</ns0:head><ns0:p>SDMs from SARS-CoV-2 population scaled cases did not appear to be distinguishable from human population density for an early point in the pandemic for the US. Future studies looking at climate's impact on this virus should, wherever possible, take into account human population density in any analyses. Manuscript to be reviewed Manuscript to be reviewed </ns0:p></ns0:div>
<ns0:div><ns0:head>Figures and Table</ns0:head><ns0:note type='other'>Figure</ns0:note><ns0:note type='other'>Figure 4</ns0:note></ns0:div><ns0:figure xml:id='fig_0'><ns0:head /><ns0:label /><ns0:figDesc>Model testing for the SARS-CoV-2 raw case dataset identified a Maxent model with linear and quadratic feature classes and a regularization multiplier of 1.5 as the best model with a high model fit (Avg. Test AUC = 0.88) but lower model transferability (Test AUC variance = 0.007) than the population scaled model. The SDM of the raw virus cases (Figure S2) fails to reconstruct the known distribution of viral cases in much of the US with strong bias towards the Pacific Northwest and Northeastern United States. The SDM for humans (Figure 3, S2) does not exactly match current human distribution for the US. This was expected, because the goal of the model is to match climates that are correlated with the places where most people live (i.e., it</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_1'><ns0:head>1 .</ns0:head><ns0:label>1</ns0:label><ns0:figDesc>Figure 1. Probability densities of SARS-CoV-2 coronavirus cases (using population scaled data; curves in red) compared to the probability densities of human populations (curves in blue) in each US county for each of seven climate variables. Probability density curves are standardized to an area of one.</ns0:figDesc><ns0:graphic coords='17,125.64,109.95,360.71,480.96' type='bitmap' /></ns0:figure>
<ns0:figure xml:id='fig_2'><ns0:head>Figure 2 .</ns0:head><ns0:label>2</ns0:label><ns0:figDesc>Figure 2. The relationship in the US between human population size and SARS-CoV-2 coronavirus cases, using (A) total viral cases and (B) population scaled viral cases. New York City, an outlying point, has been excluded for clearer visualization.</ns0:figDesc><ns0:graphic coords='18,126.00,71.93,360.00,522.15' type='bitmap' /></ns0:figure>
<ns0:figure xml:id='fig_3'><ns0:head /><ns0:label /><ns0:figDesc>Figure S3. Equivalent of Figure 4, but using raw virus data. (A) Niche Overlap and (B) similarity tests for Maxent species distribution models built with SARS-CoV-2 coronavirus case data compared to one built with human population density as occurrence data; actual model overlap indicated by a red marker in both plots. Significant p-values correspond to greater niche overlap or similarity than expected by random models.</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_4'><ns0:head>Figure 1 Figure 1 .Figure 1 .</ns0:head><ns0:label>111</ns0:label><ns0:figDesc>Figure 1</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_5'><ns0:head>Figure 2 Figure 2 .Figure 2 .</ns0:head><ns0:label>222</ns0:label><ns0:figDesc>Figure 2</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_6'><ns0:head>Figure 3 Figure 3</ns0:head><ns0:label>33</ns0:label><ns0:figDesc>Figure 3</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_7'><ns0:head>Figure 4 .</ns0:head><ns0:label>4</ns0:label><ns0:figDesc>Figure 4. (A) Niche overlap and (B) similarity tests</ns0:figDesc></ns0:figure>
<ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='19,126.00,71.67,360.00,503.94' type='bitmap' /></ns0:figure>
<ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='20,126.00,71.80,360.00,503.94' type='bitmap' /></ns0:figure>
<ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='23,125.64,109.95,360.71,480.96' type='bitmap' /></ns0:figure>
<ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='24,125.99,72.00,360.03,504.00' type='bitmap' /></ns0:figure>
<ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='25,72.00,71.28,360.00,503.95' type='bitmap' /></ns0:figure>
<ns0:figure type='table' xml:id='tab_1'><ns0:head>Table 1 .</ns0:head><ns0:label>1</ns0:label><ns0:figDesc>Evaluation of our SDM practices against the best practices that have been proposed for this field<ns0:ref type='bibr' target='#b3'>(Araújo et al. 2019a</ns0:ref>).</ns0:figDesc><ns0:table><ns0:row><ns0:cell>Guideline</ns0:cell><ns0:cell>Standard</ns0:cell><ns0:cell>Justification</ns0:cell></ns0:row><ns0:row><ns0:cell>Response</ns0:cell><ns0:cell>A) Sampling: Bronze</ns0:cell><ns0:cell>Best data available; municipalities, local</ns0:cell></ns0:row><ns0:row><ns0:cell>variables</ns0:cell><ns0:cell /><ns0:cell>governments, and states choose who to test.</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell>Positive tests only reported.</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>B) Identification: Gold</ns0:cell><ns0:cell>Assuming best practices in testing and reporting.</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>C) Spatial accuracy:</ns0:cell><ns0:cell>County assignments provide a rough georeference</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>Bronze</ns0:cell><ns0:cell>for each record, but do not precisely describe where</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell>transmission of the virus occurred. Spatial accuracy</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell>unknown. Occurrences limited to identifiable county</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell>level localities.</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>D) Environmental</ns0:cell><ns0:cell>Limiting the study area to the continental US is</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>extent: Deficient</ns0:cell><ns0:cell>unlikely to adequately test environmental</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell>boundaries.</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>E) Geographic extent:</ns0:cell><ns0:cell>Study area to include current range in the US.</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>Bronze</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>Predictor</ns0:cell><ns0:cell>A) Selection of</ns0:cell><ns0:cell>Unclear and not well documented correlations</ns0:cell></ns0:row><ns0:row><ns0:cell>variables</ns0:cell><ns0:cell>candidates:</ns0:cell><ns0:cell>between SARS-CoV-2 transmission and climate. At</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>Bronze/Deficient</ns0:cell><ns0:cell>best, distal variables with weak, indirect control on</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell>the distribution.</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>B) Spatial and temporal</ns0:cell><ns0:cell>Variables sampled from a 2.5 arcminute grid for all</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>resolution: Deficient</ns0:cell><ns0:cell>cells within 5km of each occurrence point. Mean</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell>value used for modeling. Monthly climate averages</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell>as predictors for end of March occurrence data.</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>C) Uncertainty: Bronze</ns0:cell><ns0:cell>Temporal and spatial uncertainty in occurrence data</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell>has unquantified potential effects on the model</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell>output.</ns0:cell></ns0:row><ns0:row><ns0:cell>Model</ns0:cell><ns0:cell>A) Model Complexity:</ns0:cell><ns0:cell>ENMeval for model testing and selection (maximize</ns0:cell></ns0:row><ns0:row><ns0:cell>building</ns0:cell><ns0:cell>Silver</ns0:cell><ns0:cell>testing AUC and minimize AICc in the case of ties)</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell>using internal cross validation through the block</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell>resampling method.</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>B) Treatment of</ns0:cell><ns0:cell>Internal cross validation to evaluate bias effects in</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>response bias: Silver</ns0:cell><ns0:cell>different models.</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>C) Treatment of</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell /><ns0:cell>collinearity: Bronze</ns0:cell><ns0:cell /></ns0:row></ns0:table><ns0:note>'Approximate methods are applied' -Predictor variables hand selected from monthly climate data available to avoid collinearity (i.e., used only Tmax and not Tavg or Tmin).</ns0:note></ns0:figure>
<ns0:note place='foot'>PeerJ reviewing PDF | (2020:05:48694:1:0:NEW 14 Aug 2020)</ns0:note>
<ns0:note place='foot'>PeerJ reviewing PDF | (2020:05:48694:1:0:NEW 14 Aug 2020)Manuscript to be reviewed</ns0:note>
</ns0:body>
" | "August 14, 2020
To the Editor,
We thank you for the opportunity to follow up on the comments presented in the first round of review.
Below we have addressed the editor and reviewer concerns line-by-line with our responses and changes made to the manuscript recorded in bold.
Thank you again for your consideration of this manuscript,
Robert Harbert, Seth Cunningham, Michael Tessler
Editor’s Decision: Major Revisions
The topic of the proposed manuscript is interesting but there are currently weaknesses in the use and description of methods, as highlighted by the reviewers. I believe that the substantial comments they give would allow the authors to propose a much-improved version of their study, with appropriate application and discussion of the methods, and appropriate reference to the literature. The authors have also to improve the general presentation of the study, its consistency, and the quality of the writing.
I also support the idea that SDMs are inappropriate tools for SARS-CoV-2 coronavirus distribution. But since your conclusion is 'climate may not play a central role in current US viral distribution and that human population density is likely a primary driver', I give you a chance to address reviewers' comments and also highlight the doubts/limitation about the modelling approach you used. You might cite the below paper and expand your discussion accordingly. https://www.nature.com/articles/s41559-020-1212-8
We have addressed the reviewers’ comments below, which we believe improved the description of our methods. We have added the citation you mentioned, and a few other recent studies, and expanded our discussion to better highlight additional doubts/limitations of the modeling.
We have added a paragraph in the discussion to highlight recent review articles that show how SDMs can be problematic with SARS-CoV-2: “Further caution towards SDM methodology for modeling SARS-CoV-2 in recent review literature highlights the likely limited effect climate has on a pathogen spread via direct transmission, and thus concludes that this tool is inappropriate in this situation (Carlson et al. 2020a, [b] 2020). Important confounding variables such as human interactions, public policy, and microclimate may mask any effects climate plays on this virus (Carlson et al. 2020b; Chipperfield et al. 2020). Due to this, epidemiological studies are more suited tounderstand patterns of transmission (Carlson et al. 2020b; Chipperfield et al. 2020; Carlson et al. 2020a). Importantly, given these problems with SDMs of SARS-CoV-2, there is concern that these studies may negatively affect public policy (Chipperfield et al. 2020; Carlson et al. 2020a, [b] 2020). Others continue to highlight the possible connections between climate and SARS-CoV-2 and advocate for continued research in this area (Araújo, Mestre, and Naimi 2020). ”
Furthermore, we have updated some of our wording to make it more appropriate. We have also emphasized public policy more clearly in a few places, such as: in the introduction: “Public policy decisions appear to be the main mitigating factor against this coronavirus (Leonhardt 2020).”, and in the discussion: “Furthermore, this is inline with current knowledge that public policy is the main driver for coronavirus spread in the US (Leonhardt 2020).”
[# PeerJ Staff Note: Please ensure that all review comments are addressed in a rebuttal letter and any edits or clarifications mentioned in the letter are also inserted into the revised manuscript where appropriate. It is a common mistake to address reviewer questions in the rebuttal letter but not in the revised manuscript. If a reviewer raised a question then your readers will probably have the same question so you should ensure that the manuscript can stand alone without the rebuttal letter. Directions on how to prepare a rebuttal letter can be found at: https://peerj.com/benefits/academic-rebuttal-letters/ #]
[# PeerJ Staff Note: It is PeerJ policy that additional references suggested during the peer-review process should only be included if the authors are in agreement that they are relevant and useful #]
Reviewer 1
Basic reporting
N/A
Experimental design
N/A
Validity of the findings
N/A
Comments for the author
N/A
Reviewer 2
Basic reporting
This paper seeks to determine whether climate data can contribute to predicting the range of SARS-CoV-2 in addition to human population distribution patterns. They found little evidence that climatic variables contribute much predictive power not explained by human population distribution. The paper is written with appropriate caution and the authors outline the caveats well.
However, there is much detail missing from the Methods, and the wording is so confusing that it's difficult to determine exactly what was done. This is despite the authors' own claims about the good level of methodological information they reported.
We have reworked any specific requests to the methods section, as seen in comments below to the reviewers. We have also generally gone over the methods and attempted to clarify and expand on any details that were missing.
Further, some crucial elements were missing or lacking from either the Introduction or Discussion:
1) a clear acknowledgment that viral distribution patterns are also heavily driven by regional politics (i.e., country-specific responses to public health). Doesn't this also need to be accounted for (especially in the US)? High transmission rates, regardless of climate or human behavior, may be higher somewhere just because a particular gov't did not prepare sufficiently, or because they are reporting cases more accurately, etc.
The models we produced were based on cumulative positive test data as of March 30, 2020 which was around two weeks or less from the implementation of state and national public health directives.
2) a brief review of how SDMs have been used to model pathogen distributions, the difficulties they faced, how they overcame them, etc.
We include in the Introduction an expanded, but brief, summary of some of the recent literature around applying SDMs and ecological niche modeling to pathogen distributions: “SDMs built to predict the spread of viral pathogens often do so by modeling the potential distribution of a known vectors or alternate hosts (R. H. Miller et al. 2012; Larson et al. 2010; de Almeida et al. 2019; Richman et al. 2018), but some base the models directly on pathogen occurrence and correlations with environmental, climatic, or demographic covariates (Machado-Machado 2012; Belkhiria, Alkhamis, and Martínez-López 2016; Pigott, Bhatt, et al. 2014; Messina et al. 2015; Bhatt et al. 2013). Efforts to model global disease distribution within an ecological framework frequently draw on known or hypothesized environmental variables correlated with disease occurrence to predict suitability for transmission. Output from these models can be extrapolated into areas where a disease has not yet been reported, but suitable environmental data exist (Elith and Leathwick 2009)”
As a whole, although the topic may be of interest to those studying SARS-CoV-2 distributional patterns and the paper does call out some preprints claiming that the viral distribution can be properly explained by climate, key elements are missing that make for potential confusion, and these need to be addressed.
Thank you for this summary.
Experimental design
The main research question was well-defined, but the way key results were referred to was inconsistent (please use the same names for data products throughout), key methodological details were missing, and some confusion regarding model evaluation appeared to exist. I explain in the line-by-line comments, but selection of a model via AIC does not consider cross validation results, despite the authors' claims they used spatial cross validation to choose models. Table 1 is appreciated, but these same details must be referred to somewhere in the Methods text.
Thank you for finding these issues, and thoroughly detailing them in the line-by-line comments below. We have addressed each of the issues stated here in the section below, and feel the manuscript is much improved.
Validity of the findings
This paper does help to dispel the notion that viral distributions can be predicted solely with climate, but the analysis itself does not include enough human-specific variables to make a convincing model prediction. For example, some effect of regional politics is sure to play a role in infection rates, especially in the US, but this is glossed over in a single sentence in the Discussion. Nonetheless, the findings of this study will contribute overall to the understanding of the biogeography of SARS-CoV-2, and the authors are entirely transparent about data sources.
The aim of this study is to examine the potential impact of climate on the early patterns of SARS-CoV-2 case distributions in the United States, before public policy was broadly implemented in the US. We acknowledge that our study is limited to March and discuss some of the other potential factors like public policy and human behavior. Most stay-at-home orders in the US were issued in the later half of March.
1. Abstract: “The case data were from across March, 2020, before large travel restrictions and public health policies were impacting cases across the country.”
2. Introduction: “..., particularly before policy seemed to become the principal driver of the virus.”
We have also emphasized public policy more clearly in a few places, such as:
1. Introduction: “Public policy decisions appear to be the main mitigating factor against this coronavirus (Leonhardt 2020).”
2. Discussion: “Furthermore, this is inline with current knowledge that public policy is the main driver for coronavirus spread in the US (Leonhardt 2020). It is also inline with modeling efforts that have combined epidemiological variables with climate (Baker et al. 2020).”
Comments for the author
Line-by-line comments:
L65: Change to: 'appear to explain much of the viral distribution'.
Corrected. Thank you for the suggestion.
L69-70: If SARS-CoV-2 is dependent on human hosts, what kinds of abiotic variables might shape its niche? Wouldn't most of these niche factors relate to the host's internal environment? How would knowledge of the 'abiotic preferences' of the virus help contain its spread. Please explain more clearly.
“Any improved understanding of the abiotic variables influencing the niche of SARS-CoV-2 might, in part, help prevent viral spread”
Replaced with: “Identification of abiotic variables with suitably large effect size on the distribution and spread of SARS-CoV-2 might help prevent the viral spread by identifying higher risk regions and project the seasonal variation in the risk of transmission.”
L73: Remove 'of'.
L76: Maybe 'foundation' instead of jumping-off point?
L86: lipid-enveloped
L112: Remove colon
L121: Isn't it 'to at least some extent reflect the habitat patterns of their host'?
L136: self-isolation
We have corrected each of the above specific questions. Thank you for the suggestions.
L140: Needs updating
We have updated this paragraph. Thanks for pointing this out.
L164: Change to 'interpolated climate data'. Can you at least verify that using a more up-to-date interpolated climate dataset like CHELSA would not have affected results very much? More importantly, how good is Worldclim at predicting current weather? This pandemic just occurred. Wouldn't using PRISM have made more sense?
Corrected wording to “interpolated climate data”.
We used the WorldClim version 2 which was published in 2017, and released for use more recently than that. CHELSA is also newer, but the WorldClim version 2 uses better practices and is more accurate than the older v1.4 that is commonly used in distribution modeling studies. Also, WorldClim v2 includes variables for water vapor pressure (related to available moisture/humidity) and solar radiation, both of which have been implicated in SARS-CoV-2 transmission dynamics.
Added in the Introduction: “Recent studies have also indicated that SARS-CoV-2 transmission rate may be negatively correlated with temperature and humidity (Qi et al. 2020; Wu et al. 2020; Rosario et al. 2020; Ahmadi et al. 2020). Solar radiation, or sunlight, may also reduce transmission rate by inactivating virus particles in aerosolized droplets (Sagripanti and David Lytle 2020).”
Added in Methods: “The seven climate parameters examined for the month of March, 2020 were average monthly temperature, average monthly minimum temperature, average monthly maximum temperature, average monthly precipitation, average daily solar radiation, wind speed, and water vapor pressure (a measure related to humidity). These climate parameters are consistent with possible correlates of SARS-CoV-2 transmission in several recent studies (Qi et al. 2020; Sagripanti and David Lytle 2020; Ahmadi et al. 2020; Rosario et al. 2020; Wu et al. 2020)”
L166: Min and max of what? Avg monthly? Of warmest quarter, etc.?
Changed to: “The seven climate parameters examined for the month of March, 2020 were average monthly temperature, average monthly minimum temperature, average monthly maximum temperature, average monthly precipitation, average daily solar radiation, wind speed, and water vapor pressure”
L175: What is the purpose of this section? Please outline the logic behind your methods at the end of the intro.
We have renamed this section “Climate Distribution Visualization”. We believe this better highlights what the subsection does: Produce visualizations for each climate variable, separate from our SDMs. We have also added the following to the end of the Introduction: “Here we develop a suite of data visualizations and species distribution models using both climate and human population data to determine whether the effect of climate can be appropriately disentangled from other drivers of SARS-CoV-2 transmission, given early records for the virus in the US. We feel that examining a relatively early time-point of the pandemic in the US is useful, as it in part (probably largely) predates the impact of major shifts in public health policy for the US.”
L177: Probability of what? What exactly is the function of this kernel density grid? Why does it use climate variables? It looks from Fig. 1 that you took the kernel densities of cases and human pop, not climate (which would not make sense anyway). At this stage in the paper I am confused.
We have changed this to: “Probability density distributions for SARS-CoV-2 are produced to characterize the likelihood of case occurrence given the available range of climate values. These distributions are calculated from case occurrences and corresponding climate data using a Gaussian Kernel Density Estimator and standard bandwidth estimation. These distributions can be interpreted similarly to histograms, but are normalized to be comparable between different sample sizes.”
L180: population-scaled
Corrected. Thank you for the suggestion.
L189: By 'climate record', do you mean county record associated with climate? Also, what is a 'raw SDM'? L190-193: I don't follow this at all. What does it mean to 'expand' a record? Why would any counties have fewer than one data point?
Revised the description of our record resampling and scaling method to clarify the two points above: “Occurrence data for raw SDMs were generated by expanding each county climate record by multiplying that occurrence by the total SARS-CoV-2 case count for each county. Occurrence data for population scaled SDMs were generated by expanding each county climate record by the total county SARS-CoV-2 case count divided by the county population such that no county with at least one positive case had fewer than one occurrence [i.e., cases / population x 100,000 (the expansion factor for our data)].”
L195: It is not clear what the difference between these two is. Please explain more clearly. Also, where is the description of the 'human population SDM'? Not even sure what that means, as SDMs are based on presence data, not abundance.
We have recast the paragraph to address your points: “Occurrence data for raw (i.e., total viral cases) SDMs were generated by expanding each county climate record by multiplying that occurrence by the total SARS-CoV-2 case count for each county, while occurrence data for population- scaled SDMs were generated by expanding each county climate record by the total county SARS-CoV-2 case count divided by the county population such that no county had fewer than one record. More explicitly: raw data = total viral cases; population-scaled data = cases / population x 100,000 (the expansion factor for our data). Viral host availability (i.e., total humans per county) was modeled to serve as a null distribution to be used as a control comparison for viral SDMs.”
L195-199: It looks like you selected model settings for Maxent based on two datasets (but not clear what you did if the two were different), then used these settings to make models with other datasets. This makes little methodological sense, so I think I must be misunderstanding. Please clarify. Also, was maxnet used in your ENMeval run?
Maxnet was used in the ENMeval run. This has been clarified. Text not copied here because this paragraph has undergone extensive editing.
L202: Please explain briefly why this is preferred for models with regularization. From my understanding, it is best practice for the purposes of interpretation and to avoid extrapolation errors due to changes in covariance. Also, where do you discuss model evaluation?
Revised this sentence to better explain. Thank you. -- “Models built using all temperature variables made similar models, but are not further reported on. Reducing correlated variables is generally preferable to avoid extrapolation issues that arise from local covariance between variables (Araújo et al. 2019)”
L205: Please rephrase for correct grammar and fix spelling.
We have changed the sentence to read as follows: Given recent critiques, it has become clear that when using SDMs in relation to SARS-CoV-2 it is worth documenting how close a study may come to the gold standards set for this type of modeling (Araújo et al., 2019).
L209: Please explain the niche overlap and similarity tests. What do they do? How are they different?
Revised: “Niche overlap and similarity tests were conducted with the “ecospat” library (Di Cola et al. 2017; Broennimann et al. 2012; Warren et al. 2008) to compare the climatic niche of SARS-CoV-2 and humans. The niche overlap test considers whether there is a greater than expected overlap between the climate space occupied by two population models (human total vs. SARS-CoV-2 positive) than would be expected given a null distribution sampled from both populations. The niche similarity test considers the degree of similarity between the density of occurrence between two population models relative to the same null distribution (Warren et al. 2008). A significant overlap between SARS-CoV-2 and total human population models these may indicate that the virus distribution is not necessarily constrained by climate (i.e., that the full range of human occupied climate is accessible to the virus).”
L218: record(s)
Done
L238: Please clarify 'for people'. As there are multiple variables considered, please reference them explicitly to avoid confusion.
We have changed the sentence to be more explicit. Thank you for the suggestion. It now reads: “There are some differences; for instance, the map for SARS-CoV-2 (Figure 3A) suggests Florida is less suitable than it is in the map for human population (Figure 3B).”
L271-272: The model does not indicate that raw values should not be used as inputs. You infer that after analyzing the results.
Revised to: “Other SDM studies have only used raw values as inputs (Araújo & Naimi, 2020; Bariotakis et al., 2020), which our model, again, indicates should not be used due to the risk of overfitting and low model transferability to novel regions”
L302-304: This seems like a relatively trivial exercise given you have the data in hand. Why didn't you do this and briefly compare it to your other results?
This is not a trivial exercise. The population data are not spatially continuous and are not appropriate as an input layer for distribution modeling. Doing so would require interpolating a population density surface for the US at a resolution compatible with the climate model data.
L324: If we are missing lots of data from more tropical areas, this would certainly affect global climate models. Why such uncertainty?
Revised to: “Even if we are working with incomplete data, it is hard to know if adding those data will significantly change conclusions based on preliminary models (i.e., will the SARS-CoV-2 model still be indistinguishable from a human population model).”
L325: What 'best practices' are you referring to? The lead author on the paper criticized is the same one in the best practices paper you cite. Shouldn't this person, of all people, adhere to best practices? That seems like a valid criticism.
Yes, that was a major critique of the paper (that the senior author both made the best practices and then did not follow them). It was our aim to do better and present a more transparent analysis in line with the recommendations in the cited paper. We implement several improved practices here and discuss them. Other feedback from this reviewer has helped clarify these in the Methods section.
L328: A number of these best practices are missing from your Methods.
We have added better method descriptions that cover these best practices outlined in Table 1. Thank you for pointing this out.
L351: Needs updating: https://www.who.int/news-room/commentaries/detail/bacille-calmette-gu%C3%A9rin-(bcg)-vaccination-and-covid-19
We have deleted the offending citation from this line.
Table 1. Using AICc for model selection does not consider the results of cross validation at all. So regardless of the fact that you used spatial block cross validation with ENMeval, you did not consider the cross validation results unless you also considered test stats like testing AUC or omission rates, etc. Key details like which Maxent parameters you considered, etc., are missing from the Methods.
Edited table 1 to include: “ENMeval for model testing and selection (maximize testing AUC and minimize AICc in the case of tiesAICc) using internal cross validation through the block resampling method.”
The methods for our reanalysis with proper cross validation are also updated in the methods section.
Fig 1. Nowhere in the text do you explain the relevance of wind speed, water vapor pressure, or solar radiation to SARS-CoV-2 distribution.
We have partially addressed this in our revision regarding Line 164 (“Climate Data” section in the Methods) above, please see that part of our response for more detail regarding our choice of variables for modeling.
The first paragraph of our results deals with this. Our main point is that: “The human population climate curves are visually similar compared to the population scaled SARS-CoV-2 climate curves for all seven abiotic variables...” The specifics are not all that important in our mind. We simply believe the virus is where people are, not that these results show that wind speed or another variable is found to specifically be of importance. If you still do not agree and there is something specific you would like us to add, we would be happy to accommodate.
Reviewer 3
Basic reporting
no comment
Experimental design
no comment
Validity of the findings
no comment
Comments for the author
The authors use an SDM model to compare suitability for SARS-CoV-2 with human population suitability. The aim of the study is partly to critique earlier work using similar models which found a more conclusive effect of climate on SARS-CoV-2. I think the paper is well-written, well-argued and important. I recommend the study be published by PeerJ. I have only minor comments:
92 is this “scientists” or ecologists or some other sub-group? There are lot of scientists who do not use these methods so I think you need to be more specific.
Thank you for the suggestion. We have replaced “Scientists” with “Biologists” to be more specific.139, remove “our” here as you haven’t introduced your model yet, suggest “While epidemiological models are highly uncertain, COVID-19 may peak…”
Corrected. Thank you for the suggestion.
141 This sentence is confusing. It seems you are trying to state the purpose of the paper here? “Baseline” could mean a lot of things. I think the purpose of the paper is better stated in the abstract and some version of this should be re-stated here e.g. “Here we develop a SDM model using both climate and human population data to determine whether the effect of climate can be appropriately disentangled from other drivers…” – some better version of this.
We have revised this whole paragraph to better reflect our goals and have incorporated this suggestion: “Here we develop a suite of data visualizations and species distribution models using both climate and human population data to determine whether the effect of climate can be appropriately disentangled from other drivers of SARS-CoV-2 transmission, given early records for the virus in the US. We feel that examining a relatively early time-point of the pandemic in the US is useful, as it in part (probably largely) predates the impact of major shifts in public health policy for the US.”
155 I think the authors should clearly state that they are only take one value per county: total cumulative cases as of March 30th. Are call counties reporting? Would be nice to know the sample size.
We have clarified these points: “SARS-CoV-2 case data for US counties (one record per county, with a total sample size of 1,883 counties that had available reports)...”
169 5km buffer is very small, aren’t most counties much larger than this? You could have used a county shapefile to take averages over the climate data. Or by county center do you mean the population center? If you mean the county centroid, then that isn’t necessarily representative of where people live? In general, I don’t think this will matter too much, as climate data tends to be highly correlated over short distances. Perhaps mention this to justify using the small area.
Thank you for these points. Accordingly, we have changed “county center” to “county centroid”, and more importantly we have clarified our buffer usage: “Climate values were extracted from these layers for each georeferenced county record using the average from a 5km buffer around each county centroid coordinate. These local averages should represent the county generally as climate is correlated over short distances, but these averages may not be the best representation in counties that have a large area or highly clustered population centers”
176 (similar to line 155 comment) Are you using the total cases as of March 31ST 2020? State this clearly. Lots of the climate-regression studies use the whole time series of cases and compare variations in county-level climate with county-level cases over time. I think you need to make clear that you are using a purely cross-sectional approach.
Thank you for pointing this out, we have added the date to this section: “Total cumulative positive case data for SARS-CoV-2 in each reporting county on March 30, 2020 were used to extract data for each compared to the seven abiotic variables listed above”.
Discussion: Probably worth citing this recent paper in Science suggesting even if climate can impact transmission, it will have minimal impact compared to population susceptibility. This may explain higher case numbers in Florida and other warmer locations.
Baker, Rachel E., et al. 'Susceptible supply limits the role of climate in the early SARS-CoV-2 pandemic.' Science (2020).
Thank you for pointing us to this paper, we have added it to the manuscript.
Introduction: “Recent epidemiological models have also highlighted that weather may not play a large role when most people lack immunity during the pandemic phase, but that weather has the potential to play a larger role once endemic phase (Baker et al. 2020).”
Discussion: “Discussion: “It is also inline with modeling efforts that have combined epidemiological variables with climate (Baker et al. 2020).”
" | Here is a paper. Please give your review comments after reading it. |
9,785 | "<ns0:abstract xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:p>The SARS-CoV-2 coronavirus is wreaking havoc globally, yet, as a novel pathogen, knowledge of its biology is still emerging. Climate and seasonality influence the distributions of many diseases, and studies suggest at least some link between SARS-CoV-2 and weather. One such study, building species distribution models (SDMs), predicted SARS-CoV-2 risk may remain concentrated in the Northern Hemisphere, shifting northward in summer months. Others have highlighted issues with SARS-CoV-2 SDMs, notably: the primary niche of the virus is the host it infects, climate may be a weak distributional predictor, global prevalence data have issues, and the virus is not in population equilibrium. While these issues should be considered, we believe climate's relationship with SARS-CoV-2 is still worth exploring, as it may have some impact on the distribution of cases. To further examine if there is a link to climate, we build model projections with raw SARS-CoV-2 case data and population-scaled case data in the United States. The case data were from across March 2020, before large travel restrictions and public health policies were impacting cases across the country. We show that SDMs built from population-scaled case data cannot be distinguished from control models (built from raw human population data), while SDMs built on raw case data fail to predict the known distribution of cases in the U.S. from March. The population-scaled analyses indicate that climate did not play a central role in early U.S. viral distribution and that human population density was likely the primary driver. We do find slightly more population-scaled viral cases in cooler areas. Ultimately, the temporal and geographic constraints on this study mean that we cannot rule out climate as a partial driver of the SARS-CoV-2 distribution.</ns0:p><ns0:p>Climate's role on SARS-CoV-2 should continue to be cautiously examined, but at this time</ns0:p></ns0:div>
</ns0:abstract>
<ns0:body xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:head>Introduction</ns0:head></ns0:div>
<ns0:div><ns0:head>SARS-CoV-2 and climate overview</ns0:head><ns0:p>In March 2020, the United States of America had the most reported cases of COVID-19, the disease caused by the coronavirus SARS-CoV-2 (CDC NCIRD 2020). While we are now beginning to understand the global and US-level distributions of this virus, we are only starting to learn how abiotic variables affected its geographic distribution, particularly before policy seemed to become the principal driver of the virus. With over two months of SARS-CoV-2 records across U.S. counties by the end of March 2020, the role of abiotic variables had become easier to examine, especially as multiple sources began compiling data and making it public (The New York Times 2020; <ns0:ref type='bibr' target='#b25'>Dong, Du, and Gardner 2020;</ns0:ref><ns0:ref type='bibr'>The COVID Tracking Project 2020)</ns0:ref>.</ns0:p><ns0:p>Organisms are distributed within their environment based on both direct and indirect interactions with biotic and abiotic variables <ns0:ref type='bibr' target='#b26'>(Elith and Leathwick 2009</ns0:ref>) -SARS-CoV-2 is no exception. As this virus is primarily distributed by human hosts, it is constrained by human distributions and interactions (i.e., biotic variables). There are numerous documented cases on local and global scales of individual human sources for new outbreaks <ns0:ref type='bibr' target='#b37'>(Holshue et al. 2020;</ns0:ref><ns0:ref type='bibr' target='#b41'>KCDC 2020)</ns0:ref>. Still, many other viruses, including other coronaviruses, display marked seasonality and are affected by local climatic conditions <ns0:ref type='bibr' target='#b66'>(Price, Graham, and Ramalingam 2019;</ns0:ref><ns0:ref type='bibr' target='#b31'>Fisman 2012;</ns0:ref><ns0:ref type='bibr' target='#b48'>Lofgren et al. 2007;</ns0:ref><ns0:ref type='bibr' target='#b34'>Gaunt et al. 2010</ns0:ref>). This has prompted researchers to begin looking globally at how weather and climate (i.e., a subset of abiotic variables) may relate to the presence and abundance of the virus.</ns0:p><ns0:p>Modeling the SARS-CoV-2 viral distribution from the early part of the outbreak in relation to climate and other abiotic variables could help refine our understanding of its spread, hopefully adding to the knowledge gleaned from studies on human transmission dynamics <ns0:ref type='bibr' target='#b43'>(Kucharski et al. 2020;</ns0:ref><ns0:ref type='bibr' target='#b22'>Chinazzi et al. 2020</ns0:ref>), which appears to explain much of the viral distribution. Climate can also be compared directly to biotic variables, such as human population density, to disentangle their relative importances. More dense human populations can have elevated communicable disease spread, as there is more person-to-person contact <ns0:ref type='bibr' target='#b28'>(Fang et al. 2013</ns0:ref>)the principal way SARS-CoV-2 is known to spread <ns0:ref type='bibr' target='#b71'>(Rothan and Byrareddy 2020)</ns0:ref>. Still, identification of abiotic variables with suitably large effect size on the distribution and spread of SARS-CoV-2 might help prevent viral spread by identifying higher-risk regions and projecting seasonal variation in the risk of transmission. However, it is important to make clear that the models presented here, and in other recent studies <ns0:ref type='bibr'>(Araújo and Naimi 2020;</ns0:ref><ns0:ref type='bibr' target='#b10'>Bariotakis et al. 2020)</ns0:ref>, are influenced by the variability of input parameters, the variables explored, and the restrictions inherent in modeling any complex system <ns0:ref type='bibr' target='#b27'>(Elith et al. 2011)</ns0:ref>. We present this study to serve as a baseline in our understanding, given the available data, rather than as a definitive model.</ns0:p></ns0:div>
<ns0:div><ns0:head>SARS-CoV-2 and climate research</ns0:head><ns0:p>Early studies suggest that transmission of SARS-CoV-2 may have, at a minimum, a loose association with climatic features. There have been numerous reports showing a correlation of case incidence with cool temperatures and low humidity <ns0:ref type='bibr' target='#b77'>(Wang et al. 2020;</ns0:ref><ns0:ref type='bibr' target='#b29'>Ficetola and Rubolini 2020;</ns0:ref><ns0:ref type='bibr' target='#b8'>Bannister-Tyrrell et al. 2020;</ns0:ref><ns0:ref type='bibr'>Araújo and Naimi 2020;</ns0:ref><ns0:ref type='bibr' target='#b67'>Chen et al. 2020;</ns0:ref><ns0:ref type='bibr' target='#b2'>Alvarez-Ramirez and Meraz 2020;</ns0:ref><ns0:ref type='bibr' target='#b57'>Notari 2020;</ns0:ref><ns0:ref type='bibr' target='#b58'>Paez et al. 2020)</ns0:ref>. Furthermore, studies that controlled for case growth rate or demographic factors found that weather was still a significant factor in the success of SARS-CoV-2 outbreaks <ns0:ref type='bibr' target='#b15'>(Bukhari and Jameel 2020;</ns0:ref><ns0:ref type='bibr' target='#b67'>Chen et al. 2020;</ns0:ref><ns0:ref type='bibr' target='#b73'>Sajadi et al. 2020)</ns0:ref>. Unfortunately, none of these studies provide a mechanism for how abiotic variables like temperature might limit or promote the person-to-person transmission of the virus. Still, lipid-enveloped viruses (including coronaviruses) may be more stable outside of the host in lower humidity and cooler temperatures, conditions that are common in temperate areas in late winter and early spring <ns0:ref type='bibr' target='#b66'>(Price, Graham, and Ramalingam 2019)</ns0:ref>. Recent studies have also indicated that SARS-CoV-2 transmission rate may be negatively correlated with temperature and humidity <ns0:ref type='bibr' target='#b67'>(Qi et al. 2020;</ns0:ref><ns0:ref type='bibr' target='#b81'>Wu et al. 2020;</ns0:ref><ns0:ref type='bibr' target='#b70'>Rosario et al. 2020;</ns0:ref><ns0:ref type='bibr' target='#b0'>Ahmadi et al. 2020</ns0:ref>) <ns0:ref type='bibr' target='#b72'>(Sagripanti and David Lytle 2020)</ns0:ref>. However, the mechanism causing these correlations could be indirect, as humans behave differently in different seasons <ns0:ref type='bibr' target='#b11'>(Bedford et al. 2015;</ns0:ref><ns0:ref type='bibr' target='#b79'>Wesolowski et al. 2017)</ns0:ref>. Recent epidemiological models have also highlighted that weather may not play a large role when most people lack immunity during the pandemic phase but</ns0:p><ns0:p>indicate that weather has the potential to play a larger role once the endemic phase is reached <ns0:ref type='bibr'>(Baker et al. 2020)</ns0:ref>. Biologists often employ species distribution models (SDMs; alternatively called ecological niche models) to predict geographic ranges of species. SDMs employ environmental data (typically climate) to predict if geographic space is suitable for a given species or population <ns0:ref type='bibr' target='#b60'>(Peterson 2001;</ns0:ref><ns0:ref type='bibr' target='#b61'>Peterson et al. 2011</ns0:ref>). These models have proven useful in a wide variety of applications, such as invasion biology, climate change, zoonotic diseases, and speciation <ns0:ref type='bibr' target='#b26'>(Elith and Leathwick 2009;</ns0:ref><ns0:ref type='bibr' target='#b35'>Guisan and Thuiller 2005)</ns0:ref>. SDMs built to predict the spread of viral pathogens often do so by modeling the potential distribution of known vectors or alternate hosts (R. H. <ns0:ref type='bibr' target='#b55'>Miller et al. 2012;</ns0:ref><ns0:ref type='bibr' target='#b44'>Larson et al. 2010;</ns0:ref><ns0:ref type='bibr' target='#b1'>de Almeida et al. 2019;</ns0:ref><ns0:ref type='bibr' target='#b69'>Richman et al. 2018</ns0:ref>), but some base the models directly on pathogen occurrence data <ns0:ref type='bibr' target='#b49'>(Machado-Machado 2012;</ns0:ref><ns0:ref type='bibr' target='#b12'>Belkhiria, Alkhamis, and Martínez-López 2016;</ns0:ref><ns0:ref type='bibr' target='#b63'>Pigott, Bhatt, et al. 2014;</ns0:ref><ns0:ref type='bibr'>Messina et al. 2015;</ns0:ref><ns0:ref type='bibr' target='#b13'>Bhatt et al. 2013</ns0:ref>). Efforts to model global disease distribution within an ecological framework frequently draw on known or hypothesized environmental variables correlated with disease occurrence to predict suitability for transmission. Output from these models can be extrapolated into areas where a disease has not yet been reported, but where suitable environmental variables correlated with its occurrence exist <ns0:ref type='bibr' target='#b26'>(Elith and Leathwick 2009)</ns0:ref>.</ns0:p><ns0:p>Researchers early in the pandemic created SDMs of SARS-CoV-2 using climatic variables <ns0:ref type='bibr'>(Araújo and Naimi 2020;</ns0:ref><ns0:ref type='bibr' target='#b10'>Bariotakis et al. 2020)</ns0:ref>. These studies suggested that the virus is strongly constrained by global climate patterns (Araújo and Naimi 2020; <ns0:ref type='bibr' target='#b10'>Bariotakis et al. 2020)</ns0:ref>. Preliminary models also suggested that the virus would continue to be concentrated in the Northern Hemisphere, shifting northwards throughout the summer and then back towards its spring distribution in the fall and winter <ns0:ref type='bibr'>(Araújo and Naimi 2020)</ns0:ref>.</ns0:p><ns0:p>However, a research group put forth a strong rebuttal <ns0:ref type='bibr' target='#b23'>(Chipperfield et al. 2020</ns0:ref>) to some of this modeling work <ns0:ref type='bibr'>(Araújo and Naimi 2020)</ns0:ref>. Their main criticism asserts that basic assumptions underlying SDMs are violated when modeling SARS-CoV-2, due to the mode of transmission, current population disequilibrium, and failure to incorporate epidemiological data <ns0:ref type='bibr' target='#b23'>(Chipperfield et al. 2020)</ns0:ref>. Furthermore, they highlight issues pertaining to input data for global records, such as: missing data, omitted data, and single localities representing many thousands of reported cases (hospital coordinates or political centroids) across an entire country <ns0:ref type='bibr'>(Araújo and Naimi 2020)</ns0:ref>. This is exacerbated as many countries are reporting records based on varied criteria, such as multiple types of molecular tests and tests using computed tomography scans <ns0:ref type='bibr' target='#b47'>(Lippi, Simundic, and Plebani 2020)</ns0:ref>. The rebuttal additionally suggested that model evaluation and justification were insufficient. An updated draft of the original paper <ns0:ref type='bibr'>(Araújo and Naimi 2020)</ns0:ref> addressed some of these issues, and produced reasonable arguments to some others. While we agree with many arguments from the rebuttal, we believe SDMs have the potential to be useful for modeling in a number of instances, if done carefully, and have been effective when used for other diseases <ns0:ref type='bibr'>(Pigott, Golding, et al. 2014;</ns0:ref><ns0:ref type='bibr' target='#b65'>Pigott et al. 2015;</ns0:ref><ns0:ref type='bibr' target='#b19'>Carlson, Dougherty, and Getz 2016)</ns0:ref> and other instances when data are limited or incomplete <ns0:ref type='bibr' target='#b33'>(Galante et al. 2018;</ns0:ref><ns0:ref type='bibr' target='#b9'>Barbet-Massin et al. 2018;</ns0:ref><ns0:ref type='bibr' target='#b40'>Katz and Zellmer 2018;</ns0:ref><ns0:ref type='bibr' target='#b59'>Pearson et al. 2006;</ns0:ref><ns0:ref type='bibr' target='#b32'>Fois et al. 2018;</ns0:ref><ns0:ref type='bibr' target='#b36'>Hernandez et al. 2006;</ns0:ref><ns0:ref type='bibr' target='#b42'>Kiedrzyński et al. 2017)</ns0:ref>.</ns0:p><ns0:p>Still, the climate-based SDMs for SARS-CoV-2 presented recently may, to at least some extent, reflect the climatic preferences of their host. Careful consideration of host availability (human population density) and pathogen ecologies (abiotic variables related to transmission) may be necessary to frame analyses modeling the global distribution <ns0:ref type='bibr'>(Johnson, Escobar, and</ns0:ref> even town-level data would provide finer scale resolution to our analyses, these data are likely the best curated dataset for the U.S. for this time period.</ns0:p></ns0:div>
<ns0:div><ns0:head>Climate data</ns0:head><ns0:p>Interpolated climate data averaged from 1970-2000 for the month of March were accessed through the WorldClim v2.1 database <ns0:ref type='bibr' target='#b30'>(Fick and Hijmans 2017)</ns0:ref>. The seven climate parameters examined for the month of March were average monthly temperature, average monthly minimum temperature, average monthly maximum temperature, average monthly precipitation, average daily solar radiation, wind speed, and water vapor pressure (a measure related to humidity). These climate parameters are consistent with possible correlates of SARS-</ns0:p><ns0:p>CoV-2 transmission in several recent studies <ns0:ref type='bibr' target='#b67'>(Qi et al. 2020;</ns0:ref><ns0:ref type='bibr' target='#b72'>Sagripanti and David Lytle 2020;</ns0:ref><ns0:ref type='bibr' target='#b0'>Ahmadi et al. 2020;</ns0:ref><ns0:ref type='bibr' target='#b70'>Rosario et al. 2020;</ns0:ref><ns0:ref type='bibr' target='#b81'>Wu et al. 2020)</ns0:ref>. Climate data were extracted from these layers for each georeferenced county record using the average from a 5km buffer around each county centroid coordinate. These local averages should represent the county generally as climate is correlated over short distances, but these averages may not be the best representation in counties that have a large area or highly clustered population centers. While county centroids may not be ideal in all circumstances, our view is that county centroids plus this buffer will average the climate relative to most cases in the area. More precise case georeferencing is not possible without high levels of individual case and movement tracking to identify the source of infection.</ns0:p></ns0:div>
<ns0:div><ns0:head>Climate Distribution Visualization</ns0:head><ns0:p>Total cumulative positive case data for SARS-CoV-2 in each reporting county on March 30, 2020, were used to extract data for the seven abiotic variables listed above. Probability density distributions for SARS-CoV-2 were produced to characterize the likelihood of case occurrence given the available range of climate values. These distributions are calculated from case occurrences and corresponding climate data using a Gaussian Kernel Density Estimator and standard bandwidth estimation. These distributions can be interpreted similarly to histograms, but are normalized to be comparable between different sample sizes. Raw case data (i.e., total cases per county) were scaled to reflect virus cases in each county unit by dividing total cases by the county population from the 2010 U.S. Census. These populationscaled viral cases (cases / population) were used as probability density weightings, and resulting curves were standardized to an area of one. Probability densities were also calculated with raw case count data and county-level population as weightings. Probability density estimation and visualization was done with the R 'ggplot2' package <ns0:ref type='bibr' target='#b80'>(Wickham 2009)</ns0:ref>. To test whether the SARS-CoV-2 cases and human population data were correlated, we applied Spearman tests 'cor.test(method=' spearman')'.</ns0:p></ns0:div>
<ns0:div><ns0:head>Species distribution models</ns0:head><ns0:p>Occurrence data for raw (i.e., total viral cases) SDMs were generated by expanding each county climate record by multiplying that occurrence by the total SARS-CoV-2 case count for each county. Occurrence data for population-scaled SDMs were generated the same way as for the raw SDMs except that county climate records were multiplied by the total case count divided by the county population. Then population-scaled values were multiplied by an expansion factor of 100,000 so that all counties with at least one case were represented. More explicitly: raw data = total viral cases; population-scaled data = total viral cases / human population x 100,000 (the expansion factor for our data). Viral host availability (i.e., total humans per county) was modeled to serve as a null distribution to be used as a control comparison for viral SDMs.</ns0:p><ns0:p>Maxent, an algorithm for presence-only distribution models <ns0:ref type='bibr' target='#b27'>(Elith et al. 2011)</ns0:ref>, parameters relating to model complexity were tested by building a suite of SDMs for SARS-CoV-2 distribution data with occurrences generated from raw reported virus case values and population-scaled case values. Maxent model testing and cross validation was performed using the ENMeval package <ns0:ref type='bibr' target='#b56'>(Muscarella et al. 2014</ns0:ref>) considering linear and quadratic feature classes (constraints on model fit) and regularization multipliers (penalties on complexity) of 0.5, 1, 1.5, 2, 2.5, and 3. Optimal model parameters were chosen by maximizing the average test AUC calculated with cross validation using the spatial 'block' partitioning method, and minimizing AICc in the case of ties.. All models were built using the Maxent algorithm implemented in the 'maxnet' R package <ns0:ref type='bibr' target='#b62'>(Phillips 2017)</ns0:ref>, including those tested within ENMeval. Operational models using the optimal model parameters were then built for SARS-CoV-2 using all population-scaled data, raw virus data, and for the human population using the county population data <ns0:ref type='bibr' target='#b62'>(Phillips 2017)</ns0:ref>. Of the seven WorldClim variables, the three temperature variables are not independent.</ns0:p><ns0:p>Accordingly, we only used maximum temperature, as higher temperatures have been hypothesized to lower the viral distribution. Models built using all temperature variables made similar models, but are not further reported on. Reducing correlated variables is generally preferable to avoid extrapolation issues that arise from local covariance between variables <ns0:ref type='bibr' target='#b3'>(Araújo et al. 2019a</ns0:ref>).</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:05:48694:2:0:NEW 17 Sep 2020)</ns0:p><ns0:p>Manuscript to be reviewed</ns0:p><ns0:p>Given recent critiques, it has become clear that when using SDMs in relation to SARS-CoV-2, it is worth documenting how close a study may come to the gold standards set for this type of modeling <ns0:ref type='bibr'>(Araújo et al. 2019b</ns0:ref>). The mode rank for our SDMs is bronze ( <ns0:ref type='table' target='#tab_0'>1</ns0:ref>). Niche overlap and similarity tests were conducted with the 'ecospat' library (Di <ns0:ref type='bibr' target='#b24'>Cola et al. 2017;</ns0:ref><ns0:ref type='bibr' target='#b14'>Broennimann et al. 2012;</ns0:ref><ns0:ref type='bibr' target='#b78'>Warren, Glor, and Turelli 2008)</ns0:ref> to compare the climatic niche of SARS-CoV-2 and humans. The niche overlap test considers whether there is a greater than expected overlap between the climate space occupied by two population models (human total vs. SARS-CoV-2 positive) than would be expected given a null distribution sampled from both populations. The niche similarity test considers the degree of similarity between the density of occurrence between two population models relative to the same null distribution <ns0:ref type='bibr' target='#b78'>(Warren, Glor, and Turelli 2008)</ns0:ref>. A significant overlap between SARS-CoV-2 and total human population models indicates that the virus' distribution is not necessarily constrained by climate (i.e., that the full range of human occupied climate is accessible to the virus).</ns0:p></ns0:div>
<ns0:div><ns0:head>Results</ns0:head></ns0:div>
<ns0:div><ns0:head>Climate Distribution Visualizations</ns0:head><ns0:p>The human population climate curves are visually similar compared to the populationscaled SARS-CoV-2 climate curves for all seven abiotic variables for March 30 (Figure <ns0:ref type='figure' target='#fig_1'>1</ns0:ref>).</ns0:p><ns0:p>There do appear to be modest differences in the population-scaled viral curves towards cooler temperatures and lower water vapor pressure compared to the human population curves. The raw density of SARS-CoV-2 cases (Figure <ns0:ref type='figure' target='#fig_3'>S1</ns0:ref>) is visually different from the population-scaled data (Figure <ns0:ref type='figure' target='#fig_1'>1</ns0:ref>). Population-scaled SARS-CoV-2 data appear to become better fitted to the human population data over time (i.e., these curves better match on March 30 than they did on March 2 or 16, 2020; Figure <ns0:ref type='figure' target='#fig_1'>1</ns0:ref>), whereas the raw SARS-CoV-2 case data become more and more fixed on a narrow range of values for all climate variables (Figure <ns0:ref type='figure' target='#fig_3'>S1</ns0:ref>). The latter (fixed and narrow) range presumably has to do with a few virus hotspots. There is a highly significant positive relationship (Spearman's r = 0.75, p < 2e -12 ); between the number of humans in a county and the number of viral cases; however, the population-scaled viral cases do not have a strong relationship (Spearman's r = 0.05, p = 0.03) with human populations (Figure <ns0:ref type='figure' target='#fig_4'>2</ns0:ref>). Although the latter p-value is significant, this is often not all that meaningful for large datasets, whereas the weak r value better illustrates the relationship for the population-scaled data.</ns0:p></ns0:div>
<ns0:div><ns0:head>Species distribution models</ns0:head><ns0:p>Model testing for SARS-CoV-2 population-scaled case model identified a Maxent model with linear and quadratic feature classes and a regularization multiplier of 0.5 as the best model, with a reasonably high model fit and transferability under block resampling (Avg. Test AUC = 0.82; Test AUC variance = 0.0002). Population-scaled case and human population models built using these optimal model parameters appear to be highly similar, with areas of high suitability predicted for much of the West Coast and most of the eastern half of the U.S. (Figure <ns0:ref type='figure'>3</ns0:ref>). There are some differences; for instance, the population-scaled model output for SARS-CoV-2 (Figure <ns0:ref type='figure'>3A</ns0:ref>) suggests Florida is less suitable than it is in the map for human population (Figure <ns0:ref type='figure'>3B</ns0:ref>). climates that are correlated with the places where most people live (i.e., it is not simply a map of current human populations). We did this as our ultimate goal was comparing a viral model to a human model. The niche overlap and similarity tests both find that the models of SARS-CoV-2 (population-scaled viral case data) and human population have significantly higher overlap (p < 0.01) than would be expected by chance (Figure <ns0:ref type='figure' target='#fig_9'>4</ns0:ref>), confirming the visual similarities between these maps. The raw data showed less overlap: the similarity test was significant, but an order of magnitude less so than for the population-scaled data (p = 0.02), while the overlap test was insignificant (p = 0.20), as seen in Figure <ns0:ref type='figure'>S3</ns0:ref>.</ns0:p></ns0:div>
<ns0:div><ns0:head>Discussion</ns0:head></ns0:div>
<ns0:div><ns0:head>Main findings</ns0:head><ns0:p>PeerJ reviewing PDF | (2020:05:48694:2:0:NEW 17 Sep 2020)</ns0:p><ns0:p>Manuscript to be reviewed</ns0:p><ns0:p>Our results suggest that population-scaled SARS-CoV-2 coronavirus cases from March 2020 were highly linked with the human populations in the U.S. and that any influence of climate is hard to disentangle for SARS-CoV-2 cases for the U.S. during this time period (Figures <ns0:ref type='figure' target='#fig_9'>1-4</ns0:ref>).</ns0:p><ns0:p>Furthermore, this link was strengthening over the duration of March (Figure <ns0:ref type='figure' target='#fig_1'>1</ns0:ref>). This indicates that caution should be used when dealing with climate modeling of SARS-CoV-2, at least on a national scale. Our results, while broadly similar to what was found in the U.S. for global-scale models (Araújo and Naimi 2020; <ns0:ref type='bibr' target='#b10'>Bariotakis et al. 2020)</ns0:ref>, indicate that the pattern we observed is simply reflecting human population density. Based on this, at least human population density should be compared or incorporated in any modeling exercises for SARS-CoV-2. Accordingly, the current global results from other studies (Araújo and Naimi 2020; <ns0:ref type='bibr' target='#b10'>Bariotakis et al. 2020)</ns0:ref> should not be taken at face value without a critical comparison to human distributions.</ns0:p><ns0:p>Furthermore, this is in line with current knowledge that public policy is the main driver for coronavirus spread in the U.S. <ns0:ref type='bibr' target='#b45'>(Leonhardt 2020)</ns0:ref>. It is also in line with modeling efforts that have combined epidemiological variables with climate <ns0:ref type='bibr'>(Baker et al. 2020)</ns0:ref>.</ns0:p><ns0:p>It is also noteworthy that our models using the raw viral case data (Figure <ns0:ref type='figure' target='#fig_4'>S2</ns0:ref>), as opposed to population-scaled data (Figure <ns0:ref type='figure'>3</ns0:ref>), performed poorly: the raw SDMs did not highlight most of the known March range of SARS-CoV-2 in the U.S. Other distribution modeling studies have only used raw values as inputs (Araújo and Naimi 2020; <ns0:ref type='bibr' target='#b10'>Bariotakis et al. 2020</ns0:ref>), which our model shows to be at risk of overfitting and low model transferability to novel regions. Similarly, our analysis of raw SARS-CoV-2 case data shows that over time the probability density curves focus on a narrower range of climate as more data comes from the areas most affected by the outbreak (Figure <ns0:ref type='figure' target='#fig_3'>S1</ns0:ref>). In contrast, adjusting for population size with the population scaling results in SARS-CoV-2 climate curves more closely approaching that of the human population curves over time (Figure <ns0:ref type='figure' target='#fig_1'>1</ns0:ref>). Furthermore, county-level human population numbers only appear to be strongly correlated with raw cases, rather than population-scaled cases (Figure <ns0:ref type='figure' target='#fig_4'>2</ns0:ref>).</ns0:p><ns0:p>Together these observations suggest that the SARS-CoV-2 was highly successful in a few areas in March (e.g., New York City), but when scaled for the available human population in each county we can infer a much broader geographic and climatic range was available. Again, our data provide a few indications that climate may have had some relevance for SARS-CoV-2 distributions in March. Viral probability densities were slightly higher in cooler Manuscript to be reviewed human SDMs, had lower suitability for the virus in Florida and the southwestern border (Figure <ns0:ref type='figure'>3</ns0:ref>). Whereas in reality, Florida had >5000 cases in March.</ns0:p></ns0:div>
<ns0:div><ns0:head>Avoiding a presumably faulty prediction</ns0:head><ns0:p>Given that our SDMs for humans and the virus had significant niche overlap and similarity (Figure <ns0:ref type='figure' target='#fig_9'>4</ns0:ref>), we do not believe that a future projection of our SARS-CoV-2 SDMs in the U.S. using only climate data would be trustworthy at this point in time, and therefore we do not present one. A model based on March data would likely have suggested the SARS-CoV-2 prevalence would have shifted northward during the summer, which would be congruent with another early global modeling study <ns0:ref type='bibr'>(Araújo and Naimi 2020)</ns0:ref>. However, given that humans do not move northward in large numbers in the U.S. during the summer, our predictions would be based on faulty presumptions about host resources (i.e., that there are sufficient humans in the north to harbor the bulk of cases). Most importantly, we now know that the virus has spread across much of the southern U.S. during the summer. One could imagine this counterintuitive spread was, in addition to inadequate public health policy, due in part to people in the southern states staying indoors more in the summer, replicating the behavior found in the northern states during the winter. Still, it is possible that a future model projection would turn out to be correct using updated datasets, and there could be at least some residual predictive power that is not fully encompassed by human population patterns.</ns0:p></ns0:div>
<ns0:div><ns0:head>Modeling counterarguments</ns0:head><ns0:p>It has now been argued that SDMs of SARS-CoV-2 can help to predict where the virus may generally be found now and in the future <ns0:ref type='bibr'>(Araújo and Naimi 2020;</ns0:ref><ns0:ref type='bibr' target='#b10'>Bariotakis et al. 2020)</ns0:ref>,</ns0:p><ns0:p>as well as why it might be a fool's errand to conduct these analyses in the first place <ns0:ref type='bibr' target='#b23'>(Chipperfield et al. 2020)</ns0:ref>. Here are some of the main arguments that have been put forward against SARS-CoV-2 SDMs, followed by our opinion on the subject. 1) Issue: the virus is spreading and the population was not in equilibrium with climate or any other putative niche dimension in March, let alone today. Our thoughts: it was not in equilibrium and that is imperfect; however, SDMs have been useful for a number of non-equilibrium systems, like species invasions during their spreading phase <ns0:ref type='bibr' target='#b75'>(Václavík and Meentemeyer 2009)</ns0:ref>, and waiting for equilibrium will mean predictions are no longer as useful to conduct (i.e., would only be helpful for a next outbreak of this virus). 2) Issue: the virus may have been spreading heavily or underreported in the Global South as of March. Our thoughts: surely this was partially true, but this may not completely explain observable climatic correlations. Even if we are working with incomplete data, it is hard to know if adding those data will significantly change conclusions based on preliminary models (i.e., will the SARS-CoV-2 model still be indistinguishable from a human population model). Still, to mitigate this problem we focused on the more consistent USlevel data. 3) Issue: papers applying SDMs must strive for best practices to avoid common errors. Our thoughts: yes, we agree that best practices are indeed worth pursuing, even if not everyone agrees on what best practices may be. We have done our best to achieve the somewhat aspirational standards that have been put forth <ns0:ref type='bibr' target='#b3'>(Araújo et al. 2019a)</ns0:ref>. We explicitly document this and summarize shortcomings that are inherent with a newly spreading system (Table <ns0:ref type='table' target='#tab_0'>1</ns0:ref>). 4) Issue: caution should be taken in claims and dissemination of research that could impact public health policy. Our thoughts: yes, caution is advisable in terms of studies on dire subjects like SARS-CoV-2; however, cross pollination between disciplines has been important</ns0:p><ns0:p>for many breakthroughs and advances in science.</ns0:p><ns0:p>Further caution towards distribution modeling methodology for SARS-CoV-2 in recent review literature highlights the likely limited effect climate has on a pathogen spread via direct transmission, and thus concludes that this tool is inappropriate in this situation <ns0:ref type='bibr' target='#b16'>(Carlson et al. 2020a</ns0:ref><ns0:ref type='bibr' target='#b16'>(Carlson et al. , [b] 2020))</ns0:ref>. Important confounding variables such as human interactions, public policy, and microclimate may mask any effects climate plays on this virus <ns0:ref type='bibr' target='#b18'>(Carlson et al. 2020b;</ns0:ref><ns0:ref type='bibr' target='#b23'>Chipperfield et al. 2020</ns0:ref>). Due to this, epidemiological studies are more suited to understand patterns of transmission <ns0:ref type='bibr' target='#b18'>(Carlson et al. 2020b;</ns0:ref><ns0:ref type='bibr' target='#b23'>Chipperfield et al. 2020;</ns0:ref><ns0:ref type='bibr' target='#b16'>Carlson et al. 2020a)</ns0:ref>.</ns0:p><ns0:p>Importantly, given these problems with SDMs of SARS-CoV-2, there is concern that these studies may negatively affect public policy <ns0:ref type='bibr' target='#b23'>(Chipperfield et al. 2020;</ns0:ref><ns0:ref type='bibr' target='#b16'>Carlson et al. 2020a, [b]</ns0:ref> 2020). Others continue to highlight the possible connections between climate and SARS-CoV-2 and advocate for continued research in this area <ns0:ref type='bibr' target='#b4'>(Araújo, Mestre, and Naimi 2020)</ns0:ref>.</ns0:p></ns0:div>
<ns0:div><ns0:head>Caveats and unknowns</ns0:head><ns0:p>There are individuals who regularly go undetected for SARS-CoV-2 (e.g., those who lack symptoms or have mild symptoms) -currently estimated at up to 25% <ns0:ref type='bibr' target='#b51'>(Mandavilli 2020)</ns0:ref>. This is of course an issue for all studies of SARS-CoV-2, but was likely especially true for March when testing was less widely available in the U.S. Obviously it is not ideal for modeling and forecasting; however, it is inherent in any study. In fact, we have far more data available for SARS-CoV-2 than we will ever have for the vast majority of viruses and biodiversity generally.</ns0:p><ns0:p>Our results are based on data for a single, large country. While conducting an analysis on global data would have been ideal, we avoided this because this type of global analysis is known to have certain issues and inconsistencies <ns0:ref type='bibr' target='#b23'>(Chipperfield et al. 2020)</ns0:ref>. While it may</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:05:48694:2:0:NEW 17 Sep 2020)</ns0:p><ns0:p>Manuscript to be reviewed somewhat hobble our ability to forecast viral distributions at other time points, as there are surely non-analogous weather systems to come (i.e., no place in the U.S. in March was as hot as Death Valley, California is during full summer heat), we feel it was a worthy trade-off. Policy, social factors, and a variety of other variables were not included in this study, despite their obvious, known, or potential importance <ns0:ref type='bibr' target='#b50'>(Maier and Brockmann 2020;</ns0:ref><ns0:ref type='bibr' target='#b46'>Leung et al. 2020;</ns0:ref><ns0:ref type='bibr' target='#b54'>A. Miller et al. 2020)</ns0:ref>; other variables such as wearing masks are also surely important for global patterns <ns0:ref type='bibr' target='#b50'>(Maier and Brockmann 2020;</ns0:ref><ns0:ref type='bibr' target='#b46'>Leung et al. 2020)</ns0:ref>. Unfortunately, data on these variables are often lacking in public databases. Our goals were to focus on a rather macro scale, which may be harder to do with policy data that can vary substantially from county to county, state to state, and, for a global study, country to country. While policies in the U.S. probably had not been implemented for long enough to tell their impact in March data, these changes (e.g., selfisolation and mask wearing) clearly and strongly influenced the trajectory of the virus after March.</ns0:p></ns0:div>
<ns0:div><ns0:head>Data accessibility and mobilization</ns0:head><ns0:p>We believe the New York Times (The New York Times 2020) providing robust countylevel data in an accessible repository with an open license for the U.S. sets an excellent standard that should be repeated by governments, academics, and other media organizations for other parts of the world so that this type of study may be better repeated in any country or globally. In a similar vein, we have made our analytical and modeling pipeline (along with figures) available (https://github.com/rsh249/cv19_enm/releases/tag/v0.0.5). We believe it is imperative for all pipelines and scripts to be made available for any SARS-CoV-2 research to ensure that models can be improved upon and any errors can be more quickly uncovered and resolved.</ns0:p></ns0:div>
<ns0:div><ns0:head>Future directions</ns0:head><ns0:p>There are many avenues to pursue regarding SARS-CoV-2 modeling and predictions.</ns0:p><ns0:p>We are excited to see researchers from a variety of fields extending their toolkits towards understanding this virus. We hope that ecological studies like this and others can play a role without overcomplicating the research efforts put forth by epidemiologists. Still, studies should familiarize themselves with current critiques of SDMs for SARS-CoV-2 modeling and be cautious of their inputs and conclusions.</ns0:p><ns0:p>With improving data, we feel that future studies should better be able to examine the system globally while considering human populations and public policy efforts at curbing the Manuscript to be reviewed virus. We also believe that it will at some point, in the U.S. and elsewhere, be worth examining death rates across different areas, as it would be helpful to know if climate or other abiotic variables might impact this statistic. Coupling regularly updated data with automated online resources would also be particularly helpful in learning how this virus may spread.</ns0:p></ns0:div>
<ns0:div><ns0:head>Conclusions</ns0:head><ns0:p>SDMs from SARS-CoV-2 population-scaled cases did not appear to be distinguishable from human population density for an early point in the pandemic for the U.S. Future studies looking at climate's impact on this virus should, wherever possible, take into account human population density in any analyses. Manuscript to be reviewed Manuscript to be reviewed </ns0:p></ns0:div>
<ns0:div><ns0:head>Figures and Table</ns0:head><ns0:note type='other'>Figure</ns0:note><ns0:note type='other'>Figure 4</ns0:note></ns0:div><ns0:figure xml:id='fig_0'><ns0:head /><ns0:label /><ns0:figDesc>Model testing for the SARS-CoV-2 raw case dataset identified a Maxent model with linear and quadratic feature classes and a regularization multiplier of 1.5 as the optimal model with a high model fit (Avg. Test AUC = 0.88), but with a slightly lower model transferability evidenced by higher AUC variance (Test AUC variance = 0.007) than the population-scaled model. The SDMs of the raw virus cases (Figure S2) fail to reconstruct the known distribution of viral cases in much of the U.S. with strong bias towards the Pacific Northwest and Northeastern United States. The SDMs for humans (Figure 3, S2) do not exactly match current human distribution for the U.S. However, this was expected, because the goal of the model is to match</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_1'><ns0:head>locales (Figure 1 ,</ns0:head><ns0:label>1</ns0:label><ns0:figDesc>S1). The population-scaled data were particularly compelling, even if the climate difference is smaller, as they are less driven by the current SARS-CoV-2 hotspots. Furthermore, our population-scaled viral SDMs, while statistically indistinguishable from our PeerJ reviewing PDF | (2020:05:48694:2:0:NEW 17 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_2'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:05:48694:2:0:NEW 17 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_3'><ns0:head>1 .</ns0:head><ns0:label>1</ns0:label><ns0:figDesc>Figure 1. Probability densities of SARS-CoV-2 coronavirus cases (using population-scaled data; curves in red) compared to the probability densities of human populations (curves in blue) in each U.S. county for each of seven climate variables. Probability density curves are standardized to an area of one.</ns0:figDesc><ns0:graphic coords='18,125.64,109.95,360.71,480.96' type='bitmap' /></ns0:figure>
<ns0:figure xml:id='fig_4'><ns0:head>Figure 2 .</ns0:head><ns0:label>2</ns0:label><ns0:figDesc>Figure 2. The relationship in the U.S. between human population size and SARS-CoV-2 coronavirus cases, using (A) total viral cases and (B) population-scaled viral cases. New York City, an outlying point, has been excluded for clearer visualization.</ns0:figDesc><ns0:graphic coords='19,126.00,71.93,360.00,522.15' type='bitmap' /></ns0:figure>
<ns0:figure xml:id='fig_5'><ns0:head /><ns0:label /><ns0:figDesc>Figure S3. Equivalent of Figure 4, but using raw virus data. (A) Niche Overlap and (B) similarity tests for Maxent species distribution models built with SARS-CoV-2 coronavirus case data compared to one built with human population density as occurrence data; actual model overlap indicated by a red marker in both plots. Significant p-values correspond to greater niche overlap or similarity than expected by random models.</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_6'><ns0:head>Figure 1 Figure 1 .Figure 1 .</ns0:head><ns0:label>111</ns0:label><ns0:figDesc>Figure 1</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_7'><ns0:head>Figure 2 Figure 2 .Figure 2 .</ns0:head><ns0:label>222</ns0:label><ns0:figDesc>Figure 2</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_8'><ns0:head>Figure 3 Figure 3</ns0:head><ns0:label>33</ns0:label><ns0:figDesc>Figure 3</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_9'><ns0:head>Figure 4 .</ns0:head><ns0:label>4</ns0:label><ns0:figDesc>Figure 4. (A) Niche overlap and (B) similarity tests</ns0:figDesc></ns0:figure>
<ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='20,126.00,71.67,360.00,503.94' type='bitmap' /></ns0:figure>
<ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='21,126.00,71.80,360.00,503.94' type='bitmap' /></ns0:figure>
<ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='24,125.64,109.95,360.71,480.96' type='bitmap' /></ns0:figure>
<ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='25,125.99,72.00,360.02,504.00' type='bitmap' /></ns0:figure>
<ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='26,72.00,71.28,360.00,503.95' type='bitmap' /></ns0:figure>
<ns0:figure type='table' xml:id='tab_0'><ns0:head>Table 1 )</ns0:head><ns0:label>1</ns0:label><ns0:figDesc>, largely as data from an emerging disease are inherently imperfect. Our SDMs rely on incomplete (limited or underreported data) observational data that reports only positive tests for the continental United States. The SDMs also only model over climate averages for the month ofMarch (averaged from 1970March (averaged from -2000)), which may result in a truncated climate envelope. However, despite imperfect data the SDMs reported here fulfill the silver or gold standard methodology for model complexity, treatment of bias, and model evaluation (Table</ns0:figDesc><ns0:table /></ns0:figure>
<ns0:figure type='table' xml:id='tab_1'><ns0:head>Table 1 .</ns0:head><ns0:label>1</ns0:label><ns0:figDesc>Evaluation of our species distribution modeling practices against the best practices that have been proposed for this field<ns0:ref type='bibr' target='#b3'>(Araújo et al. 2019a</ns0:ref>).</ns0:figDesc><ns0:table><ns0:row><ns0:cell>Guideline</ns0:cell><ns0:cell>Standard</ns0:cell><ns0:cell>Justification</ns0:cell></ns0:row><ns0:row><ns0:cell>Response</ns0:cell><ns0:cell>A) Sampling: bronze</ns0:cell><ns0:cell>Best data available; municipalities, local</ns0:cell></ns0:row><ns0:row><ns0:cell>variables</ns0:cell><ns0:cell /><ns0:cell>governments, and states choose who to test.</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell>Positive tests only reported.</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>B) Identification: gold</ns0:cell><ns0:cell>Assuming best practices in testing and reporting.</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>C) Spatial accuracy:</ns0:cell><ns0:cell>County assignments provide a rough georeference</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>bronze</ns0:cell><ns0:cell>for each record, but do not precisely describe where</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell>transmission of the virus occurred. Spatial accuracy</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell>unknown. Occurrences limited to identifiable county</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell>level localities.</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>D) Environmental</ns0:cell><ns0:cell>Limiting the study area to the continental U.S. is</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>extent: deficient</ns0:cell><ns0:cell>unlikely to adequately test environmental</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell>boundaries.</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>E) Geographic extent:</ns0:cell><ns0:cell>Study area to include current range in the U.S.</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>bronze</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>Predictor</ns0:cell><ns0:cell>A) Selection of</ns0:cell><ns0:cell>Unclear and not well documented correlations</ns0:cell></ns0:row><ns0:row><ns0:cell>variables</ns0:cell><ns0:cell>candidates:</ns0:cell><ns0:cell>between SARS-CoV-2 transmission and climate. At</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>bronze/deficient</ns0:cell><ns0:cell>best, distal variables with weak, indirect control on</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell>the distribution.</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>B) Spatial and temporal</ns0:cell><ns0:cell>Variables sampled from a 2.5 arcminute grid for all</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>resolution: deficient</ns0:cell><ns0:cell>cells within 5km of each occurrence point. Mean</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell>value used for modeling. Monthly climate averages</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell>as predictors for end of March occurrence data.</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>C) Uncertainty: bronze</ns0:cell><ns0:cell>Temporal and spatial uncertainty in occurrence data</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell>has unquantified potential effects on the model</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell>output.</ns0:cell></ns0:row><ns0:row><ns0:cell>Model</ns0:cell><ns0:cell>A) Model Complexity:</ns0:cell><ns0:cell>ENMeval for model testing and selection (maximize</ns0:cell></ns0:row><ns0:row><ns0:cell>building</ns0:cell><ns0:cell>silver</ns0:cell><ns0:cell>testing AUC and minimize AICc in the case of ties)</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell>using internal cross validation through the block</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell>resampling method.</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>B) Treatment of</ns0:cell><ns0:cell>Internal cross validation to evaluate bias effects in</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>response bias: silver</ns0:cell><ns0:cell>different models.</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>C) Treatment of</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell /><ns0:cell>collinearity: bronze</ns0:cell><ns0:cell /></ns0:row></ns0:table><ns0:note>'Approximate methods are applied' -Predictor variables hand selected from monthly climate data available to avoid collinearity (i.e., used only Tmax and not Tavg or Tmin).</ns0:note></ns0:figure>
<ns0:note place='foot'>PeerJ reviewing PDF | (2020:05:48694:2:0:NEW 17 Sep 2020)</ns0:note>
<ns0:note place='foot'>PeerJ reviewing PDF | (2020:05:48694:2:0:NEW 17 Sep 2020)Manuscript to be reviewed</ns0:note>
</ns0:body>
" | "We thank the editor and reviewers for this second round of edits. They have most certainly led to an improved manuscript, which we are grateful for. We respond to each suggestion below. For those that we simply did what was requested, we have responded with ‘Fixed.’
-Rob Harbert, Seth Cunningham, and Michael Tessler
Reviewer #3
Basic reporting
no comment
Experimental design
no comment
Validity of the findings
no comment
Comments for the author
I am happy with the authors' response to my questions and the changes made. I suggest the authors double-check spelling/grammar before submitting the final version e.g. Line 102 'once the endemic phase is reached' and Line 315 inline should be in line.
Thank you for reviewing our revisions. We have checked spelling/grammar and fixed the ‘in line’ and changed line 102 to ‘once the endemic phase is reached’.
Reviewer #2
Basic reporting
The paper is much easier to follow now, but there are still some sentences phrased awkwardly, grammatical mistakes, and unclear statements. I made suggestions in my line-by-line edits. Much better coverage of literature now.
Experimental design
This revision made the methods more clear, but lingering questions remain (see line-by-line edits). The authors need to use consistent language when referring to results, explain the methods in a bit more detail, better outline the methods in the introduction to prepare the reader for them, and in general take care to be consistent with the order and manner in which different results are reported (else it gets confusing).
Validity of the findings
The findings here are important for the field, as the authors demonstrated that climate-based SDMs are not appropriate to model the range of a novel pathogen spread by humans, and that human population density is a much better predictor. The authors just need to be more emphatic that this is the best way forward. The conclusions are clear and there is an appropriate amount of uncertainty outlined.
Comments for the author
Good job on this latest revision -- I think the paper has improved considerably.
Thank you for your kind words and for the thorough line-by-line comments below. We have addressed them individually, and believe they have greatly improved our manuscript.
Line-by-line comments:
L29: 'not in population equilibrium'
Fixed.
L32: 'population-scaled' -- all instances need hyphens
Fixed.
L52: What is a 'basic distribution'?
Edit: changed to ‘geographic distribution’.
L70: Spelling
Fixed.
L97: Format
Fixed.
L102: Please finish the sentence.
Changed the sentence to the following: “Recent epidemiological models have also highlighted that weather may not play a large role when most people lack immunity during the pandemic phase, but indicate that weather has the potential to play a larger role once we enter the endemic phase (Baker et al. 2020).”
L103: If the abbreviation SDM refers to 'modeling', then SDMs becomes 'species distribution modelings'.
We have fixed this to be ‘species distribution models (SDMs)’, as is used in a number of studies, and uniformly now use ‘SDMs’ throughout.
L106: Please cite something more recent and general, like 2011 Peterson et al. book.
Fixed.
L109: Grammar
Fixed.
L108-113: Not clear what the difference here is. A model of potential distribution is making correlations with climate data. Please be more clear about the distinction you are drawing.
Deleted: “and correlations with environmental, climatic, or demographic covariates” to emphasize that the only distinction is between models based on host/vector distribution and those built with pathogen occurrence data as we did here.
L117: I would say 'environmental variables correlated with its occurrence exist' because you can project to unsuitable conditions too.
Fixed.
L129: No capitalization mid-sentence.
Fixed here and elsewhere.
L130: Why do you cite the standards paper here? Seems not relevant.
Good catch, we have changed this to be the correct citation from the same author (Araujo & Naimi, 2020)
L142: 'Still, the climate-based SDMs for SARS-CoV-2 presented recently...'
Fixed.
L143: Can we really use 'habitat preferences' for modern humans? Can't you say 'climatic preferences'? This is more accurate anyway because climate does not define habitat.
Fixed.
L146: '..., and may help to better ensure...'
Fixed.
L158-159: You do not mention here the niche overlap analyses. Please prepare the reader adequately for all the methods you undertake and explain why you're doing them.
I agree, but still want to keep the overview of the methods brief in the introduction. The sentence referenced has been changed to:
“Here we develop a suite of data visualizations, species distribution models, and comparative niche overlap analyses using both climate and human population data to determine whether the effect of climate can be appropriately disentangled from other drivers of SARS-CoV-2 transmission, given early records for the virus in the US.”
L204: 'were produced' -- please make sure to stay in past tense for the Methods section
Fixed.
L209: If 'total cases per county' = 'raw case data', say it like this to avoid confusion 'Raw case data (i.e., total cases per county)'
Fixed.
L215: 'we calculated non-parametric Spearman's rank correlation coefficients in R' (the code is not necessary)
If the editor prefers this change to the sentence, we are happy to make it. However, we prefer to keep the code as is (being explicit does not seem like an issue to us).
L221-223: Rephrasing like this may be less confusing, and includes the expansion factor: 'while occurrence data for population-scaled SDMs were generated the same way as for the raw SDM except that county climate records were multiplied by the total case count divided by the county population, such that no county had fewer than one record, then multiplied by an expansion factor.' Please why you use the expansion factor.
The expansion factor is there so that, once scaled by population, that all localities with at least one occurrence had a value > 1. The linear scaling of the expansion was to avoid losing data from localities with very few cases relative to the population.
To make this more clear we have changed the phrase to:
“Occurrence data for population-scaled SDMs were generated the same way as for the raw SDMs except that county climate records were multiplied by the total case count divided by the county population. Then population-scaled values were multiplied by an expansion factor of 100,000 so that all counties with at least one case were represented.”
L224: Use 'total virus cases', not just 'cases'.
Fixed.
L227: Please cite Maxent here and introduce it as an SDM algorithm for presence-only data with complexity settings that can be tuned.
Updated to: “Maxent, an algorithm for presence-only distribution models (Elith et al. 2011), model parameters relating to model complexity were tested by building a suite of SDMs for SARS-CoV-2 distribution data with occurrences generated from raw reported virus case values and population-scaled case values.”
L230: After 'feature classes', put '(constraints on model fit)' and after 'regularization multipliers', put '(penalizes complexity)' for those who don't already know.
Updated to: “Maxent model testing and cross validation was performed using the ENMeval package (Muscarella et al. 2014) considering linear and quadratic feature classes (constraints on model fit) and regularization multipliers (penalties on complexity) of 0.5, 1, 1.5, 2, 2.5, and 3.”
L231: Not 'best', but 'optimal', which emphasizes that you did not pick the absolute best but the best of the set explored.
We have changed the phrasing here to “optimal model” and updated other mentions of best models to use “optimal” throughout the manuscript.
L232: Rephrase to: 'Optimal model parameters were chosen by maximizing the average test AUC calculated with cross validation using the spatial ‘block’ partitioning method, and minimizing AICc in the case of ties.'
We accept this change. That sentence is clear about what we did.
L234: 'including those tested within ENMeval' -- not needed?
Leaving this out caused confusion for one of the reviewers in the 1st round. We choose to leave it as-is since we build test and operational models with maxnet, but it is possible to test models with ENMeval using the old Java Maxent. The readers that know this may wonder which we used if it is not stated here.
L236: Does 'human density' here mean 'human population'? Later in the Results you refer to 'human population model' -- is that this human density SDM? Please use consistent terms.
Updated: “Operational models using the optimal model parameters were then built for SARS-CoV-2 using all population-scaled data, raw virus data, and for the human population using the county population data (Phillips 2017).”
L237: What does 'clearly linked' mean? Can't you easily quantify their correlation?
Updated: “Of the seven WorldClim variables, the three temperature variables are not independent”
Yes, it is easy to quantify the correlation, but it is also clear that mean, minimum, and maximum monthly temperatures are not independent features of the climate.
L244-245: It is crucial to mention here which particular methodological decisions brought you to bronze, else readers cannot appreciate what is so commendable about your study.
See edits in response to line 245 also.
L245: 'inherently imperfect' is very vague. Can you be more specific as to why this data is imperfect for SDMs?
Updated: “Our SDMs rely on incomplete (limited or underreported data) observational data that reports only positive tests for the continental United States. The SDMs also only model over climate averages for the month of March (averaged from 1970-2000), which may result in a truncated climate envelope. However, despite imperfect data the SDMs reported here fulfill the silver or gold standard methodology for model complexity, treatment of bias, and model evaluation (Table 1).”
L260: 'during the latest date in March' -> 'for March 30'
Fixed.
L280: What are the results for the human pop model? They should be here.
As stated in the methods the human models are built using the optimal parameters from training on the SARS-CoV-2 population scaled and raw data. We did it this way to avoid spurious differences that may arise from using different model parameters to build the SARS-CoV-2 and human models.
L287: Not sure this interpretation is correct. I would interpret mean test AUC as a measure of average transferability and the variance as variation in transferability, as the AUC itself is measuring how well the trained model predicts the new data. Further, a difference of 0.0068 in variance is very small -- perhaps 'slight difference'?
Updated to: “Model testing for the SARS-CoV-2 raw case dataset identified a Maxent model with linear and quadratic feature classes and a regularization multiplier of 1.5 as the optimal model with a high model fit (Avg. Test AUC = 0.88) but with a slightly lower model transferability evidenced by higher AUC variance (Test AUC variance = 0.007) than the population-scaled model.”
L290: Please report the results for the human pop model -- they seem to be missing. What were the optimal model settings? Also, I think they belong in the preceding paragraph where you discuss the pop scaled case model.
See comment above for L280 regarding how human models were parameterized using the optimal parameters for SARS-CoV-2 datasets.
L311: 'At least some human-focused data' is not really strong enough here when you just demonstrated that the observed infection patterns reflect human population density. Shouldn't this be the main kind of variable used?
We have changed this to ‘at least human population density’.
L315-316: 'inline' should be 'in line'
Fixed.
L336: What does 'statistically identical' mean? Aren't these different models with different complexity settings and different training data?
Updated: We have changed it to read “our population-scaled viral SDMs, while statistically indistinguishable from our human SDMs, had lower suitability for the virus in Florida” to better express that these are not statistically different, but that there are still slight differences in the output.
L338: Please quantify this, as the data exists.
Updated: “Whereas in reality, Florida had >5000 cases in March”
L341: The models are different, and are certainly 'statistically distinguishable' even though the niche similarity test was significant. Figure 3 shows the differences clearly. You might say they are 'very similar'.
Updated: “Given that our SDMs for humans and the virus had significant niche overlap and similarity (Figure 4), we do not believe that a future projection of our SARS-CoV-2 SDMs in the US using only climate data would be trustworthy at this point in time, and therefore do not present one.”
L371: Which papers? Do you mean specifically those doing SDM analyses, or those doing SDM analyses on disease vectors? Please be more specific.
Updated to: “Issue: papers applying SDMs must strive for best practices to avoid common errors.”
L392: Just say 'individuals'.
Fixed.
L398: Perhaps 'while conducting an analysis on global data would have been ideal, we avoided this because...'
Fixed.
L435: What 'heartbreaking statistic'? Might want to avoid using such emotional language.
Fixed.
" | Here is a paper. Please give your review comments after reading it. |
9,786 | "<ns0:abstract xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:p>A comprehensive meta-analysis of publicly available gene expression microarray data obtained from human-derived pancreatic ductal adenocarcinoma (PDAC) tissues and their histologically matched adjacent tissue samples was performed to provide diagnostic and prognostic biomarkers, and molecular targets for PDAC. An integrative meta-analysis of four submissions (GSE62452, GSE15471, GSE62165, and GSE56560) containing 105 eligible tumor-adjacent tissue pairs revealed 344 differentially over-expressed and 168 repressed genes in PDAC compared to the adjacent-to-tumor samples. The validation analysis using TCGA combined GTEx data confirmed 98,24% of the identified up-regulated and 73,88% of the down-regulated protein-coding genes in PDAC. Pathway enrichment analysis showed that 'ECM-receptor interaction', 'PI3K-Akt signaling pathway', and 'focal adhesion' are the most enriched KEGG pathways in PDAC. Protein-protein interaction analysis identified FN1, TIMP1, and MSLN as the most highly ranked hub genes among the DEGs. Transcription factor enrichment analysis revealed that TCF7, CTNNB1, SMAD3, and JUN are significantly activated in PDAC, while SMAD7 is inhibited. The prognostic significance of the identified and validated differentially expressed genes in PDAC was evaluated via survival analysis of TCGA Pan-Cancer pancreatic ductal adenocarcinoma data. The identified candidate prognostic biomarkers were then validated in four external validation datasets (GSE21501, GSE50827, GSE57495, and GSE71729) to further improve reliability. A total of 28 up-regulated genes were found to be significantly correlated with worse overall survival in patients with PDAC. Twenty-one of the identified prognostic genes</ns0:p></ns0:div>
</ns0:abstract>
<ns0:body xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:head>INTRODUCTION</ns0:head><ns0:p>Pancreatic cancer has the highest mortality rate of all solid tumors and ranks fourth on the list of cancer-related causes of death in the world, albeit it represents only 3% of newly diagnosed tumors <ns0:ref type='bibr' target='#b51'>(Siegel et al. 2018)</ns0:ref>. According to the American Cancer Society, 57,600 people will be diagnosed with pancreatic cancer in 2020, and 47,050 deaths will be attributed to pancreatic cancer in the United States, reflecting the fatal nature of the disease. It is expected to become the 2nd leading cause of cancer-related death by 2030, surpassing breast, colorectal, and prostate cancer. <ns0:ref type='bibr' target='#b41'>(Rahib et al. 2014)</ns0:ref>. The most common form of pancreatic cancer is pancreatic ductal adenocarcinoma (PDAC), which accounts for more than 90% of all cases <ns0:ref type='bibr' target='#b26'>(He et al. 2018)</ns0:ref>.</ns0:p><ns0:p>Despite improvements in survival rates observed in most cancer types, advancements in the treatment of pancreatic cancer have remained steady for more than 40 years, as evidenced by incidence and mortality rates <ns0:ref type='bibr' target='#b51'>(Siegel et al. 2018;</ns0:ref><ns0:ref type='bibr' target='#b69'>Wu et al. 2018)</ns0:ref>. The overall 5-year survival rate for patients with pancreatic cancer remains less than 8% and one-year survival of around 18% when all stages are combined <ns0:ref type='bibr' target='#b46'>(Saad et al. 2018;</ns0:ref><ns0:ref type='bibr' target='#b51'>Siegel et al. 2018)</ns0:ref>. The lack of distinctive symptoms in the early stages of the disease, specific risk factors, and an effective screening process eventually cause delayed diagnosis <ns0:ref type='bibr' target='#b35'>(Maitra & Hruban 2008;</ns0:ref><ns0:ref type='bibr' target='#b65'>Weledji et al. 2016)</ns0:ref>. Since traditional chemotherapy has limited benefits on survival <ns0:ref type='bibr' target='#b11'>(Conroy et al. 2011)</ns0:ref>, the only treatment option offering a chance for cure remains as curative surgery for which only 10%-20% of patients are considered eligible. However, the majority of patients (50%-60%) present with metastatic disease at the time of diagnosis <ns0:ref type='bibr' target='#b18'>(Gillen et al. 2010)</ns0:ref>, and thus cannot benefit from curative surgery, improving median overall survival to 11-23 months and five-year overall survival rates to 15%-20% (La <ns0:ref type='bibr' target='#b33'>Torre et al. 2014)</ns0:ref>. Unfortunately, 60% of the patients experience local and systemic relapse within the first 12 months after curative resection (La <ns0:ref type='bibr' target='#b32'>Torre et al. 2012)</ns0:ref>, and more than 80% of the patients die of the disease due to local recurrence or distant metastasis <ns0:ref type='bibr' target='#b18'>(Gillen et al. 2010)</ns0:ref>. Together with delayed diagnosis, underlying causes for its exceptionally dismal prognosis also include poor efficacy of treatment modalities such as adjuncts to surgery, undetected micro-metastases, and the development of resistance to chemotherapy <ns0:ref type='bibr' target='#b1'>(Amrutkar & Gladhaug 2017)</ns0:ref>. Thus, there is an urgent need for the development of novel and more effective targeted therapies capable of improving survival in patients with pancreatic cancer.</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:07:51367:1:1:NEW 9 Sep 2020)</ns0:p><ns0:p>Over the last few decades, various studies have used high-throughput transcriptome profiling to expand our understanding of the underlying molecular mechanisms of pancreatic cancer and to discover novel diagnostic biomarkers and therapeutic targets <ns0:ref type='bibr' target='#b6'>(Campagna et al. 2008;</ns0:ref><ns0:ref type='bibr' target='#b55'>Tan et al. 2003</ns0:ref>). An integrative meta-analysis of the transcriptome data allows researchers to combine results from several studies to increase sample size, thereby statistical power and consistency <ns0:ref type='bibr' target='#b42'>(Ramasamy et al. 2008)</ns0:ref>. Since the first meta-analysis was published in 2005 <ns0:ref type='bibr' target='#b22'>(Grutzmann et al. 2005)</ns0:ref>, several meta-analyses were conducted to unravel molecular and clinical subtypes of pancreatic cancer <ns0:ref type='bibr' target='#b77'>(Zhao et al. 2018)</ns0:ref>, to reveal the genes involved in the prognosis of the disease <ns0:ref type='bibr' target='#b19'>(Goonesekere et al. 2014;</ns0:ref><ns0:ref type='bibr' target='#b20'>Goonesekere et al. 2018;</ns0:ref><ns0:ref type='bibr' target='#b23'>Haider et al. 2014;</ns0:ref><ns0:ref type='bibr' target='#b78'>Zheng et al. 2018)</ns0:ref> or to identify novel diagnostic biomarkers <ns0:ref type='bibr' target='#b28'>(Irigoyen et al. 2018)</ns0:ref>.</ns0:p><ns0:p>My approach to revealing novel deregulated molecular mechanisms underlying PDAC, which differentiates this study from others, is eliminating the potential influences of clinical, demographic and environmental factors on transcriptome profiles by including only microarray data obtained from human-derived pancreatic ductal adenocarcinoma tissues and their histologically matched adjacent-to-tumor tissue samples in this study. As a result of a careful and detailed examination of the microarray data meeting the stringent inclusion criteria of this study, the present work provides not only new insights into the molecular mechanisms underlying PDAC but also suggests novel biomarkers that may serve as promising indicators of prognosis and diagnosis for PDAC.</ns0:p></ns0:div>
<ns0:div><ns0:head>MATERIALS AND METHODS</ns0:head></ns0:div>
<ns0:div><ns0:head>Selection of Microarray Datasets</ns0:head><ns0:p>NCBI Gene Expression Omnibus (http://www.ncbi.nlm.nih.gov/geo/) and ArrayExpress Archive of Functional Genomics Data were systematically searched for eligible datasets using the keywords 'pancreatic ductal adenocarcinoma'. The inclusion criteria were: i) gene expression microarray data, ii) human-derived pancreatic ductal adenocarcinoma tissues and matched adjacent non-tumor tissue samples. When tumor-adjacent tissue pairs were not specified clearly in the overall design of the study or the sample description, the authors of the relevant paper were consulted for the confirmation of sample pairs. Non-confirmed or unspecified sample pairs were excluded from the meta-analysis.</ns0:p></ns0:div>
<ns0:div><ns0:head>Generation of Gene Expression Matrix Files and Evaluation of Data Quality</ns0:head><ns0:p>PeerJ reviewing PDF | (2020:07:51367:1:1:NEW 9 Sep 2020)</ns0:p><ns0:p>Manuscript to be reviewed All data processing and integration procedures were performed using ExAtlas, which is an online software tool for meta-analysis and visualization of gene expression data <ns0:ref type='bibr' target='#b49'>(Sharov et al. 2015)</ns0:ref>. Briefly, the datasets that have been selected for meta-analysis were uploaded to ExAtlas.</ns0:p><ns0:p>Unpaired samples were removed from the sample files, and then the gene expression matrix file was generated from each dataset separately. All the extracted data had been normalized using the RMA algorithm. Individual sample quality was evaluated by checking the correlation of log10transformed expression level with other data for a set of pre-selected housekeeping genes and the level of the global standard deviation. Samples, where the correlation of expression of housekeeping genes in the range from 0.5 to 0.95, and the level of standard deviation from the global mean for each set of genes grouped by the average expression is less than 0.3 were considered of good quality. Samples of low quality were removed from the datasets.</ns0:p></ns0:div>
<ns0:div><ns0:head>Standard Meta-Analysis</ns0:head><ns0:p>In the Pairwise comparison section of ExAtlas <ns0:ref type='bibr' target='#b49'>(Sharov et al. 2015)</ns0:ref>, one of the tumor gene expression profiles was added as a sample for examination, and its adjacent non-tumor tissue pair was added for baseline control. Then, the meta-analysis section was used to add more gene expression profile pairs. The random-effects method (DerSimonian & Laird 1986), which takes into account the variance of heterogeneity between studies, was used to perform the metaanalysis. False discovery rate (FDR) is less than 0.05 and the change of gene expression is ≥ 2fold were considered significant. The analysis was performed for each gene symbol, and the effects were presented as combined fold changes and combined log-ratios (log10).</ns0:p></ns0:div>
<ns0:div><ns0:head>External Validation of the Identified DEGs</ns0:head><ns0:p>The external validation was done using GEPIA Database (http://gepia.cancerpku.cn/index.html) <ns0:ref type='bibr' target='#b56'>(Tang et al. 2017)</ns0:ref> by comparing transcriptomic data from the TCGA PAAD (pancreatic adenocarcinoma), and the TCGA normal and GTEx data. For genes, whose probes are not found in the validation dataset, Logsdon pancreas dataset (available in the Oncomine Database <ns0:ref type='bibr' target='#b44'>(Rhodes et al. 2004</ns0:ref>)) including data from pancreatic ductal adenocarcinoma and healthy pancreatic tissues was used. The external validation in a study including PDAC and matched adjacent tissues could not be performed because all studies that passed the inclusion criteria were included in the meta-analysis.</ns0:p></ns0:div>
<ns0:div><ns0:head>Transcription Factor Binding Site Enrichment Analysis</ns0:head><ns0:p>PeerJ reviewing PDF | (2020:07:51367:1:1:NEW 9 Sep 2020)</ns0:p><ns0:p>Manuscript to be reviewed TFactS Database (http://www.tfacts.org) <ns0:ref type='bibr' target='#b14'>(Essaghir & Demoulin 2012;</ns0:ref><ns0:ref type='bibr' target='#b15'>Essaghir et al. 2010</ns0:ref>) was used as a tool to predict which transcription factors (TF) are regulated, inhibited, or activated based on the list of DEGs. Briefly, the up-and down-regulated genes were uploaded to the TFact database, and the DEGs were then compared with the Sign-Sensitive catalog of validated target genes of TFs. Transcription factors whose target genes show a significant overlap with the DEGs were reported. The value of <0.05 for all four indexes (p-value, q-value, E-value, and FDR) in TFactS Database was considered statistically significant.</ns0:p></ns0:div>
<ns0:div><ns0:head>Finding Prognostic Genes for PDAC</ns0:head><ns0:p>Kaplan Meier Plotter (http://kmplot.com/analysis/), an open-access database that provides log-rank P-value and Hazard ratio with 95% confidence intervals for Kaplan-Meier analysis of the correlation between mRNA expression level and patient overall survival <ns0:ref type='bibr' target='#b36'>(Nagy et al. 2018)</ns0:ref>, was employed to predict the prognostic importance of the DEGs detected in this study. The TCGA Pan-Cancer Pancreatic ductal adenocarcinoma cohort was used for the analysis. Autoselect best cutoff value was used to split the patients in survival analysis. P<0.05 and FDR≤0.05 were used as the cutoff for significance. Then, four independent external validation datasets (GSE21501, GSE50827, GSE57495, and GSE71729 including clinical and transcriptomic data from pancreatic ductal adenocarcinoma patients were analyzed using the PROGgenev2 Prognostic Database <ns0:ref type='bibr' target='#b21'>(Goswami & Nakshatri 2014)</ns0:ref> to evaluate the prognostic relevance of the identified candidate prognostic genes by Km-Plotter. The patient cohorts were divided into two equal groups based on median expression for each gene. P<0.05 was accepted as statistically significant.</ns0:p><ns0:p>The correlation between the mRNA expression of the identified prognostic genes and the pathological stages of the disease was evaluated in TCGA PAAD data using GEPIA Database. P<0.05 was accepted as statistically significant.</ns0:p></ns0:div>
<ns0:div><ns0:head>Protein-Protein Interaction Analysis</ns0:head><ns0:p>Search Tool for the Retrieval of Interacting Genes (STRING, version 11.0, https://stringdb.org/) is an online database designated to evaluate physical and functional associations of proteins <ns0:ref type='bibr' target='#b53'>(Szklarczyk et al. 2014)</ns0:ref>. STRING app in Cytoscape Software (version 3.6.1) was used to detect the interactions among the DEGs with a confidence score cut-off ˃0.9. Then, the protein-protein interaction (PPI) network was constructed using Cytoscape Software <ns0:ref type='bibr' target='#b48'>(Shannon et al. 2003)</ns0:ref>. The Molecular Complex Detection plug-in (MCODE) was used to identify clusters in PeerJ reviewing PDF | (2020:07:51367:1:1:NEW 9 Sep 2020)</ns0:p><ns0:p>Manuscript to be reviewed the PPI network with the parameters false degree cuto 2 and K-Core 2 <ns0:ref type='bibr' target='#b4'>(Bader & Hogue 2003)</ns0:ref>.</ns0:p><ns0:p>Finally, the Gene Ontology and KEGG pathway enrichment analyzes of the DEGs in the clusters with MCODE score >5 were performed with the STRING Enrichment app in Cytoscape Software by retrieving functional enrichment for selected clusters only.</ns0:p></ns0:div>
<ns0:div><ns0:head>Gene Ontology and Pathway Enrichment Analysis of the DEGs</ns0:head><ns0:p>Gene Ontology Enrichment Analysis was performed using the Functional Enrichment Analysis Tool (FunRich) <ns0:ref type='bibr' target='#b37'>(Pathan et al. 2015)</ns0:ref> to identify the biological processes, molecular functions, and cellular components that are shared by the up-or down-regulated genes separately.</ns0:p><ns0:p>The Gene Ontology Database <ns0:ref type='bibr'>(2017;</ns0:ref><ns0:ref type='bibr' target='#b3'>Ashburner et al. 2000</ns0:ref>; The Gene Ontology Consortium 2018) was selected as a background for the analysis, and a p-value of <0.05 was used as the cutoff for significance.</ns0:p><ns0:p>The list of differentially up-and down-regulated genes in PDAC was explored for functionally enriched pathways using a WEB-based GEne SeT AnaLysis Toolkit (WebGestalt 2017) (DerSimonian & Laird 1986) with a cut-off criterion of FDR<0.05. Enrichment categories were selected as Kyoto Encyclopedia for Genes and Genomes (KEGG) <ns0:ref type='bibr' target='#b30'>(Kanehisa & Goto 2000)</ns0:ref>, Reactome Pathway Knowledgebase <ns0:ref type='bibr' target='#b16'>(Fabregat et al. 2018)</ns0:ref>, PANTHER <ns0:ref type='bibr' target='#b58'>(Thomas et al. 2003)</ns0:ref>, and Wikipathways <ns0:ref type='bibr' target='#b31'>(Kelder et al. 2012</ns0:ref>). The category size was calculated based on the number of overlapping genes between the annotated genes in the category and the ranked gene list for the GSEA method. Categories with sizes smaller than 15 and greater than 1000 were removed during the analysis.</ns0:p></ns0:div>
<ns0:div><ns0:head>RESULTS</ns0:head></ns0:div>
<ns0:div><ns0:head>Microarray Datasets</ns0:head><ns0:p>The workflow and summary of the results of the presented meta-analysis is shown in Fig. <ns0:ref type='figure' target='#fig_0'>1</ns0:ref>. A systematic search of the studies was carried out up to January 2019. The search with the keyword Pancreatic Ductal Adenocarcinoma' resulted in 18 microarray gene expression datasets which had been submitted only to the ArrayExpress Archive of Functional Genomics Data. Among these, none of the datasets was found to contain adjacent normal tissue as a control group and therefore these datasets were not included in the meta-analysis. Gene Expression Omnibus Datasets were searched with the term 'pancreatic ductal adenocarcinoma' and the search results were filtered by selecting organism as homo sapiens, study type as expression profiling by array and attribute name as tissue. As a result, 54 studies were found in the GEO Database. These <ns0:ref type='table' target='#tab_1'>PeerJ reviewing PDF | (2020:07:51367:1:1:NEW 9 Sep 2020)</ns0:ref> Manuscript to be reviewed studies were carefully evaluated and eight studies (GSE62452, GSE15471, GSE62165, GSE56560, GSE60646, GSE55643, GSE18670, and GSE58561) which contain data from 144 PDAC tissues and matched adjacent non-tumor tissue samples were found to meet the inclusion criteria. After exclusion of non-confirmed sample pairs and samples with low quality, 105 PDAC and matched adjacent non-tumor tissue pairs from four datasets (GSE62452, GSE15471, GSE62165, and GSE56560) were decided to be eligible for the meta-analysis (Table <ns0:ref type='table'>1</ns0:ref>).</ns0:p></ns0:div>
<ns0:div><ns0:head>Differentially Expressed Genes in PDAC</ns0:head><ns0:p>The meta-analysis identified 344 (342 protein-coding genes) differentially over-expressed and 168 (157 protein-coding genes) repressed genes in PDAC according to the matched adjacent non-tumor samples (Table <ns0:ref type='table'>S1</ns0:ref>). A list of the top 20 differentially expressed genes (FDR<0.05) with at least a 2-fold differential expression between groups is shown in Table <ns0:ref type='table'>2</ns0:ref>. Along with the genes reported to be associated with pancreatic cancer in previous studies, the results of the meta-analysis revealed candidate over-expressed genes for PDAC, many of which have previously received little or no attention, such as KYNU (kynureninase), ITGBL1 (integrin betalike 1), and ADGRF1 (adhesion G protein-coupled receptor).</ns0:p><ns0:p>The differentially expressed genes identified in PDAC was validated in TCGA combined GTEx data containing data from PDAC tissues and pancreatic tissues from healthy individuals. This confirmation was also aimed at determining whether the DEGs identified could also distinguish PDAC tissues from healthy pancreatic tissues. Gene expression analysis of the TCGA combined GTEx data validated the up-regulation of 336 and the down-regulation of 116 protein-coding genes in PDAC (p≤0.001, Fold Change≥2). Additionally, the elevated expression of AK4 was validated in the Logsdon Pancreas Dataset in Oncomine Database due to the absence of the probes for this gene in TCGA combined GTEx data (p=1.16E-5). The identified DEGs and the results of the validation analysis are shown in Table <ns0:ref type='table'>S1</ns0:ref>.</ns0:p><ns0:p>Taken together, the expression of a significant portion of the identified DEGs was consistent with the gene expression profile from TCGA, implying that the identified genes may have the potential to discriminate PDAC from both adjacent and healthy pancreatic tissues.</ns0:p></ns0:div>
<ns0:div><ns0:head>Activated and Inhibited Transcription Factors in PDAC</ns0:head><ns0:p>To predict which transcription factors (TF) are activated or inhibited to regulate the transcription of the identified DEGs, TFs whose target genes show a significant overlap with the DEGs were identified using TFactS Database. Table <ns0:ref type='table'>S2</ns0:ref> shows the results of the TF enrichment PeerJ reviewing PDF | (2020:07:51367:1:1:NEW 9 Sep 2020)</ns0:p><ns0:p>Manuscript to be reviewed analysis. A total of five transcription factors were found to reach statistical significance (p-value, q-value, E-value, and FDR <0.05). Among these, Transcription Factor 7 (TCF7), Catenin Beta 1 (CTNNB1), Smad Family Member 3 (SMAD3) and Jun Proto-Oncogene (JUN) were found to be significantly activated in PDAC while Smad Family Member 7 (SMAD7) was detected as significantly inhibited.</ns0:p></ns0:div>
<ns0:div><ns0:head>Prognostic Values of the Identified DEGs in PDAC</ns0:head><ns0:p>The prognostic values of the identified DEGs genes were evaluated by analyzing the TCGA Pan-Cancer pancreatic ductal adenocarcinoma data and clinicopathological features of patients included in the Km-Plotter. Sixty-one up-regulated and one down-regulated genes (p≤0.05 and FDR≤0.05) were found to be significantly associated with the overall survival rate of patients with PDAC. TCGA combined GTEx data validated the up-and down regulation of the identified candidate prognostic genes in PDAC. Subsequently, four external datasets (GSE21501, GSE50827, GSE57495, and GSE71729) were analyzed to validate the prognostic significance of the identified 62 candidate genes for PDAC. The candidate prognostic gene list identified by Km-Plotter and the results of the validation analysis is shown in Table <ns0:ref type='table' target='#tab_1'>S3</ns0:ref>.</ns0:p><ns0:p>Genes whose prognostic potentials were confirmed in one of the four validation sets were included in the 'prognostic gene list' and their predicted prognostic values were ranked based on (i) validation status (validated in 1-2-3-4 dataset(s) =validation status 1-2-3-4) (ii) the average of p-values, and (iii) the average of hazard ratios. None of the genes could reach validation status 4. However, two genes, LAMC2 and ITGB6, were found to be significantly correlated with worse overall survival in three of the four validation datasets. SERPINB5, KRT7 TGM2, IGF2BP3, INPP4B, IL1RN, DCBLD2, and MPZL2 were validated in two datasets. Three genes, TMPRSS4, SEMA3C, and MGLL, were associated with increased overall survival in one of the validation datasets, therefore excluded from the prognostic gene list. Gene expression data of CDK1, PKM, AK4, AREG were not available in some validation datasets. While AK4 and PKM, which were found to be correlated with overall survival in one of the four external datasets, were included in the prognostic gene list, CDK1 and MGLL were excluded from the list due to insufficient gene expression data.</ns0:p><ns0:p>Consequently, a total of 28 up-regulated genes were individually validated in at least one of the four external validation datasets. The ranked prognostic gene list for PDAC and the results</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:07:51367:1:1:NEW 9 Sep 2020) of the Kaplan-Meier survival analysis in five datasets are shown in Table <ns0:ref type='table' target='#tab_1'>3</ns0:ref>. Kaplan-Meier survival plots for the identified prognostic genes are shown in Fig. <ns0:ref type='figure'>2</ns0:ref>.</ns0:p><ns0:p>Further analysis of TCGA pancreatic ductal adenocarcinoma data in GEPIA showed that, twenty-one of the identified prognostic genes (ITGB6, LAMC2, KRT7, SERPINB5, IGF2BP3, IL1RN, MPZL2, SFTA2, MET, LAMA3, ARNTL2, SLC2A1, LAMB3, COL17A1, EPSTI1, IL1RAP, AK4, ANXA2, S100A16, KRT19, and GPRC5A) were also correlated with the pathological stages of the disease, underlying their prognostic value for PDAC (Fig. <ns0:ref type='figure'>3</ns0:ref>).</ns0:p></ns0:div>
<ns0:div><ns0:head>Protein-Protein Interaction (PPI) Network</ns0:head><ns0:p>The PPI analysis was performed to evaluate the physical and functional associations of proteins encoded by the identified differentially expressed genes in PDAC. The PPI network was constructed by Cytoscape based on the PPI correlations from the STRING database. PPIs among the DEGs with a confidence score cut-off ˃0.9 were selected to construct the PPI network (Fig. <ns0:ref type='figure'>4</ns0:ref>). Then the degree of each node in the network was calculated by using the NetworkAnalyzer tool of Cytoscape to identify hub proteins in the PPI network. The degree of a node is the number of edges connected to the node, and it has been stated that nodes with higher degreeswhich correspond to hub proteins in the PPI network-, play an essential role in the organization of the PPI network, therefore might be more crucial and relevant than non-hub genes <ns0:ref type='bibr' target='#b62'>(Vallabhajosyula et al. 2009)</ns0:ref>. In this study, nodes with degrees >15 are considered to indicate 'hub proteins' and are presented in Table <ns0:ref type='table'>4</ns0:ref>. Next, the PPI network was analyzed by MCODE to identify clusters in the network. Each clustered protein group was then analyzed to predict their biological functions in PDAC.</ns0:p><ns0:p>Among the 27 hub proteins, ITGA2 and COL17A1 were found to be associated with unfavorable prognosis for PDAC based on the TCGA Pan-Cancer pancreatic ductal adenocarcinoma dataset in Km-Plotter (p≤0.05 and FDR≤0.05). However, this finding could not be validated in the four external datasets, except the significant correlation between the upregulation of COL17A1 and decreased overall survival found in GSE57495. Moreover, Manuscript to be reviewed this study, POSTN, was found to be located close to the collagen sub-cluster included in CLUSTER1, emphasizing their close interactions. CLUSTER1 also included up-regulated genes with unknown importance in pancreatic cancer, such as MATN3, SERPINA1, and IGFBP5.</ns0:p><ns0:p>CLUSTER3 covered known prognostic biomarkers such as TOP2A <ns0:ref type='bibr' target='#b61'>(Tsiambas et al. 2007</ns0:ref>), CDK1 <ns0:ref type='bibr' target='#b39'>(Piao et al. 2019)</ns0:ref>, and MKI67 <ns0:ref type='bibr' target='#b52'>(Striefler et al. 2016)</ns0:ref>. This cluster was surrounded by genes (CENPK, ANLN, ECT2, and FAM83D) related to the cell cycle, but it appears that they were not included in the cluster due to the stringent confidence score cut-off criteria of the analysis. Amongst, the function of FAM83D in PDAC is not yet known. However, there are previous reports that describe this protein as a probable proto-oncogene that regulates cell proliferation, growth, migration, and epithelial to mesenchymal transition <ns0:ref type='bibr' target='#b63'>(Wang et al. 2015;</ns0:ref><ns0:ref type='bibr' target='#b64'>Wang et al. 2013</ns0:ref>), suggesting that it may have similar functions in PDAC.</ns0:p><ns0:p>Furthermore, CLUSTER2 and 4, were mapped to the chemokine signaling pathway and type I interferon signaling pathway, respectively, and located close to each other in the PPI network. Although the importance of chemokine signaling in PDAC has been previously reported <ns0:ref type='bibr' target='#b17'>(Geismann et al. 2019)</ns0:ref>, the exact role of cellular innate antiviral response in the pathogenesis of PDAC remains elusive. In this study, OAS1, OAS2, and RSAD2, which play critical roles in cellular innate antiviral responses induced by type I and type II interferon, were found to be up-regulated in PDAC. Additionally, Kaplan-Meier survival analysis of the TCGA data using Km-Plotter indicated an association of high OAS1 expression with the low survival rate of PDAC patients, however, could not be validated in other four external validation datasets.</ns0:p><ns0:p>The roles of these proteins in PDAC are still unknown, and further research is needed for the evaluation of their potential as diagnostic or prognostic biomarkers for PDAC.</ns0:p><ns0:p>While CLUSTER6 shows the known interleukin-laminin-EGF crosstalk in the focal adhesion pathway in PDAC, CLUSTER9 indicated a novel down-regulated gene in PDAC, P2RX1, a ligand-gated ion channel with relatively high calcium permeability, that may be linked to apoptosis by increasing the intracellular concentration of calcium in the presence of ATP (Uniprot Database, by similarity). CLUSTER8 included cell surface antigens such as CD109, CD66e, and CD90. However, as shown in the PPI network, the interaction between these proteins and the genes that have roles in the regulation of glycolysis such as PKM, PFKP, and <ns0:ref type='table'>S4</ns0:ref>.</ns0:p></ns0:div>
<ns0:div><ns0:head>Gene Ontology Analysis of the DEGs</ns0:head><ns0:p>GO analysis of the DEGs was performed to identify enriched biological processes, molecular functions and cellular locations associated with differential gene expression in PDAC.</ns0:p><ns0:p>As shown in Fig. <ns0:ref type='figure'>5A</ns0:ref>, the analysis identified nine significantly overrepresented individual categories of GO Molecular Function for the up-regulated genes including 'extracellular matrix structural constituents' (p˂ 0.001), 'collagen binding' (p˂ 0.001) and 'integrin-binding' (p˂ 0.001). 'Extracellular matrix organization' and 'cell adhesion' were the most significantly enriched biological processes (p˂ 0.001, Fig. <ns0:ref type='figure'>5B</ns0:ref>), while most of the proteins encoded by the upregulated genes were found to be located in the 'collagen-containing extracellular matrix' (p˂ 0.001, Fig. <ns0:ref type='figure'>5</ns0:ref>.C).</ns0:p><ns0:p>For the down-regulated genes, 'serine-type endopeptidase activity' was the only GO molecular function category that reached statistical significance (p= 0.008, Fig. <ns0:ref type='figure'>5</ns0:ref>.D). Three GO Biological Processes were found to be significantly enriched for the down-regulated genes: 'proteolysis' (p= 0.001), 'cellular zinc ion homeostasis' (p=0.002), and 'cellular response to copper ion' (p=0.042), (Fig. <ns0:ref type='figure'>5</ns0:ref>.E). Moreover, most of the over-represented GO cellular locations were associated with extracellular space for the down-regulated genes (Fig. <ns0:ref type='figure'>5</ns0:ref>.F).</ns0:p></ns0:div>
<ns0:div><ns0:head>Pathway Enrichment Analysis of the DEGs</ns0:head><ns0:p>Significantly dysregulated pathways in PDAC were identified by using four different pathway databases in WebGestalt (Table <ns0:ref type='table'>S5</ns0:ref>). The obtained results from the pathway enrichment analysis of the identified DEGs underlined the importance of the crosstalk between tumor cells and extracellular matrix, including integrin and PI3K-Akt-mTOR signaling pathways, in PDAC pathogenesis. Pathways related to metabolism and pancreatic secretion were identified as negative related categories by KEGG and Reactome Pathway Knowledgebase.</ns0:p></ns0:div>
<ns0:div><ns0:head>DISCUSSION</ns0:head><ns0:p>The lack of an effective treatment option for pancreatic cancer emphasizes an absolute need for expanding our knowledge of the etiology and molecular mechanisms of the disease to discover novel drug targets. In the current study, using stringent inclusion and exclusion criteria examining patient selection and microarray quality assessment, an integrative meta-analysis of PeerJ reviewing PDF | (2020:07:51367:1:1:NEW 9 Sep 2020)</ns0:p><ns0:p>Manuscript to be reviewed transcriptome data from four studies was conducted to suggest novel multifunctional biomarkers as well as therapeutic targets for pancreatic ductal adenocarcinoma.</ns0:p><ns0:p>Comparison of tumor and adjacent tissues has advantages such as providing more reliable results due to the elimination of the variations between individuals and anatomical locations from which samples are taken. This approach allowed the identification of potential biomarkers for PDAC that may serve in the molecular evaluation of the surgical margin, which provides a more sensitive and precise assessment of the risk of cancer recurrence than solely by histopathologic examination. Although histologically normal adjacent-to-tumor tissues are generally accepted as healthy controls, there are also studies reporting that, at the molecular level, these tissues are distinct from both healthy and tumor tissues and represent an intermediate state <ns0:ref type='bibr' target='#b2'>(Aran et al. 2017;</ns0:ref><ns0:ref type='bibr' target='#b45'>Russi et al. 2019</ns0:ref>). Therefore, the potential of the defined DEGs in PDAC to be diagnostic biomarkers was evaluated in an external validation dataset including PDAC and healthy pancreatic tissues (TCGA combined GTEx data). The validation analysis verified 98,24% of the identified up-regulated and 73,88% of the identified down-regulated proteincoding genes in PDAC, suggesting that these genes may serve as promising diagnostic biomarkers that differentiate PDAC from both healthy and adjacent-to-tumor pancreatic tissues.</ns0:p><ns0:p>Kinase inhibitors constitute a significant portion of chemotherapeutic agents that are in clinical use today. The results of this meta-analysis revealed that the expression of fifteen kinaseencoding genes that have the potential to be therapeutic targets in PDAC, is higher than both healthy and adjacent pancreatic tissues. Additional survival analysis on these individual genes revealed that higher expression of DGKH (Diacylglycerol Kinase η) and AK4 (adenylate kinase 4) was associated with worse survival probabilities in at least two of the five external datasets.</ns0:p><ns0:p>DGKH plays a crucial role in promoting cell growth and activates the RAS/Raf/MEK/ERK signaling pathway induced by EGF <ns0:ref type='bibr' target='#b73'>(Yasuda et al. 2009</ns0:ref>). AK4 is a critical mitochondrial enzyme that participates in maintaining the homeostasis of cellular nucleotides and plays a role in controlling cellular ATP levels by regulating AMPK signaling <ns0:ref type='bibr' target='#b34'>(Lanning et al. 2014</ns0:ref>).</ns0:p><ns0:p>However, the functional roles and the clinicopathological significance of AK4 and DGKH in pancreatic cancer have never been investigated. Here, the results of this study showed that increased AK4 and DGKH expressions are independently correlated with a decreased overall survival rate of patients with PDAC. It is also noteworthy that, DGKH was found to be a prognostic marker in three external validation datasets (GSE21501, GSE50827, and GSE57495), Manuscript to be reviewed indicating its prognostic power as a biomarker (Supp. Fig. <ns0:ref type='figure' target='#fig_0'>S1</ns0:ref>). These results suggest that AK4 and DGKH may be potential therapeutic targets in devising a treatment for patients with pancreatic cancer and have the potential to be prognostic and diagnostic biomarkers for PDAC.</ns0:p><ns0:p>In the context of the prognostic significance of the identified DEGs, genes were analyzed using Km-Plotter Database to see whether the differentially expressed genes identified in this study had a significant effect on the survival of patients with PDAC, as survival data were not available for some GEO datasets used in the meta-analysis. To increase the reliability, the identified candidate genes whose expressions significantly correlated with the overall survival rate of patients were further validated using four distinct GEO Datasets. The results of this analysis revealed a total of twenty-eight up-regulated genes in PDAC compared to both adjacent and normal pancreatic tissues that might have the potential to be prognostic and diagnostic biomarkers for PDAC. Notably, violin plots of gene expression by pathological stages based on the TCGA PAAD data showed that twenty-one of the identified prognostic genes (ITGB6, LAMC2, KRT7, SERPINB5, IGF2BP3, IL1RN, MPZL2, SFTA2, MET, LAMA3, ARNTL2, SLC2A1, LAMB3, COL17A1, EPSTI1, IL1RAP, AK4, ANXA2, S100A16, KRT19, and GPRC5A) also significantly correlated with pathological stages of the disease), indicating that these genes may play crucial roles in the tumorigenesis of PDAC. As expected, some of these identified genes have been reported to be associated with poor overall survival of patients with pancreatic cancer <ns0:ref type='bibr' target='#b9'>(Cheng et al. 2019;</ns0:ref><ns0:ref type='bibr' target='#b27'>Hu et al. 2019;</ns0:ref><ns0:ref type='bibr' target='#b29'>Jahny et al. 2017;</ns0:ref><ns0:ref type='bibr' target='#b38'>Pei et al. 2018;</ns0:ref><ns0:ref type='bibr' target='#b43'>Reader et al. 2019;</ns0:ref><ns0:ref type='bibr' target='#b47'>Schaeffer et al. 2010;</ns0:ref><ns0:ref type='bibr' target='#b54'>Takahashi et al. 2019;</ns0:ref><ns0:ref type='bibr' target='#b68'>Wu et al. 2019;</ns0:ref><ns0:ref type='bibr' target='#b72'>Yao et al. 2016;</ns0:ref><ns0:ref type='bibr' target='#b74'>Zhang et al. 2019a</ns0:ref>). However, the functional and clinicopathological significance of IL1RN, MPZL2, SFTA2, EPSTI1, IL1RAP, AK4 and S100A16 in PDAC have not been reported. Additionally, to the best of my knowledge, a correlation between the elevated expressions of ITGB6, LAMC2, LAMA3, ARNTL2, LAMB3, COL17A1, and GPRC5A and pathological stages of the disease has yet been stated. Taken together, the results of this analysis revealed that these genes individually might have high diagnostic and prognostic values, as well as the potential to be therapeutic targets for PDAC, prompting further study.</ns0:p><ns0:p>Moreover, to predict which transcription factors (TF) are altered to regulate the transcription of the identified DEGs, a TF enrichment analysis was conducted. Among the five TFs identified, activation of SMAD3 and inhibition of SMAD7 may together indicate a possible activation of TGF-beta signaling, which is known to play a dual role as both pro-tumorigenic and PeerJ reviewing PDF | (2020:07:51367:1:1:NEW 9 Sep 2020)</ns0:p><ns0:p>Manuscript to be reviewed tumor-suppressive in pancreatic cancer, depending on tumor stage and microenvironment <ns0:ref type='bibr' target='#b50'>(Shen et al. 2017)</ns0:ref>. Additionally, Transcription factor 7 (TCF7) and Catenin Beta-1 (CTNNB1), key proteins in the Wnt Signaling Pathway, were predicted to be activated in PDAC. Although there are scientific reports on the prognostic values of SMAD3 <ns0:ref type='bibr' target='#b71'>(Yamazaki et al. 2014</ns0:ref>) and CTNNB1 <ns0:ref type='bibr' target='#b76'>(Zhang et al. 2016)</ns0:ref> in pancreatic cancer, the precise functions of JUN (Jun Oncogene) and TCF-7 in the pathogenesis of pancreatic cancer are still largely unknown and require further understanding and research.</ns0:p><ns0:p>Further analysis showed that most of the proteins encoded by the identified up/downregulated genes are components of the extracellular domain, which is in line with the fact that PDAC has a characteristically abundant desmoplastic stroma <ns0:ref type='bibr' target='#b8'>(Cannon et al. 2018)</ns0:ref>. Moreover, the gene ontology analysis revealed that the identified up-regulated genes in PDAC generally encode extracellular matrix (ECM) structural elements together with the proteins with collagen and integrin-binding properties. These genes were mapped significantly to the biological processes called ECM organization and cell adhesion, which may promote survival, proliferation, and metastasis of PDAC that result in an aggressive disease phenotype as suggested elsewhere <ns0:ref type='bibr' target='#b66'>(Weniger et al. 2018)</ns0:ref>.</ns0:p><ns0:p>Significantly enriched pathways for up-regulated genes were found to include ECM organization and receptor interaction, focal adhesion-PI3K-Akt-mTOR signaling pathway, and integrin signaling pathway, which have well-known associations with pancreatic cancer <ns0:ref type='bibr' target='#b13'>(Ebrahimi et al. 2017;</ns0:ref><ns0:ref type='bibr' target='#b60'>Topalovski & Brekken 2016)</ns0:ref>. These pathways have been known to be activated by various types of cellular stimuli and interact with each other in a variety of complex ways <ns0:ref type='bibr' target='#b25'>(Hastings et al. 2019;</ns0:ref><ns0:ref type='bibr' target='#b67'>Wu et al. 2016)</ns0:ref>, emphasizing the intricate molecular pathogenesis of pancreatic cancer. In this study, the genes assigned to the mentioned pathways included FN1 (fibronectin) gene, a member of the top three hub proteins identified by the PPI network analysis.</ns0:p><ns0:p>Fibronectin, an abundant stromal protein in PDAC, has been known to drive metastatic spread, angiogenesis, and chemoresistance of PDAC by mediating FAK dependent activation of the PI3K/AKT/mTOR pathway or RAS/RAF/MEK pathway <ns0:ref type='bibr' target='#b60'>(Topalovski & Brekken 2016)</ns0:ref>.</ns0:p><ns0:p>However, in this study FN1 expression was not found to be an independent prognostic factor for overall survival in patients with PDAC. Moreover, LAMB3, another protein assigned to these pathways, has also been shown by a recent study to mediate apoptotic, proliferative, invasive, and metastatic behaviors in pancreatic cancer by regulating the PI3K/AKT signaling pathway PeerJ reviewing PDF | (2020:07:51367:1:1:NEW 9 Sep 2020)</ns0:p><ns0:p>Manuscript to be reviewed <ns0:ref type='bibr' target='#b75'>(Zhang et al. 2019b</ns0:ref>). The other proteins in these pathways were members of the families of interleukin (ITGA2, ITGB6), laminin (LAMC2), and collagen (COL1A1/2, COL6A3), together with cartilage oligomeric matrix protein (COMP), whose role in pancreatic cancer has not yet been clarified. Amongst, increased expressions of LAMB3 and ITGB6 were significantly associated with poor prognosis of patients with pancreatic adenocarcinoma in this study, underlying the importance of the identified enriched pathways in PDAC.</ns0:p><ns0:p>Regarding the functions of the down-regulated genes in PDAC, the GSEA pathway analysis revealed these genes significantly associated with pancreatic secretion and metabolic pathways. Abnormal pancreatic secretion is known to occur frequently in patients with pancreatic cancer. However, alterations in pancreatic enzyme secretion have not yet been wellcharacterized in PDAC. In this study, 18 genes associated with pancreatic secretion were found to be down-regulated in PDAC compared to adjacent tissues, including chymotrypsin-like elastase family members (CELA2A/B, CELA3A/B), carboxypeptidases (CPA1/2, CPB1), pancreatic lipase (PNLIP), pancreatic amylase (AMY2A) and a sodium bicarbonate cotransporter; SLC4A4. The causes of abnormal pancreatic secretion other than pancreatic duct obstruction may be a decrease in the number of secretory cells or translational alterations in secretory cells, and whether these alterations contribute to malignancy needs to be investigated.</ns0:p><ns0:p>Altered metabolism and metabolic plasticity have been associated with proliferation, aggressivity, adaptability to changes in the tumor microenvironment, and drug resistance in pancreatic cancer <ns0:ref type='bibr' target='#b5'>(Biancur & Kimmelman 2018)</ns0:ref>. In this study, the genes assigned to the suppressed metabolic pathways were mostly enzyme-coding genes associated with amino acid metabolism (ABAT, GATM, GLS2, ANPEP, PSAT1, GPT2, and GAMT). The other downregulated genes related to various metabolic pathways in PDAC were ADH1B, CTH, EPHX2, AOX1, ACADL, IMPA2, ACAT1, ACSM3, UGT2A3, and CBS. The roles of these genes in pancreatic cancer are not yet known; therefore, further studies addressing this issue may reveal therapeutic, diagnostic, or prognostic values of altered metabolism in PDAC. Moreover, Regucalcin (RGN), which is a highly conserved calcium-binding protein, was also assigned to the metabolic pathways in this study. Overexpression of Regucalcin has been demonstrated to suppress proliferation, cell death, and migration in an in vitro model of pancreatic cancer in a previous study <ns0:ref type='bibr' target='#b70'>(Yamaguchi et al. 2016)</ns0:ref>. Therefore, the identified down-regulation of RGN in PDAC may emphasize the tumor suppressor property of this protein.</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:07:51367:1:1:NEW 9 Sep 2020)</ns0:p></ns0:div>
<ns0:div><ns0:head>Manuscript to be reviewed</ns0:head><ns0:p>In conclusion, this study provided a list of multifunctional biomarkers that have the potential to distinguish PDAC from both adjacent-to-tumor tissues and healthy pancreatic tissues, as well as correlated with overall survival rate of patients and the pathological stages of the disease. Future investigations are necessary to additionally validate the combined prognostic and diagnostic value of the identified biomarkers in this study. The functional significance of some of these identified dysregulated genes in the pathogenesis of PDAC is not yet known and deserves further investigation. Moreover, the results of this study provided insights into the molecular basis of the difference between PDAC and adjacent-to-tumor tissues which may be useful in the histopathological examination of PDAC and the development of more effective targeted therapies.</ns0:p></ns0:div>
<ns0:div><ns0:head>Abbreviations List</ns0:head><ns0:p>TCGA: The Cancer Genome Atlas, PDAC: Pancreatic Ductal Adenocarcinoma, GTEx: The Genotype-Tissue Expression, ECM: Extracellular matrix, GEO: Gene Expression Omnibus, RMA: Robust Multi-array Average, GEPIA: Gene Expression Profiling Interactive Analysis, PAAD: Pancreatic Adenocarcinoma, TF: Transcription Factor, KEGG: Kyoto Encyclopedia of Genes and Genomes, DEGs: Differentially expressed genes, PPI: Protein-protein interaction, GO: Gene Ontology, EGF: Epidermal Growth Factor.</ns0:p></ns0:div>
<ns0:div><ns0:head>Gene symbol</ns0:head></ns0:div>
<ns0:div><ns0:head>Gene name</ns0:head><ns0:p>Logratio combined Manuscript to be reviewed The ranked prognostic gene list for PDAC and the results of the Kaplan-Meier survival analysis in five datasets.</ns0:p><ns0:p>HR: hazard ratio, NS: nonsignificant, p≥ 0.05; N/A: not available, p: p value, Val. Stat: validation status, PDAC: Pancreatic ductal adenocarcinoma.</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:07:51367:1:1:NEW 9 Sep 2020)</ns0:p><ns0:p>Manuscript to be reviewed </ns0:p></ns0:div><ns0:figure xml:id='fig_0'><ns0:head>Fibronectin 1 (</ns0:head><ns0:label>1</ns0:label><ns0:figDesc>FN1), tissue inhibitor of metalloproteinases 1 (TIMP1), and fibrillin 1 (MSLN) constituted the top three proteins with degrees exceeding 20. All these three proteins were members of CLUSTER1, which includes proteins whose functions mostly associated with extracellular matrix signaling pathways and structural organization, underlying the importance of extracellular dynamics in the pathogenesis of PDAC. The most significantly up-regulated gene in PeerJ reviewing PDF | (2020:07:51367:1:1:NEW 9 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_1'><ns0:head /><ns0:label /><ns0:figDesc>ENO2 was intriguing and may indicate stemness associated interactions in PDAC. The other clusters were significantly mapped to those specific KEGG pathways: 'Mucin-Type O-glycan PeerJ reviewing PDF | (2020:07:51367:1:1:NEW 9 Sep 2020) Manuscript to be reviewed Biosynthesis' (CLUSTER5) and 'Estrogen Signaling Pathway' (CLUSTER7), indicating once more the high expression of keratins and mucin-type O-glycans in pancreatic tumors. The proteins assigned to the pathways are shown in Table</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_2'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:07:51367:1:1:NEW 9 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='30,42.52,275.62,525.00,348.00' type='bitmap' /></ns0:figure>
<ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='31,42.52,255.37,525.00,331.50' type='bitmap' /></ns0:figure>
<ns0:figure type='table' xml:id='tab_0'><ns0:head>Fold change p- value FDR</ns0:head><ns0:label /><ns0:figDesc /><ns0:table><ns0:row><ns0:cell>POSTN</ns0:cell><ns0:cell>periostin, osteoblast specific</ns0:cell><ns0:cell>1,0004</ns0:cell><ns0:cell>10,009</ns0:cell><ns0:cell>0</ns0:cell><ns0:cell>0</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>factor</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>CEACAM5</ns0:cell><ns0:cell>carcinoembryonic antigen-</ns0:cell><ns0:cell>0,8855</ns0:cell><ns0:cell>7,683</ns0:cell><ns0:cell>0</ns0:cell><ns0:cell>0</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>related cell adhesion</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell /><ns0:cell>molecule 5</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>SLC6A14</ns0:cell><ns0:cell>solute carrier family 6 (amino</ns0:cell><ns0:cell>0,8611</ns0:cell><ns0:cell>7,263</ns0:cell><ns0:cell>0</ns0:cell><ns0:cell>0</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>acid transporter), member 14</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>CEACAM6</ns0:cell><ns0:cell>carcinoembryonic antigen-</ns0:cell><ns0:cell>0,8452</ns0:cell><ns0:cell>7,002</ns0:cell><ns0:cell>0</ns0:cell><ns0:cell>0</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>related cell adhesion</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell /><ns0:cell>molecule 6 (non-specific</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell /><ns0:cell>cross reacting antigen)</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>SULF1</ns0:cell><ns0:cell>sulfatase 1</ns0:cell><ns0:cell>0,8347</ns0:cell><ns0:cell>6,835</ns0:cell><ns0:cell>0</ns0:cell><ns0:cell>0</ns0:cell></ns0:row><ns0:row><ns0:cell>LAMC2</ns0:cell><ns0:cell>laminin subunit gamma 2</ns0:cell><ns0:cell>0,8279</ns0:cell><ns0:cell>6,728</ns0:cell><ns0:cell>0</ns0:cell><ns0:cell>0</ns0:cell></ns0:row><ns0:row><ns0:cell>FN1</ns0:cell><ns0:cell>fibronectin 1</ns0:cell><ns0:cell>0,8083</ns0:cell><ns0:cell>6,432</ns0:cell><ns0:cell>0</ns0:cell><ns0:cell>0</ns0:cell></ns0:row><ns0:row><ns0:cell>COL11A1</ns0:cell><ns0:cell>collagen, type XI, alpha 1</ns0:cell><ns0:cell>0,7918</ns0:cell><ns0:cell>6,191</ns0:cell><ns0:cell>0</ns0:cell><ns0:cell>0</ns0:cell></ns0:row><ns0:row><ns0:cell>INHBA</ns0:cell><ns0:cell>inhibin beta A</ns0:cell><ns0:cell>0,7713</ns0:cell><ns0:cell>5,907</ns0:cell><ns0:cell>0</ns0:cell><ns0:cell>0</ns0:cell></ns0:row><ns0:row><ns0:cell>VCAN</ns0:cell><ns0:cell>versican</ns0:cell><ns0:cell>0,7644</ns0:cell><ns0:cell>5,813</ns0:cell><ns0:cell>0</ns0:cell><ns0:cell>0</ns0:cell></ns0:row><ns0:row><ns0:cell>ALB</ns0:cell><ns0:cell>albumin</ns0:cell><ns0:cell>-0,8658</ns0:cell><ns0:cell>-7,342</ns0:cell><ns0:cell>0</ns0:cell><ns0:cell>0</ns0:cell></ns0:row><ns0:row><ns0:cell>SERPINI2</ns0:cell><ns0:cell>serpin peptidase inhibitor,</ns0:cell><ns0:cell>-0,783</ns0:cell><ns0:cell cols='2'>-6,068 8,88E-</ns0:cell><ns0:cell>4,96E-14</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>clade I (pancpin), member 2</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell>16</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>PNLIPRP1</ns0:cell><ns0:cell>pancreatic lipase-related</ns0:cell><ns0:cell>-0,7683</ns0:cell><ns0:cell cols='2'>-5,866 9,40E-</ns0:cell><ns0:cell>4,71E-10</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>protein 1</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell>12</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>ERP27</ns0:cell><ns0:cell>endoplasmic reticulum</ns0:cell><ns0:cell>-0,7367</ns0:cell><ns0:cell cols='2'>-5,454 6,66E-</ns0:cell><ns0:cell>3,75E-14</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>protein 27</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell>16</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>PNLIPRP2</ns0:cell><ns0:cell>pancreatic lipase-related</ns0:cell><ns0:cell>-0,7359</ns0:cell><ns0:cell cols='2'>-5,444 3,66E-</ns0:cell><ns0:cell>1,85E-10</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>protein 2</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell>12</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>CTRL</ns0:cell><ns0:cell>chymotrypsin-like</ns0:cell><ns0:cell>-0,7199</ns0:cell><ns0:cell cols='2'>-5,247 3,55E-</ns0:cell><ns0:cell>1,94E-13</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell>15</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>PDIA2</ns0:cell><ns0:cell>protein disulfide isomerase</ns0:cell><ns0:cell>-0,7025</ns0:cell><ns0:cell>-5,041</ns0:cell><ns0:cell>0</ns0:cell><ns0:cell>0</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>family A member 2</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>GP2</ns0:cell><ns0:cell>glycoprotein 2 (zymogen</ns0:cell><ns0:cell>-0,7005</ns0:cell><ns0:cell cols='2'>-5,018 3,06E-</ns0:cell><ns0:cell>1,51E-09</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>granule membrane)</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell>11</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>CELA2B</ns0:cell><ns0:cell>chymotrypsin like elastase</ns0:cell><ns0:cell>-0,685</ns0:cell><ns0:cell cols='2'>-4,842 7,62E-</ns0:cell><ns0:cell>3,83E-10</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>family member 2B</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell>12</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>IAPP</ns0:cell><ns0:cell>islet amyloid polypeptide</ns0:cell><ns0:cell>-0,6768</ns0:cell><ns0:cell cols='2'>-4,751 8,88E-</ns0:cell><ns0:cell>4,96E-14</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell>16</ns0:cell><ns0:cell /></ns0:row></ns0:table><ns0:note>1PeerJ reviewing PDF | (2020:07:51367:1:1:NEW 9 Sep 2020)</ns0:note></ns0:figure>
<ns0:figure type='table' xml:id='tab_1'><ns0:head>Table 3 (on next page)</ns0:head><ns0:label>3</ns0:label><ns0:figDesc /><ns0:table /></ns0:figure>
<ns0:note place='foot' n='1'>PeerJ reviewing PDF | (2020:07:51367:1:1:NEW 9 Sep 2020)</ns0:note>
</ns0:body>
" | "
September 2, 2020
Dear Editor,
I thanked the reviewers for their time and valuable comments on the manuscript. I have taken the comments on board to improve the content and readibility of the manuscript. Please find below a detailed point-by-point response to all comments (reviewers’ comments in black, my replies in dark blue). I referred to the line numbers in the revised manuscript. All revisions I made are trackable and highlighted in the tracked changes version of the manuscript.
I hope I was able to address the reviewers' concerns and this version of the article is adequate for publication in PeerJ.
Sincerely,
Sevcan ATAY
Reviewer 1.
Basic reporting
-
Experimental design
-
Validity of the findings
-
Comments for the author
The current manuscript by Sevcan Atay presents a transcriptome meta-analysis of pancreatic cancer tissues. GEO, TCGA combined with GTEx data sets and others were utilized to identify pancreatic cancer specific genes. Pathway enrichment analysis was carried out as well as survival analysis using TCGA Pan-Cancer data. Genes specific for pancreatic cancer, correlating with survival of pancreatic cancer patients were identified.
This is an interesting analysis that is solely based on publicly available data and data analysis tools. The novelty of the findings is rather limited. The utilized methods are sound and valid. However, there are further concerns that should be addressed:
It is difficult to follow, which datasets were used as training and validation cohorts. Obviously, these datasets are all different, therefore validation is not trivial. Could the author please comment?
Reply: Thank you for your time and valuable comments on my manuscript. I agree that it may be difficult for readers to follow the datasets used in the analyses throughout the article. I made the following changes to improve the readability of the article:
• I included more information about the workflow and results of the study in Figure 1.
• I tried to mention more clearly which data sets are used for what purpose in the abstract and main manuscript.
In this study, in addition to TCGA and GTEx data, four external validation datasets were used to evaluate and/or validate the prognostic or diagnostic value of the identified differentially expressed genes in PDAC. Here, I can briefly explain in which analysis and for what purpose datasets are used:
After identification of the differentially expressed genes (DEGs) between PDAC and adjacent to tumor tissues (GSE62452, GSE15471, GSE62165, and GSE56560), external validation of the diagnostic potential of the mRNA levels of these genes was performed using TCGA PAAD combined GTEx data.
The prognostic significance of the identified DEGs in PDAC was evaluated in TCGA-Pan-Cancer Pancreatic Ductal Adenocarcinoma Data. The result of this analysis revealed the candidate prognostic genes for PDAC.
Then, four independent external validation datasets (GSE21501, GSE50827, GSE57495, and GSE71729 including transcriptomic and clinical data from patients with pancreatic ductal adenocarcinoma) were analyzed to evaluate the prognostic power of the identified candidate prognostic genes for PDAC. The candidate prognostic genes whose prognostic potentials were confirmed in at least one of the four validation sets were included in the prognostic gene list for PDAC and their predicted prognostic values were ranked based on validation status, p value and hazard ratio.
After that, I evaluated the correlation between the mRNA expression of the genes in the prognostic gene list and the pathological stages of the disease using TCGA PAAD data.
Overall, the results of these sequential analyzes revealed a total of twenty-one promising multifunctional biomarkers that have the potential to differentiate PDAC from both healthy and adjacent pancreatic tissues and also significantly correlate with the overall survival rate of patients and pathological stages of the disease.
Using an inclusion criterion such as “high quality gene expression microarray data” does not necessarily mean that these data are of higher quality than other published microarray data.
Reply: The quality of microarray data is the sole responsibility of the person uploading the data to a public database. Thus, GEO includes low-quality microarray data as well as high-quality microarray data. In this study, the microarray data to be included in the study were passed through strict quality control as explained in the 'Generation of Gene Expression Matrix Files and Evaluation of Data Quality' part of the manuscript. Poor quality microarray data were omitted from datasets and not included in the meta-analysis. Therefore, ‘high-quality microarray data’ was actually one of the inclusion criteria of this study. However, I can understand the reviewer’s viewpoint here and changed the ‘high-quality gene expression microarray data’ statement with 'gene expression microarray data ' as an inclusion criterion (Line 84). Thus, 'Poor quality' remained specified as an exclusion criterion in the data quality part of the methods section. I also deleted ‘high quality’ statement in the abstract (Line 2) and introduction (Line 78).
Sample pairs can make sense but need better explanation. In pancreatic cancer, most areas next to caner contain chronic pancreatitis like changes and not ‘normal’ pancreatic tissue. This should be discussed.
Reply: In the manuscript, I used the term ‘adjacent-to-tumor normal tissues’ to define histologically normal non-tumor tissues that adjacent to the tumor (not healthy pancreatic tissues), as has also been similarly used in several studies (1-3). Although these control tissues were histologically normal, they may have undergone epigenetic and transcriptomic changes induced not only by chronic pancreatitis like changes (or inflammation) but also by the tumor, tumour-microenvironment, duct obstruction, cancer-related metabolic alterations in pancreas etc. I had mentioned in the discussion section of the article that these tissues are distinct from both healthy and tumor tissues (Lines 344-347). To avoid misunderstandings, I changed the term ‘adjacent-to-tumor normal tissues’ to ‘adjacent-to-tumor tissues’or ‘adjacent non-tumor tissues’ throughout the article.
1. Badea L, Herlea V, Dima SO, Dumitrascu T, Popescu I. Combined gene expression analysis of whole-tissue and microdissected pancreatic ductal adenocarcinoma identifies genes specifically overexpressed in tumor epithelia. Hepatogastroenterology. 2008;55(88):2016-2027.
2. Coleman O, Henry M, OʼNeill F, et al. Proteomic Analysis of Cell Lines and Primary Tumors in Pancreatic Cancer Identifies Proteins Expressed Only In Vitro and Only In Vivo. Pancreas. 2020;49(8):1109-1116. doi:10.1097/MPA.0000000000001633
3. Li HB, Zhou J, Zhao F, Yu J, Xu L. Prognostic Impact of DHRS9 Overexpression in Pancreatic Cancer. Cancer Manag Res. 2020;12:5997-6006. Published 2020 Jul 20. doi:10.2147/CMAR.S251897
The work provides “a list of multifunctional biomarkers that have the potential to distinguish PDAC from both adjacent-to-tumor tissues and healthy pancreatic tissues”. This statement is true, and the methodology is up-to-date. However, there are numerous studies with similar aims and results. Validation and functional characterization remain most important.
Reply: In this study, the identified differentially expressed genes in PDAC were validated in TCGA combined GTEx data (tumor n=179, healthy n=171). The identified candidate prognostic genes were individually validated in four external validation datasets including data from a total of 321 patients with PDAC. However, some of the identified biomarkers were novel and their functional roles in PDAC pathogenesis are not known yet. I had mentioned this in the discussion and conclusion part of the manuscript. I edited the conclusion part to particularly indicate the necessity of further additional validation of the prognostic and diagnostic value of the identified biomarkers in PDAC (Lines 464-465).
Reviewer 2.
Basic reporting
-
Experimental design
-
Validity of the findings
-
Comments for the author
This is a very nice bioinformatics work and a well written manuscript. Using many public databases and analysis tool that all of them are freely available Dr Atay comes to some interesting findings on biomarkers for PDAC. I only wonder if a more detailed analysis that could link biomarkers to staging of PDAC would be able. This could help to identify biomarkers for the early detection of this deadly cancer which is may be the biggest clinical problem.
Reply: Thank you for your generous and valuable comments. The discovery of early detection biomarkers requires a comparison of omic data from patients with stage I pancreatic ductal adenocarcinoma and control tissues. However, since late-stage diagnosis is very common in pancreatic cancer, the number of patients with stage I PDAC is very limited in most studies, which significantly decreases the statistical power of that kind of analysis. In this study, the clinical information of patients including staging and grading were not available in a significant portion of the samples included in the meta-analysis. In datasets that reported clinical data of the patients, the number of samples attributed to the stage I PDAC was very low (For example: only 2 patients with grade 1 PDAC in GSE62542). Thus, it was not possible to generate an analysis to identify early detection biomarkers in this study using data from studies included in the meta-analysis. Moreover, using publicly available data -other than those included in this study- to identify early detection biomarkers would require the generation of another integrative transcriptome meta-analysis due to low sample size (stage I) in studies. This was not preferred as it would be another study in itself. However, in this study, 21 of the identified prognostic biomarkers were found to be significantly corralated with the pathological stages of the disease (Fig.3). I added a paragraph about the correlation analysis to the ‘Finding prognostic genes for PDAC’ section of the materials and methods section of the manuscript (Lines 142-144). This was mainly indicated in the results part before.
Also it would be very helpful I believe for the reader Dr Atay to include a paragraph where all abbreviations will be explained (there are a lot!) and a table with all the databases and tools used for the meta-analysis.
Reply: I included an abbreviations list (Lines: 471-477) and rearrenged Figure 1 to include all databases and tools used for the meta-analysis as well as the results of the study.
" | Here is a paper. Please give your review comments after reading it. |
9,787 | "<ns0:abstract xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:p>Background: Colon cancer is one of the deadliest tumors worldwide. Stromal cells and immune cells play important roles in cancer biology and microenvironment across different types of cancer. This study aimed to identify the prognostic value of stromal/immune cell-associated genes for colon cancer in The Cancer Genome Atlas (TCGA) database using bioinformatic technology.</ns0:p></ns0:div>
<ns0:div><ns0:head>Methods:</ns0:head><ns0:p>The gene expression data and corresponding clinical information of colon cancer were downloaded from TCGA database. Stromal and immune scores were estimated based on the ESTIMATE algorithm. Sanger software was used to identify the differentially expressed genes (DEGs) and prognostic DEGs based on stromal and immune scores. External validation of prognostic biomarkers was conducted in Gene Expression Omnibus (GEO) database. Gene ontology (GO) analysis, pathway enrichment analysis, and gene set enrichment analysis (GSEA) were used for functional analysis. STRING and Cytoscape were used to assess Protein-protein interaction (PPI) network and screen hub genes. Quantitative real-time PCR (qRT-PCR) was used to validate the expression of hub genes in clinical tissues. Synaptosomal-associated protein 25 (SNAP25) was selected for analyzing its correlations with tumor-immune system in TISIDB database.</ns0:p></ns0:div>
</ns0:abstract>
<ns0:body xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:p>Background: Colon cancer is one of the deadliest tumors worldwide. Stromal cells and immune cells play important roles in cancer biology and microenvironment across different types of cancer. This study aimed to identify the prognostic value of stromal/immune cellassociated genes for colon cancer in The Cancer Genome Atlas (TCGA) database using bioinformatic technology. Methods: The gene expression data and corresponding clinical information of colon cancer were downloaded from TCGA database. Stromal and immune scores were estimated based on the ESTIMATE algorithm. Sanger software was used to identify the differentially expressed genes (DEGs) and prognostic DEGs based on stromal and immune scores. External validation of prognostic biomarkers was conducted in Gene Expression Omnibus (GEO) database. Gene ontology (GO) analysis, pathway enrichment analysis, and gene set enrichment analysis (GSEA) were used for functional analysis.</ns0:p><ns0:p>STRING and Cytoscape were used to assess Protein-protein interaction (PPI) network and screen hub genes. Quantitative real-time PCR (qRT-PCR) was used to validate the expression of hub genes in clinical tissues. Synaptosomal-associated protein 25 (SNAP25) was selected for analyzing its correlations with tumor-immune system in TISIDB database.</ns0:p><ns0:p>Results: Worse overall survivals of colon cancer patients were found in high stromal score group <ns0:ref type='bibr'>(2963 vs. 1930 days, log-rank test P=0.038</ns0:ref>) and high immune score group (2894 vs. 2230 days, log-rank test P=0.076). 563 up-regulated and 9 down-regulated genes were identified as stromal-immune score-related DEGs. 70 up-regulated DEGs associated with poor outcomes were identified by COX proportional hazard regression model, and 15 hub genes were selected later. Then, we verified aquaporin 4 (AQP4) and SNAP25 as prognostic biomarkers in GEO database. qRT-PCR results revealed that AQP4 and SNAP25 were significantly elevated in colon cancer tissues compared with adjacent normal tissues</ns0:p></ns0:div>
<ns0:div><ns0:head>Introduction</ns0:head><ns0:p>Colon cancer is the third most common malignant tumor and one of the leading causes of cancer-related death worldwide. Since risk factors have been investigated, earlier detection, cancer prevention, surgical techniques, radiotherapy and chemotherapy treatment have been improved, the incidence and mortality of colon cancer has been slowly declined. A model-based estimate showed that 104610 new cases of colon cancer would be diagnosed and 53200 patients would die of colon cancer in United States <ns0:ref type='bibr' target='#b16'>(Siegel et al. 2020)</ns0:ref>.However, the incidence of colorectal cancer in patients aged <50 years increased by 22% from 2000-2013 <ns0:ref type='bibr' target='#b15'>(Siegel et al. 2017)</ns0:ref>. Thus, more attention should be focused on early diagnosis through screening and accurately predicting the survival outcome of patients with colon cancer.</ns0:p><ns0:p>Stromal cells and immune cells form the major fraction of colon cancer tissue and are associated with tumor progress, inflammatory and metabolic disorders <ns0:ref type='bibr' target='#b3'>(Ghesquière et al. 2014;</ns0:ref><ns0:ref type='bibr' target='#b11'>Nilendu et al. 2018</ns0:ref>). An increasing amount of studies have highlighted the importance of stromal cells and immune cells in cancer biology and microenvironment across different types of cancers <ns0:ref type='bibr' target='#b1'>(Barros et al. 2018;</ns0:ref><ns0:ref type='bibr' target='#b2'>Garcia-Gomez et al. 2018;</ns0:ref><ns0:ref type='bibr' target='#b23'>Zhan et al. 2017)</ns0:ref>. ESTIMATE is a new method that infers stromal and immune cells in malignant tumors using gene expression signatures <ns0:ref type='bibr' target='#b22'>(Yoshihara et al. 2013)</ns0:ref>, and has been conducted in acute myeloid leukemia <ns0:ref type='bibr' target='#b20'>(Yan et al. 2019)</ns0:ref>, gastric cancer <ns0:ref type='bibr' target='#b17'>(Wang et al. 2019)</ns0:ref>, and glioblastoma <ns0:ref type='bibr' target='#b6'>(Jia et al. 2018)</ns0:ref>.</ns0:p></ns0:div>
<ns0:div><ns0:head>Abstract</ns0:head></ns0:div>
<ns0:div><ns0:head>  </ns0:head><ns0:p>In the current study, we obtained immune and stromal scores of colon cancer based on ESTIMATE. To help elucidate the stromal-immune score-based genes with prognostic value in colon cancer, we obtained gene expression dates from The Cancer Genome Atlas (TCGA), and verified the survival value in a different colon cancer cohort available from the Gene Expression Omnibus (GEO) database. Two hub genes were validated to be prognosis biomarkers and selected for further analysis. We investigated the potential underlying mechanisms of synaptosomal-associated protein 25 (SNAP25) in cancer-related signaling pathway, immunity and metabolism progresses through gene set enrichment analysis (GSEA) and TISIDB database.</ns0:p></ns0:div>
<ns0:div><ns0:head>Materials & Methods</ns0:head></ns0:div>
<ns0:div><ns0:head>Database and estimation of stromal and immune scores</ns0:head><ns0:p>The TCGA level 3 gene expression data and corresponding clinical information for colon cancer patients were obtained from Genomic Date Commons of the National Cancer Institute (http://portal.gdc.cancer.gov). Only patients with gene expression data, follow up information and clinicopathologic information were included in this study. For normalization, the RNA-seq data of all patients was transformed to transcripts per million (TPM) values (https://pubmed.ncbi.nlm.nih.gov/30379987/). The stromal and immune scores for each TCGA sample were conducted by R 3.6.2 using the R package 'estimate' (https://pubmed.ncbi.nlm.nih.gov/24113773/).</ns0:p><ns0:p>We obtained gene expression profiles and clinical information of 430 colon cancer patients from TCGA database. Among them, 231 (53.7%) cases were male and 199 (46.3%) cases were female. The average age of patients at initial pathological diagnosis was 66.3 years (range: 31-90 years). Histologic diagnosis included 369 (85.5%) cases of colon adenocarcinoma and 57 (13.3%) cases of colon mucinous adenocarcinoma, 4 (0.9%) cases were not classified. The tumor stage was stage I in 17.4%, stage II in 38.6%, stage III in 29.1, stage IV in 14.4% of cases, and 2 cases (0.5%) were of unknown stage. The tumors were located in the left (40%) or right (55.3%) colon according to their anatomic neoplasm subdivision, with 4.7% unknown. Based on ESTIMATE algorithm, we obtained stromal scores (range: -2262.07~1999.52) and immune scores (range: -954.97~3035.59) for all these colon cancer patients.</ns0:p></ns0:div>
<ns0:div><ns0:head>Correlations between clinicopathologic data and stromal/immune score</ns0:head><ns0:p>The correlations between clinicopathologic data and stromal/immune score were analyzed by SPSS 22.0 software (SPSS, Inc., Chicago, IL, USA). Patients with colon cancer were divided into high stromal/immune score (the fourth quartile) and low stromal/immune score groups (quartile 1-3). The stromal/immune score of different clinicopathologic groups was compared by Mann-Whitney U test, and overall survival was estimated by the Kaplan-Meier method and compared by log-rank tests. A value of P<0.05 was considered statistically significant.</ns0:p></ns0:div>
<ns0:div><ns0:head>Identification of differentially expressed genes (DEGs)</ns0:head><ns0:p>DEGs were identified based on immune and stromal scores (high stromal/immune score group vs. low stromal/immune score group) by Sanger_V1.0.8 software (https://shengxin.ren/ ).</ns0:p><ns0:p>Genes with log2 (fold change) >1.5 or <-1 combined with a P value <0.01 were defined as DEGs. The volcano plot of the DEGs was drawn by Sanger_V1.0.8 software, and the venn diagram was drawn on a website Venny 2.1.0 (https://bioinfogp.cnb.csic.es/tools/venny/index.html).</ns0:p></ns0:div>
<ns0:div><ns0:head>Gene ontology (GO) and pathway enrichment analyses</ns0:head><ns0:p>Cellular component (CC), molecular function (MF), biological process (BP) and pathway enrichment analyses were conducted using FunRich 3.1.3 (https://pubmed.ncbi.nlm.nih.gov/25921073/). A value of P<0.05 was considered as the screening condition.</ns0:p></ns0:div>
<ns0:div><ns0:head>Survival analysis</ns0:head><ns0:p>A COX proportional hazards model was applied to illuminate prognostic DEGs of colon cancer obtained from TCGA and a P value <0.01 was considered significant. An open source web tool PrognoScan (https://pubmed.ncbi.nlm.nih.gov/19393097/) was conducted to verify the survival outcomes between prognostic DEGs identified and colon cancer patients from GEO database.</ns0:p></ns0:div>
<ns0:div><ns0:head>Protein-protein interaction (PPI) network construction</ns0:head><ns0:p>The PPI network for DEG-encoded proteins was performed by STRING database (https://pubmed.ncbi.nlm.nih.gov/30476243/) and reconstructed by Cytoscape 3.7.2 (https://pubmed.ncbi.nlm.nih.gov/14597658/). The most significant modular analysis was identified by Molecular Complex Detection (MCODE) plugin of Cytoscape, and the plug-in Biological Networks Gene Oncology (BiNGO) of Cytoscape was applied to analysis GO term of hub genes. <ns0:ref type='table'>PDF | (2020:06:50322:1:1:NEW 11 Sep 2020)</ns0:ref> Manuscript to be reviewed</ns0:p></ns0:div>
<ns0:div><ns0:head>PeerJ reviewing</ns0:head></ns0:div>
<ns0:div><ns0:head>Heatmap and clustering analysis</ns0:head><ns0:p>Heatmap and clustering analysis were completed by 'heatmap' package.</ns0:p></ns0:div>
<ns0:div><ns0:head>Quantitative real-time PCR (qRT-PCR)</ns0:head><ns0:p>Total RNA was extracted with Trizol reagent (TaKaRa Bio Inc. Shiga, Japan) from colon cancer and adjacent normal tissues. cDNA was synthesized with PrimeScript TM RT reagent kit (TaKaRa, RR036A). Quantitative real-time PCR (qRT-PCR) was carried out using the TB Green TM Premix Ex Taq TM kit (TaKaRa, RR420A) on ABI step one Real-Time PCR system. The primers were as follows: AQP4, sense strand 5'-GAGCAGGAATCCTCTATC-3', antisense strand 5'-AGTGACATCAGTCCGTTT-3'; SNAP25, sense strand 5'-GTAGTGGACGAACGGGAGC-3', antisense strand 5'-CCTGTCGATCTGGCGATT-3'; GAPDH, sense strand 5'-GTCAACGGATTTGGTCTGTATT-3', antisense strand 5'-AGTCTTCTGGGTGGCAGTGAT-3'.</ns0:p><ns0:p>The Institutional Medical Ethics Review Board of Taizhou first people's hospital in Zhejiang Province approval to carry out the study within its facilities (Ethical Application Ref:</ns0:p><ns0:p>2019-KY009-03).</ns0:p></ns0:div>
<ns0:div><ns0:head>Gene set enrichment analysis (GSEA)</ns0:head><ns0:p>According to the expression level of AQP4 (or SNAP25), samples of the complete cohort from TCGA were divided into 2 groups, and implemented using GSEA by Sanger_V1.0.8 software. The KEGG gene set biological process database (c2.cp.kegg.v6) were chosen for enrichment analysis. Terms with both P value <0.01 and false discovery rate (FDR) <0.01 were identified.</ns0:p></ns0:div>
<ns0:div><ns0:head>Mining the immune-related mechanism of SNAP25.</ns0:head><ns0:p>TISIDB (https://pubmed.ncbi.nlm.nih.gov/14597658/) is a friendly web portal integrates 988 immune-related anti-tumor genes from 7 databases <ns0:ref type='bibr' target='#b14'>(Ru et al. 2019)</ns0:ref>. The correlations between immune features and any gene may be explored in 30 TCGA cancer types. In this study, the TISIDB database was used to investigate the associations between expression (or methylation) of SNAP25 and tumor-infiltrating lymphocytes (TILs). However, the relations between TILs and expression (or methylation) of AQP4 in colon cancer were not integrated in TISIDB database. A value of P<0.05 was considered statistically significant.</ns0:p></ns0:div>
<ns0:div><ns0:head>Results</ns0:head></ns0:div>
<ns0:div><ns0:head>The relationships between stromal/immune score and clinical features</ns0:head><ns0:p>The stromal and immune scores were variously distributed between adenocarcinoma and mucinous adenocarcinoma (Figure <ns0:ref type='figure'>1A, C</ns0:ref>). Both stromal score and immune score of colon adenocarcinoma cases were significantly lower than those of mucinous adenocarcinoma cases (P=0.001, 0.011, respectively). In addition, right colon tumors yielded higher immune scores than those left colon cases (Figure <ns0:ref type='figure'>1D</ns0:ref>, P<0.001), though no significant differences between left and right colon were found for the stromal scores (Figure <ns0:ref type='figure'>1B</ns0:ref>, P=0.818).</ns0:p><ns0:p>The potential association of overall survival and stromal/immune score was explored by classifying 430 colon cancer patients into high and low score groups based on their stromal or immune scores. As shown in Figure <ns0:ref type='figure'>1E</ns0:ref>, F, the median overall survival of patients with a low stromal score was longer than those in high score group (2963 vs. 1930 days, log-rank test P=0.038); consistently, the median overall survival of patients with a low immune score was longer than those in high score group (2894 vs. 2230 days, log-rank test P=0.076), although there was no statistically significant difference. However, patients with both a high stromal score and a high immune score were found to have significantly worse survival than those with low scores (1891 vs. 2974 days, log-rank test P=0.039) (Figure <ns0:ref type='figure'>1G</ns0:ref>).</ns0:p></ns0:div>
<ns0:div><ns0:head>Identification of DEGs by stromal and immune scores in colon cancer</ns0:head><ns0:p>After standardization of the RNA-Seq data for all 430 colon cancer patients obtained from TCGA database, we identified 4881 and 1512 DEGs based on stromal and immune scores, respectively. As shown in the volcano plots of DEGs for stromal/immune score (Figure <ns0:ref type='figure'>2A, B</ns0:ref>), 4791 genes were up-regulated and 90 genes were down-regulated for the comparison based on stromal score. Similarly, 1113 genes were up-regulated and 399 genes were down-regulated based on high immune score group vs. low immune score group. Through Venn diagram (Figure <ns0:ref type='figure'>2C</ns0:ref>, D) analysis, 563 shared up-regulated DEGs and 9 shared down-regulated DEGs from stromal score and immune score groups were identified and selected for subsequent analysis.</ns0:p></ns0:div>
<ns0:div><ns0:head>Functional and pathway enrichment analyses</ns0:head><ns0:p>Go and pathway enrichment analyses of the above 572 genes were performed (Figure <ns0:ref type='figure'>2E, F, G, H</ns0:ref>). For CC, DEGs were mainly associated with plasma membrane, extracellular and extracellular space. With regard to MF, genes were mainly clustered in receptor activity, cell adhesion molecule activity and B cell receptor activity. DEGs in the BP category primarily enriched in immune response, cell communication and signal transduction. The pathway enrichment analysis showed genes were mainly enriched in epithelial-to-mesenchymal transition, peptide ligand-binding receptors and GPCR ligand binding.</ns0:p></ns0:div>
<ns0:div><ns0:head>Identification of prognostic DEGs in colon caner</ns0:head><ns0:p>The COX proportional hazard regression model was constructed to identify potential prognostic DEGs in colon cancer. Among the 563 shared up-regulated DEGs and 9 shared downregulated DEGs, 70 up-regulated DEGs associated with poor outcomes were shown in Figure <ns0:ref type='figure'>3</ns0:ref>.</ns0:p></ns0:div>
<ns0:div><ns0:head>PPI network construction and hub gene analysis among prognostic DEGs</ns0:head><ns0:p>To further explore the interplay among the 70 identified prognostic DEGs, we conducted a PPI network containing 30 nodes and 53 edges based on STRING tool and Cytoscape software (Figure <ns0:ref type='figure' target='#fig_5'>4A</ns0:ref>). Module analysis using MCODE was constructed, and 15 hub genes were selected (Figure <ns0:ref type='figure' target='#fig_5'>4A</ns0:ref>). CC, MF, BP analyses of the total 15 hub genes were performed using BiNGO (Figure <ns0:ref type='figure' target='#fig_5'>4B</ns0:ref>). The 15 hub genes were mainly associated with plasma membrane, transporter activity, secretion and channel activity.</ns0:p></ns0:div>
<ns0:div><ns0:head>Heatmap and clustering analysis of 15 hub genes</ns0:head><ns0:p>The expression level of 15 hub genes in 'dead' and 'alive' groups was shown in Figure <ns0:ref type='figure' target='#fig_5'>4C</ns0:ref>.</ns0:p></ns0:div>
<ns0:div><ns0:head>Verifying the survival outcomes of hub genes in the GEO database</ns0:head><ns0:p>The survival outcomes of 15 hub genes were identified on the PrognoScan web tool, which provided overall survival of GSE12945, GSE17536 and GSE17537 datasets for colorectal cancer. Then, AQP4 and SNAP25 were verified (Figure <ns0:ref type='figure'>5</ns0:ref>) to be significantly associated with overall survival according to both the log-rank test and COX proportional hazards regression analysis (all P <0.05).</ns0:p></ns0:div>
<ns0:div><ns0:head>The expression of two hub genes in 20 colon cancer and adjacent normal tissue sample set using qRT-PCR.</ns0:head><ns0:p>The expression of AQP4 and SNAP25 was validated in 20 pairs of clinical tissues using qRT-PCR. Interestingly, the mRNA relative expression levels of both AQP4 and SNAP25 were significantly elevated in colon cancer tissues compared with adjacent normal tissues (P=0.003, 0.001) (Figure <ns0:ref type='figure' target='#fig_7'>6</ns0:ref>).</ns0:p></ns0:div>
<ns0:div><ns0:head>GSEA using TCGA database</ns0:head><ns0:p>To further investigate the underlying mechanism of AQP4 and SNAP25 in colon cancer, KEGG pathway enrichment analysis was performed by GSEA. For AQP4, 'vascular smooth muscle contraction' (NES=2.17, P<0.001, FDR=0.006) gene set was prominently enriched. For SNAP25, 57 gene sets were enriched, including 11 gene sets which were cancer-related processes (Figure <ns0:ref type='figure' target='#fig_8'>7</ns0:ref>). Besides, high expression of SNAP25 might also be involved in 'B cell receptor signaling pathway' (NES=2.06, P=0.002, FDR=0.002), 'cell adhesion molecules cams' (NES=2.05, P=0.008, FDR=0.002), 'chemokine signaling pathway' (NES=2.07, P=0.002, FDR=0.002), 'complement and coagulation cascades' (NES=2.08, P<0.001, FDR=0.002), 'T cell receptor signaling pathway' (NES=2.06, P<0.001, FDR=0.002), 'adipocytokine signaling pathway' (NES=1.96, P<0.001, FDR=0.005), 'aldosterone regulated sodium reabsorption' (NES=2.02, P<0.001, FDR=0.003), 'glycosaminoglycan biosynthesis heparan sulfate' (NES=2.00, P<0.001, FDR=0.004), and 'insulin signaling pathway' (NES=1.94, P=0.002, FDR=0.007). This suggested that immunity and metabolism may be as well involved in the underlying mechanism of SNAP25 in colon cancer.</ns0:p></ns0:div>
<ns0:div><ns0:head>Regulation of immune molecules by SNAP25</ns0:head><ns0:p>The spearman's correlations between lymphocytes and expression, methylation of SNAP25 were performed in TISIDB database (Figure <ns0:ref type='figure'>8</ns0:ref>). The associations between the expression of SNAP25 and immune-related signatures of TILs types were shown in Figure <ns0:ref type='figure'>8A</ns0:ref>, and the greatest correlations including natural killer cell (NK; r=0.493, P<2.2e-16), macrophage (r=0.45, P<2.2e-16), mast cell (r=0.448, P<2.2e-16), and natural killer T cell (NKT; r=0.447, P<2.2e-16) were shown in Figure <ns0:ref type='figure'>8B</ns0:ref>. The correlations between methylation of SNAP25 and lymphocytes were shown in Figure <ns0:ref type='figure'>8C</ns0:ref>, and Figure <ns0:ref type='figure'>8D</ns0:ref> displayed the remarkable negative correlations including plasmacytoid dendritic cell (pDC; r=-0.419, P=2.96e-14), type 1 T helper cell (Th1; r=-0.406, P=4.06e-13), T follicular helper cell <ns0:ref type='bibr'>(Tfh;</ns0:ref>. Therefore, the potential underlying mechanism of SNAP25 in colon cancer may be involved in the regulation of the above TILs.</ns0:p></ns0:div>
<ns0:div><ns0:head>Discussion</ns0:head><ns0:p>Solid tumor tissues comprise not only malignant cells but also tumor microenvironment, including immune cells, stromal cells, epithelial cells, fibroblasts, vascular cells and signaling molecules. Accumulating evidence clarifies that tumor microenvironment plays a crucial role in tumor growth, progression, metastasis, prognosis, and treatment <ns0:ref type='bibr' target='#b13'>(Petitprez et al. 2018;</ns0:ref><ns0:ref type='bibr' target='#b19'>Wu & Dai 2017)</ns0:ref>. In the current study, we focused on stromal and immune scores, which reflect the microenvironment of tumor and hence contribute to survival prediction in colon cancer.</ns0:p><ns0:p>Meanwhile, our results were in accordance with previous specific insights <ns0:ref type='bibr' target='#b8'>(Koi & Carethers 2017;</ns0:ref><ns0:ref type='bibr' target='#b24'>Zhang et al. 2018)</ns0:ref>, and might provide extra data in the mining of interaction between tumor and environment in colon cancer.</ns0:p><ns0:p>Then, we screened out 572 microenvironment-related DEGs, and found they were mainly enriched in plasma membrane (CC), receptor activity (MF), immune response (BP), and epithelial-to-mesenchymal transition (pathway). Afterwards, AQP4, ASTN1, ATP2B3, CADM3, CD22, CD37, CD79B, CLVS2, CXCR5, FCRLA, GABRG2, MS4A1, NRXN1, SNAP25, and SYT4 were identified as prognostic hub genes, and two of them were verified to be prognosis biomarkers in GEO database. qRT-PCR results revealed that AQP4 and SNAP25 were significantly elevated in colon cancer tissues compared with adjacent normal tissues (P=0.003, 0.001). Next, we investigated the underlying mechanism of AQP4 and SNAP25 in colon cancer by GSEA, and found that the high expression of SNAP25 might be involved in cancer-related signaling pathway, immunity and metabolism processes. Further researches in TISIDB database indicated greatest positive correlations between SNAP25 expression and TILs (NK, NKT, macrophage, mast), and greatest negative correlations between SNAP25 methylation and TILs (pDC, Th1, Tfh, NKT). TILs are associated with prognosis for the survival of various tumors in many previous studies. TILs are also reported as a predictive biomarker in colon cancer <ns0:ref type='bibr' target='#b25'>(Zhao et al. 2019)</ns0:ref>. However, some studies found no significant association between TILs and overall survival. It remains controversial on the prognostic value of TILs in colon cancer may be due to different TIL responses or subsets, different patient populations, different study designs, and the heterogeneity of immune infiltrate. Further studies should be designed to provide more insight into the study of TILs and survival. SNAP25 belongs to the soluble N-ethylmaleimide-sensitive factor attachment protein receptor (SNARE) complex, which is essential for neurotransmitter release, synaptic, secretory vesicle exocytosis, minimal fusion machinery, cell-to-cell signaling, and the regulation of ion channels <ns0:ref type='bibr' target='#b0'>(Baker & Hughson 2016;</ns0:ref><ns0:ref type='bibr' target='#b18'>Wang et al. 2017;</ns0:ref><ns0:ref type='bibr' target='#b21'>Yoon & Munson 2018)</ns0:ref>. As reported, SNAP25 is potentially important for normal vesicle fusion and lysosomal trafficking <ns0:ref type='bibr' target='#b9'>(Manca et al. 2014;</ns0:ref><ns0:ref type='bibr' target='#b10'>Mu et al. 2018)</ns0:ref>. Kyoko Kobayashi et al <ns0:ref type='bibr' target='#b7'>(Kobayashi et al. 2016)</ns0:ref> found that SNAP25 protein was expressed in 46% (77/168) of diffuse large B-cell lymphoma (DLBCL) patients and associated with CD5 expression (P=0.018). Huang CJ et al <ns0:ref type='bibr' target='#b5'>(Huang et al. 2017</ns0:ref>) displayed the expression and functional significance of SNAP25 in medulloblastoma. Manuscript to be reviewed Although the clinical significances and functions of SNAP25 for colon cancer have not been previously reported, it may serve as prognosis biomarker according to this study. To our greatest interest, SNAP25 is the most highly interconnected nodes (Figure <ns0:ref type='figure' target='#fig_5'>4A</ns0:ref>), and involves in many cancer-related processes consists of adherens junction, calcium signaling pathway, cytokine receptor interaction, ECM receptor interaction, hedgehog signaling pathway, JAK-STAT signaling pathway, MAPK signaling pathway, TGF beta signaling pathway, Toll-like receptor signaling pathway, VEGF signaling pathway, and WNT signaling pathway (Figure <ns0:ref type='figure' target='#fig_8'>7</ns0:ref>).</ns0:p><ns0:p>Besides, SNAP25 is also involved in immunity and metabolism processes such as B cell receptor signaling pathway, cell adhesion molecules cams, chemokine signaling pathway, complement and coagulation cascades, T cell receptor signaling pathway, adipocytokine signaling pathway, aldosterone regulated sodium reabsorption, glycosaminoglycan biosynthesis heparan sulfate and insulin signaling pathway. Those may bring novel insights into the potential underlying mechanism of SNAP25 in colon cancer. In addition, this paper reveals the significant correlations between SNAP25 and lymphocytes (NK, macrophage, mast cell, NKT), which indicates the potential association of tumor microenvironment and SNAP25. Recently, a lot of attention has been paid to tumor microenvironment and immune evasion for future diagnoses and treatments of malignant tumors. Immunotherapies have been revolutionizing tumor treatment, although immunological response in different patients is heterogeneous. NK cells participate in tumor immunosurveillance, and are one of the most promising therapies for various types of cancer. However, NK cell populations may shape with altered reactivity in malignant tumors <ns0:ref type='bibr' target='#b4'>(Hofer & Koehl 2017)</ns0:ref>. In colorectal cancer, a high level of mast cells was confirmed with poor survival, and the density of innate immune cells (macrophages, mast cells, neutrophils, and immature dendritic cells) increased with tumor stage <ns0:ref type='bibr' target='#b8'>(Koi & Carethers 2017)</ns0:ref>. Therefore, SNAP25 is extremely closely associated with various types of TILs, and has the potential to serve as a prognosis biomarker and an immunotherapeutic target for colon cancer.</ns0:p><ns0:p>However, our study presents a number of limitations. One major limitation is that the present research was mainly based on previous data from TCGA and GEO, therefore, future investigations in vivo and vitro are needed to investigate the effect of SNAP25 in colon cancer.</ns0:p><ns0:p>A second issue is that since the integrated bioinformatics analysis was focus on immune microenvironment for colon cancer, and the DEGs were identified based on immune and stromal scores, there was no validation for these DEGs based on control samples. Finally, we are lacking of the specimens of colon cancer and adjacent normal tissues. So, we only validate the expressions of AQP4 and SNAP25 in 20 pairs of tissues by qRT-PCR.</ns0:p></ns0:div>
<ns0:div><ns0:head>Conclusions</ns0:head><ns0:p>In summary, SNAP25 is a microenvironment-related and immune-related gene that can predict poor outcomes in colon cancer. Bioinformatic analysis suggests that SNAP25 is involved in cancer-related signaling pathway, immunity and metabolism processes, which may provide a new target for investigating the underling mechanism of colon cancer. Manuscript to be reviewed Manuscript to be reviewed Manuscript to be reviewed </ns0:p><ns0:note type='other'>Figure 4</ns0:note><ns0:note type='other'>Figure 6</ns0:note><ns0:note type='other'>Figure 7</ns0:note></ns0:div><ns0:figure xml:id='fig_0'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:06:50322:1:1:NEW 11 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_1'><ns0:head>Figure 1 Figure 1 .</ns0:head><ns0:label>11</ns0:label><ns0:figDesc>Figure 1</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_2'><ns0:head>Figure 2 Figure 2 .</ns0:head><ns0:label>22</ns0:label><ns0:figDesc>Figure 2</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_3'><ns0:head>(</ns0:head><ns0:label /><ns0:figDesc>A) Volcano plot showing up-regulated DEGs in red and down-regulated DEGs in blue for the comparison based on high and low stromal score groups. (B) Volcano plot showing upregulated DEGs in red and down-regulated DEGs in green for the comparison based on high and low immune score groups. (C, D) Venn diagrams showing 563 shared up-regulated DEGs (C) and 9 shared down-regulated DEGs (D) from stromal score and immune score groups. (E, F, G, H) Top six Go terms and pathways enriched by DEGs.</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_4'><ns0:head>Figure 3 Figure 3 .</ns0:head><ns0:label>33</ns0:label><ns0:figDesc>Figure 3</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_5'><ns0:head>Figure 4 .</ns0:head><ns0:label>4</ns0:label><ns0:figDesc>Figure 4. PPI network, GO analysis and expression of hub gene.</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_6'><ns0:head>Figure 5 Figure 5 .</ns0:head><ns0:label>55</ns0:label><ns0:figDesc>Figure 5</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_7'><ns0:head>Figure 6</ns0:head><ns0:label>6</ns0:label><ns0:figDesc>Figure 6 The relative mRNA expression levels of AQP4 (A) and SNAP25 (B) in colon cancer tissues and adjacent normal tissues were confirmed by qRT-PCR.</ns0:figDesc><ns0:graphic coords='23,42.52,199.12,525.00,193.50' type='bitmap' /></ns0:figure>
<ns0:figure xml:id='fig_8'><ns0:head>Figure 7 .</ns0:head><ns0:label>7</ns0:label><ns0:figDesc>Figure 7. GSEA enrichment plots showed that eleven gene sets related to tumor signaling pathways were enriched in the high SNAP25 expression group.</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_9'><ns0:head>Figure 8 Figure 8 .</ns0:head><ns0:label>88</ns0:label><ns0:figDesc>Figure 8</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_10'><ns0:head>(</ns0:head><ns0:label /><ns0:figDesc>A) Relations between the SNAP25 expression and abundance of TILs across human cancers. (B) Top 4 greatest positive correlations between SNAP25 expression and TILs. (C) Relations between the SNAP25 methylation and abundance of TILs across human cancers. (D) Top 4 greatest negative correlations between SNAP25 methylation and TILs.</ns0:figDesc></ns0:figure>
</ns0:body>
" | "Dear Prof. Uversky
Thank you very much for your letter and for the reviewers’ comments concerning our manuscript entitled “Mining the potential prognostic value of synaptosomal-associated protein 25 (SNAP25) in colon cancer based on stromal-immune score” (ID: 50322).
We have revised the paper, and would like to re-submit it for your consideration. We have addressed the comments raised by the reviewer, and the amendments are highlighted in red in the revised manuscript. The responds to the reviewer’s comments are as flowing:
Responds to the reviewer’s comments:
Reviewer 1:
Comment: The current manuscript described bioinformatic analysis of the correlation of several protein expression with prediction of colon cancer. However, actual experiment to confirm the current study is missing, which greatly reduced the power of the study. The author should find some stained slides or perform WB or qPCR to confirm the current findings.
Response: We appreciate this important comment. It is really true as reviewer suggested that actual experiment to confirm the current study is necessary. However, we are lacking of the specimens of colon cancer and adjacent normal tissues. So, we only validate the expressions of AQP4 and SNAP25 in 20 pairs of tissues by qPCR (Figure 6). Line 130-142, Page 5; Line 221-226, Page 8.
Reviewer 2:
Comment: Lot of tools and software are not referenced in peerj format, change them to the required format.
Response: Thank you for your suggestion. We have changed most of tools and softwares to the required format, such as R package, FunRich, PrognoScan, STRING, Cytoscape, and TISIDB. However, two of them (Sanger_V1.0.8 software and Venny 2.1.0) are not reported in pubmed database.
Comment: Relationship between TILs and Survival should be outlined more to drive the point of why correlation between TILs and gene is important.
Response: Thanks for this important and constructive comment. According to reviewer’s instruction, the relationship between TILs and survival has been added into the Discussion section. Line 276-282, Page 9-10.
Comment: Conclusion should also include with limitation of the methods used.
Response: We appreciate this important suggestion. The limitation of the methods used has been added into the last paragraph of Discussion section. Line 317-324, Page 11.
Comment: Database and estimation of stromal and immune scores - lacks information about the data such as filters to obtain specific data, you should shift the description of the data from this subheading 'The relationships between stromal/immune score and clinical features' to here
Response: Thanks for this important comment. We have shifted the description of the data from this subheading 'The relationships between stromal/immune score and clinical features' to that subheading “Database and estimation of stromal and immune scores“ as the reviewer suggested. Line 85-94, Page 3.
Comment: GEO data- why was only two GSE series was chosen? Also it would be good to show all the 15 genes survival outcomes of hub genes thus your selection of the two genes is justified
Response: This is a very meaningful advice. We verified the survival outcomes of 15 hub genes on the PrognoScan web tool based on GEO database. However, this web tool only provides overall survival of GSE12945, GSE17536 and GSE17537 datasets for colorectal cancer. It is really true as reviewer suggested that show all the 15 genes survival outcomes would be good to justified the selection of two genes. We have redrawn figure 5 according to the reviewer’s suggestion. Line 216-218, Page 7.
Comment: There is an information gap of how the RNASeq data was normalized?
Response: We appreciate this important suggestion. For normalization, the RNA-seq data of all patients was transformed to transcripts per million (TPM) values (https://pubmed.ncbi.nlm.nih.gov/30379987/). Line 80-82, Page 3.
Comment: why was the DEG from RNAseq (using counts or RPKM) values was also not considered to see how the 15 hub gene are expressed?
Response: Thank you for your comment. The expression level of 15 hub genes in “dead” and “alive” groups was shown in Figure 4C. Line 128-129, Page 5; Line 213-214, Page 7.
Comment: Also it would be good to instead information on the control samples that shows the DEGs are really DEGs in normal.
Response: We appreciate this important suggestion. It should be better according to the reviewer’s advice to instead information on the control samples. However, this paper was focus on immune microenvironment for colon cancer, and the DEGs were identified based on immune and stromal scores, there was no validation for these DEGs based on control samples. We are very sorry that we failed to show really DEGs in normal. Thank you very much for the reviewer’s valuable comment, we may greatly benefit from it when considering further study in the future. Line 320-322, Page 11.
Comment: GSEA could have been done on all 15 hub gene, thus APQ4 and SNAP25 are prominent in survival.
Response: This is an important suggestion. As Figure 5 shows, AQP4 and SNAP25 were prominent in survival, so further investigation (qRT-PCR and GSEA) was done only on these two hub genes.
Comment: The manuscripts lacks the clarity of how only APQ4 and SNAP25 are best in survival outcomes - compared to other 13?
Response: Thanks for this important comment. It has been added into the paper (in the revised Figure 5). Line 216-220, Page 7-8.
Comment: The transition from two gene APQ4 and SNAP25 to only SNAP25 focus is unclear? The GSEA was done on both, why was APQ4 not investigated further?
Response: Thank you for your comment. I am sorry that this part was not clear in the original manuscript. The relations between TILs and expression (or methylation) of AQP4 in colon cancer were not integrated in TISIDB database, so AQP4 was not investigated further. Line 154-156, Page 5.
Comment: The usage of stormal/immune scores for 430 colon cancer patients is concluded with no statistically significant difference. This contradicts the usage of these score and the dataset? (line 145-146)
Response: We appreciate this important suggestion. As we mentioned in the paper, the median overall survival of patients with a low stromal score was longer than those in high score group (2963 vs. 1930 days, log-rank test P=0.038); consistently, the median overall survival of patients with a low immune score was longer than those in high score group (2894 vs. 2230 days, log-rank test P=0.076), although there was no statistically significant difference. However, patients with both a high stromal score and a high immune score were found to have significantly worse survival than those with low scores (1891 vs. 2974 days, log-rank test P=0.039) (Figure 1G). Generally, the usage of stromal/immune scores for 430 colon cancer patients is concluded with statistically significant difference. Line 182-184, Page 6.
Comment: There are minor errors to correct: - Line 50: please remove the 2020 in front of the literature citation; - Line 89: please change flod to fold
Response: Thank you for your patience. We have revised in the paper. Line 54, Page 2; Line 105, Page 4.
Comment: - Line 158-163: please include the long form of CC, MF, BP (cellular component, molecular function, biological process/ biological pathway) because this is the first time the short form is used in the main text
Response: Thanks for your suggestion. However, we have included the long form of CC, MF, BP in Line 110.
Comment: One suggestion to improve data interpretation:
- Figure 5: it is not explained what the dotted blue/ red lines are. Are they the standard deviation of the Kaplan-Meier plot? Please indicate it on the Figure description or the main text.
Response: Thanks for this important comment. The dotted blue/ red lines indicated 95% confidence intervals of overall survival probability. The information has been incorporated into the Figure description according to reviewer’s instructions.
We would like to express our great appreciation to you and reviewer for comments on our paper. Looking forward to hearing from you.
Sincerely,
Sha sha Cai
E-mail: tyycss@163.com
" | Here is a paper. Please give your review comments after reading it. |
9,788 | "<ns0:abstract xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:p>Recent marine spatial planning efforts, including the management and monitoring of marine protected areas (MPAs), increasingly focus on the importance of stakeholder engagement. For nearly 15 years, the California Collaborative Fisheries Research Program (CCFRP) has partnered volunteer anglers with researchers, the fishing industry, and resource managers to monitor groundfishes in California's network of MPAs. While the program has succeeded in generating sustained biological observations, we know little about volunteer angler demography or the impact of participation on their perceptions and opinions on fisheries data or MPAs. In this study we surveyed CCFRP volunteers to learn about (a) volunteer angler demographics and attitudes toward groundfish management and stock health, (b) volunteer angler motivations for joining and staying in the program, and (c) whether participation in the program influenced volunteer angler opinions on MPAs and the quality of data used in resource management in California. CCFRP volunteers were older and had higher fishing avidity than average within the California recreational angling community. Many self-identified as more conservation-minded than their peers and had positive views of California groundfish management and stock health. Participation in science and giving back to fisheries resources were major motivating factors in their decision to become and remain CCFRP volunteers. Angler opinions toward MPAs were more positive after volunteering with CCFRP. Those who had volunteered for seven or more years with CCFRP were more likely than not to gain a positive opinion of MPAs. Our survey results provide evidence that long-term engagement of stakeholders in collaborative research positively influences stakeholder opinions regarding marine resource management, and highlights CCFRP's success in engaging citizen science stakeholders in collaborative fisheries research.</ns0:p></ns0:div>
</ns0:abstract>
<ns0:body xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:head>Introduction</ns0:head><ns0:p>Stakeholder engagement is an important part of marine resource protection and management <ns0:ref type='bibr' target='#b28'>(Pomeroy & Douvere, 2008)</ns0:ref>. Benefits of such engagement include the incorporation of local knowledge into policy, and the potential to build stakeholder trust in management decisions <ns0:ref type='bibr' target='#b42'>(Yochum, Starr & Wendt, 2011</ns0:ref>). California's Marine Life Protection Act (MLPA) of 1999 <ns0:ref type='bibr' target='#b38'>(Fish and</ns0:ref><ns0:ref type='bibr'>Game Code § 2850-2863)</ns0:ref> directed the state to redesign California's marine protection areas (MPAs) to function as a network and increase protection of the state's sense of ownership and trust among participants in the data being collected <ns0:ref type='bibr' target='#b42'>(Yochum, Starr & Wendt, 2011)</ns0:ref>.</ns0:p><ns0:p>The CCFRP program capitalizes on the expertise and knowledge of the fishing industry, angling public, participating scientists, and managers; together, these constituents work toward a common goal (e.g., measuring MPA effectiveness) that ultimately gives the group a shared purpose <ns0:ref type='bibr' target='#b40'>(Wendt & Starr, 2009;</ns0:ref><ns0:ref type='bibr' target='#b42'>Yochum, Starr & Wendt, 2011)</ns0:ref>. CCFRP relies on relationship building and transparency to create buy-in of MPA monitoring, evaluation, and management by involving stakeholders in all aspects of the program, from study design to data collection and sharing. An example of these efforts is the annual Volunteer Appreciation and Data Workshop hosted by the coordinating staff from California Polytechnic University, San Luis Obispo (Cal Poly) and Moss Landing Marine Labs (MLML), where survey results and trip highlights are shared with volunteer participants and other partners. To date, the impact of these events on angler opinions of MPAs has not been evaluated.</ns0:p><ns0:p>Human dimensions such as effective engagement, honesty, trust, and transparency can impact the success of MPAs <ns0:ref type='bibr' target='#b6'>(Gall & Rodwell, 2016;</ns0:ref><ns0:ref type='bibr' target='#b24'>Ordoñez-Gauger et al., 2018)</ns0:ref>. However, research on public knowledge, attitudes, and perceptions of California's MPA network is sparse, and studies that do exist vary widely across geographic region and composition of study populations <ns0:ref type='bibr' target='#b1'>(Baldassare et al., 2007</ns0:ref><ns0:ref type='bibr' target='#b0'>(Baldassare et al., , 2017;;</ns0:ref><ns0:ref type='bibr' target='#b18'>Loper, 2008;</ns0:ref><ns0:ref type='bibr' target='#b24'>Ordoñez-Gauger et al., 2018)</ns0:ref>. While the success of CCFRP in generating valuable monitoring data is clear <ns0:ref type='bibr' target='#b40'>(Wendt & Starr, 2009;</ns0:ref><ns0:ref type='bibr' target='#b33'>Starr et al., 2015)</ns0:ref>, the degree to which CCFRP participation has influenced volunteer perceptions of California's MPAs is less clear. In addition, the essential context for understanding this influence -(a) the demographics and characteristics of the CCFRP volunteer anglers and (b) their perceptions on the health of groundfish stocks, the data quality used to manage those stocks, and the effectiveness of MPAs relative to groundfish management measures (e.g. depth restrictions, bag limits, size limits) --has also not been assessed. Although California's network of MPAs were not specifically designed as a fishery management tool, any beneficial fisheries impacts of MPAs are important in the evaluation of overall MPA effectiveness. Thus, state resource managers would be well served by learning about the volunteer anglers who help monitor California's MPAs as well as their respective opinions on MPAs.</ns0:p><ns0:p>Given the longevity of CCFRP, and extensive volunteer participation, the program provides a valuable opportunity to measure outcomes of long-term stakeholder engagement. We used an online survey of current and former CCFRP volunteer anglers to learn about (a) volunteer angler demographics and attitudes toward groundfish management and stock health, (b) volunteer angler motivations for joining and staying in the program, and (c) whether participation in the program influenced volunteer angler opinions on MPAs and the quality of data used in resource management in California. By characterizing the population of CCFRP angler volunteers and their perceptions in relation to their volunteer efforts, our intent is to characterize the realized benefits of CCFRP as a collaborative research program, beyond the fisheries data it yields.</ns0:p></ns0:div>
<ns0:div><ns0:head>Materials & Methods</ns0:head></ns0:div>
<ns0:div><ns0:head>Survey</ns0:head><ns0:p>We distributed an online survey to 722 volunteer anglers who participated in CCFRP with Cal Poly and MLML between 2007 and 2018. This group represented a subgroup of the entire volunteer population during that time period (N=901), as 179 volunteer anglers had previously opted out of receiving communications from CCFRP and two other groups, individuals without e-mail addresses and anglers under the age of 18 years, were not contacted.</ns0:p><ns0:p>We used Qualtrics, an online survey platform, to deliver the survey questionnaire. Respondents We distributed the survey via a series of e-mails sent to subjects over a two-week period in Spring 2018. The first e-mail invited subjects to participate in the survey, and two subsequent e-mails sent seven and 12 days into the study period reminded subjects to complete the survey.</ns0:p><ns0:p>Each e-mail contained a description of the study, a letter of consent, and a link to the online questionnaire.</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:04:47639:1:0:NEW 30 Jul 2020)</ns0:p></ns0:div>
<ns0:div><ns0:head>Manuscript to be reviewed</ns0:head></ns0:div>
<ns0:div><ns0:head>Acceptable survey response rate and margin of error</ns0:head><ns0:p>An acceptable survey response rate for categorical response surveys is dependent in part on the confidence level and the maximum margin of error that the surveyor is willing to accept <ns0:ref type='bibr' target='#b2'>(Bartlett, Kotrlik & Higgins, 2001)</ns0:ref>. We calculated and report the acceptable minimum survey response rates according to the minimum sample sizes needed for an allowance of 5% and 10% margins of error (MOE 95 ) as described in <ns0:ref type='bibr' target='#b2'>Bartlett, Kotrlik & Higgins (2001)</ns0:ref>. Five and 10% MOE 95 are equivalent to ± 0.25 and 0.5 points respectively, on a categorical ordinal response scale from 1 to 5.</ns0:p></ns0:div>
<ns0:div><ns0:head>Measures taken to address potential limitations with survey design</ns0:head><ns0:p>Potential limitations with the survey design included 1) the possibility a biased population of CCFRP volunteer anglers responded to the survey (i.e. nonresponse bias) <ns0:ref type='bibr' target='#b2'>(Bartlett, Kotrlik & Higgins, 2001)</ns0:ref>, 2) potential bias in respondent responses due to surveyor association with the CCFRP program (i.e. response bias) and 3) reliance on a respondent's ability to accurately recall the history and influence on their opinion change or lack thereof (e.g., response bias due to reflexive counterfactual study design) <ns0:ref type='bibr'>(Smallhorn-West et al. 2009)</ns0:ref>.</ns0:p><ns0:p>Due to the survey being anonymous, we could not test for nonresponse bias. However, steps taken to help limit nonresponse bias (i.e. increase survey response rate) included providing (a) respondent anonymity, (b) a short survey (e.g., ~ 15 minute completion time) (c) a seamless online submission format, and (d) reminder emails. In addition, the survey questions provided a means to check whether respondents represented an unexpected demographic (e.g., mostly young anglers) as well as to compare the distribution of MPA opinion change responses by age, angler avidity, conservation-mindedness, level of engagement, etc.</ns0:p><ns0:p>With respect to potential surveyor influence on responses, the solicitation and reminder emails were sent via CCFRP field technicians (to keep volunteer emails confidential), but subjects were informed the survey itself was independently formulated by researchers at Scripps Institution of Oceanography, UC San Diego.</ns0:p><ns0:p>To address caveats associated with the reflexive counterfactual study design, questions related to opinion change and volunteer participation were included in separate sections of the survey so that these responses were made independent of each other. Although we could not measure how respondent opinions on MPAs would have changed had they not participated in PeerJ reviewing PDF | (2020:04:47639:1:0:NEW 30 Jul 2020)</ns0:p><ns0:p>Manuscript to be reviewed CCFRP, we were able to test how different levels of participation with CCFRP related to angler change in opinion (see Measures of Volunteer Participation and Opinion Change). The longest time frame for which a respondent was asked to recall their opinions was dependent on the length of time since joining CFFRP, which at most, was 11 years <ns0:ref type='bibr'>(e.g., 2007 -2017)</ns0:ref>.</ns0:p></ns0:div>
<ns0:div><ns0:head>Angler demographics and characteristics</ns0:head><ns0:p>Age and gender comprised the survey's demographic categories; other characteristics included years of fishing experience, frequency of fishing, degree of conservation-mindedness, whether anglers had any prior work experience in marine resource management or the recreational or commercial fishing industries, whether anglers had fished in MPA sites prior to those areas being designated MPAs, and whether anglers had ever participated in the MLPA planning process. We categorized angling avidity (i.e. a relative measure of the enthusiasm an angler has for the sport) into three avidity levels -low, medium, or high -based on the number of saltwater angling trips they took per year, outside of CCFRP surveys, with low being <4 days, medium 4-23 days, and high >23 days per year. Angler avidity ranges were based on the National Oceanographic and Atmospheric Administration (NOAA) West Coast Fishing Avidity categories <ns0:ref type='bibr' target='#b31'>(Rubio, Brinson & Wallmo, 2014)</ns0:ref>.</ns0:p></ns0:div>
<ns0:div><ns0:head>Volunteer perceptions of groundfish management and stock health</ns0:head><ns0:p>In addition to characterizing the general opinion of anglers on the health of groundfish stocks and the effectiveness of specific regulations, we also compared the percentage of respondent opinions across related work experience categories to gauge the relative degree of consensus in opinions among these groups. Work experience categories included marine resource management, commercial or recreational fishing industry, and no experience.</ns0:p></ns0:div>
<ns0:div><ns0:head>Measures of volunteer participation and opinion change</ns0:head><ns0:p>Our primary focus for assessing change in CCFRP volunteer angler opinions was on MPAs in California, and secondarily, on the quality of data used for fishery management. Given that MPAs are known to elicit stronger opinions with the angling public than fishery management data, questions regarding fishery management data were limited to whether PeerJ reviewing PDF | (2020:04:47639:1:0:NEW 30 Jul 2020)</ns0:p><ns0:p>Manuscript to be reviewed volunteers had an opinion of these data before CCFRP participation, and of those that did, whether their opinion changed either positively or negatively after having volunteered with CCFRP.</ns0:p><ns0:p>With respect to MPAs, we were interested in (a) capturing the distribution of opinions on MPAs both before and after CCFRP participation, (b) characterizing the extent of opinion change across the group, and (c) examining whether opinion change is mediated by the extent of program participation. Respondents answered questions on an ordinal scale. They could report an opinion of 'Positive,' 'Somewhat positive,' 'Somewhat negative,' 'Negative,' or 'No opinion.' We coded the answers 1, 2, 3, 4, and 5 respectively.</ns0:p><ns0:p>To capture the overall proportion of respondents having a change in opinion on MPAs after volunteering with CCFRP, we subtracted the answer code corresponding to their opinion after CCFRP participation from the answer code corresponding to their opinion before volunteering with CCFRP. Results that were positive indicated a positive change in opinion of MPAs, results that were negative indicated a negative change in opinion of MPAs, and results that were '0' represented no change in opinion. We coded these differences numerically into a single variable representing change in opinion, with positive change coded as '1', no change coded as '2,' and negative change coded as '3.' We report the survey estimate ± MOE 95 (i.e. confidence interval; CI) for the proportions of positive MPA opinions before and after CCFRP participation based on a 95% confidence level (CI 95 ) <ns0:ref type='bibr' target='#b8'>(Gilliland & Melfi, 2010)</ns0:ref>. We calculated the CI 95 for our MPA opinion change survey responses using a 1.96 z-score and the standard error for a binomial distribution, . Confidence intervals for which the (𝑝 × (1 -𝑝))/𝑛 calculated MOE 95 was not within the minimum MOE 95 for our reported survey response rate were adjusted accordingly.</ns0:p><ns0:p>To evaluate volunteer opinion change relative to levels of volunteer participation, we focused on three measures of CCFRP volunteer angler participation: (a) number of years since becoming a volunteer angler; (b) the number of CCFRP Volunteer Angler Appreciation and Data Workshops an angler had attended; and (c) approximate number of CCFRP sampling trips attended. We calculated the number of years since an angler became a volunteer by subtracting the year the respondent started volunteering from the year 2018 (the year we conducted the survey). Given the survey was anonymous, we calculated the approximate number of sampling trips a respondent went on throughout their time with CCFRP by multiplying the number of years a volunteer participated in CCFRP by the average number of trips they went on per year. We used the nnet, broom, scales, and car packages <ns0:ref type='bibr' target='#b39'>(Venables & Ripley, 2002;</ns0:ref><ns0:ref type='bibr' target='#b5'>Fox & Weisberg, 2019;</ns0:ref><ns0:ref type='bibr' target='#b41'>Wickham & Seidel, 2019;</ns0:ref><ns0:ref type='bibr' target='#b30'>Robinson & Hayes, 2020)</ns0:ref> in R version 6.3.1 (R-Core-Team, 2019) to run a multinomial logistic regression model and test the effect of each measure of volunteer participation on respondents having a positive, negative, or no opinion change on the creation of MPAs (Data S1). We used the MNLpred package <ns0:ref type='bibr' target='#b22'>(Neumann, 2020)</ns0:ref> to construct predicted probabilities and 95% confidence intervals for each opinion change category over levels of participation (Data S1). For questions on the quality of data used in resource management, we used the glm function within the stats package in R version 6.3.1 (R-Core-Team, 2019) to run a binomial logistic regression model to test the effect of each measure of volunteer participation on respondents having either a positive opinion change or no opinion change as there were too few (n = 1) negative opinion change responses to include in a multinomial logistic regression model (Data S1).</ns0:p><ns0:p>Demographics and characteristics of respondents were also compared across MPA opinion change categories.</ns0:p></ns0:div>
<ns0:div><ns0:head>Results</ns0:head><ns0:p>Of the 722 current and former volunteer anglers contacted about the survey, one-hundred and twelve (112) completed and submitted a survey (Data S1), for a response rate of 15%. One respondent had not yet volunteered on sampling trips, leaving 111 surveys included in the analysis. The acceptable minimum response rate, given a 10% MOE 95 was 12% (n = 85); for a 5% MOE 95 , the minimum response rate was 35% (n = 251). Given our survey response rate, the minimum MOE 95 for reporting was 9%. Excluding one outlier (24.7 hr), the average time respondents took to complete and submit the survey was 12.4 min (± 6.4 SD).</ns0:p></ns0:div>
<ns0:div><ns0:head>Volunteer demographics and characteristics</ns0:head><ns0:p>The distribution of respondent age was skewed left, with nearly one third of respondents being between 65 and 74 years of age, and the next largest age bracket being 55-64 years old (17%, Table <ns0:ref type='table' target='#tab_0'>1</ns0:ref>). Most respondents were male (86%); twelve percent (12%) were female. Two percent (2%) of respondents chose the option 'I prefer not to say.' Respondents had medium to high avidity for saltwater angling, with 40% taking between four and 23 trips a year and 35% taking more than 23 trips a year. Eighteen percent (18%) of respondents said they participated in the MLPA planning process. Of those participating in the MLPA planning process, 90% were characterized as having high or medium angling avidity (45% each).</ns0:p><ns0:p>Seventy percent (70%) of volunteer anglers who responded to the survey considered themselves to be more conservation minded than their peers in the recreational fishing community, and an additional 25% thought they were similarly conservation minded with their peers (Table <ns0:ref type='table' target='#tab_0'>1</ns0:ref>). Sixty-five percent (65%) of respondents did not have any experience in marine resource management, or the recreational or commercial fishing industries. Twenty-four percent (24%) had experience working in the fishing industry sector, and 14% had some experience with marine resource management (there was some overlap between these groups). Forty-five percent (45%) of the respondents had previously fished in areas that are now MPAs. Seventy-four percent (74%) of respondents participated in surveys of both MPA and reference areas during their time as volunteers with CCFRP.</ns0:p></ns0:div>
<ns0:div><ns0:head>Volunteer perceptions of groundfish management and stock health</ns0:head><ns0:p>Many respondents (52%) believed California groundfish stocks were healthy, while 25% thought they were somewhat unhealthy or very unhealthy; the distribution of responses was similar across related work experience category (Fig. <ns0:ref type='figure'>2A</ns0:ref>). Nearly four out of five respondents (79%) thought that California groundfish stocks were very well managed, well managed, or adequately managed; 14% believed they were poorly managed, and 2% thought they were very poorly managed. The distribution of responses regarding the management of stocks varied by related work experience (Fig. <ns0:ref type='figure'>2B</ns0:ref>). Respondents having no related work experience and respondents with marine resource management experience had a more positive response toward the management of groundfish stocks than respondents having worked in the fishing industry (Fig. <ns0:ref type='figure'>2B</ns0:ref>).</ns0:p><ns0:p>Eighty-five percent (85%) of respondents thought seasonal closures and bag limits were effective fisheries management tools for groundfish stocks, while spatial closures and depth restrictions were considered relatively less effective (Fig. <ns0:ref type='figure'>3A</ns0:ref>). Negative responses toward the effectiveness of spatial closures and depth restrictions were comprised mostly of respondents with experience working in the fishing industry (Fig. <ns0:ref type='figure'>3B</ns0:ref>). Respondents with no experience and PeerJ reviewing PDF | (2020:04:47639:1:0:NEW 30 Jul 2020)</ns0:p><ns0:p>Manuscript to be reviewed those with fishing industry experience were least certain about depth restrictions; this type of regulation comprised the highest percentage of 'Not sure' responses in these groups (Fig. <ns0:ref type='figure'>3C</ns0:ref>).</ns0:p></ns0:div>
<ns0:div><ns0:head>Motivations for volunteering with CCFRP</ns0:head><ns0:p>Nearly all respondents said they plan to continue volunteering with CCFRP (93%). Of those who plan to continue volunteering (N=102), the reasons why they joined the program were the same as the reasons why they continue volunteering with the program (Fig. <ns0:ref type='figure'>4</ns0:ref>). The most frequently selected reason for continuing to volunteer with CCFRP was the opportunity to participate in science (75%). Sixty-eight percent (68%) selected 'giving back to fisheries resources,' and 58% selected 'enjoying a day of fishing provided by CCFRP' as key reasons why they both joined the program and why they stay involved (Fig. <ns0:ref type='figure'>4</ns0:ref>). Several respondents who responded 'Other' described desires to help fisheries or help marine resource managers gather data to use in management. Three respondents replied that the opportunity to learn was important to why they joined the program, and an additional three respondents cited learning new information as a reason for continuing to volunteer.</ns0:p><ns0:p>Eight respondents (7%) said that they do not plan to continue volunteering with CCFRP.</ns0:p><ns0:p>Reasons included lack of available volunteer spots (n = 2), personal health (n = 2), seasickness (n = 1), old age (n = 1), turned into a job (n = 1), and a lack of extra time to volunteer (n =1).</ns0:p></ns0:div>
<ns0:div><ns0:head>Change in opinions on data quality, MPAs</ns0:head><ns0:p>Of the 111 respondents surveyed, 61% reported having no opinion of the quality of data used for resource management before volunteering with CCFRP; roughly equal portions reported having either no change (18%) or change in opinion (21%) after volunteering with CCFRP. Of the 23 anglers reporting a change in opinion after participation with CCFRP, 96% reported a positive change (16 had a more positive opinion, six had a negative to positive opinion change); one angler reported the change was more negative (Fig. <ns0:ref type='figure' target='#fig_5'>5</ns0:ref>). None of the measures of volunteer participation were significantly related to having a positive change in opinion (versus no change) on data quality (Table <ns0:ref type='table'>2</ns0:ref>).</ns0:p><ns0:p>Sixty percent (60%) of volunteer anglers surveyed said that they had positive or somewhat positive opinions of the creation of MPAs before they began volunteering (Fig. <ns0:ref type='figure' target='#fig_6'>6</ns0:ref>).</ns0:p><ns0:p>The MOE 95 for these responses was ± 9% (CI 95 = 51 -69%), which was within the MOE 95 for our survey. Twenty-eight percent (28%) said they had somewhat negative or negative opinions of MPA creation in California before volunteering, while 15% of respondents said they did not have any opinion of MPAs before joining the program (Fig. <ns0:ref type='figure' target='#fig_6'>6</ns0:ref>). When volunteers were asked what their opinions were after volunteering with CCFRP, 89% said they had a positive or somewhat positive opinion of MPAs (Fig. <ns0:ref type='figure' target='#fig_6'>6A</ns0:ref>). The MOE 95 for these responses was ± 6% (CI 95 = 83 -95%); however, after adjusting for the minimum MOE 95 (± 9%) for our survey response rate (15%), the CI 95 becomes 80 -98%. The proportion of respondents having no change of opinion on the creation of MPAs after volunteering with CCFRP was 49%; these respondents comprised 95% of those having a positive or somewhat positive opinion before participating with CCFRP. Of those respondents having a change of opinion, 47% had a positive change and 5% of respondents had a negative change in opinion of MPAs after volunteering with CCFRP (Fig. <ns0:ref type='figure' target='#fig_6'>6B</ns0:ref>). The sample size for those who expressed a negative change of opinion toward MPAs consisted of five respondents (Table <ns0:ref type='table'>4</ns0:ref>). Of these, none participated in the MLPA planning process or worked previously in marine resource management. Three had fished in both MPAs and reference sites with CCFRP, but none had visited the same MPA sites with CCFRP that they had fished in before the implementation of MPAs in 2007.</ns0:p></ns0:div>
<ns0:div><ns0:head>Discussion</ns0:head><ns0:p>Our results provide evidence that long-term engagement of stakeholders in collaborative research can positively change angler opinions on MPAs. At the outset, the creators of CCFRP postulated that collaborative research was a 'potent mechanism' that could (among other listed benefits) build trust in fisheries management and develop a more accurate consensus about resource status <ns0:ref type='bibr' target='#b40'>(Wendt & Starr, 2009)</ns0:ref>. These anticipated outcomes are directly linked to the collaborative nature of the program, where participants are working together toward a shared goal <ns0:ref type='bibr' target='#b40'>(Wendt & Starr, 2009;</ns0:ref><ns0:ref type='bibr' target='#b42'>Yochum, Starr & Wendt, 2011)</ns0:ref>. CCFRP straddles two modes of public engagement in science: collaborative fisheries research and citizen science. In so doing, it draws from a long history of scientists partnering with members of the fishing industry to study fish populations or develop management tools <ns0:ref type='bibr' target='#b14'>(Hartley & Robertson, 2009;</ns0:ref><ns0:ref type='bibr' target='#b21'>Mireles, Nakamura & Wendt, 2012;</ns0:ref><ns0:ref type='bibr' target='#b11'>Gleason, Iudicello & Caselle, 2017)</ns0:ref>. Citizen science -also called communitybased or participatory science -involves members of the public who are not scientists by trade <ns0:ref type='bibr' target='#b19'>(Mckinley et al., 2017)</ns0:ref>, and it differs from collaborative fisheries research in that the volunteers are not necessarily part of the fishing industry. While the partnership between CCFRP and CPFVs follows a more traditional collaborative fisheries research model, the inclusion of the CCFRP volunteers are mostly older, avid anglers CCFRP volunteers who responded to our survey were representative of fresh and saltwater anglers in California (mostly men); however, they were relatively older. Forty-nine percent (49%) of the larger angling community are between 18 and 44 years old and less than 4% are 65 or older (US Fish and Wildlife Service, 2011). In contrast, 18 to 44-year-olds made up less than a third of our angler respondents, and 40% were over the age of 65. CCFRP surveys occur only on weekdays, of which older, retired adults are more likely to be free for volunteering compared to younger anglers. This older demographic may have influenced the proportional distributions of certain volunteer characteristics such as angler avidity (i.e. more time for fishing opportunities) and perceptions that could be influenced by having a more historical perspective (e.g., stock health). Our survey did not include questions regarding household income or ethnicity.</ns0:p><ns0:p>Relative to saltwater recreational anglers on the West Coast of the United States <ns0:ref type='bibr' target='#b31'>(Rubio, Brinson & Wallmo, 2014)</ns0:ref>, CCFRP volunteer anglers surveyed in our study had higher fishing avidity, having on average participated in a higher number of fishing trips (non CCFRP-related) in the last year. Anglers with high fishing avidity have a greater stake in fisheries management decisions. For instance, in a 2014 survey of saltwater recreational anglers, angler avidity was positively correlated with perceived importance of ensuring 'that the opinions of all recreational fisheries stakeholders are considered in policy-making' <ns0:ref type='bibr' target='#b31'>(Rubio, Brinson & Wallmo, 2014)</ns0:ref>. While our volunteers were not asked to report their opinions on the importance of stakeholder input in policy making, we found that avid CCFRP volunteer angler respondents were more likely to have participated in the MLPA planning process. Levels of public participation in the California MLPA planning process were very high, with over 4,000 members of the public attending planning-related events and over 70,000 public comments submitted during the process and environmental review <ns0:ref type='bibr' target='#b9'>(Gleason et al., 2013)</ns0:ref>. Still, with over 39 million residents in California ('United States Census Bureau QuickFacts: California,' 2017), this is a relatively small proportion of participants. In our study, one in five CCFRP survey respondents participated in the MLPA in some form, making them more engaged PeerJ reviewing PDF | (2020:04:47639:1:0:NEW 30 Jul 2020)</ns0:p><ns0:p>Manuscript to be reviewed than the average resident. Perhaps not surprisingly, about one third of our respondents who participated in the MLPA were marine resource managers, however not all of those who had worked in marine resource management participated in the MLPA.</ns0:p><ns0:p>We found that CCFRP has successfully engaged members of the general public, as two thirds of respondents had no work experience related to either marine management or the fishing industry. Although the general audience targeted for volunteer angler recruitment was recreational anglers <ns0:ref type='bibr' target='#b40'>(Wendt & Starr, 2009)</ns0:ref>, the experience of fishing side-by-side with people from different professional backgrounds may aid in the relationship-building that is an important cornerstone of the program. Low survey response rates can introduce nonresponse bias in survey results if the respondents are not characteristic of the overall survey population <ns0:ref type='bibr' target='#b2'>(Bartlett, Kotrlik & Higgins, 2001)</ns0:ref>. In this survey, the dominant characteristics of respondents (e.g., older men with high fishing avidity and no related work experience), are not atypical of the general CCFRP volunteer population.</ns0:p></ns0:div>
<ns0:div><ns0:head>CCFRP volunteer anglers are motivated by science and conservation</ns0:head><ns0:p>Part of CCFRP's success in relationship building is evidenced by the willingness of anglers to want to continue to participate in the program year after year. Most respondents said they plan to continue volunteering. The reasons respondents chose to stay with the program were the same three reasons they cited for joining CCFRP in the first place: (a) to participate in science; (b) to give back to fisheries resources; and (c) to enjoy a day of fishing provided by CCFRP. These responses demonstrate that CCFRP anglers are not solely driven by the novelty of fishing inside MPAs, but by their interest in being involved in fisheries research. A handful of respondents were high school teachers who responded that learning was a motivator for why they joined. Three other respondents listed learning as a motivator for why they stayed. Across marine and coastal citizen science projects, increasing knowledge is often a frequent motivation for volunteering <ns0:ref type='bibr' target='#b35'>(Thiel et al., 2014)</ns0:ref>.</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:04:47639:1:0:NEW 30 Jul 2020)</ns0:p><ns0:p>Manuscript to be reviewed</ns0:p></ns0:div>
<ns0:div><ns0:head>Volunteer angler consensus on groundfish health and management</ns0:head><ns0:p>A lack of transparency between fishery managers and the fishing community has often led to angler distrust of fishery assessments and management measures <ns0:ref type='bibr' target='#b42'>(Yochum, Starr & Wendt, 2011)</ns0:ref>; thus, one goal of collaborative fisheries research is garnering accurate consensus among the fishing community and fisheries scientists regarding resource health (i.e. everyone's perception of stock health reflects reality). In this study, most CCFRP respondents, regardless of their related work experience (including no related experience), believed groundfish stocks were somewhat healthy. This is a relatively accurate assessment considering most species of fish comprising the groundfish fishery in California are rockfishes, of which many stocks have rebuilt or are rebuilding from an overfished status <ns0:ref type='bibr' target='#b23'>(NOAA Fisheries, 2019;</ns0:ref><ns0:ref type='bibr'>Pacific Fishery Management Council, 2008)</ns0:ref>. The agreement that groundfish stocks are somewhat healthy, regardless of related work experience, suggests that there is accurate consensus of resource status among these groups. Although not explicitly addressed in our survey, it seems likely that CCFRP volunteer participation influenced these angler perceptions over time. It is also possible CCFRP volunteer perceptions regarding groundfish stock health are influenced by historical perspectives, as older respondents are more likely to have participated in groundfish fishing prior to the collapse and subsequent recovery of many rockfish stocks. In another survey of the iconic saltwater bass fishery in southern California, fishermen with more years of experience (and typically older in age) were more likely to have an accurate perception of stock health <ns0:ref type='bibr' target='#b3'>(Bellquist et al., 2017)</ns0:ref>. In our survey, the proportion of younger respondents who had a 'neutral' opinion regarding groundfish stocks was higher than that of the older respondents.</ns0:p><ns0:p>Most (79%) respondents thought groundfish stocks were well-managed. Many (65%) believed spatial closures (including MPAs) were effective in ensuring healthy groundfish stocks in California, though catch limits and season closures had higher support (85% each); 21% were 'unsure' and 14% believed spatial management to be 'not effective.' Most of the uncertainty and negative opinion of spatial management was by respondents having worked in the fishing industry. However, depth restrictions were least popular among all related work experience categories and garnered the greatest amount of uncertainty. Depth restrictions for groundfish in central California prohibit fishing in waters greater than 50 fathoms (91.4 m) and were intended to assist in rebuilding overfished rockfish stocks such as Canary Rockfish (Sebastes pinniger) and Yelloweye Rockfish (Sebastes ruberrimus). However, fishing these depths for other popular recreational groundfishes in central California (e.g., Lingcod (Ophiodon elongatus), Cabezon (Scorpaenichthys marmoratus), and Greenlings (Family Hexagrammidae)) is also precluded by this regulation, and could be driving some of the uncertainty among respondents. In addition, although Canary Rockfish was rebuilt in 2015 <ns0:ref type='bibr' target='#b36'>(Thorson & Wetzel, 2015)</ns0:ref>, Yelloweye Rockfish remains in rebuilding status <ns0:ref type='bibr' target='#b7'>(Gertseva & Cope, 2017)</ns0:ref>. Interestingly, except for depth restrictions, the relative proportion of respondents stating groundfish management measures are effective was similar across regulations and related work experience categories.</ns0:p><ns0:p>The focus of CCFRP is not to educate anglers on groundfish management and regulations. However, because groundfish regulations include mandatory release of overfished rockfish species, CCFRP does actively work to increase angler awareness of the susceptibility of rockfishes to pressure-related (i.e. depth-related) injuries associated with angling and the utility of recompression (i.e. releasing fish back to depth). Generally, fishing deeper results in an increased susceptibility to barotrauma and decreased survival rates of rockfishes; thus, CCFRP protocol has always restricted captains to fish areas in depths less than 36.7 m (120 ft).</ns0:p><ns0:p>Additionally, CCFRP science crew release fish showing signs of barotrauma back to depth with descending devices since recompression alleviates signs of barotrauma and significantly increases release survival of many rockfishes <ns0:ref type='bibr' target='#b15'>(Jarvis & Lowe, 2008;</ns0:ref><ns0:ref type='bibr' target='#b13'>Hannah, Rankin & Blume, 2012)</ns0:ref>. These measures ultimately promote ethical rockfish angling practices.</ns0:p></ns0:div>
<ns0:div><ns0:head>Volunteers less opiniated on data quality</ns0:head><ns0:p>In addition to outreach, a typical day on the water provides CCFRP volunteers opportunities to observe how data are collected. Important survey protocol details are relayed to CCFRP volunteer anglers on each day's pre-survey briefing. At the end of the day, the science crew debriefs the anglers on overall fish count, fish counts by angler, and biggest and smallest fish caught, etc. Thus, although the anglers do not assist with recording data, the anglers are immediately able to informally verify the data collected that day, based on their own observations and recollections.</ns0:p><ns0:p>Unlike the topic of MPAs, most respondents (61%) stated they did not have an opinion of the data used in resource management prior to volunteering for CCFRP. After participation with CCFRP, opinion change was mostly positive, but it remains unclear the degree to which this has PeerJ reviewing PDF | (2020:04:47639:1:0:NEW 30 Jul 2020)</ns0:p><ns0:p>Manuscript to be reviewed to do with CCFRP. Although none of the metrics of angler participation were significantly related to positive change (versus no change) in opinion of data quality, our analysis was limited by a reduced sample size because (unlike the MPA analysis) only anglers who stated they had an opinion before volunteering with CCFRP were asked about their opinion change. Thus, we do not know whether anglers who had no opinion on data quality before volunteering with CCFRP eventually gained a positive or negative opinion, or what the opinions were of those not having an opinion change. Nevertheless, the mostly positive opinion change suggests CCFRP participation may be a factor, regardless of the level of engagement. Building trust in the quality of data used for management is an important step toward increasing angler perceptions of groundfish management measures, including MPAs. It is also worth noting that anglers with high avidity serve CCFRP by providing highly experienced angling services, and likely relatively high consistency in angler skill levels, all positively influencing data quality.</ns0:p></ns0:div>
<ns0:div><ns0:head>CCFRP positively influences opinions on MPAs</ns0:head><ns0:p>A significantly higher percentage of volunteer anglers surveyed had positive opinions of the creation of MPAs after volunteering with CCFRP. Although there is potential for nonresponse bias in our survey, the distribution of anglers across categories of angler avidity, conservation-mindedness, and related work-experience were similar, regardless of the direction of MPA opinion change. For example, although the majority of CCFRP volunteers responding to our survey identified themselves as being more conservation-minded than their peers, about half of them gained a positive opinion of the establishment of MPAs after volunteering with CCFRP. Thus, even those considering themselves to be conservation-minded did not necessarily have strong positive opinions of MPAs before participating with CCFRP. We also found that respondents varied in their level of engagement with CCFRP across all three different measures of participation; thus, survey respondents are not likely to be more engaged than the overall population of CCFRP volunteers. In fact, the wide range of engagement among respondents allowed us to test how different levels of participation related or not to MPA opinion change.</ns0:p><ns0:p>The increase in positive perceptions of MPAs of CCFRP volunteers mirrors the perceptions of California's general public. In 2017, more than three in four Californians said that it was very important that California have MPAs; a 20 point increase since 2006 <ns0:ref type='bibr' target='#b1'>(Baldassare et al., 2007</ns0:ref><ns0:ref type='bibr' target='#b0'>(Baldassare et al., , 2017))</ns0:ref>. While the overall increase in support for MPAs across the state in the last ten Although not a stated goal of the study, we tested a posteriori whether any of the different measures of volunteer participation were perhaps related to a volunteer's willingness to continue participating (or not) with CCFRP (e.g., were volunteers who participated in more trips more likely to state they would continue volunteering?). While volunteers were significantly more likely to state they would continue volunteering with CCFRP than not continue (~ 13x more likely), none of the measures of participation were significant predictors of their willingness to continue with the program (Data S1). This would suggest that even newly recruited and less engaged volunteer anglers are enthusiastic in their support of CCFRP.</ns0:p><ns0:p>In 2017, CCFRP was expanded statewide, and now includes a partnership of six academic institutions that lead and organize surveys to actively monitor 14 MPAs in California (Moss Landing Marine Laboraties, 2020). Between 2017 and 2019, eight-hundred and ninetyeight (898) CCFRP volunteer anglers assisted science crew and CPFV captains/crew in surveying 77,202 fish representing 94 species statewide <ns0:ref type='bibr'>(R Brooks, 2020, pers. comm.)</ns0:ref>. This large expansion of the program offers additional opportunity to learn about (a) demographics and characteristics of the fishing industry sector of CCFRP (CFPV captains and crew), (b) how demographics and characteristics compare by region within and among stakeholder groups, and (c) whether CCFRP has had differential influence on MPA perceptions across stakeholder groups. Bringing increased awareness of the human dimensions of stakeholders involved in collaborative fisheries research can only serve to continue to build relationships, create buy-in on management measures, and offer insights into areas of outreach that may need improvement.</ns0:p></ns0:div>
<ns0:div><ns0:head>Conclusions</ns0:head><ns0:p>Our survey highlights CCFRP as a model for incorporating citizen science into collaborative fisheries research by capturing the realized benefits of collaborating with the angling public. We have a clearer view of who CCFRP volunteers are as a group, and how participation in the program has shaped their perspectives. CCFRP volunteers are older and have a higher fishing avidity than the broader recreational angling community in California. Although they represent a heterogeneous group in terms of experience with related industry sectors, their perceptions of groundfish stock health and management are generally in agreement. Overall, these volunteers have a positive view of the data collected for resource management and the MPAs they help to monitor. This can be attributed, in part, to long-term participation in the program. Most notably, a positive change in opinion on MPAs was more likely to occur only after considerable time engaged with CCFRP (i.e. 7+ years). Future endeavors to develop new citizen science partnerships with collaborative fisheries research programs, in which to achieve similar benefits as CCFRP (e.g., building stewardship and advocacy), should focus not only on recruiting as many volunteers as possible, but in retaining those volunteers for as long as possible. Manuscript to be reviewed Manuscript to be reviewed a The 1 respondent who did not answer this question was not included. b The 5 respondents who did not answer this question were not included. c The 2 respondents who had incomplete answers for these questions were not included. a The 1 respondent who did not answer this question was not included. b The 5 respondents who did not answer this question were not included. c The 2 respondents who had incomplete answers for these questions were not included.</ns0:p></ns0:div><ns0:figure xml:id='fig_0'><ns0:head /><ns0:label /><ns0:figDesc>provided written consent by agreeing to participate in the survey. The questionnaire consisted of 29 questions arranged into four sections: (a) CCFRP volunteering; (b) fisheries management and health of California groundfish stocks; (c) MPAs; and (d) demographics and miscellaneous questions (Article S1). We included multiple question types (yes/no, multiple-response, ordinal scale, and free-response) and designed the survey so that respondents could complete their responses in approximately 15 minutes. The University of California, San Diego Institutional Review Board (IRB) certified this study of volunteer anglers as exempt from IRB review.</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_1'><ns0:head /><ns0:label /><ns0:figDesc>One hundred and seven (107) of the respondents answered all questions related to MPA opinion change and the three calculated measures of participation (length of time since joining the program, number of Volunteer Appreciation and Data Workshops attended, and total number of trips attended). The number of years volunteers participated with CCFRP was nearly uniformly distributed; volunteers who had been with the program since 2007 made up the highest percentage (15%) and newly recruited volunteers (in 2017) followed behind at 12%. Fifty-four percent (54%) of volunteers surveyed never attended an annual Volunteer Appreciation and Data Workshop. Of the 46% who had, most attended one to four workshops. Six percent (6%) of respondents attended five or more workshops. The estimated number of CCFRP trips attended ranged from one trip to 154 trips (median = 8 trips, mean = 17 trips). Seventeen percent (17%) of respondents attended one sampling trip.Length of time since joining CCFRP was the only significant predictor of having a change in opinion regarding MPAs (Table3). In general, as the time since joining CCFRP increased, a volunteer angler was more likely to have a positive change in opinion on MPAs than having no change in opinion (RRR = 0.82 (reference category = positive change in opinion), 95% CI = 0.72, 0.92, z = -0.293, p = 0.003; Fig.7).PeerJ reviewing PDF | (2020:04:47639:1:0:NEW 30 Jul 2020)Manuscript to be reviewedMPA opinion change by volunteer characteristicsThe distribution of respondents within different volunteer characteristics, including angler avidity, conservation mindedness, and related work experience, were similar across MPA opinion change categories (Table4); however, respondents who expressed no opinion change of MPAs tended to be younger than those who had a positive change in opinion of MPAs. Those respondents who had previously worked in marine resource management were split between having no change in (positive) opinion and having a positive change in opinion of MPAs.</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_2'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:04:47639:1:0:NEW 30 Jul 2020)Manuscript to be reviewed angling public distinguishes CCFRP as having successfully integrated citizen science into collaborative fisheries research.</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_3'><ns0:head /><ns0:label /><ns0:figDesc>years might be considered a counterfactual outcome suggesting no effect of CCFRP on volunteer opinions of MPAs (Smallhorn-West et al. 2019), our study results indicate the time spent volunteering for CCFRP was influential in volunteer opinion change. Time with CCFRP influences positive change in opinion of MPAs A positive change of opinion toward MPAs was directly related to the number of years since respondents joined CCFRP. Other measures of participation, including the number of Volunteer Appreciation and Data Workshops or the number of CCFRP trips attended, were not significantly related to MPA opinion change, indicating that change in angler perceptions takes time. In this study, the length of time necessary to achieve a greater than fifty percent (50%) probability of having a positive change in opinion on MPAs was about seven years since joining CCFRP. Long-term stakeholder engagement with CCFRP corresponds with a longer period directly and indirectly gaining knowledge and awareness of MPAs through participation in survey trips and through CCFRP communications, including e-mails, e-newsletters, and posts on social media. Although Volunteer Appreciation and Data Workshops are arguably an important part of CCFRP's relationship building and outreach tools, it is often lived experiences that are more salient and have more impact on people's knowledge, attitudes, and perceptions.</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_4'><ns0:head>Figure 1 Marine</ns0:head><ns0:label>1</ns0:label><ns0:figDesc>Figure 1</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_5'><ns0:head>Figure 5</ns0:head><ns0:label>5</ns0:label><ns0:figDesc>Figure 5</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_6'><ns0:head>Figure 6 CCFRP</ns0:head><ns0:label>6</ns0:label><ns0:figDesc>Figure 6</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_7'><ns0:head>Figure 7</ns0:head><ns0:label>7</ns0:label><ns0:figDesc>Figure 7</ns0:figDesc></ns0:figure>
<ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='31,42.52,275.62,525.00,405.00' type='bitmap' /></ns0:figure>
<ns0:figure type='table' xml:id='tab_0'><ns0:head>Table 1 (on next page)</ns0:head><ns0:label>1</ns0:label><ns0:figDesc>Demographics and characteristics of CCFRP volunteer angler survey respondents. reviewing PDF | (2020:04:47639:1:0:NEW 30 Jul 2020)</ns0:figDesc><ns0:table><ns0:row><ns0:cell>PeerJ</ns0:cell></ns0:row></ns0:table></ns0:figure>
</ns0:body>
" | "UNIVERSITY OF CALIFORNIA, SAN DIEGO UCSD
MARINE BIOLOGY RESEARCH DIVISION 9500 GILMAN DRIVE
SCRIPPS INSTITUTION OF OCEANOGRAPHY LA JOLLA, CALIFORNIA 92093
(858) 534-7313 FAX
July 10, 2020
PeerJ
Re: Manuscript Revisions for #2020:04:47639:0:1:REVIEW
Dear Editor,
We appreciate the thoughtful comments provided by the reviewers. We have revised the manuscript to address all concerns (please see Author’s Rebuttal below). In particular, the manuscript now reflects the following major additions/changes:
• A Methods section, “Acceptable response rate and margin of error”.
• A Methods section, “Measures to address potential bias in survey design”.
• Clarification on the analysis of volunteer opinions on data quality, including revising Fig. 5 and a new Discussion section, “Volunteers less opiniated about data quality”.
• Removal of data quality from the manuscript title.
• The inclusion of a new table (Table 2) showing results of a logistic regression on data quality and different measures of volunteer participation.
• Inclusion of 95% Confidence Intervals for MPA opinions before and after participation with CCFRP. Though these statistics were not requested by the reviewers, we believe they help bolster our survey results.
• Table 3 (previously Table 2) and Fig. 7 were updated to reflect the maximum number of years a volunteer could have participated. The authors noticed the error during the revision process; it did not change the overall results of the analysis.
Sincerely,
Erica T. Jarvis Mason
Graduate Student, PhD Marine Biology Program
Scripps Institution of Oceanography
on behalf of all co-authors
Author’s Rebuttal
Editor’s comments:
This is a well written article and provides useful insight to an interesting topic. Overall both reviewers have provided positive comments, but have also provided some feedback that would best if it was addressed and then resubmitted for further review.
Please pay particular attention to the following reviewer comments:
Reviewer 1 commented on your conclusions about whether long term participation improves opinions on fisheries data quality (51% changed opinion). While some moderation of language may be necessary, I wonder if a bar plot could be incorporated to support Fig 5, similar to the bar plot incorporated into Fig 6 (the results of which this reviewer found more convincing).
We agree this was not the most convincing conclusion of the paper. Unfortunately, we did not ask anglers what their opinions of data quality were before volunteering with CCFRP (only if they had an opinion beforehand); thus, we could not display the data in the same way that we did for MPAs (e.g., before and after bar plot). We now address data quality and this caveat in the Discussion in a separate section titled, Volunteers less opiniated on data quality. Nevertheless, we think it is noteworthy that of those anglers that did change their opinions, the majority change was positive (22 positive vs 1 negative). Figure 5 has been updated to reflect this finding more directly. Please also see response to reviewer #1, comment #1.
Both reviewers suggested the need for a greater mention of caveats/assumptions/shortcomings of the data generated by your study approach. For example, is it possible that more conservation-minded volunteered were more likely to respond to the survey, the study relies on respondents understanding causal links to their change in opinions and having memory of their opinions too? As suggested by Reviewer 1, these suggestions might present an opportunity to incorporate a section in the discussion focusing on the limitations of the study. This might also allow for a broader incorporation of the literature to support this section, and also to develop the Discussion overall.
We agree with both reviewers. However, rather than add to the already lengthy Discussion, we have chosen to include two new sections in the Methods, 1) Acceptable survey response rate and margin of error and 2) Measures taken to address potential limitations with survey design. Our hope is that these two sections will serve to preemptively address reader’s concerns about nonresponse and response bias and to bolster the reader’s confidence in the survey results. We also now explicitly address caveats associated with specific results in the Discussion.
Reviewer #1
Major comments:
1. First and foremost, I don’t think your data supports your conclusions about whether long term participation improves opinions on fisheries data quality. Looking at figure 5, it’s hard to really accept your conclusions when only 51% of those with preexisting opinions changed towards the positive. The nature, and limitations, of survey data like this is that it is difficult to analyze (i.e. there are no means and SD here that you can test for this specific question), therefore I don’t think it’s really fair to say that the program improved opinions on fisheries data quality. Given there were 111 participants, and how close these numbers are, there is no way to be certain that if more people were surveyed, or less, that the 51 vs 49% wouldn’t change. I therefore think the entire manuscript, from the title to the discussion, needs to be adjusted accordingly. Figure 6, and the opinions about MPA looks solid, but not the results presented in figure 5. And as your title claims, this is effectively half of the story.
Thank you, we agree this was not the most convincing conclusion of the paper. Unfortunately, we did not ask anglers what their opinions of data quality were before volunteering with CCFRP (only if they had an opinion beforehand); thus, we could not display the data in the same way that we did for MPAs in Figure 6. We now address this caveat in the Discussion. In the Methods section of the submitted draft we tried to frontload this distinction… “Given that MPAs are known to elicit stronger opinions with the angling public than fishery management data, questions regarding fishery management data were limited to…” Nevertheless, we think it is noteworthy that of those anglers that did change their opinions, the majority change was positive (22 positive vs 1 negative). Figure 5 has been updated to reflect this finding more directly. We also used a binomial logistic regression to test whether the three measures of volunteer participation were related to having a positive opinion change versus no change in opinion (similar to the MPA analysis), and we found no relationship, though this may be largely due to the limited sample size. We have dropped any reference to a significant connection between volunteer participation and opinions of data quality throughout the manuscript, including the title. In the Discussion, we now moved discussion of the data quality results in a new section titled, Volunteers less opiniated on data quality.
2. The second major point has to do with limitations from these kinds of studies, which I don’t think was adequately addressed anywhere in the manuscript. I would like to see a section, perhaps in the discussion, mentioning the caveats or limitations of this approach and data. The issue has to do with attributing causation that changes in opinions occurred as a direct result of the program, and not other confounding factors. This study uses a reflexive counterfactual (See reference below for definition, although there are other papers which may be more suitable to reference), whereby the study relies on each respondent having a clear understanding of causal links. i.e. the issue here is that after the fact, asking people whether a program has been the thing which has changed their opinion relies on them understanding other potential factors that may have also changed their opinion, and also accounting for whether they were already conservation minded. In addition to the fact that the program itself is the one doing the asking, which may bias peoples answers. In a perfect world your hypothesis would have been tested by sampling behaviors (or opinions of management, but not opinions about changes in opinions to management) prior to and following their participation in the program to look for differences. In this case this wasn’t possible, which is okay, but I think this needs to be addressed as a limitation of this kind of data. It is relying on peoples sometimes imperfect memories, after the fact and asked by a group with a potential conflict of interest. Within this, a second caveat is the fact that a biased group of people likely answered the survey, they had to be engaged enough to answer it, so it could have easily been a biased population. I don’t think all of this warrants disregarding the results, but these limitations should be articulated, and the conclusions adjusted for these caveats.
Smallhorn-West P, Weeks R, Gurney G, Pressey B (2019) Ecological and socioeconomic impacts of marine protected areas in the South Pacific: assessing the evidence base. Biodiversity and conservation
Thank you for your insightful comments – they were very helpful in our revision. We now address all the potential biases you mention with a new section in the Methods titled, Measures Taken to Address Potential Bias with Survey Design. We also address potential bias with specific results in the Discussion, and we have cited the referenced paper. In addition, we now include the margin of error for the survey along with 95% Confidence Intervals related to MPA opinions before and after CCFRP participation.
Minor comments:
• Reference for figures should be at the start of when they are mentioned, not at the end (throughout, but e.g. fig 2b should be on line 233, not 236) - Addressed
• Figures 2 and 3: I think you could change the shading order. If darkness is meant to be hypothesized conservation knowledge, then dark would be marine resource management, mid fishing industry and light no experience or vice versa. – Shading changed.
• Figure 3: If possible put effective, Not sure and ineffective in a,b and c somewhere to see in the graph not just the caption - Addressed
• Figure 6b: % should be same as in the text- Addressed
• Figure 7: Title should have relative to time at the end of the sentence. Also not sure why only positive graph has the dashed line. The title has been updated. The influence of time was only significant for a positive change in opinion; therefore, we only include a y-axis reference for a greater than 50% probability on the positive change graph.
• Table 1: include other gender categories - Addressed
• Line 97 MPA network, not MPA. - Addressed
• Figure 5 vs. 6: again if you remove 5 this won’t be an issue, but I don’t understand why the same method, and same graph, wasn’t used for both fisheries data and MPA opinions. – Please see response to comment #1.
• In future, it could be useful to have specific things about MPAs that were being asked about, as well as overall opinion (e.g. performance, social impacts…) – Agreed.
• Line 275: This sentence seems redundant, as a positive change necessitates the others going down. – Sentence now removed.
• Line 313: First sentence of discussion doesn’t mention the fisheries data opinions at all, just MPAs. These seem like they were of equal importance so would include both (since both are in the title). But again may need to be adjusted if you keep the fisheries data sections. Please see response to comment #1.
• Line 391 is confusing, why does the fact that they are rockfish imply the stocks are healthy? – Agreed. This has now been clarified.
It’s great having the time analysis, I think this is a solid conclusion from this work and a really useful point to make. Great job! At the end of the day, I think the conclusions about the MPAs are sound, but again not so sure about the fisheries data ones.
Thank you!
Reviewer #2
Comments for the Author:
Line 50: Is “(California)” necessary after “1999”? “California” is mentioned multiple times in this sentence.
We agree this is confusing. It was intended as a citation for California Fish and Game Code; the citation has been updated.
Line 158: What are the work experience categories? I appears to be defined for the first time in lines 219-220.
The categories are now specified.
Line 202: Can you briefly characterize the 15% response rate relative to other online survey efforts found in the literature? It is unclear whether 15% is average, high, or low for an online survey in general or one that is fisheries or environmentally-focused. Perhaps survey response rates are mentioned in one of the papers already referenced? E.g., Baldassare et al., Loper, Ordoñez-Gauger et al., Rubio et al.
Thank you, we agree we needed to address this. Rather than report the typical or expected online survey response rate (which can depend on a variety of factors that can not necessarily be controlled), we have decided to report on the acceptable response rate for our survey. The acceptable response rate is dependent on the confidence level and the maximum margin of error the researcher is willing to accept. Based on our response rate, the minimum margin of error we can report is 9%. We include references for these calculations and address survey response rate in a new section in the Methods titled, Acceptable Survey Response Rate and Margin of Error.
Lines 284-286: the order the three calculated measures of participation are listed here as year, workshops, trips but in a different order in lines 183-185 (year, trips, workshops). Small detail but I suggest using a consistent ordering throughout the paper. Also in lines 460-462 as well as in Table 2. Easiest fix would be to re-order in lines 183-185, as the rest of the paper has them ordered as year, workshops, trips.
Addressed.
Line 296: to aid interpretation of the statistics (e.g., RRR value) in line 297,it would be helpful to mention that “positive change in opinion” is the reference category (or referent group).
Addressed.
Line 356: rather than comparing the number of participants in the MLPA process to the total CA population (39 million), it may be more useful to compare to the number of coastal county residents (for example) or to the number of people living within a certain distance from the coast (e.g., 100 miles). These may be a more “fair” comparison due to their easier access to the coast and therefore to the MLPA process which, I assume, occurred primarily on the coast? However, using these other values will not change your primary point: that a relatively small proportion of Californians participated in the MLPA process. So this is definitely a suggestion but not a necessary change.
Thank you, we appreciate this distinction. After some careful consideration of your suggestion, we have nevertheless decided not to use the number of coastal county residents because the MLPA process was open to all Californians and because we have no indication of how many participants were coastal versus non-coastal residents (although it is probable the majority were coastal, we cannot assume all non-coastal residents would choose not to participate).
Lines 372-375: I found the beginning of this paragraph and these lines in particular very confusing. Which “two programs”? And why is it “… likely that the 111 respondents in our study represent an accurate sample of those who plan to continue with the program”? How this conclusion was reached is very unclear.
Agreed. The referenced section has been removed as it did not add to the discussion.
Line 388: was unsure what “accurate consensus” meant until I reached line 398. Perhaps define “accurate consensus” up front so that we understand it to mean consistency between what is believed and what actually is the state of things (if I understood it correctly!).
Your interpretation is correct. We have now clarified this.
Line 417: missing a “)” after “Hexagrammidae)”
Addressed.
Lines 426-427: “i.e.” vs. “i.e.,” Throughout the paper, “i.e.” seems to be used but “i.e.,” shows up here. Choose one for consistency.
Addressed.
Lines 449-450: Is it possible that more conservation-minded volunteers were also more likely to respond to your survey? This gets at possible bias in your response rate, e.g., those who responded may have been more conservation-minded than those who did not respond (nonresponse bias). Is there any data on those who did not respond to your survey that could be used to test for nonresponse bias? If not, it would be worth mentioning directly in the same paragraph as the reported 15% response rate (line 202), when you characterize your response rate relative to other online survey efforts.
Unfortunately, due to the survey being anonymous, we could not test for nonresponse bias. However, we now address steps taken to help minimize nonresponse bias. Related to nonresponse bias, we also describe how several survey questions were used to characterize the participants and examine how this related to responses on MPA opinion change (including conservation-mindedness), as well as identify whether respondents were atypical of the volunteer demographic. This is now included in a new section of the Methods titled, Measures Taken to Address Potential Bias with Survey Design. We now also address nonresponse bias with respect to angler conservation-mindedness in the Discussion.
Lines 469-471: is it possible that volunteer anglers who participated in the Volunteer Appreciation and Data Workshops were more likely to return in subsequent years to volunteer? Did you look at this? (i.e., how participation in these workshops or average number of trips might be related to continued participation)
Great question, and one we had not thought about. We mention and discuss this in the Discussion as an a posteriori test.
Lines 492-495: awkward sentence, please restructure and/or split into two sentences. E.g., “Overall, these volunteers have a positive view of the data collected for resource management and the MPAs they help to monitor. This can be attributed, in part, to long-term participation in the program.”
Restructured as suggested.
Lines 496-500: this is an important conclusion and contribution to this literature. Sustained volunteerism can contribute to attitude change.
Thank you!
Table 1 is not referenced in the text.
Addressed.
Articles S2 and S3 are not referenced in the text.
These supplemental files are requested by the journal but are not meant for publication.
" | Here is a paper. Please give your review comments after reading it. |
9,789 | "<ns0:abstract xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:p>Recent marine spatial planning efforts, including the management and monitoring of marine protected areas (MPAs), increasingly focus on the importance of stakeholder engagement. For nearly 15 years, the California Collaborative Fisheries Research Program (CCFRP) has partnered volunteer anglers with researchers, the fishing industry, and resource managers to monitor groundfishes in California's network of MPAs. While the program has succeeded in generating sustained biological observations, we know little about volunteer angler demography or the impact of participation on their perceptions and opinions on fisheries data or MPAs. In this study we surveyed CCFRP volunteers to learn about (a) volunteer angler demographics and attitudes toward groundfish management and stock health, (b) volunteer angler motivations for joining and staying in the program, and (c) whether participation in the program influenced volunteer angler opinions on the quality of fisheries data used in resource management and the establishment of MPAs in California. CCFRP volunteers were older and had higher fishing avidity than average within the California recreational angling community. Many self-identified as more conservationminded than their peers in the recreational fishing community and had positive views of California groundfish management and stock health. Participation in science and giving back to fisheries resources were major motivating factors in their decision to become and remain CCFRP volunteers. Angler opinions toward MPAs were more positive after volunteering with CCFRP. Those who had volunteered for seven or more years with CCFRP were more likely than not to gain a positive opinion of MPAs. Our survey results provide evidence that long-term engagement of stakeholders in collaborative research positively influences stakeholder opinions regarding marine resource management, and highlights CCFRP's success in engaging citizen science stakeholders in collaborative fisheries research.</ns0:p></ns0:div>
</ns0:abstract>
<ns0:body xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:head>Introduction</ns0:head><ns0:p>Stakeholder engagement is an important part of marine resource protection and management <ns0:ref type='bibr' target='#b30'>(Pomeroy & Douvere, 2008)</ns0:ref>. Benefits of such engagement include the incorporation of local knowledge into policy, and the potential to build stakeholder trust in management decisions <ns0:ref type='bibr' target='#b43'>(Yochum, Starr & Wendt, 2011</ns0:ref>). California's Marine Life Protection Act (MLPA) of 1999 <ns0:ref type='bibr' target='#b39'>(Fish and</ns0:ref><ns0:ref type='bibr'>Game Code § 2850-2863)</ns0:ref> directed the state to redesign California's sense of ownership and trust among participants in the data being collected <ns0:ref type='bibr' target='#b43'>(Yochum, Starr & Wendt, 2011)</ns0:ref>.</ns0:p><ns0:p>The CCFRP program capitalizes on the expertise and knowledge of the fishing industry, angling public, participating scientists, and managers; together, these constituents work toward a common goal (e.g., measuring MPA effectiveness) that ultimately gives the group a shared purpose <ns0:ref type='bibr' target='#b41'>(Wendt & Starr, 2009;</ns0:ref><ns0:ref type='bibr' target='#b43'>Yochum, Starr & Wendt, 2011)</ns0:ref>. CCFRP relies on relationship building and transparency to create buy-in of MPA monitoring, evaluation, and management by involving stakeholders in all aspects of the program, from study design to data collection and sharing. An example of these efforts is the annual Volunteer Appreciation and Data Workshop hosted by the coordinating staff from California Polytechnic University, San Luis Obispo (Cal Poly) and Moss Landing Marine Labs (MLML), where survey results and trip highlights are shared with volunteer participants and other partners. To date, the impact of these events on angler opinions of MPAs has not been evaluated.</ns0:p><ns0:p>Human dimensions such as effective engagement, honesty, trust, and transparency can impact the success of MPAs <ns0:ref type='bibr' target='#b8'>(Gall & Rodwell, 2016;</ns0:ref><ns0:ref type='bibr' target='#b26'>Ordoñez-Gauger et al., 2018)</ns0:ref>. However, research on public knowledge, attitudes, and perceptions of California's MPA network is sparse, and studies that do exist vary widely across geographic region and composition of study populations <ns0:ref type='bibr' target='#b1'>(Baldassare et al., 2007</ns0:ref><ns0:ref type='bibr' target='#b0'>(Baldassare et al., , 2017;;</ns0:ref><ns0:ref type='bibr' target='#b20'>Loper, 2008;</ns0:ref><ns0:ref type='bibr' target='#b26'>Ordoñez-Gauger et al., 2018)</ns0:ref>. While the success of CCFRP in generating valuable monitoring data is clear <ns0:ref type='bibr' target='#b41'>(Wendt & Starr, 2009;</ns0:ref><ns0:ref type='bibr' target='#b35'>Starr et al., 2015)</ns0:ref>, the degree to which CCFRP participation has influenced volunteer perceptions of California's MPAs is less clear. In addition, the essential context for understanding this influence -(a) the demographics and characteristics of the CCFRP volunteer anglers and (b) their perceptions on the health of groundfish stocks, the data quality used to manage those stocks, and the effectiveness of MPAs relative to groundfish management measures (e.g. depth restrictions, bag limits, size limits) --has also not been assessed. Although California's network of MPAs were not specifically designed as a fishery management tool, any beneficial fisheries impacts of MPAs are important in the evaluation of overall MPA effectiveness. Thus, state resource managers would be well served by learning about the volunteer anglers who help monitor California's MPAs as well as their respective opinions on MPAs.</ns0:p><ns0:p>Given the longevity of CCFRP, and extensive volunteer participation, the program provides a valuable opportunity to measure outcomes of long-term stakeholder engagement. We used an online survey of current and former CCFRP volunteer anglers to learn about (a) volunteer angler demographics and attitudes toward groundfish management and stock health, (b) volunteer angler motivations for joining and staying in the program, and (c) whether participation in the program influenced volunteer angler opinions on the quality of fisheries data used in resource management and the creation of MPAs in California. By characterizing the population of CCFRP angler volunteers and their perceptions in relation to their volunteer efforts, our intent is to characterize the realized benefits of CCFRP as a collaborative research program, beyond the fisheries data it yields.</ns0:p></ns0:div>
<ns0:div><ns0:head>Materials & Methods</ns0:head></ns0:div>
<ns0:div><ns0:head>Survey</ns0:head><ns0:p>We distributed an online survey to 722 volunteer anglers who participated in CCFRP with Cal Poly and MLML between 2007 and 2018. This group represented a subgroup of the entire volunteer population during that time period (N=901), as 179 volunteer anglers had previously opted out of receiving communications from CCFRP and two other groups, individuals without e-mail addresses and anglers under the age of 18 years, were not contacted.</ns0:p><ns0:p>We used Qualtrics, an online survey platform, to deliver the survey questionnaire. Respondents Manuscript to be reviewed</ns0:p></ns0:div>
<ns0:div><ns0:head>Acceptable survey response rate and margin of error</ns0:head><ns0:p>An acceptable survey response rate for categorical response surveys is dependent in part on the confidence level and the maximum margin of error that the surveyor is willing to accept <ns0:ref type='bibr' target='#b2'>(Bartlett, Kotrlik & Higgins, 2001)</ns0:ref>. We calculated and report the acceptable minimum survey response rates according to the minimum sample sizes needed for an allowance of 5% and 10% margins of error (MOE 95 ) as described in <ns0:ref type='bibr' target='#b2'>Bartlett, Kotrlik & Higgins (2001)</ns0:ref>. </ns0:p></ns0:div>
<ns0:div><ns0:head>Measures taken to address potential limitations with survey design</ns0:head><ns0:p>Potential limitations with the survey design included 1) the possibility a biased population of CCFRP volunteer anglers responded to the survey (i.e. nonresponse bias) <ns0:ref type='bibr' target='#b6'>(Fisher et al. 1996;</ns0:ref><ns0:ref type='bibr' target='#b2'>Bartlett, Kotrlik & Higgins, 2001)</ns0:ref>, 2) potential bias in respondent responses due to surveyor association with the CCFRP program (i.e. response bias) and 3) reliance on a respondent's ability to accurately recall the history and influence on their opinion change or lack thereof (e.g., response bias due to the subjectivity of a reflexive counterfactual study design) <ns0:ref type='bibr' target='#b7'>(Franks et al. 2014)</ns0:ref>.</ns0:p><ns0:p>Steps taken to increase the survey response rate and aim for a large and representative sample included providing respondents (a) assurance of confidentiality, (b) a short, welldesigned survey (e.g., ~ 15 minute completion time) (c) a seamless online submission format, and (d) reminder emails <ns0:ref type='bibr' target='#b6'>(Fisher et al. 1996)</ns0:ref>. Due to the anonymity of the survey we were unable to test (or adjust) for nonresponse bias. However, the survey questions provided a means to check whether respondents represented an unexpected demographic (e.g., mostly young anglers) as well as to compare the distribution of MPA opinion change responses by age, angler avidity, conservation-mindedness, level of engagement, etc. Questions related to opinion change and volunteer participation were included in separate sections of the survey so that these responses were made independent of each other. The longest time frame for which a respondent was asked to recall their opinions was dependent on the length of time since joining CFFRP, which at most, was 11 years <ns0:ref type='bibr'>(e.g., 2007 -2017)</ns0:ref>.</ns0:p></ns0:div>
<ns0:div><ns0:head>Angler demographics and characteristics</ns0:head><ns0:p>Age and gender comprised the survey's demographic categories; other characteristics included years of fishing experience, frequency of fishing, degree of conservation-mindedness, whether anglers had any prior work experience in marine resource management or the recreational or commercial fishing industries, whether anglers had fished in MPA sites prior to those areas being designated MPAs, and whether anglers had ever participated in the MLPA planning process. We categorized angling avidity (i.e. a relative measure of the enthusiasm an angler has for the sport) into three avidity levels -low, medium, or high -based on the number of saltwater angling trips they took per year, outside of CCFRP surveys, with low being <4 days, medium 4-23 days, and high >23 days per year. Angler avidity ranges were based on the National Oceanographic and Atmospheric Administration (NOAA) West Coast Fishing Avidity categories <ns0:ref type='bibr' target='#b33'>(Rubio, Brinson & Wallmo, 2014)</ns0:ref>.</ns0:p></ns0:div>
<ns0:div><ns0:head>Volunteer perceptions of groundfish management and stock health</ns0:head><ns0:p>In addition to characterizing the general opinion of anglers on the health of groundfish stocks and the effectiveness of specific regulations, we also compared the percentage of respondent opinions across related work experience categories to gauge the relative degree of consensus in opinions among these groups. Work experience categories included marine resource management, commercial or recreational fishing industry, and no experience. Manuscript to be reviewed</ns0:p></ns0:div>
<ns0:div><ns0:head>Volunteer opinion change on fisheries data quality and MPAs</ns0:head><ns0:p>Questions regarding the quality of fisheries data were limited to whether volunteers had an opinion of these data before CCFRP participation, and of those that did, whether their opinion changed either positively or negatively after having volunteered with CCFRP.</ns0:p><ns0:p>Given that MPAs were expected to elicit strong opinions with the angling public we were interested in (a) capturing the distribution of opinions on MPAs both before and after CCFRP participation, (b) characterizing the extent of opinion change across the group, and (c) examining whether opinion change is mediated by the extent of program participation. Respondents answered questions on an ordinal scale. They could report an opinion of 'Positive,' 'Somewhat positive,' 'Somewhat negative,' 'Negative,' or 'No opinion.' We coded the answers 1, 2, 3, 4, and 5 respectively.</ns0:p><ns0:p>To capture the overall proportion of respondents having a change in opinion on MPAs after volunteering with CCFRP, we subtracted the answer code corresponding to their opinion after CCFRP participation from the answer code corresponding to their opinion before volunteering with CCFRP. Results that were positive indicated a positive change in opinion of MPAs, results that were negative indicated a negative change in opinion of MPAs, and results that were '0' represented no change in opinion. We coded these differences numerically into a single variable representing change in opinion, with positive change coded as '1', no change coded as '2,' and negative change coded as '3.' We report the survey estimates ± MOE 95 for the proportions of positive MPA opinions before and after CCFRP participation. Confidence intervals for which the calculated MOE 95 was less than the minimum MOE 95 for our reported survey response rate were adjusted accordingly.</ns0:p></ns0:div>
<ns0:div><ns0:head>Volunteer opinion change relative to measures of participation</ns0:head><ns0:p>To evaluate volunteer opinion change relative to levels of volunteer participation, we focused on three measures of CCFRP volunteer angler participation: (a) number of years since becoming a volunteer angler; (b) the number of CCFRP Volunteer Angler Appreciation and Data Workshops an angler had attended; and (c) approximate number of CCFRP sampling trips attended. We calculated the number of years since an angler became a volunteer by subtracting the year the respondent started volunteering from the year 2018 (the year we conducted the survey). Given the survey was anonymous, we calculated the approximate number of sampling trips a respondent went on throughout their time with CCFRP by multiplying the number of years a volunteer participated in CCFRP by the average number of trips they went on per year. We used the glm function within the stats package in R version 6.3.1 (R-Core-Team, 2019) to run a binomial logistic regression model and test the effect of each measure of volunteer participation on respondents having either a positive opinion change or no opinion change on the quality of data used in resource management. There were too few (n = 1) negative opinion change responses to include this category within a multinomial logistic regression model (Data</ns0:p></ns0:div>
<ns0:div><ns0:head>S1).</ns0:head><ns0:p>We used the nnet, broom, scales, and car packages <ns0:ref type='bibr' target='#b40'>(Venables & Ripley, 2002;</ns0:ref><ns0:ref type='bibr' target='#b5'>Fox & Weisberg, 2019;</ns0:ref><ns0:ref type='bibr' target='#b42'>Wickham & Seidel, 2019;</ns0:ref><ns0:ref type='bibr' target='#b32'>Robinson & Hayes, 2020)</ns0:ref> in R version 6.3.1 (R-Core-Team, 2019) to run a multinomial logistic regression model and test the effect of each measure of volunteer participation on respondents having a positive, negative, or no opinion change on the creation of MPAs (Data S1). We used the MNLpred package <ns0:ref type='bibr' target='#b23'>(Neumann, 2020)</ns0:ref> to construct predicted probabilities and 95% confidence intervals for each opinion change category over levels of participation (Data S1).</ns0:p><ns0:p>Demographics and characteristics of respondents were also compared across MPA opinion change categories.</ns0:p></ns0:div>
<ns0:div><ns0:head>Results</ns0:head><ns0:p>Of the 722 current and former volunteer anglers contacted about the survey, one-hundred and twelve (112) completed and submitted a survey (Data S1), for a response rate of 15%. One respondent had not yet volunteered on sampling trips, leaving 111 surveys included in the analysis. The acceptable minimum response rate, given a 10% MOE 95 was 12% (n = 85); for a 5% MOE 95 , the minimum response rate was 35% (n = 251). Given our survey response rate, the minimum MOE 95 for reporting was 9%. Excluding one outlier (24.7 hr), the average time respondents took to complete and submit the survey was 12.4 min (± 6.4 SD).</ns0:p></ns0:div>
<ns0:div><ns0:head>Volunteer demographics and characteristics</ns0:head><ns0:p>The distribution of respondent age was skewed left, with nearly one third of respondents being between 65 and 74 years of age, and the next largest age bracket being 55-64 years old (17%, Table <ns0:ref type='table' target='#tab_0'>1</ns0:ref>). Most respondents were male (86%); twelve percent (12%) were female. Two percent (2%) of respondents chose the option 'I prefer not to say.' Respondents had medium to high avidity for saltwater angling, with 40% taking between four and 23 trips a year and 35% taking more than 23 trips a year. Eighteen percent (18%) of respondents said they participated in the MLPA planning process. Of those participating in the MLPA planning process, 90% were characterized as having high or medium angling avidity (45% each).</ns0:p><ns0:p>Seventy percent (70%) of volunteer anglers who responded to the survey considered themselves to be more conservation minded than their peers in the recreational fishing community, and an additional 25% thought they were similarly conservation minded with their peers (Table <ns0:ref type='table' target='#tab_0'>1</ns0:ref>). Sixty-five percent (65%) of respondents did not have any experience in marine resource management, or the recreational or commercial fishing industries. Twenty-four percent (24%) had experience working in the fishing industry sector, and 14% had some experience with marine resource management (there was some overlap between these groups). Forty-five percent (45%) of the respondents had previously fished in areas that are now MPAs. Seventy-four percent (74%) of respondents participated in surveys of both MPA and reference areas during their time as volunteers with CCFRP.</ns0:p></ns0:div>
<ns0:div><ns0:head>Volunteer perceptions of groundfish management and stock health</ns0:head><ns0:p>Many respondents (52%) believed California groundfish stocks were healthy, while 25% thought they were somewhat unhealthy or very unhealthy; the distribution of responses was similar across related work experience category (Fig. <ns0:ref type='figure' target='#fig_12'>2A</ns0:ref>). Nearly four out of five respondents (79%) thought that California groundfish stocks were very well managed, well managed, or adequately managed; 14% believed they were poorly managed, and 2% thought they were very poorly managed. The distribution of responses regarding the management of stocks varied by related work experience (Fig. <ns0:ref type='figure' target='#fig_12'>2B</ns0:ref>). Respondents having no related work experience and respondents with marine resource management experience had a more positive response toward the management of groundfish stocks than respondents having worked in the fishing industry (Fig. <ns0:ref type='figure' target='#fig_12'>2B</ns0:ref>).</ns0:p><ns0:p>Eighty-five percent (85%) of respondents thought seasonal closures and bag limits were effective fisheries management tools for groundfish stocks, while spatial closures and depth restrictions were considered relatively less effective (Fig. <ns0:ref type='figure'>3A</ns0:ref>). Negative responses toward the effectiveness of spatial closures and depth restrictions were comprised mostly of respondents with experience working in the fishing industry (Fig. <ns0:ref type='figure'>3B</ns0:ref>). Respondents with no experience and those with fishing industry experience were least certain about depth restrictions; this type of regulation comprised the highest percentage of 'Not sure' responses in these groups (Fig. <ns0:ref type='figure'>3C</ns0:ref>).</ns0:p></ns0:div>
<ns0:div><ns0:head>Motivations for volunteering with CCFRP</ns0:head><ns0:p>Nearly all respondents said they plan to continue volunteering with CCFRP (93%). Of those who plan to continue volunteering (N=102), the reasons why they joined the program were the same as the reasons why they continue volunteering with the program (Fig. <ns0:ref type='figure'>4</ns0:ref>). The most frequently selected reason for continuing to volunteer with CCFRP was the opportunity to participate in science (75%). Sixty-eight percent (68%) selected 'giving back to fisheries resources,' and 58% selected 'enjoying a day of fishing provided by CCFRP' as key reasons why they both joined the program and why they stay involved (Fig. <ns0:ref type='figure'>4</ns0:ref>). Several respondents who responded 'Other' described desires to help fisheries or help marine resource managers gather data to use in management. Three respondents replied that the opportunity to learn was important to why they joined the program, and an additional three respondents cited learning new information as a reason for continuing to volunteer.</ns0:p><ns0:p>Eight respondents (7%) said that they do not plan to continue volunteering with CCFRP.</ns0:p><ns0:p>Reasons included lack of available volunteer spots (n = 2), personal health (n = 2), seasickness (n = 1), old age (n = 1), turned into a job (n = 1), and a lack of extra time to volunteer (n =1).</ns0:p></ns0:div>
<ns0:div><ns0:head>Opinion change on fisheries data quality and MPAs</ns0:head><ns0:p>Sixty-one percent (61%) reported having no opinion of the quality of fisheries data used for resource management before volunteering with CCFRP; roughly equal portions reported having either no change (18%) or a positive change in opinion (20%) after volunteering with CCFRP; 1% reported a negative change in opinion (Fig. <ns0:ref type='figure' target='#fig_13'>5</ns0:ref>).</ns0:p><ns0:p>Sixty percent (60%) of volunteer anglers surveyed said that they had positive or somewhat positive opinions of the creation of MPAs before they began volunteering (Fig. <ns0:ref type='figure' target='#fig_14'>6A</ns0:ref>).</ns0:p><ns0:p>The MOE 95 for these responses was ± 9% (CI 95 = 51 -69%), which was within the MOE 95 for our survey. Twenty-eight percent (28%) said they had somewhat negative or negative opinions of MPA creation in California before volunteering, while 15% of respondents said they did not have any opinion of MPAs before joining the program (Fig. <ns0:ref type='figure' target='#fig_14'>6A</ns0:ref>). When volunteers were asked what their opinions were after volunteering with CCFRP, 89% said they had a positive or somewhat positive opinion of MPAs (Fig. <ns0:ref type='figure' target='#fig_14'>6A</ns0:ref>). The MOE 95 for these responses was ± 6% (CI 95 = 83 -95%); however, after adjusting for the minimum MOE 95 (± 9%) for our survey response rate, the CI 95 becomes 80 -98%. The proportion of respondents having no change of opinion on the creation of MPAs after volunteering with CCFRP was 49%; these respondents comprised 95% of those having a positive or somewhat positive opinion before participating with CCFRP.</ns0:p><ns0:p>Of those respondents having a change of opinion (n = 57), 91% had a positive change and 9% of respondents had a negative change in opinion of MPAs after volunteering with CCFRP (Fig. <ns0:ref type='figure' target='#fig_14'>6B</ns0:ref>).</ns0:p></ns0:div>
<ns0:div><ns0:head>Measures of volunteer participation</ns0:head><ns0:p>The number of years volunteers participated with CCFRP was nearly uniformly distributed; volunteers who had been with the program since 2007 made up the highest percentage (15%) and newly recruited volunteers (in 2017) followed behind at 12%. Fifty-four percent (54%) of volunteers surveyed never attended an annual Volunteer Appreciation and Data Workshop. Of the 46% who had, most attended one to four workshops. Six percent (6%) of respondents attended five or more workshops. The estimated number of CCFRP trips attended ranged from one trip to 154 trips (median = 8 trips, mean = 17 trips). Seventeen percent (17%) of respondents attended one sampling trip.</ns0:p></ns0:div>
<ns0:div><ns0:head>Volunteer opinion change relative to measures of participation</ns0:head><ns0:p>The analysis on the quality of fisheries data used in resource management was limited to only those anglers who stated they had an opinion before volunteering with CCFRP (n = 42). This is because volunteers who did not have an opinion prior to volunteering were not asked about their opinions after volunteering with CCFRP (see Methods). None of the measures of volunteer participation were significantly related to having a positive change in opinion (versus no change) on data quality (Table <ns0:ref type='table'>2</ns0:ref>). </ns0:p></ns0:div>
<ns0:div><ns0:head>MPA opinion change by volunteer characteristics</ns0:head><ns0:p>The distribution of respondents within different volunteer characteristics, including angler avidity, conservation mindedness, and related work experience, were similar across MPA opinion change categories (Table <ns0:ref type='table'>4</ns0:ref>); however, respondents who expressed no opinion change of MPAs tended to be younger than those who had a positive change in opinion of MPAs. Those respondents who had previously worked in marine resource management were split between having no change in (positive) opinion and having a positive change in opinion of MPAs.</ns0:p><ns0:p>The sample size for those who expressed a negative change of opinion toward MPAs consisted of five respondents (Table <ns0:ref type='table'>4</ns0:ref>). Of these, none participated in the MLPA planning process or worked previously in marine resource management. Three had fished in both MPAs and reference sites with CCFRP, but none had visited the same MPA sites with CCFRP that they had fished in before the implementation of MPAs in 2007.</ns0:p></ns0:div>
<ns0:div><ns0:head>Discussion</ns0:head><ns0:p>Our results provide evidence that long-term engagement of stakeholders in collaborative research can positively change angler opinions on MPAs. At the outset, the creators of CCFRP postulated that collaborative research was a 'potent mechanism' that could (among other listed benefits) build trust in fisheries management and develop a more accurate consensus about resource status <ns0:ref type='bibr' target='#b41'>(Wendt & Starr, 2009)</ns0:ref>. These anticipated outcomes are directly linked to the collaborative nature of the program, where participants are working together toward a shared goal <ns0:ref type='bibr' target='#b41'>(Wendt & Starr, 2009;</ns0:ref><ns0:ref type='bibr' target='#b43'>Yochum, Starr & Wendt, 2011)</ns0:ref>. CCFRP straddles two modes of public engagement in science: collaborative fisheries research and citizen science. In so doing, it draws from a long history of scientists partnering with members of the fishing industry to study fish populations or develop management tools <ns0:ref type='bibr' target='#b16'>(Hartley & Robertson, 2009;</ns0:ref><ns0:ref type='bibr' target='#b22'>Mireles, Nakamura & Wendt, 2012;</ns0:ref><ns0:ref type='bibr' target='#b13'>Gleason, Iudicello & Caselle, 2017)</ns0:ref>. Citizen science -also called communitybased or participatory science -involves members of the public who are not scientists by trade Manuscript to be reviewed <ns0:ref type='bibr' target='#b21'>(Mckinley et al., 2017)</ns0:ref>, and it differs from collaborative fisheries research in that the volunteers are not necessarily part of the fishing industry. While the partnership between CCFRP and CPFVs follows a more traditional collaborative fisheries research model, the inclusion of the angling public distinguishes CCFRP as having successfully integrated citizen science into collaborative fisheries research.</ns0:p></ns0:div>
<ns0:div><ns0:head>CCFRP volunteers are mostly older, avid anglers</ns0:head><ns0:p>CCFRP volunteers who responded to our survey were representative of fresh and saltwater anglers in California (mostly men); however, they were relatively older. Forty-nine percent (49%) of the larger angling community are between 18 and 44 years old and less than 4% are 65 or older (US Fish and Wildlife Service, 2011). In contrast, 18 to 44-year-olds made up less than a third of our angler respondents, and 40% were over the age of 65. CCFRP surveys occur only on weekdays, of which older, retired adults are more likely to be free for volunteering compared to younger anglers. This older demographic may have influenced the proportional distributions of certain volunteer characteristics such as angler avidity (i.e. more time for fishing opportunities) and perceptions that could be influenced by having a more historical perspective (e.g., stock health). Our survey did not include questions regarding household income or ethnicity.</ns0:p><ns0:p>Relative to saltwater recreational anglers on the West Coast of the United States <ns0:ref type='bibr' target='#b33'>(Rubio, Brinson & Wallmo, 2014)</ns0:ref>, CCFRP volunteer anglers surveyed in our study had higher fishing avidity, having on average participated in a higher number of fishing trips (non CCFRP-related) in the last year. Anglers with high fishing avidity have a greater stake in fisheries management decisions. For instance, in a 2014 survey of saltwater recreational anglers, angler avidity was positively correlated with perceived importance of ensuring 'that the opinions of all recreational fisheries stakeholders are considered in policy-making' <ns0:ref type='bibr' target='#b33'>(Rubio, Brinson & Wallmo, 2014)</ns0:ref>. While our volunteers were not asked to report their opinions on the importance of stakeholder input in policy making, we found that avid CCFRP volunteer angler respondents were more likely to have participated in the MLPA planning process.</ns0:p><ns0:p>Levels of public participation in the California MLPA planning process were very high, with over 4,000 members of the public attending planning-related events and over 70,000 public comments submitted during the process and environmental review <ns0:ref type='bibr' target='#b11'>(Gleason et al., 2013)</ns0:ref>. Still, with over 39 million residents in California ('United States Census Bureau QuickFacts: California,' 2017), this is a relatively small proportion of participants. In our study, one in five CCFRP survey respondents participated in the MLPA in some form, making them more engaged than the average resident. Perhaps not surprisingly, about one third of our respondents who participated in the MLPA were marine resource managers, however not all of those who had worked in marine resource management participated in the MLPA.</ns0:p><ns0:p>We found that CCFRP has successfully engaged members of the public, as two thirds of respondents had no work experience related to either marine management or the fishing industry.</ns0:p><ns0:p>Although the general audience targeted for volunteer angler recruitment was recreational anglers <ns0:ref type='bibr' target='#b41'>(Wendt & Starr, 2009)</ns0:ref>, the experience of fishing side-by-side with people from different professional backgrounds may aid in the relationship-building that is an important cornerstone of the program. Low survey response rates can introduce nonresponse bias in survey results if the respondents are not characteristic of the overall survey population <ns0:ref type='bibr' target='#b6'>(Fisher 1996;</ns0:ref><ns0:ref type='bibr' target='#b2'>Bartlett, Kotrlik & Higgins, 2001)</ns0:ref>. Due to the relatively low survey response rate in this study (15%), nonresponse bias is not precluded from our results <ns0:ref type='bibr' target='#b6'>(Fisher 1996)</ns0:ref>. Nevertheless, the dominant characteristics of respondents (e.g., older men with high fishing avidity and no related work experience), are not atypical of the general CCFRP volunteer population. Many volunteer anglers reported being more conservation-minded than their peers in the recreational fishing community. This characterization is also not entirely unexpected given that CCFRP volunteer anglers are citizen scientists participating in collaborative fisheries research, of which conservation is a common motivator. Manuscript to be reviewed respondents were high school teachers who responded that learning was a motivator for why they joined. Three other respondents listed learning as a motivator for why they stayed. Across marine and coastal citizen science projects, increasing knowledge is often a frequent motivation for volunteering <ns0:ref type='bibr' target='#b36'>(Thiel et al., 2014)</ns0:ref>.</ns0:p></ns0:div>
<ns0:div><ns0:head>CCFRP volunteer anglers are motivated by science and conservation</ns0:head></ns0:div>
<ns0:div><ns0:head>Volunteer angler consensus on groundfish health and management</ns0:head><ns0:p>A lack of transparency between fishery managers and the fishing community has often led to angler distrust of fishery assessments and management measures <ns0:ref type='bibr' target='#b43'>(Yochum, Starr & Wendt, 2011)</ns0:ref>; thus, one goal of collaborative fisheries research is garnering accurate consensus among the fishing community and fisheries scientists regarding resource health (i.e. everyone's perception of stock health reflects reality). In this study, most CCFRP respondents, regardless of their related work experience (including no related experience), believed groundfish stocks were somewhat healthy. This is a relatively accurate assessment considering most species of fish comprising the groundfish fishery in California are rockfishes, of which many stocks have rebuilt or are rebuilding from an overfished status <ns0:ref type='bibr' target='#b24'>(NOAA Fisheries, 2019;</ns0:ref><ns0:ref type='bibr'>Pacific Fishery Management Council, 2008)</ns0:ref>. The agreement that groundfish stocks are somewhat healthy, regardless of related work experience, suggests that there is accurate consensus of resource status among these groups. Although not explicitly addressed in our survey, it seems likely that CCFRP volunteer participation influenced these angler perceptions over time. It is also possible CCFRP volunteer perceptions regarding groundfish stock health are influenced by historical perspectives, as older respondents are more likely to have participated in groundfish fishing prior to the collapse and subsequent recovery of many rockfish stocks. In another survey of the iconic saltwater bass fishery in southern California, fishermen with more years of experience (and typically older in age) were more likely to have an accurate perception of stock health <ns0:ref type='bibr' target='#b3'>(Bellquist et al., 2017)</ns0:ref>. In our survey, the proportion of younger respondents who had a 'neutral' opinion regarding groundfish stocks was higher than that of the older respondents.</ns0:p><ns0:p>Most (79%) respondents thought groundfish stocks were well-managed. Many (65%) believed spatial closures (including MPAs) were effective in ensuring healthy groundfish stocks in California, though catch limits and season closures had higher support (85% each); 21% were 'unsure' and 14% believed spatial management to be 'not effective.' Most of the uncertainty and negative opinion of spatial management was by respondents having worked in the fishing industry. However, depth restrictions were least popular among all related work experience categories and garnered the greatest amount of uncertainty. Depth restrictions for groundfish in central California prohibit fishing in waters greater than 50 fathoms (91.4 m) and were intended to assist in rebuilding overfished rockfish stocks such as Canary Rockfish (Sebastes pinniger) and Yelloweye Rockfish (Sebastes ruberrimus). However, fishing these depths for other popular recreational groundfishes in central California (e.g., Lingcod (Ophiodon elongatus), Cabezon (Scorpaenichthys marmoratus), and Greenlings (Family Hexagrammidae)) is also precluded by this regulation, and could be driving some of the uncertainty among respondents. In addition, although Canary Rockfish was rebuilt in 2015 <ns0:ref type='bibr' target='#b37'>(Thorson & Wetzel, 2015)</ns0:ref>, Yelloweye Rockfish remains in rebuilding status <ns0:ref type='bibr' target='#b9'>(Gertseva & Cope, 2017)</ns0:ref>. Interestingly, except for depth restrictions, the relative proportion of respondents stating groundfish management measures are effective was similar across regulations and related work experience categories.</ns0:p><ns0:p>The focus of CCFRP is not to educate anglers on groundfish management and regulations. However, because groundfish regulations include mandatory release of overfished rockfish species, CCFRP does actively work to increase angler awareness of the susceptibility of rockfishes to pressure-related (i.e. depth-related) injuries associated with angling and the utility of recompression (i.e. releasing fish back to depth). Generally, fishing deeper results in an increased susceptibility to barotrauma and decreased survival rates of rockfishes; thus, CCFRP protocol has always restricted captains to fish areas in depths less than 36.7 m (120 ft).</ns0:p><ns0:p>Additionally, CCFRP science crew release fish showing signs of barotrauma back to depth with descending devices since recompression alleviates signs of barotrauma and significantly increases release survival of many rockfishes <ns0:ref type='bibr' target='#b17'>(Jarvis & Lowe, 2008;</ns0:ref><ns0:ref type='bibr' target='#b15'>Hannah, Rankin & Blume, 2012)</ns0:ref>. These measures ultimately promote ethical rockfish angling practices.</ns0:p></ns0:div>
<ns0:div><ns0:head>Volunteers are less opiniated on fisheries data quality than MPAs</ns0:head><ns0:p>In addition to outreach, a typical day on the water provides CCFRP volunteers opportunities to observe how data are collected. Important survey protocol details are relayed to CCFRP volunteer anglers on each day's pre-survey briefing. At the end of the day, the science crew debriefs the anglers on overall fish count, fish counts by angler, and biggest and smallest fish caught, etc. Thus, although the anglers do not assist with recording data, the anglers are Manuscript to be reviewed immediately able to informally verify the data collected that day, based on their own observations and recollections.</ns0:p><ns0:p>Unlike the topic of MPAs, most respondents (61%) stated they did not have an opinion of the fisheries data used in resource management prior to volunteering for CCFRP. After participation with CCFRP, opinion change was mostly positive, but it remains unclear the degree to which this has to do with CCFRP. Although none of the metrics of angler participation were significantly related to positive change (versus no change) in opinion of fisheries data quality, our analysis was limited by a reduced sample size because (unlike the MPA analysis) only anglers who stated they had an opinion before volunteering with CCFRP were asked about their opinion change. Thus, we do not know whether anglers who had no opinion on data quality before volunteering with CCFRP eventually gained a positive or negative opinion, or what the opinions were of those not having an opinion change. Nevertheless, the mostly positive opinion change suggests CCFRP participation may be a factor, regardless of the level of engagement.</ns0:p><ns0:p>Building trust in the quality of fisheries data used for management is an important step toward increasing angler perceptions of groundfish management measures, including MPAs. It is also worth noting that anglers with high avidity serve CCFRP by providing highly experienced angling services, and likely relatively high consistency in angler skill levels, all positively influencing data quality.</ns0:p></ns0:div>
<ns0:div><ns0:head>CCFRP positively influences opinions on MPAs</ns0:head><ns0:p>A significantly higher percentage of volunteer anglers surveyed had positive opinions of the creation of MPAs after volunteering with CCFRP. We did not find that this response was biased with respect to angler characteristics; the distribution of anglers across categories of angler avidity, conservation-mindedness, and related work-experience were similar, regardless of the direction of MPA opinion change. For example, although the majority of CCFRP volunteers responding to our survey identified themselves as being more conservation-minded than their peers in the recreational fishing community, about half of them gained a positive opinion of the creation of MPAs after volunteering with CCFRP. Thus, even those considering themselves to be conservation-minded did not necessarily have strong positive opinions of MPAs before participating with CCFRP. We also found that respondents varied in their level of engagement with CCFRP across all three different measures of participation; thus, survey respondents are not Manuscript to be reviewed likely to be more engaged than the overall population of CCFRP volunteers. In fact, the wide range of engagement among respondents allowed us to test how different levels of participation related or not to MPA opinion change.</ns0:p><ns0:p>Respondent subjectivity can be a disadvantage of reflexive counterfactual survey designs (i.e. before and after opinions), in which there is no true control group and there is reliance on respondents to recall changes in their beliefs and opinions <ns0:ref type='bibr' target='#b34'>(Smallhorn-West et al. 2019</ns0:ref><ns0:ref type='bibr' target='#b7'>, Franks et al. 2014)</ns0:ref>. For example, we have no measure of how respondent opinions on MPAs would have changed had volunteers never participated in CCFRP, nor do we have an indication of the accuracy or legitimacy of subjective volunteer responses. However, the strength of our study design is that it allowed us to test what aspect of participation and the extent to which that volunteer participation was a factor in volunteer-stated opinion change. In other words, we did not solely rely on respondent before and after opinions to evaluate CCFRP influence on MPA opinions. In addition, we were interested in volunteer angler beliefs despite the potential for subjectivity. Capturing volunteer beliefs and perceptions is also important for highlighting opportunities for additional outreach and education <ns0:ref type='bibr' target='#b7'>(Franks et al. 2014)</ns0:ref>.</ns0:p><ns0:p>The increase in positive perceptions of MPAs of CCFRP volunteers mirrors the perceptions of California's public. In 2017, more than three in four Californians said that it was very important that California have MPAs; a 20 point increase since 2006 <ns0:ref type='bibr' target='#b1'>(Baldassare et al., 2007</ns0:ref><ns0:ref type='bibr' target='#b0'>(Baldassare et al., , 2017))</ns0:ref> Manuscript to be reviewed directly and indirectly gaining knowledge and awareness of MPAs through participation in survey trips and through CCFRP communications, including e-mails, e-newsletters, and posts on social media. Although Volunteer Appreciation and Data Workshops are arguably an important part of CCFRP's relationship building and outreach tools, it is often lived experiences that are more salient and have more impact on people's knowledge, attitudes, and perceptions.</ns0:p><ns0:p>Although not a stated goal of the study, we tested a posteriori whether any of the different measures of volunteer participation were perhaps related to a volunteer's willingness to continue participating (or not) with CCFRP (e.g., were volunteers who participated in more trips more likely to state they would continue volunteering?). While volunteers were significantly more likely to state they would continue volunteering with CCFRP than not continue (~ 13x more likely), none of the measures of participation were significant predictors of their willingness to continue with the program (Data S1). This would suggest that even newly recruited and less engaged volunteer anglers are enthusiastic in their support of CCFRP.</ns0:p><ns0:p>In 2017, CCFRP was expanded statewide, and now includes a partnership of six academic institutions that lead and organize surveys to actively monitor 14 MPAs in California Manuscript to be reviewed a higher fishing avidity than the broader recreational angling community in California. Although they represent a heterogeneous group in terms of experience with related industry sectors, their perceptions of groundfish stock health and management are generally in agreement. Overall, these volunteers have a positive view of the fisheries data collected for resource management and the MPAs they help to monitor. This can be attributed, in part, to long-term participation in the program. Most notably, a positive change in opinion on MPAs was more likely to occur only after considerable time engaged with CCFRP (i.e. 7+ years). Future endeavors to develop new citizen science partnerships with collaborative fisheries research programs, in which to achieve similar benefits as CCFRP (e.g., building stewardship and advocacy), should focus not only on recruiting as many volunteers as possible, but in retaining those volunteers for as long as possible. Manuscript to be reviewed Manuscript to be reviewed a The 1 respondent who did not answer this question was not included. b The 5 respondents who did not answer this question were not included. c The 2 respondents who had incomplete answers for these questions were not included. a The 1 respondent who did not answer this question was not included. b The 5 respondents who did not answer this question were not included. c The 2 respondents who had incomplete answers for these questions were not included.</ns0:p></ns0:div><ns0:figure xml:id='fig_0'><ns0:head /><ns0:label /><ns0:figDesc>provided written consent by agreeing to participate in the survey. The questionnaire consisted of 29 questions arranged into four sections: (a) CCFRP volunteering; (b) fisheries management and health of California groundfish stocks; (c) MPAs; and (d) demographics and miscellaneous questions (Article S1). We included multiple question types (yes/no, multiple-response, ordinal scale, and free-response) and designed the survey so that respondents could complete their responses in approximately 15 minutes. The University of California, San Diego Institutional Review Board (IRB) certified this study of volunteer anglers as exempt from IRB review. We distributed the survey via a series of e-mails sent to subjects over a two-week period in Spring 2018. The first e-mail invited subjects to participate in the survey, and two subsequent e-mails sent seven and 12 days into the study period reminded subjects to complete the survey. Each e-mail contained a description of the study, a letter of consent, and a link to the online questionnaire. PeerJ reviewing PDF | (2020:04:47639:2:0:CHECK 8 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_1'><ns0:head /><ns0:label /><ns0:figDesc>Five and 10% MOE 95 are equivalent to ± 0.25 and 0.5 points respectively, on a categorical ordinal response scale from 1 to 5. With respect to the survey response rate and estimates, the MOE 95 corresponds to the ± percentage points defining the range of the 95% CI, where, MOE 95 , = 𝑧 * (𝑝 × (1 -𝑝))/𝑛 -score for 95% CI, 𝑧 = 𝑧 sample proportion positive, and the second term in the equation is the standard error 𝑝 = of a binomial distribution.</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_2'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:04:47639:2:0:CHECK 8 Sep 2020) Manuscript to be reviewed With respect to potential surveyor influence on responses, the solicitation and reminder emails were sent via CCFRP field technicians (to keep volunteer emails confidential), but subjects were informed the survey itself was independently formulated by researchers at Scripps Institution of Oceanography, UC San Diego.</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_3'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:04:47639:2:0:CHECK 8 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_4'><ns0:head /><ns0:label /><ns0:figDesc>One hundred and seven (107) of the respondents answered all questions related to MPA opinion change and the three calculated measures of participation (length of time since joining the program, number of Volunteer Appreciation and Data Workshops attended, and total number of trips attended). Length of time since joining CCFRP was the only significant predictor of having a change in opinion regarding MPAs (Table3). In general, as the time since joining CCFRP increased, a volunteer angler was more likely to have a positive change in opinion on MPAs than having no change in opinion (RRR = 0.82 (reference category = positive change in opinion), 95% CI = 0.72, 0.92, z = -0.293, p = 0.003; Fig.7).</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_5'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:04:47639:2:0:CHECK 8 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_6'><ns0:head /><ns0:label /><ns0:figDesc>Part of CCFRP's success in relationship building is evidenced by the willingness of anglers to want to continue to participate in the program year after year. Most respondents said they plan to continue volunteering. The reasons respondents chose to stay with the program were the same three reasons they cited for joining CCFRP in the first place: (a) to participate in science; (b) to give back to fisheries resources; and (c) to enjoy a day of fishing provided by CCFRP. These responses demonstrate that CCFRP anglers are not solely driven by the novelty of fishing inside MPAs, but by their interest in being involved in fisheries research. A handful of PeerJ reviewing PDF | (2020:04:47639:2:0:CHECK 8 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_7'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:04:47639:2:0:CHECK 8 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_8'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:04:47639:2:0:CHECK 8 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_9'><ns0:head /><ns0:label /><ns0:figDesc>. While the overall increase in support for MPAs across the state in the last ten years might be considered a counterfactual outcome suggesting no effect of CCFRP on volunteer opinions of MPAs (Smallhorn-West et al. 2019), our study results indicate the time spent volunteering for CCFRP was influential in volunteer opinion change. Time with CCFRP influences positive change of opinion on MPAs A positive change of opinion toward MPAs was directly related to the number of years since respondents joined CCFRP. Other measures of participation, including the number of Volunteer Appreciation and Data Workshops or the number of CCFRP trips attended, were not significantly related to MPA opinion change, indicating that change in angler perceptions takes time. In this study, the length of time necessary to achieve a greater than fifty percent (50%) probability of having a positive change in opinion on MPAs was about seven years since joining CCFRP. Long-term stakeholder engagement with CCFRP corresponds with a longer period PeerJ reviewing PDF | (2020:04:47639:2:0:CHECK 8 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_10'><ns0:head>(</ns0:head><ns0:label /><ns0:figDesc>Moss Landing Marine Laboraties, 2020). Between 2017 and 2019, eight-hundred and ninetyeight (898) CCFRP volunteer anglers assisted science crew and CPFV captains/crew in surveying 77,202 fish representing 94 species statewide (R Brooks, 2020, pers. comm.). This large expansion of the program offers additional opportunity to learn about (a) demographics and characteristics of the fishing industry sector of CCFRP (CFPV captains and crew), (b) how demographics and characteristics compare by region within and among stakeholder groups, and (c) whether CCFRP has had differential influence on MPA perceptions across stakeholder groups. Bringing increased awareness of the human dimensions of stakeholders involved in collaborative fisheries research can only serve to continue to build relationships, create buy-in on management measures, and offer insights into areas of outreach that may need improvement. Conclusions Our survey highlights CCFRP as a model for incorporating citizen science into collaborative fisheries research by capturing the realized benefits of collaborating with the angling public. We have a clearer view of who CCFRP volunteers are as a group, and how participation in the program has shaped their perspectives. CCFRP volunteers are older and have PeerJ reviewing PDF | (2020:04:47639:2:0:CHECK 8 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_11'><ns0:head>Figure 1 Marine</ns0:head><ns0:label>1</ns0:label><ns0:figDesc>Figure 1</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_12'><ns0:head>Figure 2</ns0:head><ns0:label>2</ns0:label><ns0:figDesc>Figure 2</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_13'><ns0:head>Figure 5 CCFRP</ns0:head><ns0:label>5</ns0:label><ns0:figDesc>Figure 5</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_14'><ns0:head>Figure 6 CCFRP</ns0:head><ns0:label>6</ns0:label><ns0:figDesc>Figure 6</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_15'><ns0:head>Figure 7</ns0:head><ns0:label>7</ns0:label><ns0:figDesc>Figure 7</ns0:figDesc></ns0:figure>
<ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='32,42.52,275.62,525.00,405.00' type='bitmap' /></ns0:figure>
<ns0:figure type='table' xml:id='tab_0'><ns0:head>Table 1 (on next page)</ns0:head><ns0:label>1</ns0:label><ns0:figDesc>Demographics and characteristics of CCFRP volunteer angler survey respondents. reviewing PDF | (2020:04:47639:2:0:CHECK 8 Sep 2020)</ns0:figDesc><ns0:table><ns0:row><ns0:cell>PeerJ</ns0:cell></ns0:row></ns0:table></ns0:figure>
</ns0:body>
" | "UNIVERSITY OF CALIFORNIA, SAN DIEGO UCSD
MARINE BIOLOGY RESEARCH DIVISION 9500 GILMAN DRIVE
SCRIPPS INSTITUTION OF OCEANOGRAPHY LA JOLLA, CALIFORNIA 92093
(858) 534-7313 FAX
September 5, 2020
PeerJ
Re: Manuscript Revisions for #2020:04:47639:1:0:REVIEW
Dear Editor,
We thank the reviewer for providing additional clarification on suggested improvements to the manuscript. We have addressed the minor changes recommended by the reviewer (please see Author’s Rebuttal below). In particular, the manuscript now reflects the following notable additions/changes:
• Updated figures (Fig. 5 and Fig. 6)
• Clarification on MOE
• Additional clarification on the limitations of the study (i.e., reflexive counterfactual, response bias and non-response bias)
• Two additional references.
After your careful review of the incorporated changes, we hope you agree that the manuscript is now ready for publication.
Sincerely,
Erica T. Jarvis Mason
Graduate Student, PhD Marine Biology Program
Scripps Institution of Oceanography
on behalf of all co-authors
Author’s Rebuttal
Editor’s comments:
I agree with the reviewer in that the manuscript is nearly there. If you could please respond to the reviewer comments and make changes as needed and then resubmit.
The necessary changes have been made. Please see responses to reviewer comments below.
Reviewer #1 comments:
Comments for the Author
Thank you for taking the time to revise the manuscript and give detailed responses to the reviews. I think overall the manuscript is almost ready for publication. There are a few minor issues I think would be good to address first, but they shouldn’t take too much time.
Major comments:
My two main comments are with figure 5 and 6, which I still find confusing.
Figure 5 – Going through the questionnaire I can see why the figure is like this, but until I did that it was unclear. I still think this is misleading, and even more so than before as now if you aren’t careful you will think that the program has made huge changes in people opinions. But the reality is that half of those who had opinions prior weren’t changed, and unlike below you can’t tell if this is just because they were already positive. So if you are going to do a column I’d have it will negative change, no change and positive change. Or what would be simpler is to just provide the pie chart with four categories: no opinion before, negative change, no change and positive change.
The result here is that half of people’s opinions weren’t changed by the program, don’t try and hide this. But unlike below we can’t tell why that was.
We agree. The changes suggested have been incorporated.
Figure 6 – I also wonder if there is a better way to display this figure. Unless I also read the text I am left thinking that 49% of people did not change their opinion, which is true, but this is because they already had positive opinions. You should aim to let people interpret the figures without having to read the text in detail. So sort of opposite to figure 5, I wonder if here you could actually just show a pie or column of those only with a change in opinion, rather than having the no-change part in the pie chart, because unlike figure 5, they already have this information in the bars. I appreciate the difference of what is possible between these figures, so how can you make them as clear as possible?
We agree. The changes suggested have been incorporated.
Minor comments:
Line 139 – Would it be possible to add a few lines explaining what you mean by Margins of Error (MOE) and how this works? I can check the referenced paper, but a few lines would be good to provide an overview.
Addressed.
Line 154 – Smallhorn-West et al. 2019. Again though there may be better references on reflexive counterfactuals within that paper. See Franks P, Roe D, Small R, Schneider H (2014) Social assessment of protected areas: early experience and results of a participatory, rapid approach. IIED Working Paper. IIED, London
Addressed.
Line 155 – I think a-d are a bit of a stretch on limiting non-response bias. Sure the demographic questions do but not the first part. Anonymity equally means you can’t determine whether there is non-response bias easily. I’d just leave this as a limitation of the manuscript. No paper is perfect and this is a key caveat of yours. I think here, and also at least once in the discussion, you should explicitly state this caveat and its potential limitations, but end by saying something like ‘nevertheless, despite these limitations this study/these findings are still useful because…’
We agree with your point regarding anonymity and potential limitations. We now state in the Methods, “Steps taken to increase the survey response rate and aim for a large and representative sample included providing respondents (a) assurance of confidentiality, (b) a short, well-designed survey (e.g., ~ 15 minute completion time) (c) a seamless online submission format, and (d) reminder emails (Fisher et al. 1996). Due to the anonymity of the survey we were unable to test (or adjust) for nonresponse bias. However, the survey questions provided a means to check whether respondents represented an unexpected demographic (e.g., mostly young anglers) as well as to compare the distribution of MPA opinion change responses by age, angler avidity, conservation-mindedness, level of engagement, etc.” We also state in the discussion, “Low survey response rates can introduce nonresponse bias in survey results if the respondents are not characteristic of the overall survey population (Fisher 1996; Bartlett, Kotrlik & Higgins, 2001). Due to the relatively low survey response rate in this study (15%), non-response bias is not precluded from our results (Fisher 1996). Nevertheless, the dominant characteristics of respondents (e.g., older men with high fishing avidity and no related work experience), are not atypical of the general CCFRP volunteer population. Many volunteer anglers reported being more conservation-minded than their peers in the recreational fishing community. This characterization is also not entirely unexpected given that CCFRP volunteer anglers are citizen scientists participating in collaborative fisheries research, of which conservation is a common motivator.”
Line 166 – Again, it’s also okay to say this is a limitation. Plenty of studies use this approach, but just be clear about what it means.
In the methods, we state upfront all potential limitations of the study. We now provide clarification regarding the reflexive counterfactual design of the survey in the discussion, “Respondent subjectivity can be a disadvantage of reflexive counterfactual survey designs (i.e. before and after opinions), in which there is no true control group and there is reliance on respondents to recall changes in their beliefs and opinions (Smallhorn-West et al. 2019, Franks et al. 2014). For example, we have no measure of how respondent opinions on MPAs would have changed had volunteers never participated in CCFRP, nor do we have an indication of the accuracy or legitimacy of volunteer responses. However, the strength of our study design is that it allowed us to test what aspect of participation and the extent to which that volunteer participation was a factor in volunteer-stated opinion change. In other words, we did not solely rely on respondent before and after opinions to evaluate CCFRP influence on MPA opinions. In addition, we were interested in volunteer angler beliefs despite the potential for subjectivity. Capturing volunteer beliefs and perceptions is also important for highlighting opportunities for additional outreach and education (Franks et al. 2014).”
Line 266 – So this clearly shows that there is in fact a bias in the population, not by demographics that you measured above, but that respondents consider themselves more conservation minded than their peers.
We think this needs clarification. When we say “peers” we are referring to the recreational fishing community and not the respondent’s CCFRP volunteer peers. This is now clearly noted throughout the ms. We respectfully disagree that this characterization of the pool of respondents demonstrates bias. We now state in the discussion, “Many volunteer anglers reported being more conservation-minded than their peers in the recreational fishing community. This characterization is also not entirely unexpected given that CCFRP volunteer anglers are citizen scientists participating in collaborative fisheries research, of which conservation is a common motivator.” We also state in the discussion, “A significantly higher percentage of volunteer anglers surveyed had positive opinions of the creation of MPAs after volunteering with CCFRP. We did not find that this response was biased with respect to angler characteristics; the distribution of anglers across categories of angler avidity, conservation-mindedness, and related work-experience were similar, regardless of the direction of MPA opinion change. For example, although the majority of CCFRP volunteers responding to our survey identified themselves as being more conservation-minded than their peers in the recreational fishing community, about half of them gained a positive opinion of the establishment of MPAs after volunteering with CCFRP. Thus, even those considering themselves to be conservation-minded did not necessarily have strong positive opinions of MPAs before participating with CCFRP.”
Line 492 – This heading is really confusing – maybe “Why volunteers may be less opinionated about fisheries data quality than MPA effectiveness” or something like that.
To maintain consistency with the rest of the headings in the discussion (e.g., statement headings), this heading has been replaced with, “Volunteers are less opiniated on fisheries data quality than MPAs”
Lastly, as a point of clarity when reading through I’ve found that in my head it all starts becoming clear when I realize there are two things being looked at in aim c (from the introduction): fisheries data quality and opinions about MPAs. First, in the last paragraph of your introduction, make sure everything is in the same order you present it in. Second, it could help with the clarity to just restate this using similar terminology in the methods, results, and discussion about i) change in opinion on fisheries data quality, and ii) change in opinions on MPAs. E.g. Line 312 heading title is confusing. Make this heading Change in opinions and then two subheadings i) fisheries data quality (not just data quality) and ii) MPAs.
Organization of fisheries data quality and MPAs addressed throughout. To keep with the journal’s recommended format, we did not choose to use additional subheadings…updated change of opinion headings in the methods and results now mention fisheries data quality and MPAs in the same heading.
" | Here is a paper. Please give your review comments after reading it. |
9,790 | "<ns0:abstract xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:p>Occludin is a structural protein of tight junctions (TJ) in the blood-testis barrier (BTB). A 22amino-acid peptide (22AA) in the second extracellular loop can reversibly regulate TJ, but its regulatory mechanism is unknown. In this study, a 22AA-induced TJ destruction animal model was constructed to investigate the effect of 22AA on Sertoli cells (SCs) and spermatid count s and cell apoptosis at different time points using a multiplex immunofluorescence technique. The effect of 22AA on the location and distribution of occludin was analyzed via dual confocal fluorescence microscope . Western blotting was used to analyze dynamic changes in occludin expression. Real-time RT-PCR was used to analyze miR-122-5p expression changes. Sperm density counts and mating methods were used to analyze the effect of 22AA on fertility in mice. The results showed that 22AA promoted SC and spermatid apoptosis, downregulated occludin, upregulated miR-122-5p, and decreased sperm density and litter size before 27 days (27D). After 27D, the expression of occludin increased again, miR-122-5p expression decreased again, both sperm density and litter size returned to normal, apoptosis stopped, and spermatogenesis began to recover. Therefore, it can be concluded that 22AA can destroy TJ by downregulating occludin and inducing cell apoptosis. After 27D, TJ and spermatogenesis functions return to normal.</ns0:p></ns0:div>
</ns0:abstract>
<ns0:body xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:head n='1'>Introduction</ns0:head><ns0:p>In the testis, there is a blood-testis barrier (BTB) between the seminiferous tubules and blood vessels. The BTB composition includes the interstitial capillary endothelium and its basement membrane, connective tissue, and tight junctions (TJ) between the basement membrane of the seminiferous epithelium and Sertoli cells (SCs). TJ are the main structures that constitute the BTB <ns0:ref type='bibr'>[1]</ns0:ref>. One cause of male infertility is TJ abnormality in SCs, resulting in blockage of spermatogenic cell migration in seminiferous tubules <ns0:ref type='bibr' target='#b1'>[2]</ns0:ref>. The three-dimensional configuration of SCs is complex. SCs are irregularly cone-shaped, with the base closely touching the basement membrane, the top extending into the lumen, and many irregular depressions in the lateral surface and lumen surface, in which various levels of spermatogenic cells are embedded.</ns0:p><ns0:p>Membranes near the basal side of the adjacent SCs form TJ, and the seminiferous epithelium is divided into a basal compartment and adluminal compartment <ns0:ref type='bibr' target='#b2'>[3]</ns0:ref>. The basal compartment is located between the basement membrane of seminiferous epithelium and the TJ of SCs and contains spermatogonia cells (including type A and type B). The adluminal compartment, located above the TJ, connects with the lumen of seminiferous tubules and contains spermatocytes, germ cells, and sperm <ns0:ref type='bibr' target='#b3'>[4]</ns0:ref>. During spermatogenesis, the spermatocytes at the preleptotene stage and the leptotene stage differentiated from type B spermatogonia cells must move across the TJ into the abluminal compartment to complete their development <ns0:ref type='bibr' target='#b4'>[5]</ns0:ref>. Therefore, the movement of germ cells across the seminiferous epithelium during spermatogenesis is closely related to structural PeerJ reviewing PDF | (2020:07:51114:1:1:NEW 27 Aug 2020)</ns0:p><ns0:p>Manuscript to be reviewed reconstruction of TJ.</ns0:p><ns0:p>Many TJ experimental models exist for studying the BTB. The in vitro BTB experimental model uses the testicular SC primary dual-chamber culture method to determine TJ tightness by measuring transepithelial electrical resistance <ns0:ref type='bibr' target='#b5'>[6]</ns0:ref>. Glycerol <ns0:ref type='bibr' target='#b6'>[7]</ns0:ref>, CdCl2 <ns0:ref type='bibr' target='#b7'>[8]</ns0:ref>, and blocking peptides <ns0:ref type='bibr' target='#b8'>[9]</ns0:ref> can also destroy testicular TJ. However, the destruction of testicular TJ induced by glycerin injection or CdCl 2 injection is irreversible. Thus, studying the dynamic changes in TJ-related molecules during the disintegration process and reconstruction process using these two animal models is difficult.</ns0:p><ns0:p>Structural proteins of TJ include occludin <ns0:ref type='bibr' target='#b9'>[10]</ns0:ref>, zonula occludens-1 (ZO-1) <ns0:ref type='bibr' target='#b10'>[11]</ns0:ref>, ZO-2 <ns0:ref type='bibr' target='#b11'>[12]</ns0:ref>, ZO-3, the claudin multigene family, and adhesion molecules <ns0:ref type='bibr' target='#b12'>[13]</ns0:ref>. Occludin is a 65-kDa protein localized at TJ <ns0:ref type='bibr' target='#b13'>[14,</ns0:ref><ns0:ref type='bibr' target='#b14'>15]</ns0:ref> that consists of four transmembrane domains, a long carboxyl-terminal cytoplasmic domain, a short N-terminal cytoplasmic domain, two extracellular loops, and one intracellular loop <ns0:ref type='bibr' target='#b15'>[16]</ns0:ref>. Its structure is highly conserved among different mammalian species <ns0:ref type='bibr' target='#b16'>[17]</ns0:ref>.</ns0:p><ns0:p>The first extracellular loop is rich in tyrosine and glycine, accounting for approximately 60% of the amino acid residues, and is involved in intercellular adhesion <ns0:ref type='bibr' target='#b17'>[18]</ns0:ref>. The second extracellular loop is involved TJ <ns0:ref type='bibr' target='#b18'>[19]</ns0:ref>. Occludin has also been reported to induce apoptosis and apoptotic sensitization, which are regarded as antitumorigenic activities <ns0:ref type='bibr' target='#b19'>[20]</ns0:ref>. However, the mechanism underlying the specific role of occludin in cell apoptosis remains poorly understood. </ns0:p></ns0:div>
<ns0:div><ns0:head n='2'>Experimental materials and methods</ns0:head></ns0:div>
<ns0:div><ns0:head n='2.1'>Experimental animals</ns0:head><ns0:p>Specific-pathogen-free Kunming (KM) mice, 6 to 8 weeks old, with a body weight of 30-35 g, were provided by Chongqing Enswell Biotechnology Co., Ltd., China. Animals were kept at a temperature of 23-25°C with a 12 h light:12 h dark (12 h-12 h) cycle. The animals were allowed food and water ad libitum. The experiment was approved by the Experimental Animal Ethics Committee of Fuyang Normal University, China (Grant No. 20200006).</ns0:p></ns0:div>
<ns0:div><ns0:head n='2.2'>Construction of animal models</ns0:head><ns0:p>The sequence of 22AA is NH2-GSQIYTICSQFYTPGGTGLYVD-COOH (from the 209th to 230th amino acid of occludin) <ns0:ref type='bibr' target='#b8'>[9]</ns0:ref>, and 22AA was synthesized by Beijing Protein Innovation Co., Ltd., China. To prepare a 200.0 g/L solution, 22AA was dissolved in 0.9% sterile saline. Male KM mice were anesthetized with 7% chloral hydrate (0.5 mL/100 g), the scrotal skin was cut, PeerJ reviewing PDF | (2020:07:51114:1:1:NEW 27 Aug 2020)</ns0:p><ns0:p>Manuscript to be reviewed both (left and right) testicles were exposed, and a 26-gauge needle (Becton Dickinson, Rutherford, NJ) was used for injection of 7.0 μL of 22AA solution at three different locations in each testicle. The control group was injected with 0.9% sterile saline. After injection, the testicles were sutured, and the mice were put back into the cage. Thirty-six experimental animals were divided into six groups, with six mice in each group: A, normal control group; B, 7 days (7D) after injection of 22AA; C, 17D after 22AA injection; D, 27D after 22AA injection; E, 37D after 22AA injection; and F, 47D after 22AA injection. After the mice were anesthetized by intraperitoneal injection of 7% chloral hydrate (0.5 mL/100 g) at different time points, the skin of the scrotum was removed, and one testis was removed and stored at -80°C for western blotting (WB) and quantitative PCR (qPCR). The other testis was fixed with 4% paraformaldehyde for morphological examination. At the end of the experiment, the mice were euthanized using 0. Guidelines for the Euthanasia of Animals, 2013 Edition.</ns0:p></ns0:div>
<ns0:div><ns0:head n='2.3'>Multiplex immunofluorescence detection</ns0:head><ns0:p>Testicular tissues were collected at different stages, fixed and blocked using conventional methods, embedded in paraffin, deparaffinized, dehydrated, and subjected to heat-induced antigen retrieval. The types and sources of antibodies used in the experiments are shown in Table <ns0:ref type='table'>1</ns0:ref>. The sections were washed with distilled water, immersed in phosphate-buffered saline (PBS) for 5 min, and then blocked with goat serum at room temperature for 60 min. The blocking solutions were then aspirated, primary anti-Bax antibody (1:200) was added dropwise, and the sections were placed in a humidified box and incubated at 4°C overnight. The next day, the sections were incubated at room temperature for 30 min. The sections were immersed and washed in PBS three times for 3 min each. After the sections were blotted, a fluorescent Cy3labeled secondary antibody (1:800 dilution) was added dropwise, followed by incubation at 37°C</ns0:p><ns0:p>for 60 min in a humidified box. The sections were washed three times with PBS for 3 min each.</ns0:p><ns0:p>After the sections were blotted, primary anti-WT1 or anti-Prm2 antibody (1:100 dilution) was added dropwise, and the sections were incubated in a humidified box at 37°C for 60 min. The sections were washed three times with PBS for 3 min each. After the sections were blotted, a fluorescent FITC-labeled secondary antibody (1:800) was added dropwise, followed by incubation at 37°C for 60 min in a humidified box. After three washes with PBS for 3 min each, DAPI was added dropwise, followed by incubation in the dark for 15 min for nuclear staining.</ns0:p><ns0:p>Then, the sections were washed for 5 min four times. After excess DAPI was washed away, the slides were mounted in anti-fluorescence quencher mounting medium. The sections were then observed under a confocal fluorescence microscope (Leica, Germany), and images were collected. Fifty sections in different parts of each testicle were examined. Images were quantitatively analyzed using ImageJ software. The relative expression value of Bax was determined by the optical density value.</ns0:p></ns0:div>
<ns0:div><ns0:head n='2.4'>Immunofluorescence analysis of occludin localization and distribution</ns0:head><ns0:p>Testicular tissues were collected at different stages, fixed and blocked by conventional methods, embedded in paraffin, deparaffinized, dehydrated, and subjected to heat-induced antigen retrieval. The sections were cooled at room temperature for 10-20 min, rinsed with</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:07:51114:1:1:NEW 27 Aug 2020)</ns0:p><ns0:p>Manuscript to be reviewed distilled water, and immersed in PBS for 5 min. Sections received 0.5% Triton X-100 dropwise, followed by incubation at room temperature for 60 min and three PBS washes of 3 min each. The sections were blocked at room temperature for 60 min via dropwise addition of normal goat serum. The blocking solution was aspirated, and primary anti-occludin antibody (Abcam, America) diluted 1:100 was added dropwise. The sections were placed in a humidified box and incubated at 4°C overnight. The humidified box was removed and rewarmed to room temperature for 30 min. Sections were washed with PBS three times for 3 min each. After the excess liquid was removed from the sections, Alexa Fluor Cy3-labeled antibody (Abcam, America) diluted 1:800 was added dropwise, followed by incubation at 37°C for 60 min in a humidified box. Sections were washed with PBS three times for 3 min each, and DAPI (Beyotime, China) was added dropwise for nuclear staining in the dark for 15 min. After excess DAPI was washed away, the slides were mounted using anti-fluorescence quencher mounting medium. The sections were then observed under a confocal fluorescence microscope (Leica, Germany), and images were collected. Fifty sections in different parts of each testicle were examined.</ns0:p></ns0:div>
<ns0:div><ns0:head n='2.5'>Detection of occludin via WB</ns0:head><ns0:p>After extraction of total proteins from testes at different stages, the total protein content was determined using a BCA kit (Solarbio, China). After sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) at 100 V for 1.5 h, samples were transferred to a nitrocellulose (NC) membrane, which was blocked for 2 h via gentle shaking in 50.0 mL of TTBS (20.0 mmol/L Tris, pH 7.5, 0. 5 g/L Tween-220, 8.0 g/L NaCl) containing 50 g/L skim milk powder.</ns0:p><ns0:p>The NC membrane was placed in a plastic mantle and sealed after addition of 5.0 mL of rabbit anti-mouse occludin monoclonal antibody (Abcam, America) at 1:500 dilution and then incubated at 4°C overnight. The NC membrane was taken out and washed with TTBS three times for 15 min each. Then, 5.0 mL of horseradish peroxidase-conjugated goat anti-rabbit IgGI</ns0:p><ns0:p>(1:2 000 dilution with TBS) (Jackson, 111-035-008) was added and incubated with the membrane with gentle shaking for 2 h. The NC membrane was taken out and washed with TTBS three times for 10 min each. novaECL reagent was added to the front surface of the NC membrane and allowed to stand for 1 min, and then, light-sensitive films were observed in a dark room. The band density was scanned using a digital gel image analysis system, and the gray value for density was measured with Image Lab 4.1. Using β-actin as the internal control, the expression of occludin was determined by the ratio of the gray values of occludin to those of βactin.</ns0:p></ns0:div>
<ns0:div><ns0:head n='2.6'>Real-time RT-PCR analysis of miR-122-5p</ns0:head><ns0:p>Testicular tissues were collected at different time points, and RNA was extracted using a</ns0:p><ns0:p>Trizol Total RNA Extraction Kit (Shanghai Sangong, China, catalog number: B511321).</ns0:p><ns0:p>According to the kit manual, RNA was extracted and reverse-transcribed into cDNA. The primers for miR-122-5p RT-PCR were F: CCTGGAGTGTGACAATG and R:</ns0:p><ns0:p>GAGCAGGCTGGAGAA. The primers for the internal control actin were F: min; 35 cycles of 94°C for 20 sec, 60°C for 30 sec, and 72°C for 30 sec; followed by 72°C for detection. Relative expression was calculated using the 2 -△△Ct method.</ns0:p></ns0:div>
<ns0:div><ns0:head n='2.7'>Analysis of sperm density</ns0:head><ns0:p>Male mice were sacrificed via cervical dislocation at various time points. After abdominal disinfection, the abdominal wall was cut open to expose the reproductive system. The epididymis was separated via aseptic surgery, and the mesentery and fat surrounding the epididymis were removed with ophthalmic scissors and rinsed. The epididymis was shred and placed in a Petri dish containing 37°C PBS and then incubated at 37°C with 5% CO 2 and saturated humidity for 30 min. The sperm density was calculated using a cell counting plate after the sperm had spontaneously spread out.</ns0:p></ns0:div>
<ns0:div><ns0:head n='2.8'>Analysis of litter size</ns0:head><ns0:p>Male mice were co-caged with female mice (female:male = 1:1) at each time point. The time when a vaginal plug was detected was taken as the 0 day of pregnancy. On the 14th day of pregnancy, the mice were sacrificed by cervical dislocation, the uterus was removed by laparotomy, and the number of embryos in the bilateral uterus was recorded.</ns0:p></ns0:div>
<ns0:div><ns0:head n='2.9'>Statistical analysis</ns0:head><ns0:p>The experimental results are expressed as the mean ± standard deviation. GraphPad Prism 6</ns0:p><ns0:p>was used to complete the data processing. Differences between control group and experimental group were examined using a t -test. Differences among the six different experimental groups were examined using two-way ANOVA, and correlations between groups were examined with a Pearson test. The significant difference level was set as P<0.01.</ns0:p></ns0:div>
<ns0:div><ns0:head n='3'>Results</ns0:head></ns0:div>
<ns0:div><ns0:head n='3.1'>22AA promotes SC apoptosis</ns0:head><ns0:p>Multiplex immunocytochemistry was used to analyze the effect of 22AA on SC apoptosis. </ns0:p></ns0:div>
<ns0:div><ns0:head n='3.3'>22AA affects occludin localization and distribution</ns0:head><ns0:p>Dual immunofluorescence was used to analyze the effect of 22AA on the localization and PeerJ reviewing PDF | (2020:07:51114:1:1:NEW 27 Aug 2020)</ns0:p><ns0:p>Manuscript to be reviewed distribution of occludin. The multiplex immunohistochemical results for each group are shown in Figure <ns0:ref type='figure' target='#fig_5'>1C</ns0:ref>.</ns0:p><ns0:p>Occludin was mainly located at TJ between the basement membrane of seminiferous tubules and SCs in the control group. At 7D, a small amount of occludin was distributed at TJ between the basement membrane of seminiferous tubule and SCs. However, the total number of cells in seminiferous tubules was less than that in the control group. At 17D and 27D, no cells with blue nuclei were found in the seminiferous tubules. From day 17 to day 27, the expression of occludin gradually decreased. At 37D, occludin expression began to increase and was found to be distributed on the basement membrane of seminiferous tubules and SCs. At 47D, occludin expression and distribution and the morphological structure of seminiferous tubules were highly similar to those in the control group. These results suggest that 22AA can reversibly affect the location and distribution of occludin.</ns0:p></ns0:div>
<ns0:div><ns0:head n='3.4'>22AA downregulates occludin</ns0:head><ns0:p>To analyze the dynamic changes in the occludin expression level after TJ disintegration and reconstruction, total testicular proteins were extracted at 0, 7, 17, 27, 37, or 47 days for western blot analysis. Representative western blot results are shown in Figure <ns0:ref type='figure' target='#fig_6'>2A</ns0:ref>. The relative occludin expression level was calculated using the occludin/β-actin gray density ratio. The detailed values are shown in Figure <ns0:ref type='figure' target='#fig_6'>2B</ns0:ref>. In the control group, the expression value was 0.9967. From 7D to 27D, the occludin expression level in the 22AA group gradually decreased to 0.1621, which was only <ns0:ref type='bibr' target='#b15'>16</ns0:ref>.26% of that in the control group. Then, the expression level of occludin gradually increased to 0.3543 at 47D, which was approximately one-third the normal expression level (35.54%).</ns0:p><ns0:p>Occludin expression was significantly different among the six groups (P < 0.05). The results</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:07:51114:1:1:NEW 27 Aug 2020)</ns0:p><ns0:p>Manuscript to be reviewed</ns0:p><ns0:p>showed that the expression level of occludin in the 22AA group at each time was significantly different from that in the control group (P < 0.01). These results indicate that 22AA can reversibly regulate occludin expression.</ns0:p></ns0:div>
<ns0:div><ns0:head n='3.5'>22AA upregulates miR-122-5p expression, and miR-122-5p expression is negatively correlated with occludin expression before 27D</ns0:head><ns0:p>To analyze the dynamic expression of miR-122-5p during TJ disintegration and reconstruction, total RNA in testes was extracted at 0, 7, 17, 27, 37, or 47 days. RT-PCR analysis was performed after the total RNA was reverse-transcribed into cDNA. The expression levels of miR-122-5p in each group are shown in Figure <ns0:ref type='figure'>3A</ns0:ref>. The miR-122-5p expression level in the control group was 0.0408 and increased to 0.0539 at 7D. The miR-122-5p expression level in the 22AA group at 27D was the highest at 0.1293. Then, miR-122-5p expression gradually decreased to 0.0867 at 47D but was still higher than that in the control group. The results showed that the miR-122-5p expression levels were significantly different among the six groups (P < 0.01). The correlation between miR-122-5p and occludin expression in each group was analyzed using Pearson correlation coefficient. The linear relationship is shown in Figure <ns0:ref type='figure'>3B</ns0:ref>. The results indicated that miR-122-5p and occludin expression are significantly negatively correlated (R 2 = -0.4905, P < 0.01).</ns0:p></ns0:div>
<ns0:div><ns0:head n='3.6'>22AA changes the sperm count</ns0:head><ns0:p>To analyze the effect of 22AA on sperm count, the epididymis was extracted at 0, 7, 17, 27, Manuscript to be reviewed after which it gradually decreased. The sperm density decreased to 164.278 × 10 4 /mL at 27D</ns0:p><ns0:p>and then gradually increased to 283.114 x 10 4 mL at 47D. The results showed a significant difference in sperm density among the six groups (P < 0.01).</ns0:p></ns0:div>
<ns0:div><ns0:head n='3.7'>22AA affects litter size</ns0:head><ns0:p>To analyze the effect of 22AA on the litter size of males, male mice were co-caged with female mice at 0, 7, 17, 27, 37, or 47 days. The number of embryos in the uterus on the 14th day of pregnancy was taken as a measure of fertility. The results are shown in Figure <ns0:ref type='figure'>4B</ns0:ref>. The largest litter size in the control group was 12, and then, litter size decreased gradually to 3.667 at 27D.</ns0:p><ns0:p>The litter size gradually increased to seven at 47D. The results showed that the difference in litter size among the six groups was significant (P < 0.01).</ns0:p></ns0:div>
<ns0:div><ns0:head n='4'>Discussion</ns0:head><ns0:p>As a component of TJ, occludin is the structural basis for TJ formation between SCs in the seminiferous epithelium. The programmed opening/resealing of TJ ensures normal progression of spermatogenesis, and abnormal opening/resealing can affect the normal spermatogenesis process <ns0:ref type='bibr' target='#b21'>[22]</ns0:ref>. Interference with the functional status of occludin protein in testicular SCs can result in infertility. TJ of the BTB are different from TJ of the blood-brain barrier and other barriers, and the specific function of BTB TJ between SCs is related to spermatogonia cell activity and differentiation. The disintegration and reconstruction of TJ between SCs is an important process <ns0:ref type='bibr' target='#b22'>[23]</ns0:ref>.</ns0:p><ns0:p>SCs are the structural basis of TJ in the testis <ns0:ref type='bibr' target='#b23'>[24]</ns0:ref>. The main functions of SCs include providing structural support, creating the BTB, participating in germ cell movement and</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:07:51114:1:1:NEW 27 Aug 2020)</ns0:p><ns0:p>Manuscript to be reviewed ejaculation, and nurturing germ cells through the secretion process <ns0:ref type='bibr' target='#b24'>[25]</ns0:ref><ns0:ref type='bibr' target='#b25'>[26]</ns0:ref><ns0:ref type='bibr' target='#b27'>[27]</ns0:ref><ns0:ref type='bibr' target='#b28'>[28]</ns0:ref>. In the present study, to analyze the dynamic changes in SCs during the process of TJ disintegration and reconstruction, a 22AA-induced TJ destruction animal model was utilized, and an immuno-double-labeling technique was used for analysis. WTI was employed as an SC marker <ns0:ref type='bibr' target='#b30'>[29]</ns0:ref>. The results showed that the number of cells and the seminiferous tubule wall thickness were decreased in SCs at 7D compared with the control group. These results were consistent with those reported by Chung et al. <ns0:ref type='bibr' target='#b8'>[9]</ns0:ref>. No blue nuclei were found in the seminiferous tubules at 17D or 27D, and the intercellular boundary had completely disappeared. WT1 was distributed throughout the seminiferous tubules. At 37D, a few SCs began to be found in the basal layer of the seminiferous tubules, and spermatogenesis began to recover. At 47D, there was almost no difference in SCs between the 22AA group and the control group, indicating that spermatogenesis had returned to normal. These phenomena reveal for the first time the changing pattern of SCs in the process of TJ disintegration and reconstruction. Next, dynamic changes in spermatids were analyzed using Prm2 as the spermatid marker <ns0:ref type='bibr'>[30] [31]</ns0:ref>. The results of multiplex immunohistochemistry showed that there were 55 ± 5 spermatids in the seminiferous tubules of the control group. At 7D, the spermatid count decreased to 15 ± 3, suggesting that the spermatids gradually became apoptotic.</ns0:p><ns0:p>No spermatids were found in seminiferous tubules at 17D or 27D. This finding indicates that the spermatids were completely apoptotic. A small amount of Prm2 was distributed in the seminiferous tubules at 37D, suggesting that spermatogenesis had begun to recover. At 47D, recovery of spermatids in seminiferous tubules was visible, and the spermatids numbered 20 ± 4, which was consistent with the experimental results of Chung et al. <ns0:ref type='bibr' target='#b8'>[9]</ns0:ref>. This finding indicates that spermatogenesis had returned to normal. These data suggest that 22AA mainly reduced the number of spermatids in seminiferous tubules before 27D. After 27D, the normal structure of the TJ was restored, and spermatogenesis resumed.</ns0:p><ns0:p>The TJ-related structural proteins include ZO-1, ZO-2, and multiple claudin genes. Wong et al. found that the 44-amino-acid peptide in the second extracellular loop of occludin had no effect on the expression level of ZO-1, ZO-2, or cingulin in the Xenopus kidney epithelial cell line A6 <ns0:ref type='bibr' target='#b20'>[21]</ns0:ref>. Therefore, in the present study, only the dynamic expression of occludin was quantitatively analyzed when investigating the mechanism underlying the destruction and recovery of TJ induced by 22AA. The expression level of occludin in the 22AA group gradually decreased from 7D to 27D until reaching only 14% of that in the control group. Then, the expression level of occludin gradually increased. At 47D, occludin expression recovered to approximately one-third the normal expression. These results suggest that 22AA can reduce the expression level of occludin at the disintegration stage of TJ. Therefore, it can be concluded that 22AA can downregulate occludin, leading to disintegration of TJ.</ns0:p><ns0:p>There are three possible reasons for the decrease in protein expression. The first reason is protein degradation <ns0:ref type='bibr' target='#b33'>[32]</ns0:ref>. Occludin phosphorylation and ubiquitination regulate TJ <ns0:ref type='bibr' target='#b34'>[33]</ns0:ref>. The western blot results showed that the molecular weight of occludin was consistent with the actual size, suggesting that occludin did not degrade. This result indicates that 22AA did not cause occludin ubiquitination. The second potential reason is cell apoptosis because cell apoptosis prevents cells from expressing relevant proteins <ns0:ref type='bibr' target='#b35'>[34,</ns0:ref><ns0:ref type='bibr' target='#b36'>35]</ns0:ref>. Therefore, SC and spermatid apoptosis might lead to a decrease in occludin expression. The third reason could be inhibition of Manuscript to be reviewed transcription or translation by noncoding RNA <ns0:ref type='bibr' target='#b5'>[6]</ns0:ref>. To investigate the occludin downregulation mechanism, we analyzed the expression of several microRNAs (miRNAs) that can target occludin (data not shown), among which only the expression of miR-122-5p was associated with occludin expression. miR-122-5p is encoded on chromosome 18q21.31 and is derived from the hcr gene transcript. miR-122-5p plays an important role in cell cycle regulation, cell proliferation and cell apoptosis <ns0:ref type='bibr' target='#b37'>[36]</ns0:ref> and is associated with multiple diseases <ns0:ref type='bibr' target='#b37'>[36]</ns0:ref><ns0:ref type='bibr' target='#b38'>[37]</ns0:ref><ns0:ref type='bibr' target='#b39'>[38]</ns0:ref><ns0:ref type='bibr' target='#b40'>[39]</ns0:ref>. Previously, we analyzed the correlation between miR-122-5p and occludin protein, and the results showed that miR-122-5p was negatively correlated with occludin expression <ns0:ref type='bibr' target='#b42'>[40]</ns0:ref>. Our other recent results showed that miR-122-5p regulates occludin expression through the AACACTCCA sequence of the occludin 3'UTR, thereby regulating the formation and tightness of TJ between SCs (submitted). We further employed real-time RT-PCR to assess whether 22AA affects miR-122-5p expression.</ns0:p><ns0:p>The current results showed that the expression level of miR-122-5p gradually increased from 0 to 27D and then began to decrease. The change in the miR-122-5p expression level was opposite that of occludin. Correlation analysis showed a significant negative correlation between miR-122-5p and occludin expression from 0 to 47D(Figure <ns0:ref type='figure'>3B</ns0:ref>). These results indicate that 22AA increased the expression of miR-122-5p, which mediated downregulation of occludin expression, thereby causing disintegration of TJ. However, the mechanism by which 22AA regulates mir-122-5p expression remains to be further studied. Manuscript to be reviewed heterodimers with Bcl-2 and has an inhibitory effect on Bcl-2. Bax is one of the most important apoptosis-promoting genes <ns0:ref type='bibr' target='#b43'>[41]</ns0:ref>. Bax expression is also closely related to spermatogenesis <ns0:ref type='bibr' target='#b44'>[42]</ns0:ref>.</ns0:p><ns0:p>In the present study, Bax expression was found in the seminiferous tubules of the control group at a low expression level of 11.373 ± 10.532. Yan et al. also found that Bax was expressed in various types of cells in normal testicular tissues <ns0:ref type='bibr' target='#b45'>[43]</ns0:ref>. At 7D, the expression level of Bax was 16.783 ± 10.157, which was higher than that in the control group, indicating that the cells in seminiferous tubules began to undergo apoptosis. However, at this stage, the structure of SCs and spermatids in the tubules was relatively intact. The expression level of Bax increased to 20.521 ± 5.781 and 30.253 ± 12.274 at 17D and 27D, respectively. At this stage, the tubules were filled with a large amount of red Bax, no blue nuclei were observed, and the intercellular boundary had completely disappeared, indicating that all the cells in the seminiferous tubules had already undergone apoptosis. At 37D, the expression level of Bax was slightly lower than at 27D.</ns0:p><ns0:p>In the basal layer of the seminiferous tubules, blue nuclei began to appear, indicating that apoptosis had slowed and the spermatids and SCs of the seminiferous tubules had started to recover. At 47D, the expression level of Bax was not significantly different from that in the control group, and the structure and number of spermatids and SCs in seminiferous tubules were not different from those in the control group, indicating that spermatogenesis had fully recovered.</ns0:p><ns0:p>These results suggest that 22AA can induce apoptosis in seminiferous tubules before 27D.</ns0:p><ns0:p>To analyze the effect of 22AA on sperm count, the epididymis was extracted at 0, 7, 17, 27, 37, or 47 days. Sperm density was analyzed after the epididymis was shredded. The results are shown in Figure <ns0:ref type='figure'>4A</ns0:ref>. The highest sperm density in the control group was 750.114 × 10 4 /mL, and then, the density gradually decreased, dropping to 164.278 × 10 4 /mL at 27D; subsequently, the sperm density gradually increased to 283.114 × 10 4 mL at 47D. However, compared with the control group, the sperm density was greatly decreased. The decrease in sperm count may be due to the following reasons. First, high expression of Bax promotes the apoptosis of type A spermatogonial stem cells <ns0:ref type='bibr' target='#b44'>[42]</ns0:ref>, thereby reducing spermatogenesis. Second, after the BTB is destroyed, immune cells enter the seminiferous tubules and engulf many sperm <ns0:ref type='bibr' target='#b47'>[44]</ns0:ref>. Third, sperm undergo apoptosis or autophagy in the epididymis <ns0:ref type='bibr' target='#b48'>[45,</ns0:ref><ns0:ref type='bibr' target='#b49'>46]</ns0:ref>. Further analysis is needed to determine which reason explains the decrease in sperm count. To analyze the effect of 22AA on litter size, male mice were co-caged with female mice at 0, 7, 17, 27, 37, or 47 days. The number of embryos in the uterus on the 14th day of pregnancy was used to measure fertility. The highest litter size in the control group was 12, but litter size decreased gradually to seven at 27D.</ns0:p><ns0:p>Afterwards, the litter size gradually increased to 6.67 at 47D.</ns0:p><ns0:p>In summary, this study investigated the effect of 22AA on TJ by using a 22AA-induced TJ destruction animal model. The results showed that before 27D, 22AA promoted SC and spermatid apoptosis, downregulated occludin, upregulated miR-122-5p, and decreased sperm density and litter size. After 27D, the occludin expression increased, miR-122-5p expression decreased, both sperm density and litter size rebounded, cell apoptosis stopped, and spermatogenesis began to recover. Therefore, it can be concluded that 22AA destroys TJ by downregulating occludin and inducing cell apoptosis. After 27D, TJ and spermatogenesis functions return to normal.</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:07:51114:1:1:NEW 27 Aug 2020)</ns0:p><ns0:p>Manuscript to be reviewed Occludin was mainly located at TJ between the basement membrane of seminiferous tubules and SCs in the control group(M). At 7D, a small amount of occludin was distributed at TJ between the basement membrane of seminiferous tubule and SCs. However, the total number of cells in seminiferous tubules was less than that in the control group(N). At 17D</ns0:p><ns0:note type='other'>Figure 1</ns0:note><ns0:p>PeerJ reviewing PDF | (2020:07:51114:1:1:NEW 27 Aug 2020)</ns0:p><ns0:p>Manuscript to be reviewed and 27D, no cells with blue nuclei were found in the seminiferous tubules(O,P). From day 17 to day 27, the expression of occludin gradually decreased. At 37D, occludin expression began to increase and was found to be distributed on the basement membrane of seminiferous tubules and SCs(Q). At 47D, occludin expression and distribution and the morphological structure of seminiferous tubules were highly similar to those in the control group(R). Bar-, 0.2 μm.</ns0:p><ns0:note type='other'>Figure 2</ns0:note><ns0:p>Effect of 22AA on the expression of occludin.</ns0:p><ns0:p>Representative western blot results are shown in Figure <ns0:ref type='figure' target='#fig_6'>2A</ns0:ref>. The relative occludin expression level was calculated using the occludin/β-actin gray density ratio. The detailed values are shown in Figure <ns0:ref type='figure' target='#fig_6'>2B</ns0:ref>. In the control group, the expression value was 0.9967. From 7D to 27D, the occludin expression level in the 22AA group gradually decreased to 0.1621, which was only 16.26% of that in the control group. Then, the expression level of occludin gradually increased to 0.3543 at 47D, which was approximately one-third the normal expression level (35.54%). Occludin expression was significantly different among the six groups (P < 0.05).</ns0:p><ns0:p>The results showed that the expression level of occludin in the 22AA group at each time was significantly different from that in the control group. ＊P < 0.01.</ns0:p><ns0:note type='other'>Figure 3</ns0:note><ns0:p>Effect of artificially synthesized 22AA on the expression of miR-122-5.</ns0:p><ns0:p>The expression levels of miR-122-5p in each group are shown in Figure <ns0:ref type='figure'>3A</ns0:ref>. The miR-122-5p expression level in the control group was 0.0408 and increased to 0.0539 at 7D. The miR-122-5p expression level in the 22AA group at 27D was the highest at 0.1293. Then, miR-122-5p expression gradually decreased to 0.0867 at 47D but was still higher than that in the control group. The results showed that the miR-122-5p expression levels were significantly different among the six groups (P < 0.01). The correlation between miR-122-5p and occludin expression in each group was analyzed using Pearson correlation coefficient.</ns0:p><ns0:p>The linear relationship is shown in Figure <ns0:ref type='figure'>3B</ns0:ref>. The results indicated that miR-122-5p and occludin expression are significantly negatively correlated ( R Manuscript to be reviewed</ns0:p><ns0:note type='other'>Figure 4</ns0:note><ns0:p>Effects of 22AA on sperm density and litter size.</ns0:p><ns0:p>The sperm density was analyzed after the epididymis was shredded. The results are shown in </ns0:p></ns0:div><ns0:figure xml:id='fig_0'><ns0:head /><ns0:label /><ns0:figDesc>Wong and Gumbiner et al. used an in vitro cell model to add an artificially synthesized 44amino-acid short peptide identical to the second extracellular loop sequence of occludin into the Xenopus kidney epithelial cell line A6, thereby reducing the tightness of TJ [21]. Chung et al. confirmed that a 22-amino-acid peptide (22AA) in the second extracellular loop of occludin could reversibly regulate TJ [9]. However, Chung et al. only analyzed the effect of 22AA on the morphological structure of the seminiferous tubules. What is the mechanism underlying the regulation of TJ by 22AA? Is the expression and localization of occludin affected by 22AA? How are SC and spermatid counts in the seminiferous tubules dynamically affected by 22AA? Can 22AA affect the number of offspring? The above questions should be studied in depth. To further investigate the mechanism underlying the regulation of TJ by 22AA, this study analyzed the effect of 22AA on occludin expression and localization, SC and spermatid apoptosis, and mouse fertility using a TJ damage animal model.</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_1'><ns0:head /><ns0:label /><ns0:figDesc>GAGACCTTCAACACCCCAGC and R: ATGTCACGCACGATTTCCC. The BR Green I protocol of the SYBR Green I method was used. The real-time fluorescence PCR kit TransStart Green qPCR SuperMix (catalog no. AQ131-01) was used for PCR amplification on a LightCycler 96 system (Roche, US). The reaction system was as follows: 25.0 μL of 2× PCR buffer, 5.0 μL of primers (25.0 pmol/μL), 0.5 μL of SYBR green I (20×), 2.0 μL of template (cDNA), and 21.5 μL of DEPC water. The amplification conditions were as follows: 94°C for 4</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_2'><ns0:head>37 ,</ns0:head><ns0:label>37</ns0:label><ns0:figDesc>or 47 days. The sperm density was analyzed after the epididymis was shredded. The results are shown in Figure 4A. The highest sperm density in the control group was 750.144 × 10 4 /mL, PeerJ reviewing PDF | (2020:07:51114:1:1:NEW 27 Aug 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_3'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:07:51114:1:1:NEW 27 Aug 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_4'><ns0:head /><ns0:label /><ns0:figDesc>To investigate the causes of SC and spermatid dysfunction, Bax was used as an apoptosis marker protein to analyze cell apoptosis in seminiferous tubules. Bax, belonging to the Bcl-2 gene family, is the most important apoptotic gene in humans. The encoded Bax protein forms PeerJ reviewing PDF | (2020:07:51114:1:1:NEW 27 Aug 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_5'><ns0:head>Figure 1 .</ns0:head><ns0:label>1</ns0:label><ns0:figDesc>Figure 1. Effects of 22AA on the apoptosis of SCs, sperm cells and the localization and distribution of occludin .</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_6'><ns0:head>2 =</ns0:head><ns0:label>2</ns0:label><ns0:figDesc>-0.4905, P < 0.01). ＊P < 0.01.PeerJ reviewing PDF | (2020:07:51114:1:1:NEW 27 Aug 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_7'><ns0:head>Figure 4A . 4 / 4 /</ns0:head><ns0:label>4A44</ns0:label><ns0:figDesc>Figure 4A. The highest sperm density in the control group was 750.144 × 10 4 /mL, after</ns0:figDesc></ns0:figure>
<ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='27,42.52,70.87,323.01,672.95' type='bitmap' /></ns0:figure>
<ns0:figure type='table' xml:id='tab_1'><ns0:head /><ns0:label /><ns0:figDesc>to the expression level of Bax at 17D, indicating that apoptosis had begun to terminate; a few SCs began to appear in the basal layer of seminiferous tubules; and TJ reconstruction started. At 47D, the number of SCs and the Bax expression level were not significantly different from those in the control group. These results demonstrate that 22AA promoted SC apoptosis.Multiplex immunofluorescence was used to analyze the effect of 22AA on spermatids. Bax was used as the apoptosis marker protein to analyze spermatid apoptosis in the seminiferous tubules. Prm2 was used as the spermatid marker to determine the spermatid count. The multiplex immunohistochemical results for each group are shown in Figure 1B. Nearly 55 ± 5 spermatids were found in the seminiferous tubules of the control group, and the expression level of Bax was 11.245 ± 4.868. At 7D, the spermatid count decreased to 15 ± 3, and the expression of Bax increased to 19.569 ± 6.158, suggesting gradual spermatid apoptosis. No spermatid was found in the seminiferous tubules at 17D or 27D, and the expression levels of Bax increased to 23.467 ± 5.327 and 31.353 ± 13.139, respectively, indicating complete spermatid apoptosis. At 37D, a small amount of Prm2 was distributed in the seminiferous tubules, and the expression level of Bax decreased to 16.362 ± 3.267, indicating that apoptosis had begun to terminate. At 47D, 20 ± 4 spermatids were rediscovered in the seminiferous tubules, and the expression level of Bax was 10.176 ± 1.682, which was not different from the control level. This result indicates that 22AA can reversibly regulate spermatid apoptosis.</ns0:figDesc><ns0:table><ns0:row><ns0:cell>3.2 22AA reversibly regulates spermatid apoptosis</ns0:cell></ns0:row><ns0:row><ns0:cell>were filled with a large amount of red Bax, and the expression levels of Bax were 20.521 ± 5.781</ns0:cell></ns0:row><ns0:row><ns0:cell>and 30.253 ± 12.274 at 17D and 27D, respectively, which were significantly higher than those in</ns0:cell></ns0:row></ns0:table><ns0:note>Bax was used as the cell apoptosis marker protein to calculate the SC apoptosis rate in seminiferous tubules. WT1 was used as a marker of SCs. The multiplex immunohistochemical results for each group are shown in Figure1A. Many SCs were present in the seminiferous tubules of the control group, with clear intercellular boundaries and a small amount of red Bax distribution, and the expression level of Bax was relatively low at 11.373 ± 10.532. At 7D, the SC count in the seminiferous tubules was not significantly different from that in the control group, and the expression level of Bax was 16.783 ± 10.157, which was significantly higher than that in the control group (P < 0.01), suggesting that cell apoptosis started in the seminiferous tubules. At 17D and 27D, no cells with blue nuclei were found in the seminiferous tubules, and the intercellular boundary completely disappeared. The WT1 patch was scattered throughout the seminiferous tubules, and no cell structure was found in the tubules. The seminiferous tubules the control group (P < 0.01), indicating that at this stage all cells in the seminiferous tubules were already apoptotic. At 37D, the expression level of Bax was 20.862 ± 3.243, which was roughlyPeerJ reviewing PDF | (2020:07:51114:1:1:NEW 27 Aug 2020)Manuscript to be reviewed equivalent</ns0:note></ns0:figure>
<ns0:figure type='table' xml:id='tab_2'><ns0:head /><ns0:label /><ns0:figDesc>Effects of 22AA on the apoptosis of SC. Many SCs were present in the control group, with clear intercellular boundaries and a small amount of red Bax distribution(A). The morphology and number of SCs at 7D were not significantly different from those of the control group(B).At 17D and 27D, there were no blue nuclei in the seminiferous tubules, and the intercellular boundary completely disappeared(C, D). The WT1 patch was scattered throughout the seminiferous tubules, with no cells present in the tubules. The tubules were filled with a large amount of red Bax. A few SCs began to appear in the basal layer of the seminiferous tubules at 37D (E). At 47D, the morphology, structure and number of SCs in the seminiferous tubules were not different from those in the control group(F). Effect of 22AA on the apoptosis of</ns0:figDesc><ns0:table><ns0:row><ns0:cell>sperm cells. Nearly 55 ± 5 spermatids were found in the seminiferous tubules of the control</ns0:cell></ns0:row><ns0:row><ns0:cell>group(J), and the expression level of Bax was 11.245 ± 4.868. At 7D, the spermatid count</ns0:cell></ns0:row><ns0:row><ns0:cell>decreased to 15 ± 3(H), and the expression of Bax increased to 19.569 ± 6.158. No</ns0:cell></ns0:row><ns0:row><ns0:cell>spermatid was found in the seminiferous tubules at 17D (I) or 27D(J), and the expression</ns0:cell></ns0:row><ns0:row><ns0:cell>levels of Bax increased to 23.467 ± 5.327 and 31.353 ± 13.139, respectively. At 37D, a small</ns0:cell></ns0:row><ns0:row><ns0:cell>amount of Prm2 was distributed in the seminiferous tubules(K), and the expression level of</ns0:cell></ns0:row><ns0:row><ns0:cell>Bax decreased to 16.362 ± 3.267. At 47D, 20 ± 4 spermatids were rediscovered in the</ns0:cell></ns0:row><ns0:row><ns0:cell>seminiferous tubules(L), and the expression level of Bax was 10.176 ± 1.682, which was not</ns0:cell></ns0:row><ns0:row><ns0:cell>different from the control level. Effect of 22AA on the localization and distribution of occludin.</ns0:cell></ns0:row></ns0:table></ns0:figure>
<ns0:note place='foot'>PeerJ reviewing PDF | (2020:07:51114:1:1:NEW 27 Aug 2020)</ns0:note>
</ns0:body>
" | "Dear editor
Thank you for your email. According to the suggestions and requirements of you and the reviewers, I have revised the article. The revised partes are marked in the text. The following is a respond to the reviewer's comments.
Reviewer 1
Basic reporting
Some more background introduction of occludin will be nice, especially known mechanisms of occludin in cell apoptosis and human fertility.
The words” Occludin has also been reported to induce apoptosis and apoptotic sensitization, which are regarded as antitumorigenic activities [20]. However, the mechanism underlying the specific role of occludin in cell apoptosis remains poorly understood.” were inserted into Introduction.
Experimental design
In statistical analysis of methods, the significant difference level was set as P<0.05. However, in every figure legend, the significant difference level was set as P<0.01. Please correct it.
P<0.05 was revised to P<0.01.
Validity of the findings
In discussion part, authors explained “The western blot images in this study showed only one band for occludin, and thus, no degradation was found”. This is a completely wrong conclusion. Protein degradation could be affected by many ways, e.g., ubiquitination, SUMOylating, autophagy-related protein degradation, etc. Authors should re-write this sentence or do protein degradation assay to exclude this possibility.
Similarly, in discussion part, author stated “Therefore, SC and spermatid apoptosis might lead to a decrease in occludin expression.” However, at 47D, spermatogenesis had returned to normal, while expression level of occludin was approximately one-third the normal expression level. Why did the expression of occludin not return to normal? Authors should explain this.
The sentence “The western blot images in this study showed only one band for occludin, and thus, no degradation was found” was revised to“The western blot results showed that the molecular weight of occludin was consistent with the actual size, suggesting that occludin did not degrade”.
“However, at 47D, spermatogenesis had returned to normal, while expression level of occludin was approximately one-third the normal expression level.” There are two possible reasons for this phenomenon. First, occludin is not the only structural protein of TJ. Second, occludin functional silencing only partially affects spermatogenesis.
Comments for the author
In the paper, authors investigated the mechanism of 22-amino-acid peptide in regulating tight junctions through occludin and cell apoptosis. The study gives an interesting new prospective of a mechanism not deeply studied for male fertility. The topic presented is of interest and the data are quite important; however, the manuscript in this form presents several problems.
1. In figure 2A, why did authors use one β-actin for two different groups?
2. From figure 2 to figure3, it is too weird. Authors analyzed the expression of several microRNAs (miRNAs) that can target occluding, but did not show the results. I think that reporting this data is fundamental.
3. Could miR-122-5p target other structural proteins of TJ?
1. When we did the laboratory, WB was done for the experimental group, the control group and the β-actin respectively, so there were three pictures. In order to make the pictures clean and beautiful, we put them together.
2. We analyzed the expression of several microRNAs (miRNAs) that can target occluding, but the results were not shown in this paper. This part of the research was not closely related to this paper. Therefore, we put it into the research papers on the regulation mechanism of microRNA on occludin. This paper is under review. If necessary, I will put this paper in the attachment.
3. After bioinformatics analysis and literature retrieval, we found that Mir-122-5p could not regulate other proteins in TJ. Of course, this needs to be confirmed by further experiments.
Reviewer: Sourav Kolay
Basic reporting
The authors present the results that provide experimental support for the role of 22AA peptide in SC and spermatid apoptosis and in sperm density regulation. The figures are relevant but need some modifications that is mentioned in later section. I commend the authors for doing a good literature review with a well written introduction section. The experiments are mostly well executed with control; the exceptions are mentioned in general comments. Some minor additional experiments can be done to improve the overall quality of the paper.
Experimental design
The study is of importance and within the scope of Peer J. The research plan is well defined and relevant experiments are performed. Some modifications and addition of better picture can improve the clarity of the paper. Additionally, some minor experiments can be done to improve the overall quality of the paper to meet the standard of Peer J.
Validity of the findings
The study is fairly robust and looks statistically sound. Conclusions are well supported by results except the ones that I have mentioned in the general comments.
Comments for the author
This is an interesting study and the strength of the study is the use of in-vivo model system. The manuscript is concise and clear. Although, there are certain topics that need attention as followed.
1. The abstract claims (line 7-8) that distribution of occluding was analyzed by immunofluorescence electron microscopy but I do not see any such data in the manuscript.
“immunofluorescence electron microscopy” was revised to “confocal fluorescence microscope”.
2. Fig 1 is a very important piece of evidence but there is some ambiguity in the data. The authors state that they counted sertoli cells and spermatids from this imaging data but the picture shows no specific staining that can be used to differentiate sertoli cells from spermatids. I only see blue staining (nucleus) and red (Bax). I see in the method section, the authors mentioned WT1 and Prm2 immunostaining but no such staining is clearly visible in these pictures (Fig 1). They should use images with better quality and contrast. I would also suggest to add different panels of single stain and then last one with merged. Additionally, The picture shows a single seminiferous tubule, I would recommend to add pictures with multiple seminiferous tubules in one field for each condition.
I would also suggest adding biochemical data to consolidate this imaging result. One experiment that can be performed to measure functional outcome of loss of sertoli cells is to measure Anti-müllerian hormone (AMH) level. AMH is produced by sertoli cells in early developmental stage. If there is loss of sertoli cells, a drop in AMH level would be expected. Considering the importance of this figure, the above mentioned points need to be addressed.
Figure 1 was the results of multiplex immunohistochemical of the apoptosis of SC, sperm cells, and the localization and distribution of occludin. Each figure was merged from different color. In the previous manuscript, I devided Figure 1 into three figures, and each figure has a large area and different panels of single stain, which can clearly show the color difference. However, the previous reviewers and editors thought there were too many pictures, and they suggested me to compress the three pictures into one. So I took their advice.
Thank you for your good idea of measuring Anti-müllerian hormone (AMH) level. But the experimental animals have been disposed of, so it is not convenient for supplementary testing.
3. For Fig 2 –Fig 4, the figure sub labeling (A/B) can be placed at the left top of each figs.
The figure sub labeling of Fig 2-Fig 4 were placed at the left top of each figs.
4. Fig 2 shows quantification of WB result; the details of quantification methods including the software used need to be given in the method section.
The band density was scanned using a digital gel image analysis system, and the gray value for density was measured with Image Lab 4.1. Using β-actin as the internal control, the expression of occludin was determined by the ratio of the gray values of occludin to those of β-actin.
5. Fig 3: This result is not really convincing. I see an increase in expression of miR-122-5p from 0 to 0.1! Is this minor increase physiologically relevant? This result can be excluded. The increase shown here is very marginal and the result is not adding any substance to the overall conclusion of the paper.
Although the increase of mir-122-5p was small, the difference was significant compared with the control group.
6. The results show that the effect of 22AA disappeared post 27D but the authors have not given much details of this incongruity in the discussion section. One line is written saying that the degradation of 22AA happens post 27D but I don’t see any data in the manuscript supporting this hypothesis. If any data exists, it should be included or else if the comment is based on previous observation that should be mentioned.
The sentence “With the degradation and consumption of 22AA in vivo” was removed from Abstract and Discussion.
7. In table 1: The table is labeled as ‘Antibody types and sources’ but this table also enlists other reagents like PBS, so I would suggest labeling it as- Table of resources/Source of reagents.
Title of table 1 was revised to “Source of reagents in this study”.
8. Fig 1 legend: Consider changing the title of the legend to make it more objective.
Fig 1 legend was revised to “Effects of 22AA on the apoptosis of SCs, sperm cells and the localization and distribution of occludin”.
9. There are typos that need to be corrected, some examples are as followed:
Line 42: remove ‘.’ before TJ
Line 60: at the end of the line ‘??’ is there. One’?’ to be removed.
Line 179: add a ‘-’ in t test.
The ‘.’ before TJ was removed in line 42. One’?’ was removed in line 60. A ‘-’ was added in t test in line 179
Reviewer 3
Basic reporting
The study investigates a relevant research question to understand the mechanisms underlying the regulation of tight junctions by 22AA. How 22AA affects the expression and localization of occludin? The manuscript covers appropriate content (literature, description of methodology and results) in specific sections.
Experimental design
Experimental design is well described now.
Validity of the findings
Findings appear to be valid.
Comments for the author
Language errors need correction:
In vivo and in vitro should be mentioned in italic font throughout the text.
L33: adluminal
L34: lumen
L42: punctuation error
In vivo and in vitro were mentioned in italic font throughout the text.
“abluminal” in line 33 was revised to “adluminal”.
“lumens” in line 34 was revised to “lumen”.
The ‘.’ before TJ was removed in line 42.
Yours
De-Yu Chen
Prof of Fuyang Normal University
West Qinghe Road 741, Fuyang, Anhui, China,236037
Tel & Fax. +86-558-2596113(O)
Mobile/Cell number. +86-18269992453
E-mail: chendeyu7104@aliyun.com
" | Here is a paper. Please give your review comments after reading it. |
9,791 | "<ns0:abstract xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:p>Occludin is a structural protein of tight junctions (TJ) in the blood-testis barrier (BTB). A 22amino-acid peptide (22AA) in the second extracellular loop can reversibly regulate TJ, but its regulatory mechanism is unknown. In this study, a 22AA-induced TJ destruction animal model was constructed to investigate the effect of 22AA on Sertoli cells (SCs) and spermatid count s and cell apoptosis at different time points using a multiplex immunofluorescence technique. The effect of 22AA on the location and distribution of occludin was analyzed via dual confocal fluorescence microscope . Western blotting was used to analyze dynamic changes in occludin expression. Real-time RT-PCR was used to analyze miR-122-5p expression changes. Sperm density counts and mating methods were used to analyze the effect of 22AA on fertility in mice. The results showed that 22AA promoted SC and spermatid apoptosis, downregulated occludin, upregulated miR-122-5p, and decreased sperm density and litter size before 27 days (27D). After 27D, the expression of occludin increased again, miR-122-5p expression decreased again, both sperm density and litter size returned to normal, apoptosis stopped, and spermatogenesis began to recover. Therefore, it can be concluded that 22AA can destroy TJ by downregulating occludin and inducing cell apoptosis. After 27D, TJ and spermatogenesis functions return to normal.</ns0:p></ns0:div>
</ns0:abstract>
<ns0:body xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:head n='1'>Introduction</ns0:head><ns0:p>In the testis, there is a blood-testis barrier (BTB) between the seminiferous tubules and blood vessels. The BTB composition includes the interstitial capillary endothelium and its basement membrane, connective tissue, and tight junctions (TJ) between the basement membrane of the seminiferous epithelium and Sertoli cells (SCs). TJ are the main structures that constitute the BTB <ns0:ref type='bibr'>[1]</ns0:ref>. One cause of male infertility is TJ abnormality in SCs, resulting in blockage of spermatogenic cell migration in seminiferous tubules <ns0:ref type='bibr' target='#b1'>[2]</ns0:ref>. The three-dimensional configuration of SCs is complex. SCs are irregularly cone-shaped, with the base closely touching the basement membrane, the top extending into the lumen, and many irregular depressions in the lateral surface and lumen surface, in which various levels of spermatogenic cells are embedded.</ns0:p><ns0:p>Membranes near the basal side of the adjacent SCs form TJ, and the seminiferous epithelium is divided into a basal compartment and adluminal compartment <ns0:ref type='bibr' target='#b2'>[3]</ns0:ref>. The basal compartment is located between the basement membrane of seminiferous epithelium and the TJ of SCs and contains spermatogonia cells (including type A and type B). The adluminal compartment, located above the TJ, connects with the lumen of seminiferous tubules and contains spermatocytes, germ cells, and sperm <ns0:ref type='bibr' target='#b3'>[4]</ns0:ref>. During spermatogenesis, the spermatocytes at the preleptotene stage and the leptotene stage differentiated from type B spermatogonia cells must move across the TJ into the abluminal compartment to complete their development <ns0:ref type='bibr' target='#b4'>[5]</ns0:ref>. Therefore, the movement of germ cells across the seminiferous epithelium during spermatogenesis is closely related to structural reconstruction of TJ.</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:07:51114:2:1:NEW 18 Sep 2020)</ns0:p><ns0:p>Manuscript to be reviewed Many TJ experimental models exist for studying the BTB. The in vitro BTB experimental model uses the testicular SC primary dual-chamber culture method to determine TJ tightness by measuring transepithelial electrical resistance <ns0:ref type='bibr' target='#b5'>[6]</ns0:ref>. Glycerol <ns0:ref type='bibr' target='#b6'>[7]</ns0:ref>, CdCl2 <ns0:ref type='bibr' target='#b7'>[8]</ns0:ref>, and blocking peptides <ns0:ref type='bibr' target='#b8'>[9]</ns0:ref> can also destroy testicular TJ. However, the destruction of testicular TJ induced by glycerin injection or CdCl 2 injection is irreversible. Thus, studying the dynamic changes in TJ-related molecules during the disintegration process and reconstruction process using these two animal models is difficult.</ns0:p><ns0:p>Structural proteins of TJ include occludin <ns0:ref type='bibr' target='#b9'>[10]</ns0:ref>, zonula occludens-1 (ZO-1) <ns0:ref type='bibr' target='#b10'>[11]</ns0:ref>, ZO-2 <ns0:ref type='bibr' target='#b12'>[12]</ns0:ref>, ZO-3, the claudin multigene family, and adhesion molecules <ns0:ref type='bibr' target='#b13'>[13]</ns0:ref>. Occludin is a 65-kDa protein localized at TJ <ns0:ref type='bibr' target='#b14'>[14,</ns0:ref><ns0:ref type='bibr' target='#b15'>15]</ns0:ref> that consists of four transmembrane domains, a long carboxyl-terminal cytoplasmic domain, a short N-terminal cytoplasmic domain, two extracellular loops, and one intracellular loop <ns0:ref type='bibr' target='#b16'>[16]</ns0:ref>. Its structure is highly conserved among different mammalian species <ns0:ref type='bibr' target='#b17'>[17]</ns0:ref>.</ns0:p><ns0:p>The first extracellular loop is rich in tyrosine and glycine, accounting for approximately 60% of the amino acid residues, and is involved in intercellular adhesion <ns0:ref type='bibr' target='#b18'>[18]</ns0:ref>. The second extracellular loop is involved TJ <ns0:ref type='bibr' target='#b19'>[19]</ns0:ref>. Occludin has also been reported to induce apoptosis and apoptotic sensitization, which are regarded as antitumorigenic activities <ns0:ref type='bibr' target='#b20'>[20]</ns0:ref>. However, the mechanism underlying the specific role of occludin in cell apoptosis remains poorly understood. </ns0:p></ns0:div>
<ns0:div><ns0:head n='2'>Experimental materials and methods</ns0:head></ns0:div>
<ns0:div><ns0:head n='2.1'>Experimental animals</ns0:head><ns0:p>Specific-pathogen-free Kunming (KM) mice, 6 to 8 weeks old, with a body weight of 30-35 g, were provided by Chongqing Enswell Biotechnology Co., Ltd., China. Animals were kept at a temperature of 23-25°C with a 12 h light:12 h dark (12 h-12 h) cycle. The animals were allowed food and water ad libitum. The experiment was approved by the Experimental Animal Ethics Committee of Fuyang Normal University, China (Grant No. 20200006).</ns0:p></ns0:div>
<ns0:div><ns0:head n='2.2'>Construction of animal models</ns0:head><ns0:p>The sequence of 22AA is NH2-GSQIYTICSQFYTPGGTGLYVD-COOH (from the 209th to 230th amino acid of occludin) <ns0:ref type='bibr' target='#b8'>[9]</ns0:ref>, and 22AA was synthesized by Beijing Protein Innovation Co., Ltd., China. To prepare a 200.0 g/L solution, 22AA was dissolved in 0.9% sterile saline. Male KM mice were anesthetized with 7% chloral hydrate (0.5 mL/100 g), the scrotal skin was cut, both (left and right) testicles were exposed, and a 26-gauge needle (Becton Dickinson, Manuscript to be reviewed Rutherford, NJ) was used for injection of 7.0 μL of 22AA solution at three different locations in each testicle. The control group was injected with 0.9% sterile saline. After injection, the testicles were sutured, and the mice were put back into the cage. Thirty-six experimental animals were divided into six groups, with six mice in each group: A, normal control group; B, 7 days (7D) after injection of 22AA; C, 17D after 22AA injection; D, 27D after 22AA injection; E, 37D after 22AA injection; and F, 47D after 22AA injection. After the mice were anesthetized by intraperitoneal injection of 7% chloral hydrate (0.5 mL/100 g) at different time points, the skin of the scrotum was removed, and one testis was removed and stored at -80°C for western blotting (WB) and quantitative PCR (qPCR). The other testis was fixed with 4% paraformaldehyde for morphological examination. At the end of the experiment, the mice were euthanized using 0.3 mL 7% chloral hydrate (0.5 mL/100 g) according to the American Veterinary Medical Association (AVMA)</ns0:p><ns0:p>Guidelines for the Euthanasia of Animals, 2013 Edition.</ns0:p></ns0:div>
<ns0:div><ns0:head n='2.3'>Multiplex immunofluorescence detection</ns0:head><ns0:p>Testicular tissues were collected at different stages, fixed and blocked using conventional methods, embedded in paraffin, deparaffinized, dehydrated, and subjected to heat-induced antigen retrieval. The types and sources of antibodies used in the experiments are shown in Table <ns0:ref type='table'>1</ns0:ref>. The sections were washed with distilled water, immersed in phosphate-buffered saline (PBS) for 5 min, and then blocked with goat serum at room temperature for 60 min. The blocking solutions were then aspirated, primary anti-Bax antibody (1:200) was added dropwise, and the sections were placed in a humidified box and incubated at 4°C overnight. The next day, the sections were incubated at room temperature for 30 min. The sections were immersed and washed in PBS three times for 3 min each. After the sections were blotted, a fluorescent Cy3labeled secondary antibody (1:800 dilution) was added dropwise, followed by incubation at 37°C</ns0:p><ns0:p>for 60 min in a humidified box. The sections were washed three times with PBS for 3 min each.</ns0:p><ns0:p>After the sections were blotted, primary anti-WT1 or anti-Prm2 antibody (1:100 dilution) was added dropwise, and the sections were incubated in a humidified box at 37°C for 60 min. The sections were washed three times with PBS for 3 min each. After the sections were blotted, a fluorescent FITC-labeled secondary antibody (1:800) was added dropwise, followed by incubation at 37°C for 60 min in a humidified box. After three washes with PBS for 3 min each, DAPI was added dropwise, followed by incubation in the dark for 15 min for nuclear staining.</ns0:p><ns0:p>Then, the sections were washed for 5 min four times. After excess DAPI was washed away, the slides were mounted in anti-fluorescence quencher mounting medium. The sections were then observed under a confocal fluorescence microscope (Leica, Germany), and images were collected. Fifty sections in different parts of each testicle were examined. Images were quantitatively analyzed using ImageJ software. The relative expression value of Bax was determined by the optical density value.</ns0:p></ns0:div>
<ns0:div><ns0:head n='2.4'>Immunofluorescence analysis of occludin localization and distribution</ns0:head><ns0:p>Testicular tissues were collected at different stages, fixed and blocked by conventional methods, embedded in paraffin, deparaffinized, dehydrated, and subjected to heat-induced antigen retrieval. The sections were cooled at room temperature for 10-20 min, rinsed with distilled water, and immersed in PBS for 5 min. Sections received 0.5% Triton X-100 dropwise,</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:07:51114:2:1:NEW 18 Sep 2020)</ns0:p><ns0:p>Manuscript to be reviewed followed by incubation at room temperature for 60 min and three PBS washes of 3 min each. The sections were blocked at room temperature for 60 min via dropwise addition of normal goat serum. The blocking solution was aspirated, and primary anti-occludin antibody (Abcam, America) diluted 1:100 was added dropwise. The sections were placed in a humidified box and incubated at 4°C overnight. The humidified box was removed and rewarmed to room temperature for 30 min. Sections were washed with PBS three times for 3 min each. After the excess liquid was removed from the sections, Alexa Fluor Cy3-labeled antibody (Abcam, America) diluted 1:800 was added dropwise, followed by incubation at 37°C for 60 min in a humidified box. Sections were washed with PBS three times for 3 min each, and DAPI (Beyotime, China) was added dropwise for nuclear staining in the dark for 15 min. After excess DAPI was washed away, the slides were mounted using anti-fluorescence quencher mounting medium. The sections were then observed under a confocal fluorescence microscope (Leica, Germany), and images were collected. Fifty sections in different parts of each testicle were examined.</ns0:p></ns0:div>
<ns0:div><ns0:head n='2.5'>Detection of occludin via WB</ns0:head><ns0:p>After extraction of total proteins from testes at different stages, the total protein content was determined using a BCA kit (Solarbio, China). After sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) at 100 V for 1.5 h, samples were transferred to a nitrocellulose (NC) membrane, which was blocked for 2 h via gentle shaking in 50.0 mL of TTBS (20.0 mmol/L Tris, pH 7.5, 0. 5 g/L Tween-220, 8.0 g/L NaCl) containing 50 g/L skim milk powder.</ns0:p><ns0:p>The NC membrane was placed in a plastic mantle and sealed after addition of 5.0 mL of rabbit Manuscript to be reviewed anti-mouse occludin monoclonal antibody (Abcam, America) at 1:500 dilution and then incubated at 4°C overnight. The NC membrane was taken out and washed with TTBS three times for 15 min each. Then, 5.0 mL of horseradish peroxidase-conjugated goat anti-rabbit IgGI</ns0:p><ns0:p>(1:2 000 dilution with TBS) (Jackson, 111-035-008) was added and incubated with the membrane with gentle shaking for 2 h. The NC membrane was taken out and washed with TTBS three times for 10 min each. novaECL reagent was added to the front surface of the NC membrane and allowed to stand for 1 min, and then, light-sensitive films were observed in a dark room. The band density was scanned using a digital gel image analysis system, and the gray value for density was measured with Image Lab 4.1. Using β-actin as the internal control, the expression of occludin was determined by the ratio of the gray values of occludin to those of βactin.</ns0:p></ns0:div>
<ns0:div><ns0:head n='2.6'>Real-time RT-PCR analysis of miR-122-5p</ns0:head><ns0:p>Testicular tissues were collected at different time points, and RNA was extracted using a</ns0:p><ns0:p>Trizol Total RNA Extraction Kit (Shanghai Sangong, China, catalog number: B511321).</ns0:p><ns0:p>According to the kit manual, RNA was extracted and reverse-transcribed into cDNA. The primers for miR-122-5p RT-PCR were F: CCTGGAGTGTGACAATG and R:</ns0:p><ns0:p>GAGCAGGCTGGAGAA. The primers for the internal control actin were F: Manuscript to be reviewed buffer, 5.0 μL of primers (25.0 pmol/μL), 0.5 μL of SYBR green I (20×), 2.0 μL of template (cDNA), and 21.5 μL of DEPC water. The amplification conditions were as follows: 94°C for 4 min; 35 cycles of 94°C for 20 sec, 60°C for 30 sec, and 72°C for 30 sec; followed by 72°C for detection. Relative expression was calculated using the 2 -△△Ct method.</ns0:p></ns0:div>
<ns0:div><ns0:head n='2.7'>Analysis of sperm density</ns0:head><ns0:p>Male mice were sacrificed via cervical dislocation at various time points. After abdominal disinfection, the abdominal wall was cut open to expose the reproductive system. The epididymis was separated via aseptic surgery, and the mesentery and fat surrounding the epididymis were removed with ophthalmic scissors and rinsed. The epididymis was shred and placed in a Petri dish containing 37°C PBS and then incubated at 37°C with 5% CO 2 and saturated humidity for 30 min. The sperm density was calculated using a cell counting plate after the sperm had spontaneously spread out.</ns0:p></ns0:div>
<ns0:div><ns0:head n='2.8'>Analysis of litter size</ns0:head><ns0:p>Male mice were co-caged with female mice (female:male = 1:1) at each time point. The time when a vaginal plug was detected was taken as the 0 day of pregnancy. On the 14th day of pregnancy, the mice were sacrificed by cervical dislocation, the uterus was removed by laparotomy, and the number of embryos in the bilateral uterus was recorded.</ns0:p></ns0:div>
<ns0:div><ns0:head n='2.9'>Statistical analysis</ns0:head><ns0:p>The experimental results are expressed as the mean ± standard deviation. GraphPad Prism 6</ns0:p><ns0:p>was used to complete the data processing. Differences between control group and experimental group were examined using a t -test. Differences among the six different experimental groups Manuscript to be reviewed were examined using two-way ANOVA, and correlations between groups were examined with a Pearson test. The significant difference level was set as P<0.01.</ns0:p></ns0:div>
<ns0:div><ns0:head n='3'>Results</ns0:head></ns0:div>
<ns0:div><ns0:head n='3.1'>22AA promotes SC apoptosis</ns0:head><ns0:p>Multiplex immunocytochemistry was used to analyze the effect of 22AA on SC apoptosis. Manuscript to be reviewed</ns0:p><ns0:note type='other'>Figure 1C.</ns0:note><ns0:p>Occludin was mainly located at TJ between the basement membrane of seminiferous tubules and SCs in the control group. At 7D, a small amount of occludin was distributed at TJ between the basement membrane of seminiferous tubule and SCs. However, the total number of cells in seminiferous tubules was less than that in the control group. At 17D and 27D, no cells with blue nuclei were found in the seminiferous tubules. From day 17 to day 27, the expression of occludin gradually decreased. At 37D, occludin expression began to increase and was found to be distributed on the basement membrane of seminiferous tubules and SCs. At 47D, occludin expression and distribution and the morphological structure of seminiferous tubules were highly similar to those in the control group. These results suggest that 22AA can reversibly affect the location and distribution of occludin.</ns0:p></ns0:div>
<ns0:div><ns0:head n='3.4'>22AA downregulates occludin</ns0:head><ns0:p>To analyze the dynamic changes in the occludin expression level after TJ disintegration and reconstruction, total testicular proteins were extracted at 0, 7, 17, 27, 37, or 47 days for western blot analysis. Representative western blot results are shown in Figure <ns0:ref type='figure' target='#fig_14'>2A</ns0:ref>. The relative occludin expression level was calculated using the occludin/β-actin gray density ratio. The detailed values are shown in Figure <ns0:ref type='figure' target='#fig_14'>2B</ns0:ref>. In the control group, the expression value was 0.9967. From 7D to 27D, the occludin expression level in the 22AA group gradually decreased to 0.1621, which was only <ns0:ref type='bibr' target='#b16'>16</ns0:ref>.26% of that in the control group. Then, the expression level of occludin gradually increased to 0.3543 at 47D, which was approximately one-third the normal expression level (35.54%).</ns0:p><ns0:p>Occludin expression was significantly different among the six groups (P < 0.05). The results</ns0:p><ns0:p>showed that the expression level of occludin in the 22AA group at each time was significantly Manuscript to be reviewed different from that in the control group (P < 0.01). These results indicate that 22AA can reversibly regulate occludin expression.</ns0:p></ns0:div>
<ns0:div><ns0:head n='3.5'>22AA upregulates miR-122-5p expression, and miR-122-5p expression is negatively correlated with occludin expression before 27D</ns0:head><ns0:p>To analyze the dynamic expression of miR-122-5p during TJ disintegration and reconstruction, total RNA in testes was extracted at 0, 7, 17, 27, 37, or 47 days. RT-PCR analysis was performed after the total RNA was reverse-transcribed into cDNA. The expression levels of miR-122-5p in each group are shown in Figure <ns0:ref type='figure' target='#fig_5'>3A</ns0:ref>. The miR-122-5p expression level in the control group was 0.0408 and increased to 0.0539 at 7D. The miR-122-5p expression level in the 22AA group at 27D was the highest at 0.1293. Then, miR-122-5p expression gradually decreased to 0.0867 at 47D but was still higher than that in the control group. The results showed that the miR-122-5p expression levels were significantly different among the six groups (P < 0.01). The correlation between miR-122-5p and occludin expression in each group was analyzed using Pearson correlation coefficient. The linear relationship is shown in Figure <ns0:ref type='figure' target='#fig_5'>3B</ns0:ref>. The results indicated that miR-122-5p and occludin expression are significantly negatively correlated (R 2 = -0.4905, P < 0.01).</ns0:p></ns0:div>
<ns0:div><ns0:head n='3.6'>22AA changes the sperm count</ns0:head><ns0:p>To analyze the effect of 22AA on sperm count, the epididymis was extracted at 0, 7, 17, 27, Manuscript to be reviewed and then gradually increased to 283.114 x 10 4 mL at 47D. The results showed a significant difference in sperm density among the six groups (P < 0.01).</ns0:p></ns0:div>
<ns0:div><ns0:head n='3.7'>22AA affects litter size</ns0:head><ns0:p>To analyze the effect of 22AA on the litter size of males, male mice were co-caged with female mice at 0, 7, 17, 27, 37, or 47 days. The number of embryos in the uterus on the 14th day of pregnancy was taken as a measure of fertility. The results are shown in Figure <ns0:ref type='figure'>4B</ns0:ref>. The largest litter size in the control group was 12, and then, litter size decreased gradually to 3.667 at 27D.</ns0:p><ns0:p>The litter size gradually increased to seven at 47D. The results showed that the difference in litter size among the six groups was significant (P < 0.01).</ns0:p></ns0:div>
<ns0:div><ns0:head n='4'>Discussion</ns0:head><ns0:p>As a component of TJ, occludin is the structural basis for TJ formation between SCs in the seminiferous epithelium. The programmed opening/resealing of TJ ensures normal progression of spermatogenesis, and abnormal opening/resealing can affect the normal spermatogenesis process <ns0:ref type='bibr' target='#b22'>[22]</ns0:ref>. Interference with the functional status of occludin protein in testicular SCs can result in infertility. TJ of the BTB are different from TJ of the blood-brain barrier and other barriers, and the specific function of BTB TJ between SCs is related to spermatogonia cell activity and differentiation. The disintegration and reconstruction of TJ between SCs is an important process <ns0:ref type='bibr' target='#b23'>[23]</ns0:ref>.</ns0:p><ns0:p>SCs are the structural basis of TJ in the testis <ns0:ref type='bibr' target='#b24'>[24]</ns0:ref>. The main functions of SCs include providing structural support, creating the BTB, participating in germ cell movement and ejaculation, and nurturing germ cells through the secretion process <ns0:ref type='bibr' target='#b25'>[25]</ns0:ref><ns0:ref type='bibr' target='#b26'>[26]</ns0:ref><ns0:ref type='bibr' target='#b27'>[27]</ns0:ref><ns0:ref type='bibr' target='#b28'>[28]</ns0:ref>. In the present study, Manuscript to be reviewed to analyze the dynamic changes in SCs during the process of TJ disintegration and reconstruction, a 22AA-induced TJ destruction animal model was utilized, and an immuno-double-labeling technique was used for analysis. WTI was employed as an SC marker <ns0:ref type='bibr' target='#b30'>[29]</ns0:ref>. The results showed that the number of cells and the seminiferous tubule wall thickness were decreased in SCs at 7D compared with the control group. These results were consistent with those reported by Chung et al. <ns0:ref type='bibr' target='#b8'>[9]</ns0:ref>. No blue nuclei were found in the seminiferous tubules at 17D or 27D, and the intercellular boundary had completely disappeared. WT1 was distributed throughout the seminiferous tubules. At 37D, a few SCs began to be found in the basal layer of the seminiferous tubules, and spermatogenesis began to recover. At 47D, there was almost no difference in SCs between the 22AA group and the control group, indicating that spermatogenesis had returned to normal. These phenomena reveal for the first time the changing pattern of SCs in the process of TJ disintegration and reconstruction. Next, dynamic changes in spermatids were analyzed using Prm2 as the spermatid marker <ns0:ref type='bibr'>[30] [31]</ns0:ref>. The results of multiplex immunohistochemistry showed that there were 55 ± 5 spermatids in the seminiferous tubules of the control group. At 7D, the spermatid count decreased to 15 ± 3, suggesting that the spermatids gradually became apoptotic.</ns0:p><ns0:p>No spermatids were found in seminiferous tubules at 17D or 27D. This finding indicates that the spermatids were completely apoptotic. A small amount of Prm2 was distributed in the seminiferous tubules at 37D, suggesting that spermatogenesis had begun to recover. At 47D, recovery of spermatids in seminiferous tubules was visible, and the spermatids numbered 20 ± 4, which was consistent with the experimental results of Chung et al. <ns0:ref type='bibr' target='#b8'>[9]</ns0:ref>. This finding indicates that spermatogenesis had returned to normal. These data suggest that 22AA mainly reduced the Manuscript to be reviewed number of spermatids in seminiferous tubules before 27D. After 27D, the normal structure of the TJ was restored, and spermatogenesis resumed.</ns0:p><ns0:p>The TJ-related structural proteins include ZO-1, ZO-2, and multiple claudin genes. Wong et al.</ns0:p><ns0:p>found that the 44-amino-acid peptide in the second extracellular loop of occludin had no effect on the expression level of ZO-1, ZO-2, or cingulin in the Xenopus kidney epithelial cell line A6 <ns0:ref type='bibr' target='#b21'>[21]</ns0:ref>. Therefore, in the present study, only the dynamic expression of occludin was quantitatively analyzed when investigating the mechanism underlying the destruction and recovery of TJ induced by 22AA. The expression level of occludin in the 22AA group gradually decreased from 7D to 27D until reaching only 14% of that in the control group. Then, the expression level of occludin gradually increased. At 47D, occludin expression recovered to approximately one-third the normal expression. These results suggest that 22AA can reduce the expression level of occludin at the disintegration stage of TJ. Therefore, it can be concluded that 22AA can downregulate occludin, leading to disintegration of TJ. The 22AA -induced disruption in the TJ barrier is possibly mediated by one of the following mechanisms. First, it might be possible that SCs were using these peptides as building blocks for TJ assembly. However, because they did not have the structural confirmation of the entire molecule to reinforce the TJ functionality, TJs became perturbed and disrupted. Second, homotypic interactions of the synthetic peptides with other intact occludin molecules between two neighboring SCs caused the recruitment of intact and functional occludin to the same site to become impossible. Thus, the TJ permeability barrier became disrupted.</ns0:p><ns0:p>There are three possible reasons for the decrease in protein expression. The first reason is Manuscript to be reviewed protein degradation <ns0:ref type='bibr' target='#b33'>[32]</ns0:ref>. Occludin phosphorylation and ubiquitination regulate TJ <ns0:ref type='bibr' target='#b34'>[33]</ns0:ref>. The western blot results showed that the molecular weight of occludin was consistent with the actual size, suggesting that occludin did not degrade. This result indicates that 22AA did not cause occludin ubiquitination. The second potential reason is cell apoptosis because cell apoptosis prevents cells from expressing relevant proteins <ns0:ref type='bibr' target='#b35'>[34,</ns0:ref><ns0:ref type='bibr' target='#b36'>35]</ns0:ref>. Therefore, SC and spermatid apoptosis might lead to a decrease in occludin expression. The third reason could be inhibition of transcription or translation by noncoding RNA <ns0:ref type='bibr' target='#b5'>[6]</ns0:ref>. To investigate the occludin downregulation mechanism, we analyzed the expression of several microRNAs (miRNAs) that can target occludin (data not shown), among which only the expression of miR-122-5p was associated with occludin expression. miR-122-5p is encoded on chromosome 18q21.31 and is derived from the hcr gene transcript. miR-122-5p plays an important role in cell cycle regulation, cell proliferation and cell apoptosis <ns0:ref type='bibr' target='#b37'>[36]</ns0:ref> and is associated with multiple diseases <ns0:ref type='bibr' target='#b37'>[36]</ns0:ref><ns0:ref type='bibr' target='#b38'>[37]</ns0:ref><ns0:ref type='bibr' target='#b39'>[38]</ns0:ref><ns0:ref type='bibr' target='#b40'>[39]</ns0:ref>. Previously, we analyzed the correlation between miR-122-5p and occludin protein, and the results showed that miR-122-5p was negatively correlated with occludin expression <ns0:ref type='bibr' target='#b42'>[40]</ns0:ref>. Our other recent results showed that miR-122-5p regulates occludin expression through the AACACTCCA sequence of the occludin 3'UTR, thereby regulating the formation and tightness of TJ between SCs (submitted). We further employed real-time RT-PCR to assess whether 22AA affects miR-122-5p expression.</ns0:p><ns0:p>The current results showed that the expression level of miR-122-5p gradually increased from 0 to 27D and then began to decrease. The change in the miR-122-5p expression level was opposite that of occludin. Correlation analysis showed a significant negative correlation between miR-122-5p and occludin expression from 0 to 47D(Figure <ns0:ref type='figure' target='#fig_5'>3B</ns0:ref>). These results indicate that 22AA</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:07:51114:2:1:NEW 18 Sep 2020)</ns0:p><ns0:p>Manuscript to be reviewed increased the expression of miR-122-5p, which mediated downregulation of occludin expression, thereby causing disintegration of TJ. However, the mechanism by which 22AA regulates mir-122-5p expression remains to be further studied.</ns0:p><ns0:p>To investigate the causes of SC and spermatid dysfunction, Bax was used as an apoptosis marker protein to analyze cell apoptosis in seminiferous tubules. Bax, belonging to the Bcl-2 gene family, is the most important apoptotic gene in humans. The encoded Bax protein forms heterodimers with Bcl-2 and has an inhibitory effect on Bcl-2. Bax is one of the most important apoptosis-promoting genes <ns0:ref type='bibr' target='#b43'>[41]</ns0:ref>. Bax expression is also closely related to spermatogenesis <ns0:ref type='bibr' target='#b44'>[42]</ns0:ref>.</ns0:p><ns0:p>In the present study, Bax expression was found in the seminiferous tubules of the control group at a low expression level of 11.373 ± 10.532. Yan et al. also found that Bax was expressed in various types of cells in normal testicular tissues <ns0:ref type='bibr' target='#b45'>[43]</ns0:ref>. At 7D, the expression level of Bax was 16.783 ± 10.157, which was higher than that in the control group, indicating that the cells in seminiferous tubules began to undergo apoptosis. However, at this stage, the structure of SCs and spermatids in the tubules was relatively intact. The expression level of Bax increased to 20.521 ± 5.781 and 30.253 ± 12.274 at 17D and 27D, respectively. At this stage, the tubules were filled with a large amount of red Bax, no blue nuclei were observed, and the intercellular boundary had completely disappeared, indicating that all the cells in the seminiferous tubules had already undergone apoptosis. At 37D, the expression level of Bax was slightly lower than at 27D.</ns0:p><ns0:p>In the basal layer of the seminiferous tubules, blue nuclei began to appear, indicating that apoptosis had slowed and the spermatids and SCs of the seminiferous tubules had started to recover. At 47D, the expression level of Bax was not significantly different from that in the Manuscript to be reviewed control group, and the structure and number of spermatids and SCs in seminiferous tubules were not different from those in the control group, indicating that spermatogenesis had fully recovered.</ns0:p><ns0:p>These results suggest that 22AA can induce apoptosis in seminiferous tubules before 27D.</ns0:p><ns0:p>To analyze the effect of 22AA on sperm count, the epididymis was extracted at 0, 7, 17, 27, 37, or 47 days. Sperm density was analyzed after the epididymis was shredded. The results are</ns0:p><ns0:p>shown in Figure <ns0:ref type='figure'>4A</ns0:ref>. The highest sperm density in the control group was 750.114 × 10 4 /mL, and then, the density gradually decreased, dropping to 164.278 × 10 4 /mL at 27D; subsequently, the sperm density gradually increased to 283.114 × 10 4 mL at 47D. However, compared with the control group, the sperm density was greatly decreased. The decrease in sperm count may be due to the following reasons. First, high expression of Bax promotes the apoptosis of type A spermatogonial stem cells <ns0:ref type='bibr' target='#b44'>[42]</ns0:ref>, thereby reducing spermatogenesis. Second, after the BTB is destroyed, immune cells enter the seminiferous tubules and engulf many sperm <ns0:ref type='bibr' target='#b47'>[44]</ns0:ref>. Third, sperm undergo apoptosis or autophagy in the epididymis <ns0:ref type='bibr' target='#b48'>[45,</ns0:ref><ns0:ref type='bibr' target='#b49'>46]</ns0:ref>. Further analysis is needed to determine which reason explains the decrease in sperm count. To analyze the effect of 22AA on litter size, male mice were co-caged with female mice at 0, 7, 17, 27, 37, or 47 days. The number of embryos in the uterus on the 14th day of pregnancy was used to measure fertility. The highest litter size in the control group was 12, but litter size decreased gradually to seven at 27D.</ns0:p><ns0:p>Afterwards, the litter size gradually increased to 6.67 at 47D. Image A, G and M were control panels, others image were different treatment panels.</ns0:p><ns0:p>Effects of 22AA on the apoptosis of SCs. Many SCs were present in the control group, with clear intercellular boundaries and a small amount of red Bax distribution(A). The morphology and number of SCs at 7D were not significantly different from those of the control group(B). At 17D and 27D, there were no blue nuclei in the seminiferous tubules, and the intercellular boundary completely disappeared(C, D). The WT1 patch was scattered throughout the seminiferous tubules, with no cells present in the tubules. The tubules were filled with a large amount of red Bax. A few SCs began to appear in the basal layer of the seminiferous tubules at 37D (E). At 47D, the morphology, structure and number of SCs in the seminiferous tubules were not different from those in the control group(F).</ns0:p><ns0:p>Effect of 22AA on the apoptosis of sperm cells. Nearly 55 ± 5 spermatids were found in the seminiferous tubules of the control group(G), and the expression level of Bax was 11.245 ± 4.868. At 7D, the spermatid count decreased to 15 ± 3(H), and the expression of Bax increased to 19.569 ± 6.158. No spermatid was found in the seminiferous tubules at 17D (I) or 27D(J), and the expression levels of Bax increased to 23.467 ± 5.327 and 31.353 ± 13.139, respectively. At 37D, a small amount of Prm2 was distributed in the seminiferous tubules(K), and the expression level of Bax decreased to 16.362 ± 3.267. At 47D, 20 ± 4 spermatids were rediscovered in the seminiferous tubules(L), and the expression level of Bax was 10.176 ± 1.682, which was not different from the control level.</ns0:p><ns0:p>Effect of 22AA on the localization and distribution of occludin. Occludin was mainly located at TJ between the basement membrane of seminiferous tubules and SCs in the control group(M). At 7D, a small amount of occludin was distributed at TJ between the basement membrane of seminiferous tubule and SCs. However, the total number of cells in seminiferous tubules was less than that in the control group(N). At 17D and 27D, no cells with blue nuclei were found in the seminiferous tubules(O,P). From day 17 to day 27, the expression of occludin gradually decreased. At 37D, occludin expression began to increase and was found to be distributed on the basement membrane of seminiferous tubules and SCs(Q). At 47D, occludin expression and distribution and the morphological structure of seminiferous tubules were highly similar to those in the control group(R). Bar-, 0.2 μm.</ns0:p><ns0:note type='other'>Figure 2</ns0:note><ns0:p>Effect of 22AA on the expression of occludin.</ns0:p><ns0:p>Representative western blot results are shown in Figure <ns0:ref type='figure' target='#fig_14'>2A</ns0:ref>. The relative occludin expression level was calculated using the occludin/β-actin gray density ratio. The detailed values are shown in Figure <ns0:ref type='figure' target='#fig_14'>2B</ns0:ref>. In the control group, the expression value was 0.9967. From 7D to 27D, the occludin expression level in the 22AA group gradually decreased to 0.1621, which was only 16.26% of that in the control group. Then, the expression level of occludin gradually increased to 0.3543 at 47D, which was approximately one-third the normal expression level (35.54%). Occludin expression was significantly different among the six groups (P < 0.05).</ns0:p><ns0:p>The results showed that the expression level of occludin in the 22AA group at each time was significantly different from that in the control group. ＊P < 0.01.</ns0:p><ns0:note type='other'>Figure 3</ns0:note><ns0:p>Effect of artificially synthesized 22AA on the expression of miR-122-5.</ns0:p><ns0:p>The expression levels of miR-122-5p in each group are shown in Figure <ns0:ref type='figure' target='#fig_5'>3A</ns0:ref>. The miR-122-5p expression level in the control group was 0.0408 and increased to 0.0539 at 7D. The miR-122-5p expression level in the 22AA group at 27D was the highest at 0.1293. Then, miR-122-5p expression gradually decreased to 0.0867 at 47D but was still higher than that in the control group. The results showed that the miR-122-5p expression levels were significantly different among the six groups (P < 0.01). The correlation between miR-122-5p and occludin expression in each group was analyzed using Pearson correlation coefficient.</ns0:p><ns0:p>The linear relationship is shown in Figure <ns0:ref type='figure' target='#fig_5'>3B</ns0:ref>. The results indicated that miR-122-5p and occludin expression are significantly negatively correlated ( R Manuscript to be reviewed</ns0:p><ns0:note type='other'>Figure 4</ns0:note><ns0:p>Effects of 22AA on sperm density and litter size.</ns0:p><ns0:p>The sperm density was analyzed after the epididymis was shredded. The results are shown in </ns0:p></ns0:div><ns0:figure xml:id='fig_0'><ns0:head /><ns0:label /><ns0:figDesc>Wong and Gumbiner et al. used an in vitro cell model to add an artificially synthesized 44amino-acid short peptide identical to the second extracellular loop sequence of occludin into the Xenopus kidney epithelial cell line A6, thereby reducing the tightness of TJ [21]. Chung et al. confirmed that a 22-amino-acid peptide (22AA) in the second extracellular loop of occludin PeerJ reviewing PDF | (2020:07:51114:2:1:NEW 18 Sep 2020) Manuscript to be reviewed could reversibly regulate TJ [9]. However, Chung et al. only analyzed the effect of 22AA on the morphological structure of the seminiferous tubules. What is the mechanism underlying the regulation of TJ by 22AA? Is the expression and localization of occludin affected by 22AA? How are SC and spermatid counts in the seminiferous tubules dynamically affected by 22AA? Can 22AA affect the number of offspring? The above questions should be studied in depth. To further investigate the mechanism underlying the regulation of TJ by 22AA, this study analyzed the effect of 22AA on occludin expression and localization, SC and spermatid apoptosis, and mouse fertility using a TJ damage animal model.</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_1'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:07:51114:2:1:NEW 18 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_2'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:07:51114:2:1:NEW 18 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_3'><ns0:head /><ns0:label /><ns0:figDesc>GAGACCTTCAACACCCCAGC and R: ATGTCACGCACGATTTCCC. The BR Green I protocol of the SYBR Green I method was used. The real-time fluorescence PCR kit TransStart Green qPCR SuperMix (catalog no. AQ131-01) was used for PCR amplification on a LightCycler 96 system (Roche, US). The reaction system was as follows: 25.0 μL of 2× PCR PeerJ reviewing PDF | (2020:07:51114:2:1:NEW 18 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_4'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:07:51114:2:1:NEW 18 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_5'><ns0:head>3. 3</ns0:head><ns0:label>3</ns0:label><ns0:figDesc>22AA affects occludin localization and distribution Dual immunofluorescence was used to analyze the effect of 22AA on the localization and distribution of occludin. The multiplex immunohistochemical results for each group are shown in PeerJ reviewing PDF | (2020:07:51114:2:1:NEW 18 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_6'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:07:51114:2:1:NEW 18 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_7'><ns0:head>37 ,</ns0:head><ns0:label>37</ns0:label><ns0:figDesc>or 47 days. The sperm density was analyzed after the epididymis was shredded. The results are shown in Figure 4A. The highest sperm density in the control group was 750.144 × 10 4 /mL, after which it gradually decreased. The sperm density decreased to 164.278 × 10 4 /mL at 27D PeerJ reviewing PDF | (2020:07:51114:2:1:NEW 18 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_8'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:07:51114:2:1:NEW 18 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_9'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:07:51114:2:1:NEW 18 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_10'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:07:51114:2:1:NEW 18 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_11'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:07:51114:2:1:NEW 18 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_12'><ns0:head /><ns0:label /><ns0:figDesc>In summary, this study investigated the effect of 22AA on TJ by using a 22AA-induced TJ destruction animal model. The results showed that before 27D, 22AA promoted SC and spermatid apoptosis, downregulated occludin, upregulated miR-122-5p, and decreased spermPeerJ reviewing PDF | (2020:07:51114:2:1:NEW 18 Sep 2020)Manuscript to be reviewed density and litter size. After 27D, the occludin expression increased, miR-122-5p expression decreased, both sperm density and litter size rebounded, cell apoptosis stopped, and spermatogenesis began to recover. Therefore, it can be concluded that 22AA destroys TJ by downregulating occludin and inducing cell apoptosis. After 27D, TJ and spermatogenesis functions return to normal.</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_13'><ns0:head>Figure 1 Figure 1 .</ns0:head><ns0:label>11</ns0:label><ns0:figDesc>Figure 1</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_14'><ns0:head>2 =</ns0:head><ns0:label>2</ns0:label><ns0:figDesc>-0.4905, P < 0.01). ＊P < 0.01. PeerJ reviewing PDF | (2020:07:51114:2:1:NEW 18 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_15'><ns0:head>Figure 4A . 4 / 4 /</ns0:head><ns0:label>4A44</ns0:label><ns0:figDesc>Figure 4A. The highest sperm density in the control group was 750.144 × 10 4 /mL, after</ns0:figDesc></ns0:figure>
<ns0:figure type='table' xml:id='tab_0'><ns0:head /><ns0:label /><ns0:figDesc>At 47D, the number of SCs and the Bax expression level were not significantly different from those in the control group. These results demonstrate that 22AA promoted SC apoptosis. Multiplex immunofluorescence was used to analyze the effect of 22AA on spermatids. Bax was used as the apoptosis marker protein to analyze spermatid apoptosis in the seminiferous tubules. Prm2 was used as the spermatid marker to determine the spermatid count. The multiplex immunohistochemical results for each group are shown in Figure 1B. Nearly 55 ± 5 spermatids were found in the seminiferous tubules of the control group, and the expression level of Bax was 11.245 ± 4.868. At 7D, the spermatid count decreased to 15 ± 3, and the expression of Bax increased to 19.569 ± 6.158, suggesting gradual spermatid apoptosis. No spermatid was found in the seminiferous tubules at 17D or 27D, and the expression levels of Bax increased to 23.467 ± 5.327 and 31.353 ± 13.139, respectively, indicating complete spermatid apoptosis. At 37D, a small amount of Prm2 was distributed in the seminiferous tubules, and the expression level of Bax decreased to 16.362 ± 3.267, indicating that apoptosis had begun to terminate. At 47D, 20 ± 4 spermatids were rediscovered in the seminiferous tubules, and the expression level of Bax was 10.176 ± 1.682, which was not different from the control level. This result indicates that 22AA can reversibly regulate spermatid apoptosis.</ns0:figDesc><ns0:table><ns0:row><ns0:cell>3.2 22AA reversibly regulates spermatid apoptosis</ns0:cell></ns0:row><ns0:row><ns0:cell>521 ± 5.781</ns0:cell></ns0:row><ns0:row><ns0:cell>and 30.253 ± 12.274 at 17D and 27D, respectively, which were significantly higher than those in</ns0:cell></ns0:row><ns0:row><ns0:cell>the control group (P < 0.01), indicating that at this stage all cells in the seminiferous tubules were</ns0:cell></ns0:row></ns0:table><ns0:note>Bax was used as the cell apoptosis marker protein to calculate the SC apoptosis rate in seminiferous tubules. WT1 was used as a marker of SCs. The multiplex immunohistochemical results for each group are shown in Figure1A. Many SCs were present in the seminiferous tubules of the control group, with clear intercellular boundaries and a small amount of red Bax distribution, and the expression level of Bax was relatively low at 11.373 ± 10.532. At 7D, the SC count in the seminiferous tubules was not significantly different from that in the control group, and the expression level of Bax was 16.783 ± 10.157, which was significantly higher than that in the control group (P < 0.01), suggesting that cell apoptosis started in the seminiferous tubules. At 17D and 27D, no cells with blue nuclei were found in the seminiferous tubules, and the intercellular boundary completely disappeared. The WT1 patch was scattered throughout the seminiferous tubules, and no cell structure was found in the tubules. The seminiferous tubules were filled with a large amount of red Bax, and the expression levels of Bax were 20.already apoptotic. At 37D, the expression level of Bax was 20.862 ± 3.243, which was roughly equivalent to the expression level of Bax at 17D, indicating that apoptosis had begun to terminate;PeerJ reviewing PDF | (2020:07:51114:2:1:NEW 18 Sep 2020)Manuscript to be reviewed a few SCs began to appear in the basal layer of seminiferous tubules; and TJ reconstruction started.</ns0:note></ns0:figure>
</ns0:body>
" | "Dear editor
Thank you for your email. According to the suggestions and requirements of you and the reviewers, I have revised the article. The revised partes are marked in the text. The following is a respond to the reviewer's comments.
Reviewer 1
Thanks for your review.
Reviewer 2
1.I would suggest adding some of the short and long term medical consequences of the 22AA mediated effects.
Thank you for your good advice. It had been some time since the experiment ended, and the animals had been disposed of. If the experiment is redesigned and the treatment time of 22AA on experimental animals is increased or shortened, it will cost me too much time and money. In the future when similar experimental design, I will design more perfect.
2. Fig 1: Images need to be properly labeled. I would specifically suggest labeling control and 22AA peptide treatment panels.
Figure 1 was relabeled. The sentence “Image A, G and M were control panels, others image were different treatment panels.”was inserted to the legend of Figure 1.
3. The author should discuss in detail the reason of disappearance of the 22AA mediated phenotypes post 27 days.
The 22AA -induced disruption in the TJ barrier is possibly mediated by one of the following mechanisms. First, it might be possible that SCs were using these peptides as building blocks for TJ assembly. However, because they did not have the structural confirmation of the entire molecule to reinforce the TJ functionality, TJs became perturbed and disrupted. Second, homotypic interactions of the synthetic peptides with other intact occludin molecules between two neighboring SCs caused the recruitment of intact and functional occludin to the same site to become impossible. Thus, the TJ permeability barrier became disrupted.
This part of content was inserted into LL384-392.
Yours
De-Yu Chen
Prof of Fuyang Normal University
West Qinghe Road 741, Fuyang, Anhui, China,236037
Tel & Fax. +86-558-2596113(O)
Mobile/Cell number. +86-18269992453
E-mail: chendeyu7104@aliyun.com
" | Here is a paper. Please give your review comments after reading it. |
9,792 | "<ns0:abstract xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:p>Smad nuclear interacting protein 1 (SNIP1) is a nuclear protein and involved in essential biological processes. MicroRNAs are effective regulators of tumorigenesis and cancer progression via targeting multiple genes. In present study, we aimed to investigate the function of SNIP1 and identify novel miRNA-SNIP1 axis in the development of cervical cancer. The results showed for the first time that silencing of the SNIP1 gene inhibited the migration and proliferation in HeLa cells significantly. Bioinformatics analysis and dual luciferase reporter assay demonstrated that miR-29a-3p could target 3'UTR of SNIP1 directly. The mRNA and protein expression levels of SNIP1 were negative regulated by miR-29a-3p according to the RT-qPCR and Western blot analysis, respectively. Furthermore, functional studies showed that over-expression of miR-29a-3p restrained HeLa cells migration and proliferation, and the mRNA expression of SNIP1 downstream genes (HSP27, c-Myc, and cyclin D1) were down-regulated by miR-29a-3p. Together, we concluded that miR-29a-3p suppressed the migration and proliferation in HeLa cells by directly targeting SNIP1. The newly identified miR-29a-3p/SNIP1 axis could provide new insight into the development of cervical cancer.</ns0:p></ns0:div>
</ns0:abstract>
<ns0:body xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:head>Introduction</ns0:head><ns0:p>Cervical cancer is the fourth most prevalent malignancy and remains the major cause of cancer death among females worldwide <ns0:ref type='bibr' target='#b3'>(Bray et al., 2018)</ns0:ref>.Despite there are considerable improvements of treatment strategies in the therapy of cervical cancer, the overall 5-year survival rates of patients remain poor for metastasis <ns0:ref type='bibr' target='#b11'>(Forouzanfar et al., 2011;</ns0:ref><ns0:ref type='bibr' target='#b40'>Tewari et al., 2014;</ns0:ref><ns0:ref type='bibr' target='#b24'>Li, Wu, & Cheng, 2016)</ns0:ref>. Therefore, understanding the pathogenesis and progression of cervical cancer is quite necessary and may facilitate identification of effective therapeutic targets for cervical cancer treatment.</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:06:49760:1:0:NEW 14 Sep 2020)</ns0:p><ns0:p>Manuscript to be reviewed Smad nuclear interacting protein 1 (SNIP1)，as an evolutionarily conserved nuclear protein, is involved in essential biological processes such as cell proliferation <ns0:ref type='bibr' target='#b37'>(Roche et al,. 2004;</ns0:ref><ns0:ref type='bibr'>Fujii et al., 2006)</ns0:ref>, small RNA biogenesis <ns0:ref type='bibr' target='#b52'>(Yu et al., 2008)</ns0:ref>, DNA damage response <ns0:ref type='bibr'>(Chen et al., 2018)</ns0:ref>, and several signaling pathways <ns0:ref type='bibr' target='#b21'>(Kim et al., 2000;</ns0:ref><ns0:ref type='bibr'>Kim et al., 2001)</ns0:ref>. In patients with non-small cell lung cancer or tongue squamous cell carcinoma, SNIP1 might be a reliable prognostic indicator <ns0:ref type='bibr' target='#b27'>(Liang et al., 2011;</ns0:ref><ns0:ref type='bibr' target='#b20'>Jeon et al., 2013)</ns0:ref>. Recently, SNIP1 was considered to be targeted by microRNA-335 and involved in osteosarcoma proliferation and metastasis <ns0:ref type='bibr' target='#b44'>(Xie et al., 2019)</ns0:ref>.</ns0:p><ns0:p>However, the exact role of SNIP1 in the development of cervical cancer remains obscure.</ns0:p><ns0:p>MicroRNAs (miRNAs) are a group of non-coding RNAs with 19-25 nucleotides and have closely relationships with the occurrence and progression of human cancers <ns0:ref type='bibr' target='#b38'>(Romano et al., 2017)</ns0:ref>. MicroRNAs regulate gene expression mainly through binding to the 3'-untranslated region (UTR) of target mRNAs <ns0:ref type='bibr' target='#b0'>(Bartel, 2004)</ns0:ref>. As oncogene or tumor suppressor in different cancers, miRNAs play essential roles in basic biological processes such as cell proliferation, apoptosis, differentiation, migration and invasion <ns0:ref type='bibr' target='#b0'>(Bartel, 2004;</ns0:ref><ns0:ref type='bibr' target='#b30'>Pardini et al., 2018)</ns0:ref>.</ns0:p><ns0:p>Accumulating studies have shown that miR-29a was abnormally expressed in various cancers, including cervical cancer <ns0:ref type='bibr' target='#b34'>(Pei, Lei, & Liu, 2016;</ns0:ref><ns0:ref type='bibr' target='#b46'>Yang et al., 2017;</ns0:ref><ns0:ref type='bibr' target='#b16'>Gong et al., 2019)</ns0:ref>. However, the detailed role and underlying mechanism of miR-29a in cervical cancer is still largely unclear.</ns0:p><ns0:p>In this research, we investigated the biological role of SNIP1 in the progression of cervical cancer. Furthermore, we predicted and demonstrated that miR-29a-3p inhibited the transcription and protein expression of SNIP1 by targeting 3'UTR directly, and suppressed the proliferation and migration of cervical cancer cells.</ns0:p></ns0:div>
<ns0:div><ns0:head>Materials & Methods</ns0:head><ns0:p>PeerJ reviewing PDF | (2020:06:49760:1:0:NEW 14 Sep 2020)</ns0:p></ns0:div>
<ns0:div><ns0:head>Manuscript to be reviewed</ns0:head></ns0:div>
<ns0:div><ns0:head>Cell culture</ns0:head><ns0:p>The human cervical cancer HeLa cells were purchased from American Type Culture Collection (Manassas, VA), and maintained in Dulbecco's modified Eagle's medium (Hyclone, USA) containing 10% fetal bovine serum (Gibco, USA) under a humidified environment with 5% CO 2 at 37 o C.</ns0:p></ns0:div>
<ns0:div><ns0:head>Transfection</ns0:head><ns0:p>Three small interfering RNAs (siRNAs) targeting SNIP1 (siSNIP1-330, siSNIP1-871, siSNIP1-1059) were purchased from GenePharma (Suzhou, China). The miR-29a-3p mimics and negative control (NC) were manufactured by RiboBio (Guangzhou, China). All transfections were performed using siRNA-Mate reagent (GenePharma, China) in accordance with the instruction manual. The cells were collected after 48 h of transfection for further experiments.</ns0:p></ns0:div>
<ns0:div><ns0:head>Quantitative RT-PCR (RT-qPCR)</ns0:head><ns0:p>Total RNA was harvested from HeLa cells by EZNA Total RNA Kit (Omega BioTek, USA).</ns0:p><ns0:p>The first-strand cDNA was generated using HiScript II Q RT SuperMix (Vazyme, China). RT-qPCR was performed to quantify relative RNA levels using ChamQ Universal SYBR qPCR Master Mix (Vazyme, China) on a CFX96 Touch (Bio-rad, USA). The 2 -ΔΔCt method was used to measure the relative expression level, and GAPDH served as the internal reference. The primers used for RT-qPCR were presented in Table1.</ns0:p></ns0:div>
<ns0:div><ns0:head>Western blot analysis</ns0:head><ns0:p>The cultured cells were lysed with RIPA buffer (Beyotime, China). The protein concentration was quantified by Enhanced BCA Assay kit (Beyotime, China). Total protein samples were separated with 10% SDS-PAGE gel and transferred onto PVDF membrane (Millipore, USA).</ns0:p><ns0:p>After blocked with 5% skim milk, the membrane was probed with primary antibodies against</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:06:49760:1:0:NEW 14 Sep 2020)</ns0:p><ns0:p>Manuscript to be reviewed </ns0:p></ns0:div>
<ns0:div><ns0:head>Dual luciferase reporter assay</ns0:head><ns0:p>To determine the binding affinity between SNIP1 and miR-29a-3p, the recombinant psiCHECK-2 vectors (Promega, USA) with the wild type (WT) or mutant of SNIP1 gene 3'-UTR were constructed. Then, the recombinant vectors (WT or mutant) and miR-29a-3p mimics (or NC)</ns0:p><ns0:p>were co-transfected in HeLa cells. Dual-Luciferase Reporter Assay system (Promega, USA) was used to calculate the luciferase activity after transfection according to the manual instruction.</ns0:p></ns0:div>
<ns0:div><ns0:head>Scratch assay</ns0:head><ns0:p>Transfected HeLa cells were cultured in 6-well plates until the confluence reached 100%. Then, a sterile pipet tip was used to scrape on the bottom of culture plates. Cell migration was observed at 12 h under an inverted microscope (Olympus, Japan) and images were captured for each sample. The scratch area was measured and analyzed by ImageJ software (NIH, MD, USA).</ns0:p></ns0:div>
<ns0:div><ns0:head>Cell Counting Kit-8 assay</ns0:head><ns0:p>Hela cells (3000 cells/well) were seeded into 96-well plates after transfection, and cultured for 24 h, 48 h, 72 h and 96 h. Fresh medium with 10 % CCK-8 (Genview Scientific, AUS) was mixed carefully, and the absorbance values of 450 nm wavelength were detected at least three times by a spectrophotometric plate reader (Hitachi, Japan).</ns0:p></ns0:div>
<ns0:div><ns0:head>Transwell migration assay</ns0:head><ns0:p>Transfected cells (5×10 4 ) were placed into the top chamber of the transwell inserts (Corning, USA) and maintained with serum-free medium. Then the lower chamber was filled with</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:06:49760:1:0:NEW 14 Sep 2020)</ns0:p><ns0:p>Manuscript to be reviewed complete medium. And the migrated cells were fixed and stained 24 hours later. Finally, the cells stained in more than three visual fields were randomly selected and photographed with an inverted microscope (Olympus, Japan) and counted.</ns0:p></ns0:div>
<ns0:div><ns0:head>Statistical analysis</ns0:head><ns0:p>Statistical analysis was carried out using the GraphPad Prism 8 (GraphPad Software, SanDiego, USA) program. All results were presented as mean ± standard error (SD) of three independent experiments. Comparison among multiple groups was assessed by Student's t test. A statistically significant difference was defined as P < 0.05.</ns0:p></ns0:div>
<ns0:div><ns0:head>Results</ns0:head></ns0:div>
<ns0:div><ns0:head>Knockdown of SNIP1 reduced migration and proliferation in cervical cancer cells</ns0:head><ns0:p>To address the biological role of SNIP1 in the progression of cervical cancer, three small interference RNAs (siRNAs) targeting SNIP1 were synthesized and transfected into HeLa cells.</ns0:p><ns0:p>Among them, siSNIP1-330 showed the best silencing effect (Fig. <ns0:ref type='figure' target='#fig_2'>1A and B</ns0:ref>), which was subsequently chosen for further analysis. After transfected with siSNIP1-330, the wound area ratio increased (Fig. <ns0:ref type='figure' target='#fig_2'>1C</ns0:ref>) and the number of migrated cells decreased (Fig. <ns0:ref type='figure' target='#fig_2'>1D</ns0:ref>), which demonstrated the migration of HeLa cells was suppressed. Furthermore, the cell proliferation rate declined after transfection for 48 h (Fig. <ns0:ref type='figure' target='#fig_2'>1E</ns0:ref>). In addition, the mRNA expression levels of several migration-related genes (MMP9, VIM, MAPK1, N-cadherin and E-cadherin) and proliferationrelated genes (CDK2) in HeLa cells <ns0:ref type='bibr'>(Wang & Chen, 2019)</ns0:ref> can also be downregulated or upregulated (Fig. <ns0:ref type='figure' target='#fig_2'>1F</ns0:ref>). Hence, SNIP1 knockdown could reduce migration and proliferation in cervical cancer HeLa cells.</ns0:p></ns0:div>
<ns0:div><ns0:head>SNIP1 was directly targeted by miR-29a-3p</ns0:head><ns0:p>PeerJ reviewing PDF | (2020:06:49760:1:0:NEW 14 Sep 2020)</ns0:p><ns0:p>Manuscript to be reviewed</ns0:p><ns0:p>To further confirm the exact miRNA that can directly target SNIP1, three different bioinformation tools starBase, TargetScan and miRanda, were performed to calculate the possibility scores. Intersection of these three sets showed that there were 12 candidate miRNAs which target SNIP1(Fig. <ns0:ref type='figure' target='#fig_3'>2A</ns0:ref>), and miR-29a-3p was chosen for the highest score (Table <ns0:ref type='table' target='#tab_0'>2</ns0:ref>).</ns0:p><ns0:p>Moreover, the expression relationship between miR-29a-3p and SNIP1 is negative correlation (P<0.05) in cervical squamous cell carcinoma and endocervical adenocarcinoma (CESC) samples from starBase (Fig. <ns0:ref type='figure' target='#fig_3'>2B</ns0:ref>, Table <ns0:ref type='table' target='#tab_0'>2</ns0:ref>). Compared with the control group, the luciferase activity was inhibited significantly in HeLa cells when co-transfected with SNIP1 wild type 3'-UTR vector and miR-29a-3p mimics (Figs. <ns0:ref type='figure' target='#fig_3'>2C and D</ns0:ref>). Furthermore, both mRNA and protein expression levels of SNIP1 in HeLa cells were declined after transfection with miR-29a-3p mimics (Figs. <ns0:ref type='figure' target='#fig_3'>2E and F</ns0:ref>).Taken together, these results demonstrated that miR-29a-3p targeted SNIP1 via directly binding its 3' UTR region and negatively regulated SNIP1 expression in cervical cancer.</ns0:p></ns0:div>
<ns0:div><ns0:head>MiR-29a-3p inhibited migration and proliferation in cervical cancer cells</ns0:head><ns0:p>To evaluate the regulatory roles of miR-29a-3p in HeLa cells, the scratch assay and transwell assay for migration were performed. As results shown in Figs. <ns0:ref type='figure'>3A and C</ns0:ref>, transfection with miR-29a-3p mimics restrained migration in HeLa cells. Moreover, miR-29a-3p mimics also significantly decreased the relative cell viability in HeLa cells (Fig. <ns0:ref type='figure'>3B</ns0:ref>). Additionally, miR-29a-3p also regulated the mRNA expression levels of genes associated with migration or proliferation in HeLa cells (Fig. <ns0:ref type='figure'>3D</ns0:ref>). Therefore, these data indicated that miR-29a-3p inhibited migration and proliferation in cervical cancer cells.</ns0:p></ns0:div>
<ns0:div><ns0:head>MiR-29a-3p regulated the mRNA expression of SNIP1 downstream genes</ns0:head><ns0:p>PeerJ reviewing PDF | (2020:06:49760:1:0:NEW 14 Sep 2020)</ns0:p><ns0:p>Manuscript to be reviewed</ns0:p><ns0:p>To investigate whether miR-29a-3p would have effects on the downstream of SNIP1, the mRNA levels of these downstream genes (HSP27, c-Myc and Cyclin D1) <ns0:ref type='bibr'>(Fujii et al., 2006;</ns0:ref><ns0:ref type='bibr' target='#b1'>Bracken et al., 2008;</ns0:ref><ns0:ref type='bibr' target='#b56'>Zhu et al., 2010)</ns0:ref> were detected by RT-qPCR after transfection with miR-29a-3p mimics or siSNIP1-330. The data suggested that miR-29a-3p and knockdown of SNIP1 both reduced the expression of those genes significantly (Fig. <ns0:ref type='figure'>4A</ns0:ref>).</ns0:p></ns0:div>
<ns0:div><ns0:head>Discussion</ns0:head><ns0:p>SNIP1 is a transcription regulator contains a nuclear localization sequence, and plays a key role in tumor development and progression <ns0:ref type='bibr' target='#b21'>(Kim et al., 2000;</ns0:ref><ns0:ref type='bibr'>Kim et al., 2001;</ns0:ref><ns0:ref type='bibr'>Fujii et al., 2006;</ns0:ref><ns0:ref type='bibr' target='#b1'>Bracken et al., 2008)</ns0:ref>. Over-expression of SNIP1 promoted cell invasion and migration in osteosarcoma cells <ns0:ref type='bibr' target='#b44'>(Xie et al., 2019)</ns0:ref>. And the knockdown of SNIP1 restrained the anchorageindependent growth of lung cancer cells <ns0:ref type='bibr' target='#b20'>(Jeon et al., 2013)</ns0:ref>. However, the function of SNIP1 in cervical cancer development is poorly understood. In our analysis, the migration and proliferation of cervical cancer cells was significantly suppressed after siRNA-mediated silencing of SNIP1 (Figs. <ns0:ref type='figure' target='#fig_2'>1C, D and E</ns0:ref>). These results suggested that SNIP1 was involved in the development of HeLa cells as an oncogene.</ns0:p><ns0:p>As a member of the miR-29 family, miR-29a is considered to play a crucial role in the regulation of multiple cancers <ns0:ref type='bibr' target='#b42'>(Wang et al., 2018)</ns0:ref>. It was up-regulated and promoted epithelialmesenchymal transition, migration and invasion in breast cancer cells <ns0:ref type='bibr' target='#b43'>(Wu et al., 2019)</ns0:ref>. In contrast, other studies have shown that miR-29a was down-regulated and inhibited the progression of cancers such as colon cancer <ns0:ref type='bibr' target='#b43'>(Shi et al., 2019)</ns0:ref>, non-small cell lung cancer <ns0:ref type='bibr' target='#b19'>(Hu et al., 2016)</ns0:ref> and adenocarcinoma <ns0:ref type='bibr' target='#b54'>(Zhang et al., 2018)</ns0:ref>. Previous studies have shown that miR-29a was low expression in cervical cancer, as a tumor suppressor, miR-29a can target multiple genes, such as CDC42, HSP47, SIRT1 and DNMT1 <ns0:ref type='bibr' target='#b32'>(Park et al., 2009;</ns0:ref><ns0:ref type='bibr'>Yamamoto et al., 2013;</ns0:ref><ns0:ref type='bibr' target='#b16'>Gong et</ns0:ref> PeerJ reviewing PDF | (2020:06:49760:1:0:NEW 14 Sep 2020)</ns0:p><ns0:p>Manuscript to be reviewed <ns0:ref type='bibr'>al., 2019;</ns0:ref><ns0:ref type='bibr' target='#b29'>Nan et al., 2019)</ns0:ref>. Nonetheless, the detailed biological function of miR-29a in cervical cancer has not yet been completely revealed. In this research, we showed for the first time that SNIP1 was targeted by miR-29a-3p in cervical HeLa cells. The results indicated that miR-29a-3p could down-regulate SNIP1 expression levels (Figs. <ns0:ref type='figure' target='#fig_3'>2E and F</ns0:ref>), and the direct binding site of SNIP1 mRNA 3'UTR was confirmed by dual luciferase reporter assay (Fig. <ns0:ref type='figure' target='#fig_3'>2D</ns0:ref>). Furthermore, our data also supported previous reports <ns0:ref type='bibr'>(Yamamoto et al., 2013)</ns0:ref>, indicating that miR-29a-3p can suppress the migration and proliferation of HeLa cells (Figs. <ns0:ref type='figure'>3A, B and C</ns0:ref>). Sharing the same seed region in miR-29 family (miR-29a/b/c), we inferred that SNIP1 may also be targeted by other miR-29s.</ns0:p><ns0:p>It has been reported that SNIP1 could improve the transcriptional activity of c-Myc <ns0:ref type='bibr'>(Fujii et al., 2006)</ns0:ref>, and regulate the stability of Cyclin D1 mRNA <ns0:ref type='bibr' target='#b1'>(Bracken et al., 2008)</ns0:ref>. Otherwise, SNIP1 could down-regulate the transcription of HSP27 <ns0:ref type='bibr' target='#b56'>(Zhu et al., 2010)</ns0:ref>. These downstream genes (c-Myc, Cyclin D1 and HSP27) regulated by SNIP1 are considered to be closely associated with cell proliferation and migration <ns0:ref type='bibr' target='#b9'>(Evan et al., 1994;</ns0:ref><ns0:ref type='bibr' target='#b56'>Zhu et al., 2010;</ns0:ref><ns0:ref type='bibr' target='#b25'>Li et al., 2012;</ns0:ref><ns0:ref type='bibr' target='#b35'>Pestell, 2013)</ns0:ref>. In this research, miR-29a-3p decreased the mRNA expression levels of c-Myc, Cyclin D1 and HSP27 in HeLa cells as an upstream regulator (Fig. <ns0:ref type='figure'>4A</ns0:ref>). Therefore, miR-29a-3p may mediate the regulation of cell proliferation and migration in cervical cancer cells via downstream genes of SNIP1 (Fig. <ns0:ref type='figure'>4B</ns0:ref>). Further study is required to illustrate the underlying mechanism of miR-29a-3p/SNIP1 pathway in cervical cancer oncogenesis.</ns0:p><ns0:p>In addition, this research has some defects in the following aspects. Although it was confirmed that miR-29a-3p can target SNIP1 to inhibit the migration and proliferation of HeLa cells, whether over-expression of SNIP1 could supplement the inhibitory effect of miR-29a-3p should be further observed. It was more convinced to detect the protein levels related to migration, PeerJ reviewing PDF | (2020:06:49760:1:0:NEW 14 Sep 2020)</ns0:p><ns0:p>Manuscript to be reviewed proliferation as well as downstream of SNIP1, while the mRNA levels were reduced markedly by miR-29a-3p or SNIP1 siRNA in HeLa cells. In this study, we have used only one cell line to verify the function of miR-29a-3p/SNIP1, more cell lines and more in vitro and in vivo experiments are needed to be carried out. Finally, miR-590-3p is worth exploring as miR-29a-3p</ns0:p><ns0:p>for the Pearson correlation analysis suggested that miR-590-3p was negatively correlated with SNIP1 in CESC tissues(r = -0.149, P = 0.00883) (Table <ns0:ref type='table' target='#tab_0'>2</ns0:ref>).</ns0:p></ns0:div>
<ns0:div><ns0:head>Conclusions</ns0:head><ns0:p>In conclusion, miR-29a-3p suppressed the migration and proliferation of cervical cancer cells by directly targeting SNIP1, and could also down-regulate the mRNA expression of SNIP1 downstream genes such as c-Myc, Cyclin D1 and HSP27. The newly identified miR-29a-3p/SNIP1 axis may provide new insights into the understanding of the progression of cervical cancer, and represent an effective treatment target for cervical cancer.</ns0:p><ns0:p>Table <ns0:ref type='table'>1</ns0:ref> Primers for quantitative RT-PCR Manuscript to be reviewed GAPDH served as the internal control. *P<0.05.</ns0:p></ns0:div>
<ns0:div><ns0:head>Gene</ns0:head></ns0:div><ns0:figure xml:id='fig_0'><ns0:head>SNIP1 ( 1</ns0:head><ns0:label>1</ns0:label><ns0:figDesc>:1000, Proteintech, USA) and GAPDH (1:20000, Proteintech, USA) at 4 o C overnight and followed by incubation with secondary antibodies (1:4000, Beyotime, China). Blots were visualized by BeyoECL Plus Kit (Beyotime, China) and scanned with a ChemiDoc XRS imaging system (Bio-Rad, USA).</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_2'><ns0:head>Figure 1</ns0:head><ns0:label>1</ns0:label><ns0:figDesc>Figure 1</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_3'><ns0:head>Figure 2 miR</ns0:head><ns0:label>2</ns0:label><ns0:figDesc>Figure 2</ns0:figDesc></ns0:figure>
<ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='25,42.52,331.87,525.00,270.75' type='bitmap' /></ns0:figure>
<ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='26,42.52,280.87,525.00,156.00' type='bitmap' /></ns0:figure>
<ns0:figure type='table' xml:id='tab_0'><ns0:head>Table 2 (on next page)</ns0:head><ns0:label>2</ns0:label><ns0:figDesc>Potential miRNAs binding SNIP1 and co-expression analysis for the miRNA-SNIP1 in</ns0:figDesc><ns0:table><ns0:row><ns0:cell>CESC</ns0:cell></ns0:row><ns0:row><ns0:cell>Note: CESC, cervical squamous cell carcinoma and endocervical adenocarcinoma. 1, analysis</ns0:cell></ns0:row><ns0:row><ns0:cell>from TargetScan7.2 (http://www.targetscan.org/vert_72/). 2, analysis from starBase v3.0 pan-</ns0:cell></ns0:row><ns0:row><ns0:cell>cancer analysis(http://starbase.sysu.edu.cn/panMirCoExp.php).</ns0:cell></ns0:row></ns0:table><ns0:note>PeerJ reviewing PDF | (2020:06:49760:1:0:NEW 14 Sep 2020)Manuscript to be reviewed 1</ns0:note></ns0:figure>
<ns0:figure type='table' xml:id='tab_1'><ns0:head>Table 2</ns0:head><ns0:label>2</ns0:label><ns0:figDesc /><ns0:table><ns0:row><ns0:cell /><ns0:cell cols='4'>Potential miRNAs binding SNIP1 and</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell /><ns0:cell cols='4'>co-expression analysis for the miRNA-SNIP1 in CESC</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell /><ns0:cell>Position</ns0:cell><ns0:cell /><ns0:cell>context++ score</ns0:cell><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>miRNA</ns0:cell><ns0:cell>in the UTR 1</ns0:cell><ns0:cell>seed match 1</ns0:cell><ns0:cell>percentile 1</ns0:cell><ns0:cell>r 2</ns0:cell><ns0:cell>p-value 2</ns0:cell></ns0:row><ns0:row><ns0:cell>hsa-miR-29a-3p</ns0:cell><ns0:cell>293-299</ns0:cell><ns0:cell>7mer-m8</ns0:cell><ns0:cell>96</ns0:cell><ns0:cell>-0.165</ns0:cell><ns0:cell>3.91E-03</ns0:cell></ns0:row><ns0:row><ns0:cell>hsa-miR-542-3p</ns0:cell><ns0:cell>836-842</ns0:cell><ns0:cell>7mer-m8</ns0:cell><ns0:cell>94</ns0:cell><ns0:cell>-0.057</ns0:cell><ns0:cell>3.19E-01</ns0:cell></ns0:row><ns0:row><ns0:cell>hsa-miR-384</ns0:cell><ns0:cell>314-321</ns0:cell><ns0:cell>8mer</ns0:cell><ns0:cell>91</ns0:cell><ns0:cell>0.000</ns0:cell><ns0:cell>1.00E+00</ns0:cell></ns0:row><ns0:row><ns0:cell>hsa-miR-520d-5p</ns0:cell><ns0:cell>361-367</ns0:cell><ns0:cell>7mer-m8</ns0:cell><ns0:cell>90</ns0:cell><ns0:cell>0.023</ns0:cell><ns0:cell>6.91E-01</ns0:cell></ns0:row><ns0:row><ns0:cell>hsa-miR-371a-5p</ns0:cell><ns0:cell>504-510</ns0:cell><ns0:cell>7mer-m8</ns0:cell><ns0:cell>89</ns0:cell><ns0:cell>-0.034</ns0:cell><ns0:cell>5.59E-01</ns0:cell></ns0:row><ns0:row><ns0:cell>hsa-miR-520a-3p</ns0:cell><ns0:cell>535-541</ns0:cell><ns0:cell>7mer-1A</ns0:cell><ns0:cell>88</ns0:cell><ns0:cell>0.098</ns0:cell><ns0:cell>8.66E-02</ns0:cell></ns0:row><ns0:row><ns0:cell>hsa-miR-520b</ns0:cell><ns0:cell>535-541</ns0:cell><ns0:cell>7mer-1A</ns0:cell><ns0:cell>87</ns0:cell><ns0:cell>0.012</ns0:cell><ns0:cell>8.29E-01</ns0:cell></ns0:row></ns0:table></ns0:figure>
</ns0:body>
" | "
College of Life Science
Yangtze University
No. 266 Jingmi Road
Jingzhou Hubei 434025
People’s Republic of China
Tel: +86-716 - 806 - 6182
Fax: +1-619 - 806 - 6257
http://www.yangtzeu.edu.cn/
chenyingcy@yangtzeu.edu.cn September 14th, 2020
Dear Professor Vladimir Uversky,
We would like to thank you and the reviewers for your constructive comments and suggestions, which gave us the opportunity to improve the paper.
In the following pages, we give our point-by-point responses to each of the concerns of the reviewers. We hope that the revisions to the manuscript and our accompanying responses are sufficient to make our manuscript suitable for publication in PeerJ.
Thank you very much for all your help. We look forward to hearing from you at your earliest convenience.
With my best regards,
Ying Chen, PhD
College of Life Science,Yangtze University
On behalf of all authors.
Reviewer 1 (Anonymous)
Basic reporting
The research article by Chen et al. describes how miR-29a-3p directly targets Smad nuclear interacting protein 1 and reduces the migration and proliferation of cervical cancer HeLa cells. This is an important study as it opens up new avenue to target cervical cancer by inhibiting their proliferation and migration.
Experimental design
The experiments performed by the investigators are designed thoroughly and are interpreted clearly. There is a sufficient detail on the experimental procedures used and the author has clearly discussed the loopholes in the study.
Validity of the findings
However, I have certain queries and suggestions for the authors as given below:
1. I believe the title of the manuscript sounds vague. I would title this manuscript as “miR-29a-3p directly targets Smad nuclear interacting protein 1 is directly and affects the migration and proliferation of cervical cancer HeLa cells”.
Thanks a lot for your suggestion. The title of our manuscript has been modified to“miR-29a-3p directly targets Smad nuclear interacting protein 1 and inhibits the migration and proliferation of cervical cancer HeLa cells” on line1~3.
2. Figure 1E shows that the relative cell viability decreases significantly at 48 hrs post siSNIP1 treatment. However, at 72 hrs the viability seems higher than 48 hrs in the siSNIP1 treated groups. Please explain the possible cause of this effect?
We have detected the SNIP1 mRNA and protein level of HeLa cells after transfection with siSNIP1, and the results showed that the level of mRNA at 72 h was higher than that at 48 h (FigS 1A and B). It is suggested that RNAi induced by siSNIP1 is time-dependent, which is consistent with the trend of cell proliferation assay (Fig 1E), indicating that the proliferation of HeLa cells is affected by the level of SNIP1 expression.
Figure S1 Expression of SNIP1 in HeLa cells after transfected with siSNIP1 at 72 h (A-B) RT-qPCR and Western blot analysis of SNIP1 expression in HeLa cells after transfected with siSNIP1-330 or siNC at 48 h and 72 h, respectively.
3. Also, in the Figure 1F, E- Cadherin expression should be significant as the expression has increased by 10 folds post- siSNIP1-330 treatment and it lacks * mark on the figure.
Thanks for your advice. We did miss the asterisk in Figure1F, which has been filled.
4. Figure 2F blot is unclear and I would suggest it to be replaced with a clear one to see significant downreguation in SNIP1 expression at protein level.
Thanks for your advice. We have selected another figure to show in Fig2F. At the same time, through gray statistics, there is a significant difference (raw data - “Fig2 data” - “Fig. 2F”).
5. Figure 3D again shows all markers significantly downregulated except E-cadherin which seems significantly upregulated statistically and lacks * mark.
Thanks for your advice. In this set of data about E-cadeherin, although three repeats are all greater than 1, the difference is large and the P value is 0.0792, so there is no asterisk.
6. Can you show the effect in vivo by injecting the control vs the miRNa transfected cells in a mice model to show the effect on HeLa cell migration and proliferation?
The mice model you proposed will be a powerful confirmation of the results of this paper, but due to the limited conditions, we are not able to carry out this experiment for the time being, and we will try our best to promote live animal experiments in the future.
Comments for the Author
The manuscript definitely requires lot of corrections in terms of language and grammar. It lacks proper scientific writing, for eg: terms have been vaguely used such as in line number 135 “related genes (CDK2) in HeLa cells (Wang & Chen, 2019) can also be regulated (Fig. 1F)”. It can be rewritten as downregulated or upregulated.
Thank you for your comments. We have corrected these language and grammar errors on line 43~44, line 59, line 136~137 and line 149, respectively.
Reviewer 2 (Anonymous)
Basic reporting
In this manuscript, Chen and colleagues report a novel regulatory interaction between Smad Nuclear Interacting Protein 1 (SNIP1) and miR-29a-3p with consequences on Hela cell proliferation and migration. Overall, I commend authors on doing a good job in putting together this concise and timely piece of work. However, I would recommend a few additional changes (also check Section 3) before acceptance, which I believe could help readers grasp and appreciate the scope of the study better.
A. The foundation of the entire study relies on HeLa cells but authors pitch the potential relevance of their study in context of cervical cancer. These claims should be toned down since authors don’t provide any data on levels of SNIP1 (and miR29a) from primary cervical cancer samples. While Hela cells are of cervical origin, at this point they have been growing in cultures for more than 5 decades, so it’s at best a cell culture model more than a cancer model.
We really did not have cervical cancer samples to confirm the correlation between the expression of SNIP1 and miR-29a-3p. According to the GEPIA (http://gepia.cancer-pku.cn/) website, SNIP1 was highly expressed in cervical cancer, however, there was no significant difference compared with normal control tissues (FigS2A). HPA (https://www.proteinatlas.org/) analysis also suggested that the protein level of SNIP1 was medium in half of the patients with cervical cancer, and second only to endometrial carcinoma (FigS2B). In other tumor tissues, such as osteosarcoma (Xie et al., 2019) and non-small cell lung cancer (Jeon et al., 2013), the protein level of SNIP1 was significantly higher than that of normal tissues. In HeLa cells and SiHa cells, the mRNA expression level of SNIP1 was also moderate (FigS2C). Furthermore, the low expression of miR-29a in cervical cancer has been reported (Gong et al., 2019).
We agree with you that HeLa cells as a cell culture model. Our results lead us to think about and infer that model cells like HeLa may account for a common phenomenon in tumor cells, that is, whether SNIP1/miR-29a-3p axis exists in all kinds of tumor cells. At present, according to the analysis of starbase (http://starbase.sysu.edu.cn/), it is found that there is a negative correlation between the expression of SNIP1 and miR-29a-3p in about 80% of cancer samples (p-value < 0.05) (TableS1).
Figure S2 Expression of SNIP1 and miR-29a-3p in cancer samples.
(A) Differential expression levels of SNIP1 from cervical cancer samples vs normal samples analyzed from GEPIA (http://gepia.cancer-pku.cn/). (B) Protein expression of SNIP1 in different cancer patients analyzed from HPA(https://www.proteinatlas.org/). 5/11 patients with cervical cancer show medium expression. (C) RNA expression of SNIP1 in different cell lines analyzed from HPA(https://www.proteinatlas.org/).
B. Instead of the purported significance of the study to provide therapeutic solutions to cervical cancers, the highlight of the study could be the novel regulatory link between miR29a and SNIP and how this interaction plays a ‘master regulator’ role in controlling proliferation and migration. To me this is the most significant message from this work and authors could focus a bit more to highlight the same.
Thanks for your advice, we have made changes in the abstract (line38~40) and line 215~217, highlighting the role of our newly discovered miR29a-3p/SNIP1 axis in suppressing proliferation and migration of cervical cancer cells.
C. The manuscript is easy to read and generally written well. Introduction provides a good background of the miRNA and SNIP1 literature. But authors should provide more background on miR29a in the discussion.
Thanks a lot for your comments. We have added more background about miR-29a in the discussion on line 177~182.
Experimental design
As such, I find the primary research of this study lies within the Aims and Scope of Peer J. Regulation of SNIP1 is poorly understood and this work provides new mechanistic hints involving miRNA axis. The experiments are well planned and executed, adequate controls have been presented and raw data has been shared. Overall, the method sections provide sufficient details.
Validity of the findings
The findings are generally internally consistent with the underlying hypothesis and narrative, but I have a few concerns for authors to address:
A. In Fig 1E, viability of siSNIP1 drops significantly at 48h but recover to near-normal levels by 72h. Can authors comment or speculate on this peculiar trend. Is this because the effect of siRNA wears off by 72h. It would be good to support this finding by showing mRNA (and protein) levels at 48h v/s 72h for siSNIP1 scenario compared to control.
We have detected the SNIP1 mRNA and protein level of HeLa cells after transfection with siSNIP1, and the results showed that the level of mRNA at 72 h was higher than that at 48 h (FigS 1A and B). It is suggested that RNAi induced by siSNIP1 is time-dependent, which is consistent with the trend of cell proliferation assay (Fig 1E), indicating that the proliferation of HeLa cells is affected by the level of SNIP1 expression.
Figure S1 Expression of SNIP1 in HeLa cells after transfected with siSNIP1 at 72 h (A-B) RT-qPCR and Western blot analysis of SNIP1 expression in HeLa cells after transfected with siSNIP1-330 or siNC at 48 h and 72 h, respectively.
B. The figure legend for Fig 2 could be changed to ‘miR29a-3p directly targets SNIP1 to downregulate SNIP1 expression’. This change will reflect the data and its significance better for readers.
Thanks for your suggestion. The figure legend for Fig 2 has been changed to “miR-29a-3p directly targets SNIP1 in HeLa cells”.
C. In Fig 3a, why do the NC condition also show poor migratory recovery in the wound-healing assay. Although, statistically significant, the effect size is very negligible between NC and miR29a case. Could authors comment on this?
The reason is that the initial area of scratches was too wide, so the migration rates appeared insignificant.
Comments for the Author
The schematic in Fig 4B is not correct conceptually (at least to this reviewer). The relation between miR29a and SNIP1 (negative regulation, symbolised by ‘inverted T’) is shown correctly. But, SNIP1 is a positive regulator of HSP27, c-Myc and Cyclin D1 and they should be linked by ‘Arrow’ (indicating positive regulation) and not ‘inverted T’. Further downstream, arrows (or dashed lines) should link these 3 players to Cell Migration and proliferation. Dashed arrows/lines are best to use since authors don’t address any of these aspects in this study.
Thank you for your valuable comments. We have made corrections in Fig 4B.
" | Here is a paper. Please give your review comments after reading it. |
9,793 | "<ns0:abstract xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:p>Long non-coding RNAs (lncRNAs) have been proved to have an important role in different malignancies including clear cell renal cell carcinoma (ccRCC). However, their role in disease progression is still not clear. The objective of the study was to identify lncRNAbased prognostic biomarkers and further to investigate the role of one lncRNA LINC01234 in progression of ccRCC cells. We found that six adverse prognostic lncRNA biomarkers including LINC01234 were identified in ccRCC patients by bioinformatics analysis using The Cancer Genome Atlas (TCGA) database. LINC01234 knockdown impaired cell proliferation, migration and invasion in vitro as compared to negative control. Furthermore, the epithelial-mesenchymal transition (EMT) was inhibited after LINC01234 knockdown.</ns0:p><ns0:p>Additionally, LINC01234 knockdown impaired hypoxia-inducible factor-2a (HIF-2α) pathways，including a suppression of the expression of HIF-2α, vascular endothelial growth factor A (VEGFA), epidermal growth factor receptor (EGFR), c-Myc, Cyclin D1 and MET.</ns0:p><ns0:p>Together, these datas showed that LINC01234 was likely to regulate the progression of ccRCC by HIF-2α pathways, and LINC01234 was both a promising prognostic biomarker and a potential therapeutic target for ccRCC.</ns0:p></ns0:div>
</ns0:abstract>
<ns0:body xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:head>Introduction</ns0:head><ns0:p>In 2018, it was predicted that 403,262 new cases would be diagnosed with kidney cancer in 185 countries and 175,098 cases would be dead <ns0:ref type='bibr' target='#b1'>(Bray et al. 2018)</ns0:ref>. In 36 kinds of cancers, the morbidity and mortality of kidney cancer were 2.2% and 1.8% respectively <ns0:ref type='bibr' target='#b1'>(Bray et al. 2018)</ns0:ref>. Clear cell RCC (ccRCC) is the most common subtype of RCC and it accounts for 75% <ns0:ref type='bibr'>(Song et al. 2018)</ns0:ref>. Although surgery is still the preferred therapeutic option for the localized and locally advanced ccRCC, the long-term prognosis remains unsatisfactory and unpredictable. Current existed evaluation approach for prognosis of ccRCC is mainly based on clinicopathologic data, such as TNM staging. However, it does not reflect the biological heterogeneity of cancer <ns0:ref type='bibr' target='#b3'>(Cheng 2018)</ns0:ref>. Therefore, there is an urgent need for discovering a new prognostic model and biomarkers for ccRCC. Moreover, it is necessary to understand the molecular mechanisms of the prognostic biomarkers underlying ccRCC development. Long non-coding RNA (lncRNA) is a kind of RNA transcripts with a length of > 200 nucleotides. Unlike mRNA, it does not encode proteins. Currently, it is reported that lncRNA is engaged in numerous important biological processes and the development and progression of numerous human diseases, including ccRCC <ns0:ref type='bibr' target='#b8'>(Esteller 2011;</ns0:ref><ns0:ref type='bibr' target='#b13'>Gupta et al. 2010;</ns0:ref><ns0:ref type='bibr' target='#b16'>Jin et al. 2017;</ns0:ref><ns0:ref type='bibr' target='#b22'>Martens-Uzunova et al. 2014;</ns0:ref><ns0:ref type='bibr' target='#b26'>Ponting et al. 2009;</ns0:ref><ns0:ref type='bibr' target='#b27'>Quinn & Chang 2016;</ns0:ref><ns0:ref type='bibr' target='#b28'>Rinn & Chang 2012)</ns0:ref>. The overexpression, deficiency or mutation of lncRNA are associated with the tumor formation, progression, metastasis and prognosis in many human malignancies including ccRCC <ns0:ref type='bibr' target='#b8'>(Esteller 2011;</ns0:ref><ns0:ref type='bibr' target='#b10'>Ghaffar et al. 2018;</ns0:ref><ns0:ref type='bibr' target='#b13'>Gupta et al. 2010;</ns0:ref><ns0:ref type='bibr' target='#b14'>He et al. 2018;</ns0:ref><ns0:ref type='bibr' target='#b16'>Jin et al. 2017;</ns0:ref><ns0:ref type='bibr' target='#b22'>Martens-Uzunova et al. 2014)</ns0:ref>. However, the functions of the majority of lncRNA are not well understood. In our study, we constructed a lncRNA-based prognostic model and identified six lncRNAs as independent prognostic biomarkers in ccRCC, including LINC01234. Several studies showed that LINC01234 was upregulated and had oncogenic potentials in several cancers, such as gastric cancer <ns0:ref type='bibr' target='#b2'>(Chen et al. 2018)</ns0:ref>, esophageal cancer <ns0:ref type='bibr' target='#b10'>(Ghaffar et al. 2018)</ns0:ref>, and colorectal adenocarcinoma <ns0:ref type='bibr' target='#b14'>(He et al. 2018)</ns0:ref>. Most ccRCC are associated with loss of von Hippel-Lindau tumor suppressor (pVHL) function and deregulation of hypoxia pathways <ns0:ref type='bibr' target='#b30'>(Schödel et al. 2016)</ns0:ref>. Adaptation to hypoxia plays an important role in the progression of ccRCC <ns0:ref type='bibr' target='#b9'>(Garje et al. 2018)</ns0:ref>. Hypoxia is mediated via hypoxia-inducible factors (HIFs) HIF-1α and HIF-2α <ns0:ref type='bibr' target='#b31'>(Semenza 2012)</ns0:ref>. Recently, studies showed that HIF-2α, rather than HIF-1α, was a predominant driver in renal cancer progression <ns0:ref type='bibr' target='#b18'>(Keith et al. 2011)</ns0:ref>. Although HIF-1α can act as a ccRCC tumor suppressor, HIF-1α activity is commonly diminished by chromosomal deletion in ccRCC <ns0:ref type='bibr' target='#b30'>(Schödel et al. 2016)</ns0:ref>. Conversely, HIF-2α has emerged as the key HIF isoform acting as an oncogene that is essential for ccRCC tumor progression <ns0:ref type='bibr' target='#b23'>(Meléndez-Rodríguez et al. 2018;</ns0:ref><ns0:ref type='bibr' target='#b30'>Schödel et al. 2016)</ns0:ref>. The polymorphisms at the HIF-2α gene locus predispose to the development of ccRCC, and HIF-2α promotes tumor growth <ns0:ref type='bibr' target='#b30'>(Schödel et al. 2016)</ns0:ref>. Indeed, preclinical and clinical data have shown that pharmacological inhibitors of HIF-2α can efficiently inhibit ccRCC growth <ns0:ref type='bibr' target='#b23'>(Meléndez-Rodríguez et al. 2018)</ns0:ref>. However, the role of LINC01234, as well as the relationship between LINC01234 and HIF-2α in ccRCC remains unclear. In the present study, we showed that LINC01234 was likely to regulate the progression of ccRCC by HIF-2α pathways. Therefore, LINC01234 might serve as a promising prognostic biomarker and a potential therapeutic target for patients with ccRCC.</ns0:p></ns0:div>
<ns0:div><ns0:head>Materials & Methods</ns0:head></ns0:div>
<ns0:div><ns0:head>LncRNA expression and clinical datasets of ccRCC cases</ns0:head><ns0:p>The TCGA Research Network was available at http://cancergenome.nih.gov/ <ns0:ref type='bibr' target='#b6'>(Deng et al. 2016)</ns0:ref>. The datasets for ccRCC cases within the TCGA database were downloaded using the GDC Data Portal. The version of the dataset was: Data Release 14.0-December 18, 2018.</ns0:p></ns0:div>
<ns0:div><ns0:head>Differentially expression analysis to identify differentially expressed lncRNAs</ns0:head><ns0:p>Differentially expression analysis was performed as previously described <ns0:ref type='bibr' target='#b38'>(Yang et al. 2019)</ns0:ref>. A volcano plot was plotted for the differentially expressed lncRNAs.</ns0:p></ns0:div>
<ns0:div><ns0:head>Univariate cox regression and least absolute shrinkage and selection operator (LASSO) regression to identify key prognostic lncRNAs</ns0:head><ns0:p>The univariate cox regression was performed for the differentially expressed lncRNAs. Then, the statistically significant lncRNAs (p < 0.05) were used for LASSO regression to identify key prognostic lncRNAs. The univariate cox regression and LASSO regression were performed as previously described <ns0:ref type='bibr' target='#b38'>(Yang et al. 2019)</ns0:ref>.</ns0:p></ns0:div>
<ns0:div><ns0:head>Multivariate cox regression to establish the prognostic model</ns0:head><ns0:p>The multivariate cox regression was performed for the key prognostic lncRNAs as previously described <ns0:ref type='bibr' target='#b38'>(Yang et al. 2019)</ns0:ref>. It calculated the risk score for each patient. Based on the median of the risk score, all patients were divided into the high-risk group and low-risk group. A heatmap was plotted to present the expression levels of the key prognostic lncRNAs in the two groups. And a forest plot was plotted to present the hazard ratio (HR) and 95% confidence interval (CI) for the key prognostic lncRNAs.</ns0:p></ns0:div>
<ns0:div><ns0:head>ROC curve and C-index to evaluate the prognostic model</ns0:head><ns0:p>The 3-year and 5-year time-dependent receiver operating characteristic (ROC) curves, the area under the ROC curves (AUCs) and the C-index were performed as previously described <ns0:ref type='bibr' target='#b38'>(Yang et al. 2019)</ns0:ref>.</ns0:p></ns0:div>
<ns0:div><ns0:head>Kaplan-Meier (K-M) survival analysis to identify independent prognostic biomarkers</ns0:head><ns0:p>The R package 'survival' (cran.r-project. org/web/packages/survival/index.html) was used for K-M survival analysis. Firstly, The K-M survival analysis was performed for the high-risk group and the low-risk group. Then K-M survival curves were plotted individually for each statistically significant lncRNA from the result of the multivariate cox regression. Validation of the expression and prognostic significance of the independent prognostic biomarkers Gene Expression Profiling Interactive Analysis (GEPIA) server <ns0:ref type='bibr' target='#b35'>(Tang et al. 2017</ns0:ref>) is a newly developed interactive web server and has been running for three years. It was used for analyzing the RNA sequencing expression data computed by a standard processing pipeline. Therefore, we validated the expression levels and prognostic significance of the independent prognostic biomarkers in patients with ccRCC via GEPIA server according to their Ensembl ID.</ns0:p></ns0:div>
<ns0:div><ns0:head>Cell culture</ns0:head><ns0:p>Human RCC cells Caki-2 and A498 (Chinese Academy of Medical Sciences Shanghai Cell Bank) were cultured in RPMI 1640 medium (Gibco, USA) containing 10% fetal bovine serum (FBS) (Gibco, USA), 100 U/ml penicillin (Sigma-Aldrich, St Louis, MO) and 100 μg/ml streptomycin (Sigma-Aldrich). All cells were routinely cultured in 5% CO2 at 37°C.</ns0:p></ns0:div>
<ns0:div><ns0:head>Lentivirus-mediated shRNA transfection</ns0:head><ns0:p>The recombinant lentivirus with short hairpin RNA (shRNA) and the corresponding control lentivirus were purchased from Genechem (Shanghai, China). Transfection in vitro was performed following the manufacturer's protocols. Stable shRNA-expressing colonies were selected using puromycin (Solarbio, Beijing, China). The target sequences of shRNA were as follows: 5'-CCTCGGTCTCAGTTTCTCCATTTAT-3' (shRNA) and 5'-TTCTCCGAACGTGTCACGT-3' (control) respectively.</ns0:p></ns0:div>
<ns0:div><ns0:head>RNA extraction, reverse transcription and real-time quantitative PCR (qPCR)</ns0:head><ns0:p>RNA extraction and reverse transcription were performed as previously described <ns0:ref type='bibr' target='#b36'>(Wang et al. 2020)</ns0:ref>. QPCR was performed using SYBR Green Realtime PCR Master Mix (TOYOBO, Osaka, Japan) in the QuantStudio 5 Real-Time PCR System (Thermo Fisher Scientific, USA). The PCR primers are shown in Supplementary Table <ns0:ref type='table'>S1</ns0:ref>. The relative expression levels of genes were calculated using the 2 -ΔΔCt method relative to GAPDH.</ns0:p></ns0:div>
<ns0:div><ns0:head>CCK-8 cell proliferation assay</ns0:head><ns0:p>Cells stably expressing LINC01234 shRNA or control vector were plated into 96-well plates (2000 cells per well) and incubated at 37℃ under 5% CO2 for 1, 2, 3 or 4 days respectively. Then CCK-8 solution (Dojindo, Japan) was added into the culture medium, and the optical density (OD) at 450 nm was measured with a Microplate Reader (Bio-Rad Laboratories Inc, Hercules, CA, USA) after incubation for 1.5 h. Each group had five duplicates and the experiment was performed in triplicate.</ns0:p></ns0:div>
<ns0:div><ns0:head>Cell colony formation assay</ns0:head><ns0:p>Cells stably expressing LINC01234 shRNA or control vector were plated into 10 cm culture dish (1500 cells per dish) and incubated for 14 days. Wells were fixed with 4% paraformaldehyde and stained with 0.1% crystal violet. The cell colonies with > 50 cells were counted. Each group had three duplicates and the experiment was performed in triplicate.</ns0:p></ns0:div>
<ns0:div><ns0:head>Transwell assays</ns0:head><ns0:p>Transwell assays including migration assays and invasion assays were performed as previously described <ns0:ref type='bibr' target='#b36'>(Wang et al. 2020)</ns0:ref>. Each group had three duplicates and the experiment was performed in triplicate.</ns0:p></ns0:div>
<ns0:div><ns0:head>Western blots</ns0:head><ns0:p>Western blots were performed as described <ns0:ref type='bibr' target='#b20'>(Liu et al. 2013;</ns0:ref><ns0:ref type='bibr' target='#b39'>Yang et al. 2018)</ns0:ref>. Total cellular protein was extracted using RIPA buffer (Beyotime, Shanghai, China) with 1% of 100 mM PMSF (Solarbio, Beijing, China). Protein concentration was quantified using a BCA Protein Quantitative Kit (Beyotime, Shanghai, China). Briefly, 30 μg of protein was resolved by 10% SDS-PAGE, and transferred to a PVDF membrane (Millipore, Billerica, MA). The membrane was blocked with 5% skim milk and then probed with rabbit or mouse anti-human primary antibodies respectively. Next, the membranes were incubated with corresponding HRPconjugated goat anti-rabbit or anti-mouse IgG (1:1000 dilution) (CST, Boston, USA) and detected with Western Blotting Luminol Reagent (Santa Cruz, CA, USA). The experiment was performed in triplicate. Epithelial-Mesenchymal Transition (EMT) Antibody Sampler Kit #9782 (all 1:1000 dilution) (CST, Boston, USA) were used for western blots. Besides, mouse antihuman primary antibody HIF-1α (Abcam, Cambridge, MA, USA)，rabbit anti-human primary antibodies HIF-2α (Abcam, Cambridge, MA, USA)，VEGFA (Abcam, Cambridge, MA, USA), EGFR (Abcam, Cambridge, MA, USA), c-Myc (CST, Boston, USA), Cyclin D1 (CST, Boston, USA) and MET (CST, Boston, USA) (all 1:1000 dilution) were also detected.</ns0:p></ns0:div>
<ns0:div><ns0:head>Statistical analysis</ns0:head><ns0:p>For the datasets from TCGA database, the software Perl, R (version 3.4.4), RStudio 1.2.1335-Windows 7+ (64-bit) and R packages were used for data integration, extraction, analysis and visualization. Briefly, the R package 'edgeR' was utilized to screen differentially expressed genes (FDR < 0.05 and |log 2 FC| > 2). The univariate cox regression and the Lasso regression were performed to identify key prognostic factors. The multivariate cox regression and K-M survival curve were performed to establish the risk score model and identify independent prognostic factors. ROC curve and C-index were performed to estimate the prognostic power of the risk score model. For the data about the function of LINC01234, SPSS 22.0 (IBM, USA) and GraphPad Prism 5.01 (GraphPad Software, USA) were used for statistical analyses. The data was expressed as mean ± SD from at least three independent experiments. Cell proliferation abilities of CCK-8 assay were compared with two-way ANOVA. Cell colony, migration and invasion levels, as well as qPCR data were compared using the Student's t-test. A p < 0.05 was considered statistically significant.</ns0:p></ns0:div>
<ns0:div><ns0:head>Results</ns0:head></ns0:div>
<ns0:div><ns0:head>Identification of differentially expressed lncRNAs and key prognostic lncRNAs in patients with ccRCC</ns0:head><ns0:p>A total of 70 normal tissue samples and 541 cancer tissue samples from patients with ccRCC were collected. 11,368 lncRNAs were extracted from the transcriptome Profiling. Compared with the normal tissues, a total of 1541 lncRNAs were identified as differentially expressed lncRNAs in tumor tissues (FDR < 0.05 and |logFC| > 2), including 1075 upregulated (logFC > 2) and 466 downregulated (logFC < −2) lncRNAs (Figure <ns0:ref type='figure' target='#fig_0'>1A</ns0:ref>) (Supplementary Table <ns0:ref type='table'>S2</ns0:ref>). Preliminarily, a total of 323 statistically significant lncRNAs were considered to be related to the prognosis by the univariate cox regression (Supplementary Table <ns0:ref type='table'>S3</ns0:ref>). Next, through the LASSO regression, 13 lncRNAs were identified as key prognostic lncRNAs (Figure <ns0:ref type='figure' target='#fig_0'>1B</ns0:ref>, 1C), which were used for the further establishment of the risk score model by multivariate cox regression.</ns0:p></ns0:div>
<ns0:div><ns0:head>Establishment and evaluation of the prognostic model in ccRCC</ns0:head><ns0:p>The median cutoff point of the risk scores calculated by multivariate cox regression was 0.842. All patients were divided into the high-risk group and low-risk group. It was revealed that the patients in the high-risk group had a significantly worse overall survival rate than those in the low-risk group (p < 0.001) (Figure <ns0:ref type='figure' target='#fig_1'>2A</ns0:ref>). The AUC was 0.753 (3-year ROC curve) and 0.784 (5-year ROC curve) respectively, and the C-index was 0.753 (Figure <ns0:ref type='figure' target='#fig_1'>2B</ns0:ref>). In addition, it also presented the relationship between the survival time and the risk score for patients (the death and the alive) (Figure <ns0:ref type='figure' target='#fig_1'>2C</ns0:ref>). Moreover, a heatmap was plotted to illustrate the expression levels of the key prognostic lncRNAs in the high-risk group and low-risk group (Figure <ns0:ref type='figure' target='#fig_1'>2D</ns0:ref>).</ns0:p></ns0:div>
<ns0:div><ns0:head>Identification of independent prognostic biomarkers</ns0:head><ns0:p>The multivariate cox regression revealed HR and 95% CI for the 13 key prognostic lncRNAs with a forest plot (Figure <ns0:ref type='figure' target='#fig_2'>3</ns0:ref>). It indicated 6 statistically significant lncRNAs as the independent prognostic biomarkers, including lncRNAs AC009654.1, AC012615.3, AC092490.2, AL357507.1, LINC01234 and LINC01956. Moreover, K-M survival analysis was performed for the 6 lncRNAs. It revealed that all the 6 lncRNAs with high expression levels predicted a significantly worse overall survival rate than the low expressed one (Figure <ns0:ref type='figure' target='#fig_3'>4A-F</ns0:ref>). Therefore, they could serve as the adverse independent prognostic factors. Furthermore, we validated the expression levels and prognostic significance of the 6 lncRNAs in patients with ccRCC via GEPIA server. It suggested that AL357507.1, LINC01234, and LINC01956 were highly expressed at higher pathological stage of the disease, while LINC01234 exhibited the highest significance in terms of expression at different pathological stage of the disease (Figure <ns0:ref type='figure' target='#fig_4'>5A-F</ns0:ref>). Moreover, GEPIA server revealed the significance of LINC01234 in terms of survival time. The high expression level of LINC01234 predicted a significantly worse disease-free survival rare or overall survival rate than the low expressed one (Figure <ns0:ref type='figure' target='#fig_4'>5G</ns0:ref>, 5H). Unfortunately, GEPIA server could not provide the prognostic significance of the other 5 lncRNAs because the server showed the sample size was insufficient. LINC01234 knockdown suppressed the proliferation and clone formation of ccRCC cells Knockdown of LINC01234 was performed in Caki-2 and A498 cells by the lentivirus-mediated shRNA transfection. It suggested that the expression of LINC01234 was reduced in Caki-2 and A498 cells, which was validated by qPCR (Figure <ns0:ref type='figure' target='#fig_6'>6A</ns0:ref>). Next, the CCK-8 assay revealed the proliferations of Caki-2 and A498 cells were significantly suppressed (Figure <ns0:ref type='figure' target='#fig_6'>6B, 6C</ns0:ref>). Moreover, cell colony formation assay was performed to analyze the role of LINC01234 in the colony formation of Caki-2 and A498 cells. As shown in Figure <ns0:ref type='figure' target='#fig_6'>6D</ns0:ref>-I, the clonogenic capacities of Caki-2 and A498 cells were dramatically inhibited. Obviously, it indicated that LINC01234 played an important role in the proliferation and colony formation of Caki-2 and A498 cells.</ns0:p></ns0:div>
<ns0:div><ns0:head>LINC01234 depletion inhibited the migration and invasion of ccRCC cells</ns0:head><ns0:p>The migration capabilities of Caki-2 and A498 cells were assessed by Transwell migration assay, while the invasion capabilities of these cells were assessed by Transwell Matrigel invasion assay. The results of the Transwell assay indicated that LINC01234 knockdown significantly inhibited the migration capabilities of Caki-2 and A498 cells (Figure <ns0:ref type='figure' target='#fig_7'>7A-F</ns0:ref>). Similarly, the invasion capabilities of Caki-2 and A498 cells were also suppressed following LINC01234 depletion (Figure <ns0:ref type='figure' target='#fig_7'>7G-L</ns0:ref>). These findings demonstrated that LINC01234 played an important role in the migration and invasion capacities of ccRCC cells. LINC01234 knockdown suppressed EMT process in ccRCC cells EMT process was closely related to the migration and invasion of cancer cells. Therefore, the mRNA levels of EMT-associated genes and the levels of EMT-associated proteins were detected by RT-PCR and western blots respectively. As shown in Figure <ns0:ref type='figure' target='#fig_8'>8A</ns0:ref> and 8B, we found that the mRNA level of epithelial marker E-cadherin was increased, while the mRNA level of mesenchymal marker N-cadherin was decreased in Caki-2 and A498 cells following LINC01234 knockdown. Similarly, as shown in Figure <ns0:ref type='figure' target='#fig_8'>8C</ns0:ref>, the protein expression levels of mesenchymal markers Vimentin and N-cadherin were significantly decreased in Caki-2 and A498 cells with LINC01234 knockdown, while the protein expression level of epithelial marker E-cadherin was upregulated. Moreover, the protein expression level of the transcription factor Snail was decreased in Caki-2 and A498 cells with LINC01234 knockdown. In addition, the protein expression level of β-catenin was also inhibited in Caki-2 and A498 cells following LINC01234 depletion.</ns0:p></ns0:div>
<ns0:div><ns0:head>LINC01234 suppression suppressed HIF-2α pathways in ccRCC cells</ns0:head><ns0:p>As shown in Figure <ns0:ref type='figure' target='#fig_8'>8A</ns0:ref> and 8B, we found that the mRNA levels of HIF-2α and vascular endothelial growth factor A (VEGFA) were decreased in Caki-2 and A498 cells following LINC01234 knockdown. Similarly, as shown in Figure <ns0:ref type='figure' target='#fig_8'>8D</ns0:ref>, we found that the protein expression levels of HIF-1α and HIF-2α were significantly decreased in Caki-2 and A498 cells with LINC01234 knockdown. Additionally, we found the protein expression levels of several target genes of HIF-2α, including VEGFA, epidermal growth factor receptor (EGFR), c-myc, Cyclin D1 and MET, were also inhibited in Caki-2 and A498 cells following LINC01234 depletion.</ns0:p></ns0:div>
<ns0:div><ns0:head>Discussion</ns0:head><ns0:p>LncRNA was a kind of long RNA transcripts (> 200 nucleotides) and it had no apparent proteincoding potentials <ns0:ref type='bibr' target='#b27'>(Quinn & Chang 2016)</ns0:ref>. Even so, lncRNA possessed a wide range of biological functions involved in multiple vital cellular activities <ns0:ref type='bibr' target='#b32'>(Shen et al. 2015)</ns0:ref>. Generally, lncRNA achieved its function by regulating gene expression in the levels of epigenetics, transcription and post-transcription <ns0:ref type='bibr' target='#b19'>(Lee 2012;</ns0:ref><ns0:ref type='bibr' target='#b37'>Wang & Chang 2011)</ns0:ref>. It could serve as a molecular signal, a molecular decoy, a molecular guide, or a molecular scaffold to achieve its functions <ns0:ref type='bibr' target='#b37'>(Wang & Chang 2011)</ns0:ref>. The function of lncRNA was associated with its subcellular localization <ns0:ref type='bibr' target='#b37'>(Wang & Chang 2011)</ns0:ref>. More specifically, lncRNA might be involved in chromatin regulation, gene transcription and alternative splicing of transcripts when it was in nucleus, while if it was in cytoplasm, it might serve as a competing endogenous RNA (ceRNA), and regulated the stability or translation of mRNA <ns0:ref type='bibr' target='#b38'>(Yang et al. 2019)</ns0:ref>. Recently, more and more evidences indicated that aberrations of lncRNA, such as overexpression, deficiency or mutation, played an important role in malignant phenotypes of cancers <ns0:ref type='bibr' target='#b29'>(Schmitt & Chang 2016)</ns0:ref>, including tumor formation, progression, metastasis and poor prognosis <ns0:ref type='bibr' target='#b8'>(Esteller 2011;</ns0:ref><ns0:ref type='bibr' target='#b13'>Gupta et al. 2010;</ns0:ref><ns0:ref type='bibr' target='#b22'>Martens-Uzunova et al. 2014;</ns0:ref><ns0:ref type='bibr' target='#b40'>Yu et al. 2017;</ns0:ref><ns0:ref type='bibr' target='#b41'>Yue et al. 2016)</ns0:ref>. Some aberrant lncRNAs were also associated with lots of malignant biological behaviors of cancer cells, such as proliferation, apoptosis, migration and invasion <ns0:ref type='bibr' target='#b7'>(Ellinger et al. 2016;</ns0:ref><ns0:ref type='bibr' target='#b15'>Huang et al. 2017;</ns0:ref><ns0:ref type='bibr' target='#b41'>Yue et al. 2016)</ns0:ref>. It was also reported that some aberrant lncRNAs could serve as prognostic indicators in ccRCC, such as lncRNA Fer1L4 <ns0:ref type='bibr' target='#b5'>(Cox et al. 2020)</ns0:ref>. With the development of molecular biological techniques and bioinformatics, more and more lncRNAs were marked as novel biomarkers and prognostic signatures for ccRCC utilizing TCGA database. For example, lncRNA Fer1L4 was overexpressed in ccRCC tissues, and its high expression levels were found in higher grade, higher stage, and metastatic tumors <ns0:ref type='bibr' target='#b5'>(Cox et al. 2020)</ns0:ref>. LncRNA Fer1L4 overexpression was also an independent prognostic factor for patients with ccRCC <ns0:ref type='bibr' target='#b5'>(Cox et al. 2020)</ns0:ref>. It was also reported an 11-lncRNA signature (AC245100.1, AP002761.1, LINC00488, AC017033.1, LINC-PINT, COL5A1-AS1, AC026471.4, AL009181.1, LINC00524, HOTTIP, AL078590.3) and a 6-lncRNA signature (CTA-384D8.35, CTD-2263F21.1, LINC01510, RP11-352G9.1, RP11-395B7.2, RP11-426C22.4) were clearly linked to the overall survival (OS) rate of ccRCC patients based on TCGA database <ns0:ref type='bibr' target='#b42'>(Zeng et al. 2019;</ns0:ref><ns0:ref type='bibr' target='#b33'>Zhang et al. 2019</ns0:ref>).</ns0:p><ns0:p>In the present study, utilizing the TCGA database, we identified 1541 differentially expressed lncRNAs. More importantly, we not only constructed a 13-lncRNA-based risk score model with moderate accuracy, but also identified 6 independent adverse prognostic lncRNAs for patients with ccRCC, including lncRNA AC009654.1, AC012615.3, AC092490.2, AL357507.1, LINC01234 and LINC01956. It was similar to a recent study which suggested that lncRNA AC009654.1, AC092490.2, LINC00524, LINC01234 and LINC01885 were significantly associated with ccRCC prognosis (Zhang et al. 2020). The expression levels of the 6 lncRNAs above were upregulated in ccRCC tissues and the high expression levels of them predicted a worse overall survival in ccRCC patients. Furtherly, we investigated their expression levels at different pathological stages and validated their prognostic significance in ccRCC patients via GEPIA server. It revealed AL357507.1, LINC01234, and LINC01956 were highly expressed at higher pathological stages of the disease, while LINC01234 exhibited the highest significance in terms of expression at different pathological stages of the disease. It was a very interesting finding, because the pathological stage was closely associated with the prognosis of ccRCC patients. Moreover, GEPIA server revealed the significance of LINC01234 in terms of survival time. Unfortunately, GEPIA server could not provide the prognostic significance of the other 5 lncRNAs because the server showed the sample size was insufficient. Besides, I also referred to the recent studies and references about these six lncRNAs. Nevertheless, except limited researches for LINC01234, there are no investigations for them currently and it deserves to further researches. Therefore, we mainly focused on lncRNA LINC01234 for the subsequent experiments.</ns0:p><ns0:p>Recently, partial functions and mechanisms of LINC01234 (also known as LCAL84) were reported in cancers, such as gastric cancer <ns0:ref type='bibr' target='#b2'>(Chen et al. 2018)</ns0:ref>, esophageal cancer <ns0:ref type='bibr' target='#b10'>(Ghaffar et al. 2018)</ns0:ref>, and colorectal adenocarcinoma <ns0:ref type='bibr' target='#b14'>(He et al. 2018)</ns0:ref>. LINC01234 was upregulated and had oncogenic potentials in esophageal carcinoma cells in vitro <ns0:ref type='bibr' target='#b10'>(Ghaffar et al. 2018;</ns0:ref><ns0:ref type='bibr' target='#b14'>He et al. 2018)</ns0:ref>. LINC01234 was significantly associated with the prognosis of colorectal adenocarcinoma and the malignant biological behaviors of esophageal carcinoma cells including proliferation, migration, invasion and apoptosis <ns0:ref type='bibr' target='#b10'>(Ghaffar et al. 2018;</ns0:ref><ns0:ref type='bibr' target='#b14'>He et al. 2018)</ns0:ref>. Besides, LINC01234 expression was significantly upregulated in gastric cancer tissues and was associated with larger tumor size, advanced TNM stage, lymph node metastasis, and shorter survival time <ns0:ref type='bibr' target='#b2'>(Chen et al. 2018)</ns0:ref>. Moreover, LINC01234 could serve as ceRNA to regulate core-binding factor β (CBFB) expression by sponging miR-204-5p to regulate the apoptosis, growth arrest and tumorigenesis in gastric cancer <ns0:ref type='bibr' target='#b2'>(Chen et al. 2018)</ns0:ref>. In our study, we also explored the role of LINC01234 in ccRCC. It indicated that LINC01234 expression was upregulated in ccRCC tissues. LINC01234 was expressed increasingly as the stage increased. The high expression level of LINC01234 predicted a significantly worse disease-free survival rate or overall survival rate than the low expressed one for the patients with ccRCC. Besides, LINC01234 knockdown inhibited proliferation, migration, invasion and epithelial-mesenchymal transition (EMT) process of ccRCC cells. More importantly, LINC01234 knockdown impaired the expression of HIF-1a, HIF-2a, VEGFA, EGFR, c-Myc, Cyclin D1 and MET in Caki-2 and A498 cells following LINC01234 depletion. EMT was considered as an essential process during development whereby epithelial cells acquired mesenchymal, fibroblast-like characteristics and displayed reduced intracellular adhesion and increased motility <ns0:ref type='bibr' target='#b0'>(Aigner et al. 2007;</ns0:ref><ns0:ref type='bibr' target='#b24'>Moreno-Bueno et al. 2008)</ns0:ref>. EMT played a critical role in in the progression of primary tumors towards spread and metastasis, as well as the migration and invasion of malignant tumor cells <ns0:ref type='bibr' target='#b11'>(Gloushankova et al. 2018;</ns0:ref><ns0:ref type='bibr' target='#b25'>Peinado et al. 2007;</ns0:ref><ns0:ref type='bibr' target='#b39'>Yang et al. 2018)</ns0:ref>. Recently, an increasing number of studies supported the role of lncRNAs in the regulation of tumor progression and metastasis through the regulation of EMT <ns0:ref type='bibr' target='#b12'>(Gugnoni & Ciarrocchi 2019</ns0:ref>). In the carcinogenic progression, downregulation of cell-adhesion molecules like epithelial cadherins, occludins, claudins, certain cytokeratins, and ZO-1 together with the coordinated upregulation of mesenchymal cadherins, vimentin, fibronectin and β1 and β3 integrins, promoted loss of cell-cell adhesion and apico-basal polarity and acquisition of invasive and migratory capacity <ns0:ref type='bibr' target='#b12'>(Gugnoni & Ciarrocchi 2019;</ns0:ref><ns0:ref type='bibr' target='#b21'>Lu & Kang 2019)</ns0:ref>. A group of transcription factors including Snail, Slug, Twist, zinc finger E-box-binding homeobox 1 and 2 (ZEB1, ZEB2) were well known to regulated EMT process partially or completely <ns0:ref type='bibr' target='#b12'>(Gugnoni & Ciarrocchi 2019;</ns0:ref><ns0:ref type='bibr' target='#b39'>Yang et al. 2018)</ns0:ref>. Therefore, we detected the mRNA levels of EMT-associated genes and the expression levels of EMT-associated proteins in ccRCC cells by qPCR and western blots respectively following LINC01234 knockdown. It revealed that the mRNA level of epithelial marker E-cadherin was increased, while the mRNA level of mesenchymal marker N-cadherin was decreased in Caki-2 and A498 cells following LINC01234 knockdown. The protein expression levels of the transcription factor Snail and the epithelial markers N-cadherin and Vimentin were reduced, while the protein expression level of the mesenchymal marker Ecadherin was up-regulated in A498 and Caki-2 cells with LINC01234 knockdown. These findings indicated that the function of LINC01234 was associated with EMT process. EMT was impaired after LINC01234 knockdown. In addition, we also found that inhibition of the βcatenin pathway contributed to the EMT impairment after LINC01234 depletion. All these evidences suggested that LINC01234 knockdown could inhibit the cell proliferation, migration and invasion, as well as EMT process in ccRCC. During EMT process, LINC01234 knockdown might suppress the expression of transcription factor Snail, and further stimulate the expression of E-cadherin, and inhibit the expressions of Vimentin and N-cadherin, which might result in a inhibition of malignant biological behaviors of ccRCC cells, such as cell proliferation, migration and invasion. Hypoxia could induce ccRCC cells to undergo EMT, angiogenesis and metastasis <ns0:ref type='bibr' target='#b23'>(Meléndez-Rodríguez et al. 2018;</ns0:ref><ns0:ref type='bibr' target='#b35'>Zhang et al. 2017)</ns0:ref>. Adaptation to a hypoxic environment played an important role in the progression of ccRCC <ns0:ref type='bibr' target='#b9'>(Garje et al. 2018)</ns0:ref>. Hypoxia was mediated via hypoxia-inducible factors <ns0:ref type='bibr' target='#b31'>(Semenza 2012)</ns0:ref>. Previously, HIF-1α was supposed to be a key oncogenic factor, but recent evidence showed HIF-2α was a predominant driver in renal cancer progression <ns0:ref type='bibr' target='#b18'>(Keith et al. 2011)</ns0:ref>. Currently, HIF-1α is supposed to be a ccRCC tumor suppressor, but the activity of HIF-1α is commonly diminished by chromosomal deletion in ccRCC <ns0:ref type='bibr' target='#b30'>(Schödel et al. 2016)</ns0:ref>. Conversely, HIF-2α has emerged as an oncogene that is essential for ccRCC tumor progression <ns0:ref type='bibr' target='#b23'>(Meléndez-Rodríguez et al. 2018;</ns0:ref><ns0:ref type='bibr' target='#b30'>Schödel et al. 2016)</ns0:ref>. The polymorphisms at the HIF-2α gene locus predispose to the development of ccRCC, and HIF-2α can promote tumor growth <ns0:ref type='bibr' target='#b30'>(Schödel et al. 2016)</ns0:ref>. Indeed, preclinical and clinical data have shown that pharmacological inhibitors of HIF-2α can efficiently inhibit ccRCC growth <ns0:ref type='bibr' target='#b23'>(Meléndez-Rodríguez et al. 2018)</ns0:ref>. HIF-2α was found to be more sensitive to moderate hypoxia and showed more enduring expression in hypoxic conditions <ns0:ref type='bibr' target='#b35'>(Zhang et al. 2017)</ns0:ref>. HIF-2α could translocate to the nucleus and bind to the hypoxia response elements <ns0:ref type='bibr' target='#b9'>(Garje et al. 2018)</ns0:ref>. This binding resulted in the expression of several target genes involved in angiogenesis, proliferation, migration and invasion of cancer cells, such as VEGFA, EGFR, c-Myc, Cyclin D1 and MET <ns0:ref type='bibr' target='#b9'>(Garje et al. 2018)</ns0:ref>. VEGFA played an important role in the formation of blood vessels, which was closely associated with carcinogenesis <ns0:ref type='bibr' target='#b33'>(Shi et al. 2019)</ns0:ref>. In ccRCC, as a well-known target of HIF-2α, VEGFA also played a vital role in angiogenesis and was a key target of anti-cancer therapeutic agents <ns0:ref type='bibr' target='#b9'>(Garje et al. 2018;</ns0:ref><ns0:ref type='bibr' target='#b23'>Meléndez-Rodríguez et al. 2018)</ns0:ref>. Besides, EGFR, c-Myc and Cyclin D1 were associated with ccRCC cell cycle and proliferation <ns0:ref type='bibr' target='#b23'>(Meléndez-Rodríguez et al. 2018)</ns0:ref>. EGFR signaling could also promote ccRCC survival <ns0:ref type='bibr' target='#b23'>(Meléndez-Rodríguez et al. 2018</ns0:ref>). Moreover, MET was related to ccRCC metastasis <ns0:ref type='bibr' target='#b23'>(Meléndez-Rodríguez et al. 2018)</ns0:ref>. Based on this above, we detected HIF-2α pathways after LINC01234 depletion. It revealed that the mRNA levels of HIF-2α and VEGFA were decreased in A498 and Caki-2 cells with LINC01234 knockdown. Similarly, the expression levels of proteins HIF-2α, VEGFA, EGFR, c-Myc, Cyclin D1 and MET were reduced in A498 and Caki-2 cells with LINC01234 knockdown. In our study, LINC01234 was expressed increasingly as the stage increased and its high expression level predicted a significantly worse disease-free survival rate or overall survival rate for the patients with ccRCC. LINC01234 knockdown suppressed cell proliferation, migration and invasion of ccRCC cells. Combined with all these findings above, it suggested that LINC01234 knockdown might suppress the expression of HIF-2α, and then inhibit the expression of VEGFA, EGFR, c-Myc, Cyclin D1 and MET, which might further inhibit the proliferation, metastasis and survival of ccRCC. Unfortunately, there was several limitations in our study. Firstly, the function of lncRNA was associated with its subcellular localization, but we did not identify the subcellular localization of LINC01234 in ccRCC cell lines. Secondly, although LINC01234 functioned as ceRNA to regulate CBFB expression by sponging miR-204-5p in gastric cancer, we did not identify any miRNAs as direct targets of LINC01234 to investigate whether LINC01234 was a ceRNA for miRNAs in ccRCC. It deserves to more investigations.</ns0:p></ns0:div>
<ns0:div><ns0:head>Conclusions</ns0:head><ns0:p>In summary, we constructed a lncRNA-based prognostic model with moderate accuracy and identified LINC01234 as an independent prognostic biomarker in ccRCC. Moreover, LINC01234 knockdown might inhibit the proliferation and metastasis of ccRCC cells by suppressing HIF-2α pathways. Therefore, LINC01234 might serve as a promising prognostic biomarker and a potential therapeutic target for patients with ccRCC. </ns0:p></ns0:div><ns0:figure xml:id='fig_0'><ns0:head>Figure 1</ns0:head><ns0:label>1</ns0:label><ns0:figDesc>Figure 1</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_1'><ns0:head>Figure 2</ns0:head><ns0:label>2</ns0:label><ns0:figDesc>Figure 2</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_2'><ns0:head>Figure 3</ns0:head><ns0:label>3</ns0:label><ns0:figDesc>Figure 3</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_3'><ns0:head>Figure 4 K</ns0:head><ns0:label>4</ns0:label><ns0:figDesc>Figure 4</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_4'><ns0:head>Figure 5</ns0:head><ns0:label>5</ns0:label><ns0:figDesc>Figure 5</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_5'><ns0:head /><ns0:label /><ns0:figDesc>(A-F) Expression levels of the 6 independent prognostic lncRNAs in different stages of patients with ccRCC. It suggested that AL357507.1, LINC01234, and LINC01956 were highly expressed at higher pathological stage of the disease, while LINC01234 exhibited the highest significance in terms of expression levels at different pathological stage of the disease. (G, H) Prognostic significance of LINC01234 in patients with ccRCC. It showed the high expression level of LINC01234 predicted a significantly worse disease-free survival rate or overall survival rate than that of the low expressed (all p < 0.05). ccRCC, clear cell renal cell carcinoma; GEPIA, Gene Expression Profiling Interactive Analysis.</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_6'><ns0:head>Figure 6 LINC01234</ns0:head><ns0:label>6</ns0:label><ns0:figDesc>Figure 6</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_7'><ns0:head>Figure 7 LINC01234</ns0:head><ns0:label>7</ns0:label><ns0:figDesc>Figure 7</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_8'><ns0:head>Figure 8 EMT</ns0:head><ns0:label>8</ns0:label><ns0:figDesc>Figure 8</ns0:figDesc></ns0:figure>
<ns0:note place='foot'>PeerJ reviewing PDF | (2020:05:49159:1:1:NEW 26 Aug 2020)</ns0:note>
</ns0:body>
" | "Editor comments (Giulia Piaggio)
MAJOR REVISIONS
The authors have to answer very carefully to all the criticisms raised by the reviewers, especially 2 and 3. The revised manuscript will go back to these reviewers and only if the authors have substantially modified the paper following the suggestions can it be considered again.
[# PeerJ Staff Note: Please ensure that all review comments are addressed in a rebuttal letter and any edits or clarifications mentioned in the letter are also inserted into the revised manuscript where appropriate. It is a common mistake to address reviewer questions in the rebuttal letter but not in the revised manuscript. If a reviewer raised a question then your readers will probably have the same question so you should ensure that the manuscript can stand alone without the rebuttal letter. Directions on how to prepare a rebuttal letter can be found at: https://peerj.com/benefits/academic-rebuttal-letters/ #]
Reviewer 1 (Anonymous)
Basic reporting
The manuscript is written in a clear and unambiguous technically English language throughout the paper. The article includes sufficient introduction but background references should be improved including most recent references. Overall article structures and figures sufficiently satisfy journal guideline. Raw data have been supplied. All results relevant for the hypothesis have been included.
Response: Thanks for your constructive comments. We updated and improved some corresponding background content and references into the most recent references.
Experimental design
The present research is of interest for PeerJ and for PeerJ readers. The experimental design is properly described and conducted according to technical and ethical standard. Methods are sufficiently described to be replicated
Validity of the findings
Results are solid, consistent with hypothesis and clearly reported
Comments for the Author
1) Manuscript can be improved by the evaluation of mRNA levels of genes reported in Figure 8 (not necessarily all)
Response: Thanks for your constructive comments. We evaluated mRNA levels of E-cadherin, N-cadherin, HIF-2α and VEGFA reported in Figure 8. It revealed mRNA level of E-cadherin was upregulated, while that of N-cadherin, HIF-2α and VEGFA were decreased following LINC01234 knockdown. The corresponding result was added in line line 267-269 (clean manuscript). It was also presented in Figure 8A and 8B.
2) The article is well written and results are clearly described. However, it is not clear why authors are referring to HIF-2α pathways. They can generalize by referring to HIF-2α pathway. Moreover, it is not clear why authors focused on HIF-2α pathway, although also HIF-1α was modulated.
Response: Thanks for your constructive comments. We focused on HIF-2α pathway because recent studies showed that HIF-2α, rather than HIF-1α, was a predominant driver in renal cancer. This was mentioned in line 71-80 (clean manuscript). Besides, although HIF-1α can act as a ccRCC tumor suppressor, HIF-1α activity is commonly diminished by chromosomal deletion in ccRCC. Conversely, HIF-2α has emerged as the key HIF isoform acting as an oncogene that is essential for ccRCC tumor progression. The polymorphisms at the HIF-2α gene locus predispose to the development of ccRCC, and HIF-2α promotes tumor growth. Indeed, preclinical and clinical data have shown that pharmacological inhibitors of HIF-2α can efficiently inhibit ccRCC growth. Therefore, our major concern is HIF-2α pathways. And I also added these corresponding background content and references into the most recent references.
3) Figure 8 can be improved. The quality of some western blot images is not satisfactory
Response: Thanks for your constructive comments. I have improved the quality of some western blot images in Figure 8.
4) Figure 8 it is not mentioned in the relative Results section.
Response: In the Results section, Figure 8A-D was mentioned in line 263-285 (clean manuscript).
Reviewer 2 (Anonymous)
Basic reporting
The English language in the manuscript by Jang et al. should be improved to ensure that an International audience can clearly understand the text. Nevertheless, the article conforms to professional standards of courtesy and expression.
Experimental design
The initial experimental approach is appropriate. It is essentially based on meta-analysis of bioinformatics data derived from an overall cohort of patients (> 600) with diagnosis of Clear Cell Renal Cell Carcinoma (ccRCC). The source for data analysis of ccRCC cases was the TCGA database referring to Data Release 14.0-December 18, 2018. In the second part of their study, however, authors focused on 1 out of 6 lncRNAs emerging as relevant for the patient outcome, specifically LINC01234 using in vitro experimental cell model, the Caki2 and A498 cell lines, without a clear rationale.
The source of both cell lines used are correctly identified in Methods sections
Validity of the findings
Authors found that LINC01234 knockdown compromised the HIF-2α pathways. LINC01234 silencing represses in turns HIF-2α, VEGFA, EGFR, c-Myc, Cyclin D1 and MET expression levels. Authors claim that LINC01234 is likely to regulate the progression of ccRCC by modulating the HIF-2α pathways, known to be highly associated with this particular type of cancer and propose it, at the same time, as a biomarker and potential therapeutic target.
Comments for the Author
The manuscritpt by Yang et al aimed to identify a subset of lncRNAs with a prognostic significance in terms of disease progression in Clear Cell Renal Cell Carcinoma (ccRCC). The study aligns in a very timely research topic that strongly indicate that lncRNAs, frequently deregulated in a variety of human cancer, including RCC, may serve as biomarkers for primary diagnostics.
Major concerns:
1. Based on the Kaplan Meyer survival curves (Figure 4) it appears that among the six independently prognostic lncRNAs identified by multivariate cox regression, lncRNA AL357507.1 exhibited the highest significance in terms of survival time. It is not clear why authors focused instead on lncRNA LINC01234 for the subsequent experiments. Which was the criteria adopted in determining the decision to follow and characterize this particular lncRNAs rather that one exhibiting the highest significance in terms of multivariate cox regression analysis?
Response: Thanks for your constructive comments. We identified six significantly independent prognostic lncRNAs including LINC01234, and AL357507.1 seemed to exhibit the highest significance in terms of survival time. However, we further validated their expression in different pathological stages and prognostic significance in ccRCC patients via GEPIA server. It revealed AL357507.1, LINC01234, and LINC01956 were highly expressed at higher pathological stage of the disease, while LINC01234 exhibited the highest significance in terms of expression at different pathological stage of the disease. It was a very interesting finding, because the pathological stage was closely associated with the prognosis of ccRCC patients. Moreover, GEPIA server revealed the significance of LINC01234 in terms of survival time. Unfortunately, GEPIA server could not provide the prognostic significance of the other 5 lncRNAs because the server showed the sample size was insufficient. Besides, I also refer to the recent studies and references, little is known about the other 5 lncRNAs, except for LINC01234. Therefore, we mainly focused on lncRNA LINC01234 for the subsequent experiments. And we will continue to investigate the role of other lncRNAs in our subsequent experiment. In the Result section, these corresponding contents were mentioned in line 231-241 (clean manuscript). In the discussion section, these corresponding contents were mentioned in line 329-341 (clean manuscript). And all these results were presented in figure 5 which has been updated according to your constructive advice.
2. On the same page, authors should explain why they have pursued the analysis and validated the expression and prognostic significance of LINC01234 in patients with ccRCC via GEPIA server without a comparative analysis with at least another among those six lncRNAS proved to be independent prognostic variable as well. It appears that LINC01234 was highly expressed at higher pathological stage of the disease, and this is a very interesting finding. A similar correlation should be investigated in at least another lncRNAs among those identified by bioinformatics.
Response: Thanks for your constructive advice. The pathological stage is closely associated with the prognosis of ccRCC patients, and GEPIA server can exhibit the expression levels of the six lncRNAs at different pathological stages of ccRCC. Therefore, according to your advice, via GEPIA server, we pursued the analysis and validated the expression and prognostic significance of all the six lncRNAs in patients with ccRCC. It revealed AL357507.1, LINC01234, and LINC01956 were highly expressed at higher pathological stage of the disease, while LINC01234 exhibited the highest significance in terms of expression at different pathological stage of the disease. All these results were presented in figure 5 which has been updated according to your constructive advice.
3. The bibliography supporting the association between high expression of HIF2a and progression of ccRCC mentioned in the Introduction (lanes 64-66) is very robust but has been recently expanded and deepened. Therefore, it is important to broaden the discussion on this matter and update bibliography for a correct positioning of the study in the context of the current literature on this specific topic.
Response: Thanks for your constructive advice. I have added some recent expanded and deepened corresponding background content and references into line 72-80 (clean manuscript). The content I added was as follows: “Although HIF-1α can act as a ccRCC tumor suppressor, HIF-1α activity is commonly diminished by chromosomal deletion in ccRCC. Conversely, HIF-2α has emerged as the key HIF isoform acting as an oncogene that is essential for ccRCC tumor progression. The polymorphisms at the HIF-2α gene locus predispose to the development of ccRCC, and HIF-2α promotes tumor growth. Indeed, preclinical and clinical data have shown that pharmacological inhibitors of HIF-2α can efficiently inhibit ccRCC growth”. Moreover, I also broaden the discussion on this matter and update bibliography for a correct positioning.
Reviewer 3 (Anonymous)
Basic reporting
The authors do not cite and comment some recent papers regarding the same field of interest. Among them:
Zeng, J., Lu, W., Liang, L. et al. Prognosis of clear cell renal cell carcinoma (ccRCC) based on a six-lncRNA-based risk score: an investigation based on RNA-sequencing data. J Transl Med 17, 281 (2019).
Jianguo Shi, Datian Zhang, Zhenhai Zhong, and Wen Zhang. lncRNA ROR promotes the progression of renal cell carcinoma through the miR-206/VEGF axis. Mol Med Rep. 2019 Oct; 20(4): 3782–3792.
Jiarun Zhang, Xiaotong Zhang, Chiyuan Piao, Jianbin Bi, Zhe Zhang, Zhenhua Li, Chuize Kong. A Long Non-Coding RNA Signature to Improve Prognostic Prediction in Clear Cell Renal Cell Carcinoma. Biomed Pharmacother. 2019 Oct;118:109079.
Cox, A., Tolkach, Y., Kristiansen, G. et al. The lncRNA Fer1L4 is an adverse prognostic parameter in clear-cell renal-cell carcinoma. Clin Transl Oncol (2020).
Response:
Thank you very much for providing these valuable references regarding the same field of interest. I cite and comment them in the appropriate paragraph.
(1) In line 306-319 (clean manuscript), I cite and comment the corresponding content which is as follows: “It was also reported that some aberrant lncRNAs could serve as prognostic indicators in ccRCC, such as lncRNA Fer1L4 (Cox et al. 2020). With the development of molecular biological techniques and bioinformatics, more and more lncRNAs were marked as novel biomarkers and prognostic signatures for ccRCC utilizing TCGA database. For example, lncRNA Fer1L4 was overexpressed in ccRCC tissues, and its high expression levels were found in higher grade, higher stage, and metastatic tumors (Cox et al. 2020). LncRNA Fer1L4 overexpression was also an independent prognostic factor for patients with ccRCC (Cox et al. 2020). It was also reported an 11-lncRNA signature (AC245100.1, AP002761.1, LINC00488, AC017033.1, LINC-PINT, COL5A1-AS1, AC026471.4, AL009181.1, LINC00524, HOTTIP, AL078590.3) and a 6-lncRNA signature (CTA‑384D8.35, CTD‑2263F21.1, LINC01510, RP11‑352G9.1, RP11‑395B7.2, RP11‑426C22.4) were clearly linked to the overall survival (OS) rate of ccRCC patients based on TCGA database (Zeng et al. 2019; Zhang et al. 2019).” The corresponding references are also updated.
(2) In line 419-422 (clean manuscript), I also updated and revised some content and corresponding references. The content is as follows: “VEGFA played an important role in the formation of blood vessels, which was closely associated with carcinogenesis (Shi et al. 2019). In ccRCC, as a well-known target of HIF-2α, VEGFA also played a vital role in angiogenesis and was a key target of anti-cancer therapeutic agents (Garje et al. 2018; Meléndez-Rodríguez et al. 2018).”
Experimental design
The paper is basically divided in two sections. In the first one the authors identified differentially expressed long non coding RNAs (lncRNAs) in clear cell renal cell carcinoma samples using the existing lncRNA expression data and corresponding clinical data from TCGA, to identify novel prognostic markers. The approach is convincing but the data presented in Figures 1-2-3 are presented in a confusing manner and the analysis content is difficult to understand.
In the second section the authors downregulate the expression of LINC01234, identified as a possible marker of aggressiveness in this kind of cancer, observing reduced cell viability, clonegenic potential and migration/invasion potential in two renal cancinoma cell lines. Then they evaluate the effect of LINC01234 downregulation on the expression of different proteins involved in epithelial to mesenchimal transition and respone to hypoxia among the others. The figures are well organized and convincing, but the authors do not include details about the number of the replicates in the case of Western blot analysis (figure 8). In figure 6A-C the font size is too little.
Response:
(1) The data presented in Figures 1-2-3 are presented in the order of analysis. Firstly, we analyzed the differentially expressed lncRNAs, and the result was presented as Figures 1A. Then we identified lncRNAs associated with the prognosis by univariate cox regression from the differentially expressed lncRNAs. Furtherly, we used LASSO regression to screen key prognostic lncRNAs based on the lncRNAs identified by univariate cox regression. Figures 1B, C were the process and result of LASSO regression. Using the key prognostic lncRNAs identified by LASSO regression, we established a risk score model by multivariate cox regression, which divided all patients into high-risk and low-risk group. Then we used K-M survival curve to identify whether this model was significant (Figure 2A), and we also used ROC curve to evaluate the accuracy of the model (Figure 2B). Besides, Figure 2C presented that the risk score model divided all patients into obvious two clusters, the alive and the dead. Figure 2D presented the expression levels of the key prognostic lncRNAs in the high-risk group and low-risk group. These key prognostic lncRNAs were identified by LASSO regression and used for constructing the risk score model. Figure 3 was performed to identify independent prognostic factors by multivariate cox regression among the key prognostic lncRNAs which were identified by LASSO regression and were used for constructing the risk score model.
(2) Western blot analysis (figure 8) was performed in triplicate, which was illustrate in line 176-177 (clean manuscript). And I have adjusted the font size in figure 6A-C.
Validity of the findings
Unfortunately, the conclusions are not supported but the results presented in the second part of the manuscript. Even the title reported a statement that is not demonstrated. The downregulation of LIN01234 decreases HIF-2alpha expression but there is no evidence that this reduced expression is responsible of decreased in vitro aggressiveness. The effect of LINC01234 is merely descriptive and the authors do not explore the molecular mechanisms involved in the modulation of gene expression induced by LINC01234 downregulation. The authors do not even speculate about this aspect in Discussion section: is LINC01234 a ceRNA for miRNAs? Recently a pre-print paper showed that miR-513a-5p is a target of LINC01234.
Response: Thanks a lot for your constructive and valuable advice. We investigated the effect of LINC01234 in ccRCC by loss of function. We found the downregulation of LIN01234 inhibited ccRCC cell aggressiveness and the expression of HIF-2alpha as well as its several downstream targets in vitro. It indicated that LINC01234 played an important role in cell aggressiveness of ccRCC, and this role was associated with HIF-2alpha. Unfortunately, it is unknown about how LINC01234 downregulation influences the expression of HIF-2alpha and its downstream targets. Your constructive and valuable advice is a good guide here for our further research and we will continue to deeply investigate the molecular mechanisms of LINC01234. You spoke of a recent pre-print paper that showed miR-513a-5p is a target of LINC01234, may LINC01234 act as a ceRNA for miR-513a-5p to regulate the gene expression. It deserves to further investigations. In addition, as the limitations in our study, we also illustrate these limitations in line 439-445 (clean manuscript). Besides, the recent pre-print paper is not formally published at present, so miR-513a-5p was not discussed in our study.
" | Here is a paper. Please give your review comments after reading it. |
9,794 | "<ns0:abstract xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:p>Motivation. Long read sequencing and Bionano Genomics optical maps are two techniques that, when used together, make it possible to reconstruct entire chromosome or chromosome arms structure. However, the existing tools are often too conservative and organization of contigs into scaffolds is not always optimal.</ns0:p><ns0:p>Results. We developed BiSCoT (Bionano SCaffolding COrrection Tool), a tool that post-processes files generated during a Bionano scaffolding in order to produce an assembly of greater contiguity and quality. BiSCoT was tested on a human genome and four publicly available plant genomes sequenced with Nanopore long reads and improved significantly the contiguity and quality of the assemblies. BiSCoT generates a fasta file of the assembly as well as an AGP file which describes the new organization of the input assembly.</ns0:p></ns0:div>
</ns0:abstract>
<ns0:body xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:head>INTRODUCTION</ns0:head><ns0:p>Assembling large and repetitive genomes, such as plant genomes, is a challenging field in bioinformatics.</ns0:p><ns0:p>The appearance of short reads technologies several years ago improved considerably the number of genomes publicly available. However, a high proportion of them are still fragmented and few represent the chromosome organization of the genome. Recently, long reads sequencing techniques, like Oxford Nanopore Technologies and Pacific Biosciences, were introduced to improve the contiguity of assemblies, by sequencing DNA molecules that can range from a few kilobases to more than a megabase in size <ns0:ref type='bibr' target='#b7'>(Istace et al. (2017)</ns0:ref>; <ns0:ref type='bibr' target='#b13'>Schmidt et al. (2017)</ns0:ref>; <ns0:ref type='bibr' target='#b11'>Kim et al. (2019)</ns0:ref>; <ns0:ref type='bibr'>Shafin et al. (2019)</ns0:ref>). Nevertheless and even if the assemblies were greatly improved, the chromosome-level organization of the sequenced genome cannot be deciphered in a majority of cases. In 2017, Bionano Genomics launched its Saphyr system which was able to generate optical maps of a genome, by using the distribution of enzymatic labelling sites. These maps were used to orient and order contigs into scaffolds but the real improvement came in 2018, when Bionano Genomics introduced their Direct Label and Stain (DLS) technology that was able to produce genome maps at the chromosome-level with a N50 several times higher than previously Figure <ns0:ref type='figure'>1</ns0:ref>. The Bionano scaffolding tool does not merge contigs even if they share labels. Instead, it inserts 13 N's gap between contigs, thus artificially duplicating the shared region. a. BiSCoT merges contigs that share enzymatic labelling sites. b. If contigs do not share labels but share a genomic region, BiSCoT attempts to merge them by aligning the borders of the contigs. c. The Bionano scaffolding tool does not handle cases where contigs can be inserted into others. BiSCoT attempts to merge the inserted map with the one containing it if they share labels. variation studies. They originate from overlaps that are not fused in the input assembly and usually correspond to allelic duplications. In addition, contigs can sometimes be inserted into other contigs, these cases are not handled by the Bionano scaffolding tool that discards the inserted contigs (Figure <ns0:ref type='figure'>1</ns0:ref> case 3).</ns0:p><ns0:p>We developed BiSCoT, a python script that examinates data generated during a previous Bionano scaffolding and merges contigs separated by a 13-Ns gap if needed. BiSCoT also re-evaluates gap sizes and searches for an alignment between two contigs if the gap size is inferior to 1,000 nucleotides. BiSCoT is therefore not a traditional scaffolder since it can only be used to improve an existing scaffolding, based on an optical map.</ns0:p></ns0:div>
<ns0:div><ns0:head>METHODS</ns0:head></ns0:div>
<ns0:div><ns0:head>Mandatory files loading</ns0:head><ns0:p>During the scaffolding, the Bionano scaffolder generates a visual representation of the hybrid scaffolds that is called an 'anchor'. It also generates one '.key' file, which describes the mapping between map identifiers and contig names, several CMAP files, which contain the position of enzymatic labelling sites on contig maps and on the anchor, and a XMAP file, that describes the alignment between a contig map and an anchor.</ns0:p><ns0:p>BiSCoT first loads the contigs into memory based on the key file. Then, the anchor CMAP file and contig CMAP files are loaded into memory. Finally, the XMAP file is parsed and loaded.</ns0:p></ns0:div>
<ns0:div><ns0:head>Scaffolding</ns0:head><ns0:p>Alignments of contigs onto anchors contained in the XMAP file are first sorted by their starting position on the anchor. Then, alignments on one anchor are parsed by pairs of adjacent contigs, i.e alignment of contig C k is examined at the same time as contig C n , with C k aligned before C n on the anchor. Aligned anchor labels are extracted from these alignments and a list of shared labels L n,k is built. For the following cases, we suppose C k and C n to be aligned on the forward strand (Figure <ns0:ref type='figure'>1</ns0:ref>).</ns0:p></ns0:div>
<ns0:div><ns0:head>2/6</ns0:head><ns0:p>PeerJ reviewing <ns0:ref type='table' target='#tab_0'>PDF | (2019:12:44195:1:2:NEW 26 Jun 2020)</ns0:ref> Manuscript to be reviewed</ns0:p></ns0:div>
<ns0:div><ns0:head>Case 1: contig maps share at least one anchor label</ns0:head><ns0:p>The last label l from L n,k is extracted and the position P l of l on both contigs C k and C n is recovered from the CMAP files. In the resulting scaffold, the sequence of C k will be included up to the P l position and the sequence of C n will be included from the P l position. In this case, the gap is removed, both contigs C k and C n are fused and BiSCoT generates a single contig instead of two contigs initially separated by a gap in the input assembly. </ns0:p><ns0:formula xml:id='formula_0'>d k = Size k − Em k (2)</ns0:formula><ns0:formula xml:id='formula_1'>d n = Sm n (3)</ns0:formula><ns0:p>Finally, we can compute the gap size g with:</ns0:p><ns0:formula xml:id='formula_2'>g = n − d k − d n (4)</ns0:formula><ns0:p>If g ≤ 1000, a BLAT <ns0:ref type='bibr' target='#b10'>(Kent (2002)</ns0:ref>) alignment of the last 30kb of C k is launched against the first 30kb of Finally, if an Illumina polishing step was done before or after Bionano scaffolding, we recommend doing one additional round of polishing using Illumina reads after BiSCoT has been applied. Indeed, short reads tend to be aligned only against one copy of the duplicated regions, leaving the other copy unpolished.</ns0:p></ns0:div>
<ns0:div><ns0:head>RESULTS AND DISCUSSIONS Validation on simulated data</ns0:head><ns0:p>In order to simulate a genome assembly, we downloaded the chromosome 1 of the GRCh38.p12 human reference genome and fragmented it to create contigs. We generated 120 contigs with an N50 size of 2.4Mb and a cumulative size of 231Mb. Contigs were generated with either overlaps or gaps between them. We introduced 50 gaps with a mean length of 50kb, the smallest being 3.4kbp long and the largest 99.6kb long, and 50 overlaps with a mean size of 44kb, the smallest being 278b long and the largest 98.6kb long. We also generated five contigs, with an N50 of 254kb, that were subsequences of larger contigs, to simulate contained contigs.</ns0:p><ns0:p>Then, we used these contigs and Bionano DLE and BspQI optical maps available on the Bionano Genomics website as input to the Bionano scaffolder. We gave the results of this scaffolding to BiS-</ns0:p><ns0:p>CoT and aligned all assemblies to the chromosome 1 reference using Quast <ns0:ref type='bibr' target='#b4'>(Gurevich et al. (2013)</ns0:ref>, v5.0.2).</ns0:p><ns0:p>BiSCoT was able to resolve 39 overlaps out of the 50 we introduced (Supplementary Table <ns0:ref type='table' target='#tab_0'>1</ns0:ref>), 31 using shared labels and 8 using a Blat alignment. The 11 remaining overlaps could not be resolved</ns0:p></ns0:div>
<ns0:div><ns0:head>3/6</ns0:head><ns0:p>PeerJ reviewing PDF | (2019:12:44195:1:2:NEW 26 Jun 2020)</ns0:p><ns0:p>Manuscript to be reviewed due to contigs not sharing enough labels or the overlap being too small to produce an alignment of sufficient confidence. BiSCoT was also able to integrate all contained contigs back to their original place in the assembly. Furthermore, BiSCoT did not close any of the real gaps introduced during the assembly generation.</ns0:p><ns0:p>Regarding assembly metrics (Supplementary Table <ns0:ref type='table'>2</ns0:ref>), The N50 decreased by 1.4% in scaffolds and increased by 22% in contigs. The number of Ns in scaffolds decreased from 20.7Mb to 20.4Mb. Moreover, the number of misassemblies decreased by 68% after applying BiSCoT and the duplication ratio estimated by Quast decreased from 1.026 in Bionano scaffolds to 1.021 in BiSCoT scaffolds.</ns0:p><ns0:p>In order to estimate the accuracy of gap sizes, we compared the gap sizes we introduced in the input assembly to the ones that were estimated using optical maps (Supplementary Figure <ns0:ref type='figure'>1</ns0:ref>). We found that estimated gap sizes were very close to the reality, with a mean scaled absolute error of 0.8%.</ns0:p></ns0:div>
<ns0:div><ns0:head>Validation on real data</ns0:head><ns0:p>We downloaded genome assemblies for which a DLE optical map was available: the NA12878 human genome <ns0:ref type='bibr' target='#b9'>(Jain et al. (2018)</ns0:ref>), Brassica oleracea HDEM (PRJEB26621, <ns0:ref type='bibr' target='#b1'>Belser et al. (2018)</ns0:ref>), Brassica rapa Z1 (PRJEB26620, <ns0:ref type='bibr' target='#b1'>Belser et al. (2018)</ns0:ref>) and Musa schizocarpa (PRJEB26661, <ns0:ref type='bibr' target='#b1'>Belser et al. (2018)</ns0:ref>) and Sorghum bicolor Tx430 (PRJNA472170, <ns0:ref type='bibr' target='#b2'>Deschamps et al. (2018)</ns0:ref>).</ns0:p><ns0:p>The QUAST and BUSCO <ns0:ref type='bibr' target='#b17'>(Simão et al. (2015)</ns0:ref>, v4.0.5) tools were used respectively to evaluate the number of misassemblies to the GRCh38.p12 human reference genome and the number of conserved genes among eukaryotes.</ns0:p><ns0:p>In all cases, we first used the Bionano workflow to scaffold the draft assembly and launched BiSCoT using the files generated by the Bionano tools (Table <ns0:ref type='table' target='#tab_0'>1</ns0:ref>, Supplementary Table <ns0:ref type='table'>3, 4, 5 and 6</ns0:ref>). The output of the Bionano workflow and BiSCoT are scaffolds, but we generated a contig file for each assembly by splitting each scaffold at every position with at least one N.</ns0:p><ns0:p>Concerning the NA12878 genome, we could detect 515 overlapping regions with a mean size of 47kb and representing in total 24.5Mb of duplicated sequences. Among these 515 regions, 499 were corrected by BiSCoT using either shared labels (113 regions) or a BLAT alignment (386 regions) when no shared labels were found.</ns0:p><ns0:p>Globally, the contig NX and NGAX metrics increased drastically: the contigs NGA50 of NA12878 increased by around 10%, going from 5.8Mb to 6.3Mb. The scaffolds NGAX metrics also increased: the scaffolds NGA50 increased from 10.8Mb in Bionano scaffolds to 11.7Mb in BiSCoT scaffolds. Moreover, the number of Ns decreased marginally and the number of complete eukaryotic genes stayed the same in scaffolds. More importantly, when aligning the assemblies against the reference genome, we could detect a decrease in the number of mis-assemblies going from 1,602 in Bionano scaffolds to 1,515 in BiSCoT </ns0:p></ns0:div>
<ns0:div><ns0:head>SUMMARY</ns0:head><ns0:p>Thanks to the advent of long reads and optical maps technologies, it is now possible to obtain highquality chromosome-scale assemblies. However, the official Bionano scaffolding tool does not always perform optimally when joining two contigs. Indeed, it does not merge two sequences when they share a genomic region, creating artificial gaps in the assembly. We developed BiSCoT, a tool that corrects these problematic regions in a prior Bionano scaffolding and showed that it increased significantly contiguity metrics of the resulting assembly, while preserving its quality.</ns0:p></ns0:div><ns0:figure xml:id='fig_0'><ns0:head>Case 2 :</ns0:head><ns0:label>2</ns0:label><ns0:figDesc>contig maps do not share anchor labels Let Size k be the size of the contig C k , Sm k and Em k the start and end of an alignment on a contig map and Sa k and Ea k the corresponding coordinates on the anchor. The number n of bases between the last aligned label of C k and the first aligned label of C n is then: n = Sa n − Ea k (1) We then have to subtract the part d k of C k after the last aligned label of C k and the part d n of C n before the first aligned label of C n :</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_1'><ns0:head /><ns0:label /><ns0:figDesc>C n . If an alignment is found and if its score is higher than 5,000, C k and C n are merged at the starting position of the alignment and, as in case1, BiSCoT generates a single contig instead of two contigs initially separated by a gap in the input assembly. Otherwise, a number g of Ns is inserted between C k and C n .Case 3: insertion of small contigs Let Sm k and Em k the start and end of an alignment on a contig map. If [Sm n , Em n ] ⊂ [Sm k , Em k ], then the left-most shared label identifier l l and right-most shared label identifier l r are extracted. If C n hasmore of its labels mapped in this region than C k , the sequence of C n will be inserted between l l and l r in the scaffolds. Otherwise, the sequence of C k remains unchanged and C n will be included as a singleton sequence in the scaffolds file.</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_2'><ns0:head>Figure 2 .</ns0:head><ns0:label>2</ns0:label><ns0:figDesc>Figure 2. a. Distribution of the sizes of overlapping regions in the raw assemblies. Detection was done using either Bionano labels (Case 1) or a BLAT alignment (Case 2). b. N50 contigs of raw assemblies and assemblies before or after BiSCoT treatment.</ns0:figDesc><ns0:graphic coords='6,152.07,63.79,392.80,112.74' type='bitmap' /></ns0:figure>
<ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='3,152.07,63.78,392.89,227.05' type='bitmap' /></ns0:figure>
<ns0:figure type='table' xml:id='tab_0'><ns0:head>Table 1 .</ns0:head><ns0:label>1</ns0:label><ns0:figDesc>Metrics of the NA12878 scaffolds and contigs before or after BiSCoT treatment. Bold formatting indicates the best scoring assembly among contigs.</ns0:figDesc><ns0:table><ns0:row><ns0:cell /><ns0:cell>Nanopore contigs</ns0:cell><ns0:cell cols='2'>Bionano</ns0:cell><ns0:cell cols='2'>BiSCoT</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell>Contigs</ns0:cell><ns0:cell>Scaffolds</ns0:cell><ns0:cell>Contigs</ns0:cell><ns0:cell>Scaffolds</ns0:cell></ns0:row><ns0:row><ns0:cell>Cumulative size</ns0:cell><ns0:cell>2,818,937,673</ns0:cell><ns0:cell>2,818,997,568</ns0:cell><ns0:cell>2,878,230,106</ns0:cell><ns0:cell>2,810,480,725</ns0:cell><ns0:cell>2,868,077,379</ns0:cell></ns0:row><ns0:row><ns0:cell>N50</ns0:cell><ns0:cell>11,821,944</ns0:cell><ns0:cell>10,566,783</ns0:cell><ns0:cell>86,858,024</ns0:cell><ns0:cell>12,894,141</ns0:cell><ns0:cell>86,833,728</ns0:cell></ns0:row><ns0:row><ns0:cell>L50</ns0:cell><ns0:cell>67</ns0:cell><ns0:cell>71</ns0:cell><ns0:cell>14</ns0:cell><ns0:cell>64</ns0:cell><ns0:cell>14</ns0:cell></ns0:row><ns0:row><ns0:cell>N90</ns0:cell><ns0:cell>2,143,851</ns0:cell><ns0:cell>1,863,173</ns0:cell><ns0:cell>26,054,782</ns0:cell><ns0:cell>2,321,940</ns0:cell><ns0:cell>26,037,000</ns0:cell></ns0:row><ns0:row><ns0:cell>L90</ns0:cell><ns0:cell>280</ns0:cell><ns0:cell>301</ns0:cell><ns0:cell>36</ns0:cell><ns0:cell>254</ns0:cell><ns0:cell>36</ns0:cell></ns0:row><ns0:row><ns0:cell>auN*</ns0:cell><ns0:cell>15,164,719</ns0:cell><ns0:cell>14,547,428</ns0:cell><ns0:cell>82,760,251</ns0:cell><ns0:cell>15,977,835</ns0:cell><ns0:cell>82,474,548</ns0:cell></ns0:row><ns0:row><ns0:cell># Ns</ns0:cell><ns0:cell>0</ns0:cell><ns0:cell>0</ns0:cell><ns0:cell>59,232,538</ns0:cell><ns0:cell>0</ns0:cell><ns0:cell>57,596,654</ns0:cell></ns0:row><ns0:row><ns0:cell>NGA50</ns0:cell><ns0:cell>5,794,944</ns0:cell><ns0:cell>5,729,014</ns0:cell><ns0:cell>10,816,842</ns0:cell><ns0:cell>6,360,576</ns0:cell><ns0:cell>11,713,900</ns0:cell></ns0:row><ns0:row><ns0:cell>NGA75</ns0:cell><ns0:cell>1,511,206</ns0:cell><ns0:cell>1,495,174</ns0:cell><ns0:cell>2,701,541</ns0:cell><ns0:cell>1,596,102</ns0:cell><ns0:cell>2,938,187</ns0:cell></ns0:row><ns0:row><ns0:cell># misassemblies</ns0:cell><ns0:cell>1,356</ns0:cell><ns0:cell>1,299</ns0:cell><ns0:cell>1,602</ns0:cell><ns0:cell>1,278</ns0:cell><ns0:cell>1,515</ns0:cell></ns0:row><ns0:row><ns0:cell>Complete BUSCOs</ns0:cell><ns0:cell>235 (92.2%)</ns0:cell><ns0:cell>234 (91.8%)</ns0:cell><ns0:cell>231 (90.6%)</ns0:cell><ns0:cell>235 (92.2%)</ns0:cell><ns0:cell>231 (90.6%)</ns0:cell></ns0:row><ns0:row><ns0:cell>Duplicated BUSCOs</ns0:cell><ns0:cell>5 (2.0%)</ns0:cell><ns0:cell>4 (1.6%)</ns0:cell><ns0:cell>4 (1.6%)</ns0:cell><ns0:cell>4 (1.6%)</ns0:cell><ns0:cell>4 (1.6%)</ns0:cell></ns0:row><ns0:row><ns0:cell>Missing BUSCOs</ns0:cell><ns0:cell>11 (4.3%)</ns0:cell><ns0:cell>10 (3.9%)</ns0:cell><ns0:cell>13 (5.1%)</ns0:cell><ns0:cell>10 (3.9%)</ns0:cell><ns0:cell>13 (5.1%)</ns0:cell></ns0:row><ns0:row><ns0:cell cols='4'>(*) auN is a new metric to measure assembly contiguity (Li (2020))</ns0:cell><ns0:cell /><ns0:cell /></ns0:row></ns0:table></ns0:figure>
<ns0:note place='foot'><ns0:ref type='bibr' target='#b1'>(Belser et al. (2018)</ns0:ref>;<ns0:ref type='bibr' target='#b3'>Formenti et al. (2018)</ns0:ref>;<ns0:ref type='bibr' target='#b5'>Hu et al. (2019)</ns0:ref>).However, scaffolds generated with the tool provided by Bionano Genomics do not reach optimal contiguity. Indeed, when two contigs C 1 and C 2 are found to share labels, one could expect that the tool would merge the two sequences at the shared site. Instead, the software chooses a conservative approach and outputs the sequence of C 1 followed by a 13-Ns gap and then the C 2 sequence, thus duplicating the region that is shared by the two contigs (Figure1case 1 and 2) and in numerous cases, these duplicated regions could reach several kilobases. As an example, on the human genome we used to evaluate BiSCoT (see Results), we could detect 515 of those regions, affecting 16 genes and corresponding to around 24.5Mb of duplicated sequences, the longest being 237kb in size. These duplicated regions affect the contiguity and have to be corrected as they can be problematic for downstream analyses, like copy numberPeerJ reviewing PDF | (2019:12:44195:1:2:NEW 26 Jun 2020)Manuscript to be reviewed</ns0:note>
</ns0:body>
" | "First, we would like to thank the reviewers for their positive feedback and constructive criticism, including some very useful suggestions to assist with improving the manuscript. We have carefully considered their recommendations and edited the manuscript accordingly.
Reviewer 1
I liked the work presented by the authors in this manuscript. I have a few comments:
1. A positive control could help confirm the validity of the BiSCoT code. I am not sure if such an experiment exists.
Answer: Thank you very much for this suggestion. We simulated contigs on the chromosome 1 of the human reference genome and incorporated contig duplications and show that the usage of BiSCoT fused a large fraction of the overlapping contigs. The result are presented in the paragraph 'Validation on simulated data' of the results section. BiSCoT not only detected artifactual duplications but also decreased the number of misassemblies in scaffolds and increased the contig contiguity.
2. I am surprised that the method developed by Bionano Genomics doesn't do it as is. As I understand, most community uses the software made by the community, which makes me further speculate the magnitude of the issue brought to light by the authors. Is this a common issue when using optical maps for scaffolding?
Answer: This issue happens every time a hybrid scaffolding is done using optical maps. The consequences of this issue can be seen by comparing contig metrics before and after Bionano scaffolding. As seen in Table 1 and Supplementary Tables, the contig N50 falls drastically after scaffolding, showing that the Bionano scaffolder does not do any sequence merging but rather only conflict resolution. The Bionano users have often reported this issue at user meetings and many of them are now using our software.
3. Is there a risk in doing this contig correction analysis on unphased assemblies? If an issue, this would escalate further in plant genomes.
Answer: On diploid and heterozygous genomes, the hybrid scaffolding is more difficult. Generally, the optical maps are able to separate haplotypes as the input molecule length is of higher order than long reads. Most of the time, the assembly merged haplotypes and the hybrid scaffolder provided by Bionano Genomics is not able to place a given contigs on two different maps, leading to a phased assembly that contains a high proportion of gaps. In these cases, users should filter the optical maps and keep only a single haplotype before hybrid scaffolding.
I send my best wishes to the authors.
Reviewer 2
The manuscript does not discuss prior work, except regarding the Bionano product. There are ”traditional” scaffolders, working with paired-end and mate-pair reads, that handles overlapping contigs. There are also other scaffolders than Bionano’s, why cannot they be used?
Answer: We thank the reviewer for his comment, and realize that this aspect was not sufficiently clear in the article. Here we did not describe a new scaffolder, as BiSCoT can not handle sequencing data like paired-end and mate-pair reads but only optical maps. In contrast, actual scaffolders are not able to use optical maps produced using the Saphyr system provided by the Bionano Genomics company. Comparing BiSCoT to “traditional” scaffolders would be out of the scope of its functionalities. BiSCoT post-processes the results from a prior Bionano scaffolding and corrects problematic regions caused by the scaffolder.
I think you are re-defining the contig concept a bit, which made me confused when reading your results. My understanding of contigs is that they are the result of overlapping and merged reads (short or long) from the assembler. Then a scaffolder is further putting contigs together using auxiliary information (such as linked reads or optical maps, sometimes ever RNA evidence). It does not make sense to me to classify a scaffold as a contig just because identical sequence has been identified by the scaffolder. I can understand your thinking and will not insist on my definition (which may be outdated), but I think you should clarify this reclassification of scaffolds into contigs.
Answer: You are right, contigs are traditionally the result of merged reads. However, the assembly process is often made up of several stages which modified the inital contig set. This is why we have chosen to adopt a common definition of a contig, which is a contiguous sequence without any gaps. As BiSCoT is able to merge contigs using optical map information, the initial contigs are then modified and to be fair we choose to adopt the same definition of contig for all the assemblies that we compared. We realized that this aspect was not clear in the previous manuscript and we decided to add a sentence to explain what we call “a contig” in the methods section.
Please consider making BiSCoT available on PyPI (pipy.org) so that all a user needs to do is ”pip install BiSCoT”. It is actually pretty easy to set it up and it makes your tool so much more available.
Answer: We thank the reviewer for his suggestion, BiSCoT is now available on github and PyPI (https://pypi.org/project/biscot/).
Text issues:
* Change ”long reads sequencing” to ”long read sequencing” (first occurrence in the abstract).
Answer: We took the reviewer advice into account and made the requested change.
* The key sentence in the abstract, 'We developed BiSCoT (Bionano SCaffolding COrrection Tool), a tool that uses informations produced by a pre-existing assembly based on optical maps as input and improves the contiguity and the quality of the generated assembly.” has several deficiencies in my opinion. (1) An assembly does not produce information. It might for example ”provide” or ”contain” information, but you could also rephrase. (2) It is unclear what the actual input is. (3) The sentence does not quite state what BiSCoT does, only that it improves quality. It would be good to be more explicit and explain that is is ”post-processing of Bionano scaffolds” or something like that. (4) I was confused about what ”pre-existing assembly” would mean. One can read that as ”any assembly from a repository”, but you are actually only working with Bionano-scaffolded assemblies. (5) The phrase ”… of the generated assembly” is unclear since it is not obvious that your tool is generating an assembly. (6) Finally, ”information” should not be used in plural.
Answer: We took the reviewer advice into account and rephrased the part of the abstract that was problematic to be more easily understandable.
* The word ”apparition” is not strange in this context. ”The appearance of short read technologies” is better.
* I don’t know what an ”endonuclease nicking site” is.
* ”several folds higher” is not good English, use ”several times higher”.
* Change ”doesn’t” to ”does not”.
* ”contained into” should probably be ”contained in”.
Answer: We took the reviewer advice into account and made the requested changes.
* In ”Case 1”, what is meant by 'First,the right-most shared label identifier l is extracted”?
* Please define variables and functions clearly. Reading ”C_k (P_l(C_k))” is just confusing (see ”Case 1”). I suggest rewriting in the style ”The position P_l for a label identifier l” etc.
* What is really meant by ”The position l … on both contigs … is searched”?
Answer: We rephrased the ‘Case 1’ paragraph of the Methods section to be clearer.
* The columns in Table 1 should be justified so that it is easy to read and compare numbers. The left-most column should be left-justified and the data-columns should be justified on decimal point.
* Bold text for best values in the tables (including Supplemental tables) is not consistent. There are several places where there are several numbers that are equal but only some are in bold.
Answer: We formatted all tables following the reviewer advice and made sure the bold formatting was consistent across all columns.
* Capitalisation in the reference list should be corrected for words like ”Nanopore”, ”Promethion”, ”Solanum”, ”BLAST”.
Answer: We followed reviewer advice and checked the capitalisation of references.
Experimental design
## Code ##
I looked at your code. I would strongly recommend to structure the code by defining functions for any kind of logical block of code you have. Limit the length of a function to one screen page. You have one function, ”main”, and it encompasses the program in its entirety. There is not a single block of code that is easy to test. This makes it hard to build upon your code.
I also note that you have code like
try:
…
except:
pass
That is a bad practice. If an error occurs, the ”pass” statement simply hides it from the user. There is usually a good reason for an exception to occur, so don’t hide them — handle them.
Answer: We took the reviewer advice into account and overhauled significantly the codebase.
## Method ##
You start the method section with 'Briefly, BiSCoT loads all necessary files into memory”. That is not specific enough for me. I should not need to guess what is considered needed. In the same sentence, you refer to alignments, but at that point it is not clear where the alignments come from.
Answer: We removed this paragraph as it was redundant with the following ones.
For Case 2, for contigs that are linked (I presume, it is a bit unclear) but do no share labels, you align contig ends if the estimated distance is at most 1000 nt. Why 1000? Is that a ”safe” distance?
Answer: We rephrased portions of the Methods section to make it clear that contigs in Case 1 and 2 are aligned to the same anchor.
There is a bit of uncertainty around distances estimated using Bionano maps. In other words, due to this small uncertainty, contigs that we see ‘not overlapping’ in the alignments may in reality be overlapping. The 1,000bp distance is a safeguard to take this uncertainty into account. Moreover, we added a ‘--aggressive’ flag to BiSCoT which lets the user specify if the BLAT step should be done or not.
I am deducing that the distance between two contigs is _estimated_ using Equation 1, but I think you should make that clear. How well is that distance estimated?
Answer: The distance between two contigs is estimated using the optical map and the equation (4). First we make it clearer in the text, and we investigate how well this distance is estimated using our simulated experiments on chromosome 1 of the human genome.
It is not clear to me that ”Case 3”, merging of contigs sharing labels indicating that one contig _may_ have accidentally been excised from another contig. How can you exclude that the smaller (in your case) contig is not simply a partial duplication? What is the rationale?
Answer: The third case does not correspond to a contig ‘accidentally’ excised from another one. We believe that small contigs contained in larger contigs are the results of i) heterozygosity or ii) collapsed repeats in the assembly. If the small contig being aligned onto the larger one is indeed a partial duplication, then both contigs will share all their labels, and the sequence of the affected region won’t be altered.
I find the findings believable, but it is a weakness that there are only four test assemblies.
Answer: Following reviewer advice, we tested BiSCoT on a fourth plant genome and a simulated assembly of a human chromosome.
" | Here is a paper. Please give your review comments after reading it. |
9,795 | "<ns0:abstract xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:p>In 2005, a chikungunya virus outbreak devastated the tropical island of Reunion, infecting a third of the total population. Motivated by the Reunion Island case study, we investigate the potential for two intervention measures under both voluntary and mandatory protocols to control a vector-borne disease when there is risk of the disease becoming endemic. The first measure uses insect repellent to prevent mosquito bites, while the second involves emigrating to the neighboring Mauritius Island to avoid infection. There is a threshold on the cost of using repellent above which both voluntary and mandatory regimes find it optimal to forgo usage. Below that threshold, mandatory usage protocols will eradicate the disease; however, voluntary adoption leaves the disease at a small endemic level.</ns0:p><ns0:p>Emigrating from the island to avoid infection results in a tragedy-of-the-commons effect: while being potentially beneficial to specific susceptible individuals, the remaining islanders paradoxically face a higher risk of infection. Mandated relocation of susceptible individuals away from the epidemic is viable only if the cost of this relocation is several magnitudes lower than the cost of infection. Since this assumption is unlikely to hold for chikungunya, it is optimal to discourage such emigration for the benefit of the entire population.</ns0:p></ns0:div>
</ns0:abstract>
<ns0:body xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:head>A</ns0:head><ns0:p>. In 2005, a chikungunya virus outbreak devastated the tropical island of Reunion, infecting a third of the total population. Motivated by the Reunion Island case study, we investigate the potential for two intervention measures under both voluntary and mandatory protocols to control a vector-borne disease when there is risk of the disease becoming endemic. The first measure uses insect repellent to prevent mosquito bites, while the second involves emigrating to the neighboring Mauritius Island to avoid infection. There is a threshold on the cost of using repellent above which both voluntary and mandatory regimes find it optimal to forgo usage. Below that threshold, mandatory usage protocols will eradicate the disease; however, voluntary adoption leaves the disease at a small endemic level. Emigrating from the island to avoid infection results in a tragedy-of-the-commons effect: while being potentially beneficial to specific susceptible individuals, the remaining islanders paradoxically face a higher risk of infection. Mandated relocation of susceptible individuals away from the epidemic is viable only if the cost of this relocation is several magnitudes lower than the cost of infection. Since this assumption is unlikely to hold for chikungunya, it is optimal to discourage such emigration for the benefit of the entire population. the incidence levels of this disease remain low, its potential to cause future outbreaks in these areas is cause for concern. In this paper, we investigate the viability of voluntary participation in personal protective measures (mosquito repellent and emigration) against diseases like chikungunya on Reunion Island by constructing a game-theoretic model in which individual strategic payoffs are compared against the average population payoff.</ns0:p><ns0:p>Chikungunya virus (CHIKV) is an Alphavirus in the Togaviridae family, similar to Dengue fever and Zika virus <ns0:ref type='bibr' target='#b20'>[20,</ns0:ref><ns0:ref type='bibr' target='#b25'>25,</ns0:ref><ns0:ref type='bibr' target='#b24'>24]</ns0:ref>. It is a vector-borne virus spread through bites by the females of Aedes aegypti and Aedes albopictus mosquitoes.</ns0:p><ns0:p>After a bite, there is a latency period for both humans and mosquitoes: it can take between 2 to 6 days for symptoms to develop and for an individual to become infectious <ns0:ref type='bibr' target='#b15'>[15]</ns0:ref>. The major symptoms associated with CHIKV are fever, rash, arthritis, headache, and nausea <ns0:ref type='bibr' target='#b15'>[15]</ns0:ref>. The defining characteristic of CHIKV is the persistence of arthritis for years after the initial infection <ns0:ref type='bibr' target='#b20'>[20]</ns0:ref>.</ns0:p><ns0:p>A small percentage of people infected with CHIKV, however, never develop symptoms of the disease <ns0:ref type='bibr' target='#b12'>[12]</ns0:ref>. Humans are no longer infectious about a week and a half after the initial infection, but may still be symptomatic. Recovered individuals acquire lifelong immunity from future infections <ns0:ref type='bibr' target='#b9'>[9]</ns0:ref>. There is no vaccine to prevent or medicine to treat chikungunya virus <ns0:ref type='bibr' target='#b9'>[9]</ns0:ref>. The most effective way to prevent infection from CHIKV is to prevent mosquito bites, for example, by using insect repellent <ns0:ref type='bibr' target='#b7'>[8]</ns0:ref>.</ns0:p><ns0:p>Chikungunya was first isolated in 1952-1953 in Tanzania <ns0:ref type='bibr' target='#b28'>[27]</ns0:ref>. The name translates to the native term for 'that which bends up' <ns0:ref type='bibr' target='#b31'>[30]</ns0:ref>. There were limited outbreaks between the initial discovery of the disease and a worldwide outbreak that occurred in 2004-2005 <ns0:ref type='bibr' target='#b12'>[12]</ns0:ref>. This outbreak started in Kenya and spread to the surrounding islands including Mauritius, Rodrigues, The Seychelles, Mayotte, Madagascar and Reunion Island <ns0:ref type='bibr' target='#b27'>[26]</ns0:ref>. From these islands, it spread to other regions of the world-chikungunya virus is now present on every continent except Antarctica-most likely carried by tourists. The disease impacted Reunion Island most severely: a third of the population became infected, unusually severe forms were present, and the first occurrences of maternal-neonatal transmission were documented <ns0:ref type='bibr' target='#b3'>[4]</ns0:ref>. This severity of impact may be attributed to an increase in travel between islands and the climate of the region at the time of the epidemic <ns0:ref type='bibr' target='#b3'>[4,</ns0:ref><ns0:ref type='bibr' target='#b36'>35]</ns0:ref>.</ns0:p><ns0:p>While the severity of the Reunion chikungunya outbreak may seem like an isolated event, vector-borne diseases such as malaria and dengue are becoming an increasingly prevalent public health issue in today's society. In the United States there has been a 23-fold increase of vector-borne disease cases OPTIMAL STRATEGIES TO CONTROL CHIKUNGUNYA OUTBREAK ON REUNION ISLAND 3 in the past ten years <ns0:ref type='bibr' target='#b29'>[28]</ns0:ref>. There are now 16 vector-borne diseases widely distributed in the United States, all of which are resistant to control, and only one of these (Yellow Fever) has an FDA-approved vaccine <ns0:ref type='bibr' target='#b29'>[28]</ns0:ref>. Even though there are limited cases of CHIKV in the United States and its territories, the disease is becoming more persistent: the number of national cases and distribution are increasing, and the range of the mosquitoes that transmit CHIKV has spread to 38 states as of 2016 <ns0:ref type='bibr' target='#b29'>[28]</ns0:ref>.</ns0:p><ns0:p>Due to the transmission patterns of mosquito-borne diseases and the lack of sufficient vector control to eradicate such diseases, individuals often have to rely on voluntary participation in personal protection measures. Unfortunately, individual self-interest in protection against an infectious disease does not necessarily correspond to the desired outcome for society <ns0:ref type='bibr' target='#b16'>[16]</ns0:ref>, namely eradication of the disease. The effect of potentially selfish human behavior on the spread of infectious diseases has only recently begun to receive attention, forming a new field of behavioral epidemiology; see <ns0:ref type='bibr' target='#b21'>[21]</ns0:ref> for a review of behavioral epidemiology.</ns0:p><ns0:p>Originally designed for the field of economics <ns0:ref type='bibr' target='#b39'>[38]</ns0:ref>, game theory has since been used to model many biological phenomena <ns0:ref type='bibr' target='#b23'>[23,</ns0:ref><ns0:ref type='bibr' target='#b18'>18,</ns0:ref><ns0:ref type='bibr' target='#b11'>11,</ns0:ref><ns0:ref type='bibr' target='#b38'>37,</ns0:ref><ns0:ref type='bibr' target='#b5'>6]</ns0:ref>, including individual-level vaccination decisions <ns0:ref type='bibr' target='#b1'>[2]</ns0:ref>. In a vaccination game, a selfish individual seeks to maximize its benefit, or rather to minimize the potential loss resulting from either employing a potentially costly protective measure or facing the consequences of the disease. As the likelihood of contracting the disease is dependent upon the behavior of others within the at-risk population, the resulting strategic interactions between individuals can be modeled using game theory. Game-theoretic frameworks have been adopted to studying optimal individual vaccination strategies for smallpox <ns0:ref type='bibr' target='#b2'>[3]</ns0:ref>, influenza <ns0:ref type='bibr' target='#b16'>[16,</ns0:ref><ns0:ref type='bibr' target='#b32'>31]</ns0:ref>, measles <ns0:ref type='bibr' target='#b33'>[32]</ns0:ref>, rubella <ns0:ref type='bibr' target='#b34'>[33]</ns0:ref>, toxoplasmosis <ns0:ref type='bibr' target='#b35'>[34]</ns0:ref>, Ebola <ns0:ref type='bibr' target='#b4'>[5]</ns0:ref>, cholera <ns0:ref type='bibr' target='#b19'>[19]</ns0:ref>, and meningitis <ns0:ref type='bibr' target='#b22'>[22]</ns0:ref>. It has also been applied to other personal protective measures such as insecticide-treated cattle <ns0:ref type='bibr' target='#b10'>[10]</ns0:ref>, mosquito repellent <ns0:ref type='bibr' target='#b13'>[13]</ns0:ref>, insecticide-treated bed nets <ns0:ref type='bibr' target='#b6'>[7]</ns0:ref>, clean water <ns0:ref type='bibr' target='#b19'>[19]</ns0:ref>, and clean injecting equipment <ns0:ref type='bibr' target='#b30'>[29]</ns0:ref>. For an extensive review of behavior-linked vaccination models, see <ns0:ref type='bibr' target='#b40'>[39]</ns0:ref>.</ns0:p><ns0:p>In this paper, we investigate the potential effects of voluntary and governmentmandated participation in utilizing the insect repellent as a protective measure against a disease such as chikungunya on Reunion Island. We also analyze the effect of emigration to a neighboring island (Mauritius) on the spread of chikungunya among the remaining population of Reunion Island. We find that the latter protocol has a paradoxically worsening of outcomes for the nonparticipating population.</ns0:p></ns0:div>
<ns0:div><ns0:head n='2.'>M</ns0:head><ns0:p>We adopt a version of the epidemiological model of the chikungunya outbreak on Reunion Island by Yakob and Clements <ns0:ref type='bibr' target='#b41'>[40]</ns0:ref> by adding population dynamics (birth and death demographic processes) for both humans and mosquitoes and strategically-linked parameters so that the disease potentially may establish itself endemically. This assumption then permits us to use the framework of Dorsett et al. <ns0:ref type='bibr' target='#b13'>[13]</ns0:ref>, Amaku <ns0:ref type='bibr' target='#b0'>[1]</ns0:ref>, and Bauch and Earn <ns0:ref type='bibr' target='#b1'>[2]</ns0:ref>. All human inhabitants of the island ( ) are divided into 5 compartments: individuals susceptible to chikungunya ( ); exposed individuals ( ), who had been bitten by an infected mosquito and acquired the disease; symptomatic infectious individuals ( ), who developed the symptoms of the disease and became infectious to biting mosquitoes; asymptomatic infectious individuals ( ), who became infectious but did not develop symptoms; and recovered individuals ( ), who recovered from chikungunya and acquired immunity. The mosquito population is divided into 3 compartments: susceptible mosquitoes ( ); exposed mosquitoes ( ), who bit an infected human and were exposed to the pathogen; and infectious mosquitoes ( ), who may infect humans by biting susceptible individuals.</ns0:p><ns0:p>We did not consider in this model the full life-cycle of mosquitoes, such as egg and larval stages, because we did not incorporate mosquito population control as one of the measures to fight chikungunya.</ns0:p><ns0:p>New individuals enter the susceptible part of the population at a rate Λ 1 due to birth or immigration; there is a natural per capita human mortality 1 .</ns0:p><ns0:p>Similarly, new mosquitoes are recruited into the susceptible compartment at a rate Λ 2 , and there is a natural per capita mosquito mortality 2 . We disregard the human disease-induced mortality because it is low, and doing so allows us to compute endemic equilibria analytically.</ns0:p><ns0:p>Susceptible humans who are bitten by infectious mosquitoes become exposed. The force of infection Infectious humans (both symptomatic and asymptomatic) recover at a rate and acquire immunity from future infections. The lifespan of a mosquito is too short to recover; an infectious mosquito remains as such until it dies.</ns0:p><ns0:p>Figure <ns0:ref type='figure' target='#fig_2'>1</ns0:ref> shows a diagram for the chikungunya transmission model on Reunion Island. The parameters of the epidemiological model are summarized in table 1. The table includes the baseline value of the mosquito-to-human transmission parameter, denoted by 0 1 . Later this parameter will be affected by an intervention measure (insect repellent), and hence it will become a function of the level of insect repellent usage, given by ( <ns0:ref type='formula' target='#formula_12'>7</ns0:ref>). The dynamics of the compartment model in figure <ns0:ref type='figure' target='#fig_2'>1</ns0:ref> is described by the following system of differential equations:</ns0:p><ns0:formula xml:id='formula_0'>d d = Λ 1 − 1 − 1 , d d = 1 − 1 − 1 , d d = 1 − − 1 , d d = (1 − ) 1 − − 1 , d d = ( + ) − 1 , d d = Λ 2 − 2 ( + ) − 2 , d d = 2 ( + ) − 2 − 2 ,<ns0:label>and</ns0:label></ns0:formula><ns0:formula xml:id='formula_1'>d d = 2 − 2 .</ns0:formula><ns0:p>(</ns0:p><ns0:formula xml:id='formula_2'>)<ns0:label>1</ns0:label></ns0:formula><ns0:p>The disease-free equilibrium (DFE) of this system is given by</ns0:p><ns0:formula xml:id='formula_3'>0 , 0 , 0 , 0 , 0 , 0 , 0 , 0 = Λ 1 1 , 0, 0, 0, 0, Λ 2 2 , 0, 0 .<ns0:label>(2)</ns0:label></ns0:formula><ns0:p>To compute the basic reproduction number 0 , we use the next-generation matrix approach <ns0:ref type='bibr' target='#b37'>[36]</ns0:ref>. To simplify this computation, we temporarily combined the symptomatic and asymptomatic infectious compartments into one infectious compartment + : individuals in both and compartments have identical contributions to the dynamics of chikungunya. We order the compartments that contribute to new infections as follows: , + , , and . Then the vector of the rates of appearance of new infections in these four compartments  and the vector of the rates of transfer of existing infections between these four compartments  are given by</ns0:p><ns0:formula xml:id='formula_4'> = ⎡ ⎢ ⎢ ⎢ ⎢ ⎣ 1 0 2 ( + ) 0 ⎤ ⎥ ⎥ ⎥ ⎥ ⎦ and  = ⎡ ⎢ ⎢ ⎢ ⎢ ⎣ 1 + 1 − 1 + ( + ) + 1 ( + ) 2 + 2 − 2 + 2 ⎤ ⎥ ⎥ ⎥ ⎥ ⎦ . (<ns0:label>3</ns0:label></ns0:formula><ns0:formula xml:id='formula_5'>)</ns0:formula><ns0:p>The matrices and are the Jacobians of  and  respectively, evaluated at DFE; they are given by</ns0:p><ns0:formula xml:id='formula_6'>= ⎡ ⎢ ⎢ ⎢ ⎢ ⎣ 0 0 0 1 0 0 0 0 0 Λ 2 2 1 Λ 1 2 0 0 0 0 0 0 ⎤ ⎥ ⎥ ⎥ ⎥ ⎦ and = ⎡ ⎢ ⎢ ⎢ ⎢ ⎣ 1 + 1 0 0 0 − 1 + 1 0 0 0 0 2 + 2 0 0 0 − 2 2 ⎤ ⎥ ⎥ ⎥ ⎥ ⎦ . (<ns0:label>4</ns0:label></ns0:formula><ns0:formula xml:id='formula_7'>)</ns0:formula><ns0:p>The basic reproduction number is the spectral radius of the matrix −1 ; it is given by</ns0:p><ns0:formula xml:id='formula_8'>0 = 1 2 Λ 2 1 2 1 1 2 Λ 1 ( + 1 )( 1 + 1 )( 2 + 2 ) .<ns0:label>(5)</ns0:label></ns0:formula><ns0:p>If 0 > 1, then the system converges to the endemic equilibrium (EE) given by * = Λ</ns0:p><ns0:formula xml:id='formula_9'>1 − ( 1 + 1 ) * 1 , * = Λ 1 Λ 2 1 2 1 1 2 − Λ 2 1 2 2 ( 1 + 1 )( 2 + 2 )( + 1 ) Λ 2 1 2 1 1 2 ( 1 + 1 ) + Λ 1 2 1 2 1 ( 1 + 1 )( 2 + 2 ) , * = 1 * + 1 , * = (1 − ) 1 * + 1 , * = 1 * 1 ( + 1 ) , * = Λ 1 Λ 2 ( + 1 ) 2 1 1 * + Λ 1 2 ( + 1 ) , * = Λ 2 2 1 1 * ( 2 + 2 )( 2 1 1 * + Λ 1 2 ( + 1 ))</ns0:formula><ns0:p>, and</ns0:p><ns0:formula xml:id='formula_10'>* = Λ 2 2 1 1 2 * 2 ( 2 + 2 )( 2 1 1 * + Λ 1 2 ( + 1 )) . (<ns0:label>6</ns0:label></ns0:formula><ns0:formula xml:id='formula_11'>)</ns0:formula><ns0:p>In the game-theoretic models constructed in the next section, we will be assuming that the system has reached an endemic equilibrium. In particular, we will use the values from ( <ns0:ref type='formula' target='#formula_10'>6</ns0:ref>) for relevant compartment sizes.</ns0:p></ns0:div>
<ns0:div><ns0:head n='3.'>R</ns0:head><ns0:p>We consider two intervention measures to fight the chikungunya outbreak on Reunion Island: (1) using insect repellent to prevent mosquito bites, and (2)</ns0:p><ns0:p>emigrating to the neighboring Mauritius Island.</ns0:p><ns0:p>3.1. Optimal levels of voluntary insect repellent usage. We adopt a modeling approach of <ns0:ref type='bibr' target='#b1'>[2,</ns0:ref><ns0:ref type='bibr' target='#b13'>13]</ns0:ref>. The strategy of an individual is the proportion of the day ∈ [0, 1] the individual is protected from mosquito bites; the protection is granted by insect repellent. We assume that the repellent provides complete protection from mosquito bites while it is active. Since mosquitoes cannot bites humans while they are protected by the insect repellent, the mosquitoto-human transmission coefficient 1 becomes a function of . If no protection is used ( = 0), then 1 (0) is at its base value 0 1 (given in table <ns0:ref type='table' target='#tab_1'>1</ns0:ref>). If humans are protected at all times ( = 1), then mosquitoes cannot bite these humans at all, and hence they cannot infect humans: 1 (1) = 0. We therefore assume that the mosquito-to-human transmission coefficient is a linear function of given by</ns0:p><ns0:formula xml:id='formula_12'>1 = 0 1 (1 − ).<ns0:label>(7)</ns0:label></ns0:formula><ns0:p>If all susceptible humans in the population adopt the same strategy pop , then the basic reproduction number becomes a function of pop by substituting the expression (7) for 1 into (5). The graph of the basic reproduction number as a function of the population strategy pop is shown in figure <ns0:ref type='figure' target='#fig_4'>2</ns0:ref>. The herd immunity protection level HI is the population protection level that reaches the threshold 0 = 1 for disease eradication.</ns0:p><ns0:p>We define the utility function (expected payoff) of a susceptible individual using strategy in a population that adopted strategy pop as</ns0:p><ns0:formula xml:id='formula_13'>( , pop ) = − ( , pop ) − ,<ns0:label>(8)</ns0:label></ns0:formula><ns0:p>where is the cost of infection, is the cost of complete protection through insect repellent, and ( , pop ) is the probability of infection. The latter depends on the individual's strategy because it determines how often mosquitoes may bite the individual, and on the population strategy pop because it affects the prevalence of the disease (e.g., the number of infected mosquitoes). The outcome of a game does not change if the utility function is scaled, so we divide the right-hand side of (8) by to obtain where = / is the cost of complete protection relative to the cost of infection.</ns0:p><ns0:formula xml:id='formula_14'>( , pop ) = − ( , pop ) − ,<ns0:label>(9)</ns0:label></ns0:formula><ns0:p>We next compute the probability of getting infected and becoming symptomatic as the transition probability from the susceptible compartment to the symptomatic infectious compartment . This probability is the product of the probability that a susceptible individual becomes exposed 1 /( 1 + 1 ) multiplied by the probability that an exposed individual becomes symptomatically</ns0:p><ns0:formula xml:id='formula_15'>infected 1 /( 1 + 1 + (1 − ) 1 ) = 1 /( 1 + 1 ): ( , pop ) = 1 ( , pop ) 1 + 1 ( , pop ) 1 1 + 1 ,<ns0:label>(10)</ns0:label></ns0:formula><ns0:p>where</ns0:p><ns0:formula xml:id='formula_16'>1 ( , pop ) = 0 1 (1 − ) * *<ns0:label>(11)</ns0:label></ns0:formula><ns0:p>is the force of infection, which depends on the individual protection level and on the population protection level pop . The individual protection level determines the rate at which mosquitoes bite the individual 0 1 (1 − ). The population protection level pop affects the prevalence of the disease in the population; it determines the size of the compartment * via the substitution of the expression 0 1 (1 − pop ) for 1 into (6). In particular, * and * do not depend on the individual protection level .</ns0:p><ns0:p>To find the Nash equilibrium population protection level, we attempt to maximize the utility function (9) of a focal individual. Observe that</ns0:p><ns0:formula xml:id='formula_17'>1 ( , pop ) = (1 − ) 1 (0, pop ),<ns0:label>(12)</ns0:label></ns0:formula><ns0:p>and hence</ns0:p><ns0:formula xml:id='formula_18'>1 ( , pop ) = − 1 (0, pop ). (<ns0:label>13</ns0:label></ns0:formula><ns0:formula xml:id='formula_19'>)</ns0:formula><ns0:p>It follows that 2 2 ( , pop ) = 2 1 1 (0, pop ) 2 ( 1 + 1 ( , pop )) 3 > 0.</ns0:p><ns0:p>(</ns0:p><ns0:p>Consequently, the utility function is a convex function of , and thus it attains its maximum value at one of the endpoints: = 0 or = 1.</ns0:p><ns0:p>This conclusion can be interpreted as follows. If the population repellent usage is sufficiently high, then the probability of getting infected is very low.</ns0:p><ns0:p>A focal individual would rather bypass the potentially costly preventive measure and face the low morbidity risk instead. Hence individuals may improve their payoff by deviating from the population strategy (they should stop using repellent). On the other hand, if the population repellent usage is low, then the probability of getting infected is too high, and a focal individual should prefer to pay the cost of complete protection rather than face the high morbidity risk.</ns0:p><ns0:p>In this case, individuals may also improve their payoff by deviating from the population strategy (they should use repellent 100% of the time).</ns0:p><ns0:p>So, if the population repellent usage is too high, then individuals would do better if they stop using repellent, and hence the population repellent usage will decrease. Conversely, if the population repellent usage is too low, then individuals would do better if they start using repellent 100% of the time, and hence the population repellent usage will increase. If the population repellent usage is 'just right' (Nash equilibrium), then individuals cannot improve their payoffs by using repellent either less frequently or more frequently. We note that there is a presumption of the population-wide adoption of treatment rates in our model that is common to ESS-modeling; however, in situations such as <ns0:ref type='bibr' target='#b14'>(14)</ns0:ref>, there is a potential implication that the population should in fact separate into distinct groups with different adoption rates. As it goes beyond the framework discussed here, we leave that investigation for future research.</ns0:p><ns0:p>The Nash equilibrium protection level of the population NE is thus a solution to the equation</ns0:p><ns0:formula xml:id='formula_21'>(0, NE ) = (1, NE )<ns0:label>(15)</ns0:label></ns0:formula><ns0:p>or</ns0:p><ns0:formula xml:id='formula_22'>1 (0, NE ) 1 + 1 (0, NE ) 1 1 + 1 = . (<ns0:label>16</ns0:label></ns0:formula><ns0:formula xml:id='formula_23'>)</ns0:formula><ns0:p>The graph of the optimal (Nash equilibrium) repellent usage as a function of the relative cost of protection is shown in figure <ns0:ref type='figure'>3a</ns0:ref>. The optimal repellent usage NE reaches the herd immunity HI level only when the cost of the protective measure relative to the cost of chikungunya infection is negligible (i.e., zero mathematically). The optimal repellent usage remains very close to the herd immunity level for a range of values of the relative cost , and then drops off sharply. Once the relative cost of protection becomes too large ( max ), then everyone stops using insect repellent because its high cost forces individuals to prefer to risk the cost of infection.</ns0:p></ns0:div>
<ns0:div><ns0:head>Figure 3. (a)</ns0:head><ns0:p>The graph of the optimal level of population repellent usage NE as a function of the relative cost of protection. The optimal repellent usage reaches the herd immunity level only when = 0. Everyone stops using repellent if its relative cost is too high: larger than the threshold value max . (b) The graph of the basic reproduction number computed at the optimal population repellent usage level NE as a function of the relative cost of protection. When = 0, the optimal protection level is equal to the herd immunity threshold, so 0 = 1. When the relative cost of protection exceeds the threshold value max , the population stops using repellent, and the basic reproduction number reaches its maximum value.</ns0:p></ns0:div>
<ns0:div><ns0:head>3.2.</ns0:head><ns0:p>Optimal levels of mandatory insect repellent usage. We now consider a scenario where an organization (e.g., the government) enforces the use of insect repellent in the population to fight chikungunya. The organization must balance the cost of prevention of the disease in the population and the cost of treatment of symptomatically infected individuals. On the one hand, every individual who utilizes insect repellent 100% of the time results in a cost (same as the cost of voluntary complete protection). On the other hand, every symptomatically infected individual results in a cost .</ns0:p><ns0:p>The goal of the mandating organization is to find the repellent usage level for the population pop ∈ [0, 1] so that the expected payoff (negative of the total cost) the susceptible population with</ns0:p><ns0:formula xml:id='formula_24'>( pop ) = − * − pop *<ns0:label>(</ns0:label></ns0:formula><ns0:formula xml:id='formula_25'>d d = Λ 1 − 1 − 1 − . (<ns0:label>19</ns0:label></ns0:formula><ns0:formula xml:id='formula_26'>)</ns0:formula><ns0:p>The DFE of the modified system is given by ( 0 , 0 , 0 , 0 , 0 , 0 , 0 , 0 ) = Λ 1 1 + , 0, 0, 0, 0, Λ 2 2 , 0, 0 ,</ns0:p><ns0:p>and the corresponding basic reproduction number of the disease is</ns0:p><ns0:formula xml:id='formula_28'>0 = 1 2 Λ 2 1 2 1 2 ( 1 + ) Λ 1 ( + 1 )( 1 + 1 )( 2 + 2 ) . (<ns0:label>21</ns0:label></ns0:formula><ns0:formula xml:id='formula_29'>)</ns0:formula><ns0:p>The graph of the basic reproduction number as a function of the emigration rate is shown in figure <ns0:ref type='figure' target='#fig_7'>5</ns0:ref>. It is an increasing function of , and hence emigrating susceptible individuals paradoxically make it worse for the remaining susceptible population. When the fresh blood supply is reduced due to emigration, susceptible mosquitoes are more likely to prey upon infectious humans, increasing the disease prevalence in the vector population. Consequently, the remaining susceptible human population is at an increased risk of contracting the disease from a mosquito bite. It follows that, while being potentially beneficial to specific individuals, voluntary emigration may result in a tragedyof-the-commons effect for the remaining islanders. To further investigate the effect of voluntary emigration on the chikungunya epidemic and whether (selfish) susceptible individuals should emigrate, we compute the EE values of all compartments in the model with emigration:</ns0:p><ns0:formula xml:id='formula_30'>* = Λ 1 1 + ( 1 + 1 ) * 1 ( 1 + ) , * = Λ 1 − ( 1 + 1 ) * 1 + , * = 1 * + 1 , * = (1 − ) 1 * + 1 , * = 1 * 1 ( + 1 ) , * = Λ 2 ( + 1 )(Λ 1 1 + ( 1 + 1 ) * + 2 ( + 1 )(Λ 1 1 + ( 1 + 1 ) * ) , * = Λ 2 2 1 1 ( 1 + ) * ( 2 + 2 )[ + Λ 1 1 2 ( + 1 ) + 2 ( 1 + 1 )( + 1 ) * ]</ns0:formula><ns0:p>, and</ns0:p><ns0:formula xml:id='formula_31'>* = Λ 2 2 1 1 2 ( 1 + ) * 2 ( 2 + 2 )[ + Λ 1 1 2 ( + 1 ) + 2 ( 1 + 1 )( + 1 ) * ] ,<ns0:label>(22)</ns0:label></ns0:formula><ns0:p>where</ns0:p><ns0:formula xml:id='formula_32'>= 2 1 1 ( 1 + ) * ,<ns0:label>(23)</ns0:label></ns0:formula><ns0:p>and * is the solution to the quadratic equation</ns0:p><ns0:formula xml:id='formula_33'>2 + + = 0<ns0:label>(24)</ns0:label></ns0:formula><ns0:p>with coefficients</ns0:p><ns0:formula xml:id='formula_34'>= − 2 ( 1 + 1 ) 2 ( 2 + 2 )[ 2 1 1 ( 1 + ) + 2 ( 1 + 1 )( + 1 )], = −Λ 2 1 2 2 1 1 2 ( 1 + 1 )( 1 + ) − 2Λ 1 1 2 2 ( 1 + 1 ) 2 ( 2 + 2 )( + 1 ) − Λ 1 2 2 1 2 1 ( 1 + 1 )( 2 + 2 )( 1 + ), and = Λ 1 Λ 2 1 2 2 1 1 2 ( 2 + ) − Λ 2 1 2 1 2 2 ( 1 + 1 )( 2 + 2 )( + 1 ).<ns0:label>(25)</ns0:label></ns0:formula><ns0:p>The biologically meaningful root of this equation is given by *</ns0:p><ns0:formula xml:id='formula_35'>= − − √ 2 − 4 2 . (<ns0:label>26</ns0:label></ns0:formula><ns0:formula xml:id='formula_36'>)</ns0:formula><ns0:p>Figure <ns0:ref type='figure' target='#fig_8'>6</ns0:ref> shows the graphs of the number and proportion of symptomati- We next consider a game-theoretic model of individual migration decisions.</ns0:p><ns0:p>Suppose that the population adopted the emigration rate pop . A focal susceptible individual is presented with a choice to either migrate or not migrate. Each of the two strategic choices carries a corresponding payoff: m for migrate and nm for not migrate, given by m ( pop ) = − − pop , and</ns0:p><ns0:formula xml:id='formula_37'>nm ( pop ) = − ( pop ) ,<ns0:label>(28)</ns0:label></ns0:formula><ns0:p>where is the base (fixed) cost of migration, is the scaling cost of migration, is the cost of the (symptomatic) chikungunya infection, and ( pop ) is the probability of getting infected given the population emigration rate pop .</ns0:p><ns0:p>We assume that the cost of emigration is an increasing function of the migration rate because of the limited immigration potential of Mauritius: the more individuals migrate to Mauritius, the harder it becomes to find housing and jobs. For simplicity, we model the increasing emigration cost as a linear function of the migration rate. The probability of getting infected and incurring the cost of a symptomatic chikungunya infection if remaining on Reunion Island is the transition probability from the susceptible class to the symptomatically infectious class : </ns0:p><ns0:formula xml:id='formula_38'>( pop ) = 1 ( pop ) 1 + 1 ( pop ) 1 1 + 1 .<ns0:label>(29)</ns0:label></ns0:formula><ns0:p>where ̃ and ̃ are relative base and scaling costs of emigration, respectively.</ns0:p><ns0:p>A susceptible individual should emigrate when the relative cost of doing so is less than the probability of getting infected: ̃ + pop ̃ < ( pop ), and the individual should remain on the island otherwise. The regions in the ( ̃ , )parameter space corresponding to the best choice for a focal individual for several values of ̃ are shown in figure <ns0:ref type='figure' target='#fig_12'>8</ns0:ref>. If the scaling cost of emigration is negligible (i.e., the cost of emigration does not depend on the number of emigrating individuals), then the best strategy of a susceptible individual is to emigrate as long as the relative base cost of emigration ̃ is sufficiently small (figure <ns0:ref type='figure' target='#fig_12'>8a</ns0:ref>). On the other hand, as the relative scaling cost of emigration ̃ grows, the individual's decision to emigrate starts to depend on the emigration decisions of other individuals (figure <ns0:ref type='figure' target='#fig_12'>8b-c</ns0:ref>), until it becomes unprofitable to emigrate regardless of the relative base cost of emigration if the emigration rate is too high (figure <ns0:ref type='figure' target='#fig_12'>8d</ns0:ref>). Manuscript to be reviewed number of emigrating susceptible individuals, we consider the difference between the total population size at equilibrium without emigration ( * = Λ 1 / 1 ) and the total population size at equilibrium given the population migration rate (this expression is given in the first equation of ( <ns0:ref type='formula' target='#formula_31'>22</ns0:ref>)); we denote this difference by * .</ns0:p><ns0:p>The payoff of the emigration policy with migration rate is given by</ns0:p><ns0:formula xml:id='formula_40'>( ) = − * − ̃ * ,<ns0:label>(31)</ns0:label></ns0:formula><ns0:p>where ̃ = / is the cost of migration relative to the cost of infection. The graphs of this function for several values of ̃ are shown in figure <ns0:ref type='figure' target='#fig_13'>9</ns0:ref>. There are three qualitatively different outcomes:</ns0:p><ns0:p>1. For very low relative migration cost ( ̃ ≤ 0.00022), higher migration rates result in smallest overall costs; however, the near-optimal costs are quickly achieved by small values of migration rate ( = 0.01)-see figure 9a.</ns0:p><ns0:p>2. There is a small interval of the relative migration cost values (0.00023 ≤ ̃ ≤ 0.00025) where the optimal cost is achieved in the interior for very small values of the migration rate ( < 0.002)-see figures 9b and 9c.</ns0:p><ns0:p>3. For all sufficiently large values of the relative migration cost ( ̃ ≥ 0.00026), it is best to not allow individuals to emigrate from the islandsee figure <ns0:ref type='figure' target='#fig_13'>9d</ns0:ref>.</ns0:p><ns0:p>In practice, however, the cost of emigration (such as relocation from Reunion to Mauritius) is usually comparable to or higher than the cost of the symptomatic chikungunya infection. Therefore, the scenario shown in figure <ns0:ref type='figure' target='#fig_13'>9d</ns0:ref> is the most realistic one: it is best to not allow susceptible individuals to leave the island during the outbreak. The mandated repellent usage protocol resulted in the same outcome as the voluntary (i.e., selfishly rational) compliance scenario if the cost of the preventive measure relative to the cost of the disease was too high: it was best to bypass the repellent usage altogether. But if the relative cost of protection was sufficiently low, so that repellent usage was warranted, then the two scenarios effected different outcomes. In the voluntary compliance case, the population repellent usage fell short-albeit not by much-of the herd immunity threshold. In the mandated protocol case, reaching the herd immunity usage level and thus eradicating the disease was most effective.</ns0:p><ns0:p>That voluntary adoption of preventative measures against an infectious disease falls short of the herd immunity threshold has also been observed in other studies <ns0:ref type='bibr' target='#b17'>[17,</ns0:ref><ns0:ref type='bibr' target='#b1'>2,</ns0:ref><ns0:ref type='bibr' target='#b13'>13,</ns0:ref><ns0:ref type='bibr' target='#b4'>5,</ns0:ref><ns0:ref type='bibr' target='#b19'>19]</ns0:ref>. Yet looking at a mandated repellent usage scenario revealed that a mandatory protocol might have eliminated the epidemic if the relative cost of the preventive measure was sufficiently low.</ns0:p><ns0:p>Mandatory emigration from Reunion Island demonstrated that this preventive measure made sense for the public benefit only when the cost of relocation was significantly lower than the public cost of infection. Since this mathematical assumption is not likely to hold in practice, the model predicted that it The qualitative differences in optimal behavior under the two alternative treatment protocols invite further examination of our model's behavior and assumptions. Both evacuation/emigration of the human populace and the use of repellent reduce the pool of potential blood hosts for the mosquito population; however, they produce contrasting effects on the force of infection.</ns0:p><ns0:p>A base assumption in the model is that each insect has a consistent average number of encounters with humans over a given time span. Repellent usage directly decreases the force of infection by deterring biting upon encounterit is this feature of 'wasted' encounters that permits the development of herd immunity. In contrast, reduction in the size of the standing human population elevates the force of infection by increasing the number of encounters an individual human experiences. Secondarily, this results in increased prevalence of the disease in the vector-population as their blood hosts are more likely to be infected. We hypothesize that distinct protocol results depend upon the presence of (1) a distinct vector population; <ns0:ref type='bibr' target='#b1'>(2)</ns0:ref> an assumption of constant predation encounters for vectors; (3) the proportional allocation of encounters across humans; (4) an inability of vectors to pre-judge encounters and thereby shift towards more palatable hosts; and (5) a secondary food source to support constant recruitment of new vectors. We propose a followup study to this paper that focuses specifically on the dynamic analysis of the force of infection as these assumptions are introduced or removed.</ns0:p></ns0:div>
<ns0:div><ns0:head n='5.'>C</ns0:head><ns0:p>There are several additional directions in which our model can be improved. </ns0:p></ns0:div><ns0:figure xml:id='fig_0'><ns0:head /><ns0:label /><ns0:figDesc>OPTIMAL VOLUNTARY AND MANDATORY INSECT REPELLENT USAGE AND EMIGRATION STRATEGIES TO CONTROL THE CHIKUNGUNYA OUTBREAK ON REUNION ISLAND SYLVIA R. M. KLEIN 1 , ALEX O. FOSTER 2 , DAVID A. FEAGINS 3 , JONATHAN T. ROWELL 4 , AND IGOR V. EROVENKO ⋆4</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_1'><ns0:head /><ns0:label /><ns0:figDesc>Reunion Island is a tropical island located in the Indian Ocean 500 miles east of Madagascar and approximately 150 miles southwest of Mauritius. The island was devastated by a major chikungunya outbreak in 2005-2006, when approximately 266 thousand of the 785 thousand inhabitants were infected,</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_2'><ns0:head>Figure 1 .</ns0:head><ns0:label>1</ns0:label><ns0:figDesc>Figure 1. The compartment model of chikungunya virus transmission on Reunion Island. The human population is divided into five compartments: susceptible ( ), exposed ( ), symptomatic infectious ( ), asymptomatic infectious ( ), and recovered ( ). The mosquito population is divided into three compartments: susceptible ( ), exposed ( ), and infectious ( ). The forces of infection on human and mosquito populations, 1 and 2 , respectively, are population frequency-dependent functions of the state variables.</ns0:figDesc><ns0:graphic coords='6,135.90,219.69,340.20,225.06' type='bitmap' /></ns0:figure>
<ns0:figure xml:id='fig_3'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2019:08:40674:1:3:NEW 15 Mar 2020) Manuscript to be reviewed OPTIMAL STRATEGIES TO CONTROL CHIKUNGUNYA OUTBREAK ON REUNION ISLAND 7</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_4'><ns0:head>Figure 2 .</ns0:head><ns0:label>2</ns0:label><ns0:figDesc>Figure2. The graph of the basic reproduction number as a function of the population protection level pop . The basic reproduction number is at its maximum value-given by (5)-when no insect repellent is used by susceptible individuals ( = 0), and it becomes zero if the population employs complete protection from mosquito bites ( = 1). The threshold for disease eradication ( 0 = 1) is achieved at the herd immunity protection level HI .</ns0:figDesc><ns0:graphic coords='10,228.60,72.00,154.79,136.31' type='bitmap' /></ns0:figure>
<ns0:figure xml:id='fig_5'><ns0:head /><ns0:label /><ns0:figDesc><ns0:ref type='bibr' target='#b17'>17)</ns0:ref> is maximal. (Note that the equilibrium values * and * depend on pop .) Here we analyze the case where the mandating organization addresses mosquito-tohuman transmission by advising susceptible individuals to spray themselves with insect repellent. One may also consider an alternative scenario where the infectious individuals are using insect repellent to reduce the human-tomosquito transmission.As before, we scale the payoff function and consider( pop ) = − * − pop * ,(18)where = / is the relative cost of protection. The graphs of this function for different values of are shown in figure 4. There are two possible outcomes: (1) the susceptible individuals should adopt the repellent usage level equal to that of the herd immunity threshold HI , leading to the eradication of the disease; or (2) no insect repellent should be used, and it is more costeffective to treat symptomatically infected individuals only. The first outcome occurs for sufficiently low values of (less than 0.00024), and the second outcome occurs for greater values of (greater than 0.00024); the threshold value of separating the two outcomes was found numerically.</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_6'><ns0:head>Figure 4 . 3 . 3 .</ns0:head><ns0:label>433</ns0:label><ns0:figDesc>Figure 4. The graphs of the expected payoff function given by the equation (18). (a) = 0.0001, (b) = 0.01. The graphs show two qualitatively different outcomes. If < 0.00024 then mandating repellent usage necessary to reach the herd immunity threshold is most effective. If > 0.00024 then no insect repellent should be used, and all efforts should be devoted to treating infected individuals.</ns0:figDesc><ns0:graphic coords='13,126.00,290.77,360.01,144.81' type='bitmap' /></ns0:figure>
<ns0:figure xml:id='fig_7'><ns0:head>Figure 5 .</ns0:head><ns0:label>5</ns0:label><ns0:figDesc>Figure 5. The graph of 0 as a function of . If susceptible individuals emigrate from Reunion Island, then the remaining inhabitants face an increased spread of the disease.</ns0:figDesc><ns0:graphic coords='14,228.60,427.66,154.80,133.89' type='bitmap' /></ns0:figure>
<ns0:figure xml:id='fig_8'><ns0:head>Figure 6 .</ns0:head><ns0:label>6</ns0:label><ns0:figDesc>Figure 6. The graphs of the (a) number and (b) proportion of symptomatically infectious individuals in the population as functions of the emigration rate . Increased migration levels result in fewer infectious individuals overall but a greater proportion of infectious individuals in the population.</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_9'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2019:08:40674:1:3:NEW 15 Mar 2020) Manuscript to be reviewed This probability is an increasing function of the emigration rate because each of the remaining susceptible individuals faces a higher risk of infection (cf. figure 5); the graph is shown in figure 7.</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_10'><ns0:head>Figure 7 .</ns0:head><ns0:label>7</ns0:label><ns0:figDesc>Figure 7. The probability of symptomatic chikungunya infection for a susceptible individual on Reunion Island is increasing with the emigration rate .</ns0:figDesc><ns0:graphic coords='17,228.60,128.83,154.80,129.54' type='bitmap' /></ns0:figure>
<ns0:figure xml:id='fig_11'><ns0:head>3. 4 .</ns0:head><ns0:label>4</ns0:label><ns0:figDesc>Optimal levels of mandatory emigration. Finally, we consider the potential impacts on the chikungunya epidemic on Reunion Island of coordinated emigration efforts. A mandating organization attempts to minimize overall costs, which are comprised of the cost of treatment of symptomatically infected individuals and the relocation costs of emigrating individuals. To estimate the PeerJ reviewing PDF | (2019:08:40674:1:3:NEW 15 Mar 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_12'><ns0:head>Figure 8 .</ns0:head><ns0:label>8</ns0:label><ns0:figDesc>Figure 8. The regions in the ( ̃ , )-parameter space showing whether a focal susceptible individual should emigrate to Mauritius or remain on Reunion. Color code: white-emigrate, gray-stay. (a) ̃ = 0, (b) ̃ = 0.1, (c) ̃ = 1, (d) ̃ = 10.</ns0:figDesc><ns0:graphic coords='18,126.00,72.00,360.00,285.39' type='bitmap' /></ns0:figure>
<ns0:figure xml:id='fig_13'><ns0:head>Figure 9 .</ns0:head><ns0:label>9</ns0:label><ns0:figDesc>Figure 9. The overall cost of the mandated emigration policy as a function of the migration rate . (a) ̃ = 0.00022, (b) ̃ = 0.00023, (c) ̃ = 0.00024, (d) ̃ = 0.00026.</ns0:figDesc><ns0:graphic coords='19,126.00,249.47,360.00,282.58' type='bitmap' /></ns0:figure>
<ns0:figure xml:id='fig_14'><ns0:head /><ns0:label /><ns0:figDesc>We investigated potential implications of both voluntary and mandatory intervention measures to fight the chikungunya outbreak on Reunion Island.Susceptible individuals may either prevent the infection by using insect repellent and hence reduce the frequency of mosquito bites, or leave ReunionPeerJ reviewing PDF | (2019:08:40674:1:3:NEW 15 Mar 2020) Manuscript to be reviewed OPTIMAL STRATEGIES TO CONTROL CHIKUNGUNYA OUTBREAK ON REUNION ISLAND 19Island and emigrate to neighboring Mauritius. We adopted a version of a previous epidemiological model of the chikungunya transmission on Reunion Island<ns0:ref type='bibr' target='#b41'>[40]</ns0:ref>. The epidemiological model informed the payoff functions in the game-theoretic models of individual and centralized decisions on the level of adoption of the protective measures. We found that the two protocols resulted in qualitatively different predictions concerning optimal allocations, with the latter measure creating an additional hazard for non-participants.Voluntary participation in the two intervention measures produced opposite population-level effects. The more susceptible individuals spray themselves with insect repellent, the less likely the infectious mosquitoes generate new human infections before they die. Consequently, higher adoption levels of insect repellent usage in the population resulted in lower basic reproduction number values for the disease. Individuals using repellent provide (near) herd-immunity-effect benefits to the entire population. In contrast, if susceptible individuals vacated the island, then susceptible mosquitoes were more likely to bite infectious humans as a percentage of the remaining population, thus increasing the disease prevalence among mosquitoes. The remaining susceptible individuals subsequently faced an increased risk of contracting the infection from a mosquito bite. Increased migration levels resulted in drastically elevated basic reproduction number values. Thus, the impact of voluntary emigration is similar to the tragedy-of-the-commons effect: while being potentially beneficial to specific individuals, it hurts the remaining islanders.</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_15'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2019:08:40674:1:3:NEW 15 Mar 2020) was best to avoid migration of susceptible individuals from the island. The potentially high cost of relocating susceptible individuals away from the epidemic was not compensated by the minimal decrease in the number of infected individuals.</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_16'><ns0:head>First, we areA</ns0:head><ns0:label /><ns0:figDesc>assuming that individuals possess complete information regarding the prevalence of the disease and the costs of protection relative to infection. But individuals rarely have access to the exact disease prevalence data, and hence they may only guess the relevant numbers. Second, the cost of intervention (such as using insect repellent) and the cost of the disease must be estimated individually. These costs include both direct costs such as paying for repellent or medical treatment, and indirect costs such as potentially harmful side effects of the chemicals in repellent or morbidity risks of the infection.Additionally, different individuals may have various opinions about the risk of using repellent or getting infected with chikungunya virus. Building these uncertainties into the model should allow a broader outlook at different strategies to combat such outbreaks. PeerJ reviewing PDF | (2019:08:40674:1:3:NEW 15 Mar 2020) Manuscript to be reviewed OPTIMAL STRATEGIES TO CONTROL CHIKUNGUNYA OUTBREAK ON REUNION ISLAND 21 Moreover, our model assumes that the population has reached an equilibrium with respect to the disease dynamics. But reaching this equilibrium usually occurs on a different timescale compared to individual preventive actions.For example, individuals could be more likely to participate in preventive efforts when the epidemic is at its peak rather then when the disease reached the endemic state. A dynamic model where susceptible individuals inform their preventive decisions on the current state of the prevalence of the disease which, in turn, affects the dynamics of the disease transmission, should present a more realistic analysis of selfish individual decisions to prevent the infection.This research was conducted as part of a Research Experiences for Undergraduates program at the University of North Carolina at Greensboro in summer 2018, which was funded by the NSF grant DMS-1659646. SRMK, AOF, and DAF were undergraduate student participants, and JTR and IVE were faculty mentors. We thank J. Safley and P. Waiker, who served as graduate assistants during the REU program where this research was conducted.C I</ns0:figDesc></ns0:figure>
<ns0:figure type='table' xml:id='tab_1'><ns0:head>Table 1 .</ns0:head><ns0:label>1</ns0:label><ns0:figDesc>Summary of the model parameters</ns0:figDesc><ns0:table><ns0:row><ns0:cell cols='2'>Symbol Meaning</ns0:cell><ns0:cell>Value</ns0:cell><ns0:cell>Source</ns0:cell></ns0:row><ns0:row><ns0:cell>0 1 2 Λ 1 Λ 2 1 2 1 2</ns0:cell><ns0:cell cols='3'>Mosquito-to-human transmission Human-to-mosquito transmission Human recovery rate Human birth rate Mosquito birth rate Rate of humans becoming infectious Rate of mosquitoes becoming infectious Human natural death rate Mosquito natural death rate Proportion of hosts that develop symptoms 0.97 0.37 0.37 0.14 3.58 4.76 × 10 3 Assumed [14] [14] [40] Assumed 0.5 [40] 0.5 [40] 3.58 × 10 −5 Assumed 0.05 [40] [40]</ns0:cell></ns0:row></ns0:table></ns0:figure>
<ns0:note place='foot'>PeerJ reviewing PDF | (2019:08:40674:1:3:NEW 15 Mar 2020)</ns0:note>
</ns0:body>
" | "First of all, we would like to thank the reviewers for taking the time to carefully read the
paper and provide constructive comments and suggestions for improvement.
Below we will address the points raised by the reviewers in detail. Our answers are
highlighted in blue.
Reviewer 1
Basic reporting
The article is professionally written in good English.
While the references are adequate, the organization of the literature review is a bit
confusing. [21] is a book that outlines the state of the art of the field of behavioral
epidemiology at the time it was written; this should be made more explicit in line 66 of
page 3, which right now sounds a bit like if [21] had started it all. Line 83 at the same page
mentions [39] in the context of statistical physics, but this is just part of a flashy tile of this
article. In reality, this is really an extensive and exhaustive survey of the parts (777
references in all) of behavioral epidemiology that most closely relate to vaccination
games.
We adjusted the introduction to explicitly mention that [21] is a review of behavioral
epidemiology. We also adjusted our introduction of the reference [39]. We also made
several minor improvements of the exposition in the introduction.
It is not clear whether Section 2 contains any novelty. It should be made explicit whether
and to what extent it goes beyond a condensed review of [40]. This should be made
explicit.
The first sentence of Section 2 has been rewritten to more directly draw attention to the
modeling novelty of our paper, namely the inclusion demographic processes and
incorporating model parameters that express the strategic make-up of the population.
Also, the results of subsections 3.1 and and 3.2 should be juxtaposed more clearly,
perhaps by combining Figures 3 and 4 into one figure, so that it becomes visually obvious
that the Nash equilibrium fall short of the societally optimal vaccination coverage.
The subsection 3.1 includes a game-theoretic model, which results from strategic
interactions within at-risk population, while the subsection 3.2 includes an optimization
model, which results from the government-level decision on the optimal plan of action.
Figure 3a clearly shows that the Nash equilibrium solution to the game falls short of the
eradication threshold (the societally optimal outcome) as long as the cost of the protetive
measure is not negligible. There is no Nash equilibrium in the model of the subsection 3.2
because there is no game there. Instead, two optimal outcomes are possible: (1) use
enough repellent to eradicate the disease; or (2) do not use any repellent at all. These
scenarios are separated by a threshold value of the relative protection cost, and they are
shown in figures 4a and 4b, respectively.
Experimental design
The authors study meaningful questions and their methodology for analyzing the model
appear to be sound and professionally executed.
However, some important information is missing, for example, it is not being explicitly said
whether the graphs in Figure 4 rely on a numerical solver for finding the relevant
equilibria, as I believe they would need to.
We added a clarification at the end of subsection 3.2: “the threshold value of $C$
separating the two outcomes was found numerically.”
Validity of the findings
While my comments pertaining to points 1 and 2 can be addressed by minor modification,
here I have a major issue:
Why is it that the two interventions considered by the authors lead to two qualitatively
different predictions? The authors appear to have completely overlooked this question,
but it is an absolutely crucial one.
Let's take the point of view of a mosquito: In both cases it is faced with removal of
potential blood meals. Either because the human resource has become unapproachable
due to repellent, or has been removed to another location. In both cases the mosquito can
either continue its search for another human provider of a blood meal until it succeeds, or
content itself without a meal of human blood. These scenarios would translate differently
into differential equations. The authors have essentially shown (among other things and
under certain assumptions about the costs) that in the second scenario the societally
optimal strategy is to remove enough humans (by using insect repellent) to achieve herd
immunity, while in the first scenario the cost of removing more humans will not lower the
overall cost to society, as a higher proportion of the remaining susceptibles will eventually
experience infection. I find this quite interesting and it should be the major result of this
paper.
But this issue is not addressed at all. Instead, the authors set up their differential
equations in such a way that insecticide repellent would apparently discourage
mosquitoes from searching better-smelling humans that they might feed on [this is
essentially what (12) translates into], while having fewer humans around overall would
still make them search until they succeed [translate (19) and the first line of (22)]. I find
this highly implausible. Insect repellent does not kill or disable mosquitoes, it only drives
them away from particular humans.
In consideration of this comment and the reviewer’s final comment below, we have added
additional material to our introduction (final paragraph) and throughout the discussion to
address the contrasting model outcomes as an interpretation of the effect of repellent and
migration on mosquito feeding behavior. The repellent is not killing or disabling
mosquitoes, but rather it is rendering encounters non-effective - the mosquito must still
expend time during the pre-feeding encounter before the repellent is detected. The
primary reason for the divergence in results is in how the two treatments affect the force
of infection. We now outline a set of hypotheses about specific model features that would
support this divergence and propose a follow-up theoretical study to examine the
questions posed by the reviewer.
Therefore I cannot recommend publication of the paper in its current form.
Comments for the Author
As you saw from point 3, I find some assumptions of your model implausible and did not
recommend publication of this paper.
However, I believe that you have discovered, perhaps inadvertently, something that to the
best of my knowledge has not yet been reported in the literature: The two different
assumptions about mosquito behavior that I outlined above lead to two qualitatively
different optimal strategies. While I appreciate the care you took with using realistic
parameters for a particular biological system to the extent possible,
I believe this is a more general finding about driving mechanisms and calls for a more
general, abstract, treatment. Such a paper might be a very valuable addition to the
literature.
We appreciate the reviewer’s thoughts on this matter. As we say in our reply above, we
have made alterations to the introduction and discussion to better highlight the effect
that different protocols have on our assumptions of mosquito feeding rates. We fully
agree that a more generalized examination of this phenomenon is warranted, but we do
not feel it is appropriate for this specific paper which necessarily is more focused on the
chikungunya outbreak on Reunion. Instead, we have included in our discussion a proposal
for a follow up paper that tackles exactly what the reviewer suggests along with a testable
hypothesis of multiple factors affecting the force of infection that contribute to this result.
On the more empirical side: I don't know which assumption about mosquito behavior is
(more) realistic. Perhaps the truth lies somewhere in between; perhaps there are
empirical studies of this (in this case they should be cited); if not, such should be
performed (in which case this should be recommended in your conclusions).
At present, this is beyond the scope of the current paper, but we believe this would be an
important inclusion in the follow up paper discussed above.
Reviewer 2
Basic reporting
no comment
Experimental design
no comment
Validity of the findings
no comment
Comments for the Author
Optimal Voluntary and Mandatory Insect Repellent Usage and Emigration Strategies to
Control the Chikungunya Outbreak on Reunion Island
Sylvia R. M. Klein, Alex O. Foster, David A. Feagins, Jonathan T. Rowell, and Igor V.
Erovenko
The authors investigated the potential for two intervention measures for under voluntary
and mandatory protocols to control Chikungunya disease on Reunion Island when there is
a risk of the disease becoming endemic.
General comment: I found the manuscript easy to read and follow and can easily be
replicated. I like the fact that the authors included their codes. I recommend it for
publication subject to the revision below.
Specific comments
Line 129: remove the extra “the”
The sentence that starts in line 129 in the original manuscript reads: “Figure 1 shows a
diagram for the chikungunya transmission model on Reunion Island.” We do not see an
extra “the” to remove in this sentence.
Line 131: capitalize “table 1”
Many writing manuals urge not to capitalize “table” in scientific texts. We would rather
defer this decision to the journal’s editorial staff.
Equation 1: Give an explanation as to why disease-related dead is left out and not
accounted for. The authors noted in the introduction 200 people dead from the disease on
the Island.
We added the following sentence to the second paragraph of the Model section: “We
disregard the human disease-induced mortality because it is low, and doing so allows us to
compute endemic equilibria analytically.” We assume that the reviewer meant
disease-related death (an additional mortality rate applied to infected individuals) and not
the physical remains of the deceased who had carried the disease (which would be an
important model feature for hemorrhagic diseases such as ebola). As stated, there were
over 200 deaths during the Reunion outbreak believed to be related to the disease;
however, this was out of 266,000 cases (i.e., at most only .075% of infected died
prematurely within the period of the outbreak). Although this is a larger rate than the
baseline mortality, the clinical view of chikungunya is that it is the lifelong pains it inflicts
on its victims that is most characteristic and concerning. As such, we elected not to
complicate the model with an additional parameter to reflect this process.
Line 139: Do the authors expect to get a different result if this simplification is no made?
We do not expect different results because of the computational simplification.
Symptomatic and asymptotmatic infected individuals have a combined effect (I + I_a) on
the dynamics for recovered humans and vector populations. They are generated in
proportional numbers, and they undergo equal mortality and recovery rates. As such,
there should be no effect due to this simplification.
Line 153: Which of the two equilibria are the authors referring to?
We clarified the statement by changing “an equilibrium” to “an endemic equilibrium.”
Line 159: What is the justification for the use of repellant as a control measure?
Mosquito repellant is standard approach to dealing with vector-borne diseases, and it is
mentioned by the CDC as one protocol for dealing with Chikungunya (reference [8]).
Line 160: Revise the order of the papers cited.
Done.
Equation 10: The authors should give an explanation of this equation. It is not immediately
clear from the description given. The authors should give explanation as given in one of
Erovenko’s previous papers.
We have clarified that the probability of becoming symptomatic is the product of two
separate transition probabilities from S to E and from E to I, as defined by comparative
outflow rates from each box. We now include parentheses in equation (10) to aid the
reader in identifying these individual probabilities.
Equation 13: The derivative of equation 12 is not this simple expression. Are Z^* and
N^*constant? Are these not functions of r? They contain beta_1 = beta_1(r). Or am I
missing something?
Equation (13) is the correct partial derivative of Equation 2. Although N* and Z* are not
static values, their dependence is on the population value of r (r_pop) rather than the
individual choice of r by a focal individual. Equation 13 is about the change of focal
individual fitness by modifying the individual strategy given the background average
behavior within the population. We rephrased the text following the equation (11) to
explain more explicitly that Z* and N* depend on r_pop only and not on r.
Line 206: replace “stopping” with “stop”.
Done.
Equation 22: It is not clear how this equilibrium is greatly different from the first
equilibrium in equation 6, the model with emigration is the same as equation 1 except for
the addition omega from the S class.
In the absence of emigration (and ignoring the disease-induced death rate), the total
human population size at both DFE and EE remains constant at Lambda_1 / mu_1. In
particular, N* = Lambda_1 / mu_1, and computing analytically the values of all
compartments at EE was straightforward. However, once an extra flow out of one
compartment (S) is added, this is no longer the case, and computing the value of N*
becomes difficult. That is why it was much more difficult to obtain closed-form
expressions for the endemic equilibria values in this scenario. There is a more significant
change between Equation (22) and Equation (6) than may be first apparent. In this
modified system, the equilibrium value of the exposed class E has a non-linear
dependence on omega. This is presented in equations (23) through (26).
Line 296: “resulting in a greater proportion”. Am I missing something here? How is this so?
The magnitude of I/N is 10^(-4).
While the total number of infected individuals declines with the increasing migration rate
(figure 6a), the proportion of infected individuals I*/N* grows with the migration rate
(figure 6b) but quickly saturates (equation 27).
Line 327: “Are shown in Figure 8”. Discuss the graphs as was done for Figure 9. There are
four panels in the figure, what do they say individually?
We added an explanation of the phenomena observed in the four panels of figure 8 at the
end of subsection 3.3.
" | Here is a paper. Please give your review comments after reading it. |
9,796 | "<ns0:abstract xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:p>In 2005, a chikungunya virus outbreak devastated the tropical island of Reunion, infecting a third of the total population. Motivated by the Reunion Island case study, we investigate the theoretic potential for two intervention measures under both voluntary and mandatory protocols to control a vector-borne disease when there is risk of the disease becoming endemic. The first measure uses insect repellent to prevent mosquito bites, while the second involves emigrating to the neighboring Mauritius Island to avoid infection. There is a threshold on the cost of using repellent above which both voluntary and mandatory regimes find it optimal to forgo usage. Below that threshold, mandatory usage protocols will eradicate the disease; however, voluntary adoption leaves the disease at a small endemic level. Emigrating from the island to avoid infection results in a tragedy-of-thecommons effect: while being potentially beneficial to specific susceptible individuals, the remaining islanders paradoxically face a higher risk of infection. Mandated relocation of susceptible individuals away from the epidemic is viable only if the cost of this relocation is several magnitudes lower than the cost of infection. Since this assumption is unlikely to hold for chikungunya, it is optimal to discourage such emigration for the benefit of the entire population. An underlying assumption about the conservation of human-vector encounter rates in mosquito biting behavior informs our conclusions and may warrant additional experimental verification.</ns0:p></ns0:div>
</ns0:abstract>
<ns0:body xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:head>INTRODUCTION</ns0:head><ns0:p>Reunion Island is a tropical island located in the Indian Ocean 500 miles east of Madagascar and approximately 150 miles southwest of Mauritius. The island was devastated by a major chikungunya outbreak in <ns0:ref type='bibr'>[2005]</ns0:ref><ns0:ref type='bibr'>[2006]</ns0:ref>, when approximately 266 thousand of the 785 thousand inhabitants were infected, causing over 200 deaths <ns0:ref type='bibr' target='#b47'>(Yakob and Clements, 2013)</ns0:ref>. In the aftermath of that outbreak, the chikungunya virus spread from Africa to Europe, USA, and Australia, and although the incidence levels of this disease remain low, its potential to cause future outbreaks in these areas is cause for concern. In this paper, we investigate the viability of voluntary participation in personal protective measures (mosquito repellent and emigration) against diseases like chikungunya on Reunion Island by constructing a game-theoretic model in which individual strategic payoffs are compared against the average population payoff.</ns0:p><ns0:p>Chikungunya virus (CHIKV) is an Alphavirus in the Togaviridae family, similar to Dengue fever and Zika virus <ns0:ref type='bibr' target='#b25'>(Leao et al., 2018;</ns0:ref><ns0:ref type='bibr' target='#b30'>Prow et al., 2018;</ns0:ref><ns0:ref type='bibr' target='#b29'>Pile et al., 1999)</ns0:ref>. It is a vector-borne virus spread through bites by the females of Aedes aegypti and Aedes albopictus mosquitoes. After a bite, there is a latency period for both humans and mosquitoes: it can take between 2 to 6 days for symptoms to develop and for an individual to become infectious <ns0:ref type='bibr' target='#b20'>(Fourie and Morrison, 1979)</ns0:ref>. The major symptoms associated with CHIKV are fever, rash, arthritis, headache, and nausea <ns0:ref type='bibr' target='#b20'>(Fourie and Morrison, 1979)</ns0:ref>. The defining characteristic of CHIKV is the persistence of arthritis for years after the initial infection <ns0:ref type='bibr' target='#b25'>(Leao et al., 2018)</ns0:ref>. A small percentage of people infected with CHIKV, however, never develop symptoms of the disease <ns0:ref type='bibr' target='#b17'>(Darrigo et al., 2018)</ns0:ref>. Humans are no longer infectious about a week and a half after the initial infection, but may still be symptomatic. Recovered individuals acquire lifelong immunity from future infections (Centers for Disease Control and Prevention, b). There is no vaccine to prevent or medicine to treat chikungunya virus (Centers for Disease <ns0:ref type='bibr'>Control and Prevention, b)</ns0:ref>. The most effective way to prevent infection from CHIKV is to prevent mosquito bites, for example, by using insect repellent (Centers for Disease Control and Prevention, a).</ns0:p><ns0:p>Chikungunya was first isolated in <ns0:ref type='bibr'>[1952]</ns0:ref><ns0:ref type='bibr'>[1953]</ns0:ref> in Tanzania <ns0:ref type='bibr' target='#b32'>(Robinson, 1955)</ns0:ref>. The name translates to the native term for 'that which bends up' <ns0:ref type='bibr' target='#b37'>(Seneviratne et al., 2007)</ns0:ref>. There were limited outbreaks between the initial discovery of the disease and a worldwide outbreak that occurred in 2004-2005 <ns0:ref type='bibr' target='#b17'>(Darrigo et al., 2018)</ns0:ref>. This outbreak started in Kenya and spread to the surrounding islands including Mauritius, Rodrigues, The Seychelles, Mayotte, Madagascar and Reunion Island <ns0:ref type='bibr' target='#b31'>(Renault et al., 2012)</ns0:ref>. From these islands, it spread to other regions of the world-chikungunya virus is now present on every continent except Antarctica-most likely carried by tourists. The disease impacted Reunion Island most severely: a third of the population became infected, unusually severe forms were present, and the first occurrences of maternal-neonatal transmission were documented <ns0:ref type='bibr' target='#b6'>(Borgherini et al., 2008)</ns0:ref>. This severity of impact may be attributed to an increase in travel between islands and the climate of the region at the time of the epidemic <ns0:ref type='bibr' target='#b6'>(Borgherini et al., 2008;</ns0:ref><ns0:ref type='bibr' target='#b42'>Thiberville et al., 2013)</ns0:ref>.</ns0:p><ns0:p>While the severity of the Reunion chikungunya outbreak may seem like an isolated event, vector-borne diseases such as malaria and dengue are becoming an increasingly prevalent public health issue in today's society. In the United States there has been a 23-fold increase of vector-borne disease cases in the past ten years <ns0:ref type='bibr' target='#b33'>(Rosenberg et al., 2018)</ns0:ref>. There are now 16 vector-borne diseases widely distributed in the United States, all of which are resistant to control, and only one of these (Yellow Fever) has an FDA-approved vaccine <ns0:ref type='bibr' target='#b33'>(Rosenberg et al., 2018)</ns0:ref>. Even though there are limited cases of CHIKV in the United States and its territories, the disease is becoming more persistent: the number of national cases and distribution are increasing, and the range of the mosquitoes that transmit CHIKV has spread to 38 states as of 2016 <ns0:ref type='bibr' target='#b33'>(Rosenberg et al., 2018)</ns0:ref>.</ns0:p><ns0:p>Due to the transmission patterns of mosquito-borne diseases and the lack of sufficient vector control to eradicate such diseases, individuals often have to rely on voluntary participation in personal protection measures. Unfortunately, individual self-interest in protection against an infectious disease does not necessarily correspond to the desired outcome for society <ns0:ref type='bibr' target='#b21'>(Galvani et al., 2007)</ns0:ref>, namely eradication of the disease. The effect of potentially selfish human behavior on the spread of infectious diseases only recently begun to receive attention, forming a new field of behavioral epidemiology; see <ns0:ref type='bibr' target='#b26'>Manfredi and D'Onofrio (2013)</ns0:ref> for a review of behavioral epidemiology.</ns0:p><ns0:p>Originally designed for the field of economics <ns0:ref type='bibr' target='#b45'>(von Neumann and Morgenstern, 1944)</ns0:ref>, game theory has since been used to model many biological phenomena <ns0:ref type='bibr' target='#b28'>(Maynard Smith, 1982;</ns0:ref><ns0:ref type='bibr' target='#b23'>Hofbauer and Sigmund, 1998;</ns0:ref><ns0:ref type='bibr' target='#b16'>Cressman, 2003;</ns0:ref><ns0:ref type='bibr' target='#b44'>Vincent and Brown, 2005;</ns0:ref><ns0:ref type='bibr' target='#b9'>Broom and Rychtář, 2013)</ns0:ref>, including individual-level vaccination decisions <ns0:ref type='bibr' target='#b4'>(Bauch and Earn, 2004)</ns0:ref>. In a vaccination game, a selfish individual seeks to maximize its benefit, or rather to minimize the potential loss resulting from either employing a potentially costly protective measure or facing the consequences of the disease. As the likelihood of contracting the disease is dependent upon the behavior of others within the at-risk population, the resulting strategic interactions between individuals can be modeled using game theory. Game-theoretic frameworks have been adopted to studying optimal individual vaccination strategies for smallpox <ns0:ref type='bibr' target='#b5'>(Bauch et al., 2003)</ns0:ref>, influenza <ns0:ref type='bibr' target='#b21'>(Galvani et al., 2007;</ns0:ref><ns0:ref type='bibr' target='#b38'>Shim et al., 2012a)</ns0:ref>, rubella <ns0:ref type='bibr' target='#b40'>(Shim et al., 2009)</ns0:ref>, measles <ns0:ref type='bibr' target='#b39'>(Shim et al., 2012b)</ns0:ref>, toxoplasmosis <ns0:ref type='bibr' target='#b41'>(Sykes and Rychtář, 2015)</ns0:ref>, Ebola <ns0:ref type='bibr' target='#b8'>(Brettin et al., 2018)</ns0:ref>, cholera <ns0:ref type='bibr' target='#b24'>(Kobe et al., 2018)</ns0:ref>, meningitis <ns0:ref type='bibr' target='#b27'>(Martinez et al., 2019)</ns0:ref>, hepatitis B <ns0:ref type='bibr' target='#b14'>(Chouhan et al., 2019)</ns0:ref>, monkeypox <ns0:ref type='bibr' target='#b3'>(Bankuru et al., 2019)</ns0:ref>, poliomyelitis <ns0:ref type='bibr' target='#b13'>(Cheng et al., 2020)</ns0:ref>, and typhoid fever <ns0:ref type='bibr' target='#b1'>(Acosta-Alonzo et al., 2020)</ns0:ref>. It has also been applied to other personal protective measures such as insecticide-treated cattle <ns0:ref type='bibr' target='#b15'>(Crawford et al., 2015)</ns0:ref>, mosquito repellent <ns0:ref type='bibr' target='#b18'>(Dorsett et al., 2016)</ns0:ref>, insecticide-treated bed nets <ns0:ref type='bibr' target='#b10'>(Broom et al., 2016)</ns0:ref>, clean water <ns0:ref type='bibr' target='#b24'>(Kobe et al., 2018)</ns0:ref>, and clean injecting equipment <ns0:ref type='bibr' target='#b36'>(Scheckelhoff et al., 2019)</ns0:ref>. For an extensive review of</ns0:p></ns0:div>
<ns0:div><ns0:head>2/17</ns0:head><ns0:p>PeerJ reviewing PDF | (2019:08:40674:2:0:NEW 12 May 2020)</ns0:p><ns0:p>Manuscript to be reviewed behavior-linked vaccination models, see <ns0:ref type='bibr' target='#b46'>Wang et al. (2016)</ns0:ref>.</ns0:p><ns0:p>In this paper, we investigate the potential effects of voluntary and government-mandated participation in utilizing the insect repellent as a protective measure against a disease such as chikungunya on Reunion Island. We also analyze the effect of emigration to a neighboring island (Mauritius) on the spread of chikungunya among the remaining population of Reunion Island. We find that the latter protocol has a paradoxically worsening of outcomes for the non-participating population potentially as a consequence of the non-responsiveness in the mosquitoes feeding behavior to decreases in the human population relative to other blood sources.</ns0:p></ns0:div>
<ns0:div><ns0:head>METHODS</ns0:head><ns0:p>We adopt a version of the epidemiological model of the chikungunya outbreak on Reunion Island by <ns0:ref type='bibr' target='#b47'>Yakob and Clements (2013)</ns0:ref> by adding population dynamics (birth and death demographic processes) for both humans and mosquitoes and strategically-linked parameters so that the disease potentially may establish itself endemically. This assumption then permits us to use the framework of <ns0:ref type='bibr' target='#b18'>Dorsett et al. (2016)</ns0:ref>, <ns0:ref type='bibr' target='#b2'>Amaku et al. (2014)</ns0:ref>, and <ns0:ref type='bibr' target='#b4'>Bauch and Earn (2004)</ns0:ref>. All human inhabitants of the island (N) are divided into 5 compartments: individuals susceptible to chikungunya (S); exposed individuals (E), who had been bitten by an infected mosquito and acquired the disease; symptomatic infectious individuals (I), who developed the symptoms of the disease and became infectious to biting mosquitoes; asymptomatic infectious individuals (I a ), who became infectious but did not develop symptoms; and recovered individuals (R), who recovered from chikungunya and acquired immunity. The mosquito population is divided into 3 compartments: susceptible mosquitoes (X); exposed mosquitoes (Y ), who bit an infected human and were exposed to the pathogen; and infectious mosquitoes (Z), who may infect humans by biting susceptible individuals. We did not consider in this model the full life-cycle of mosquitoes, such as egg and larval stages, because we did not incorporate mosquito population control as one of the measures to fight chikungunya.</ns0:p><ns0:p>New individuals enter the susceptible part of the population at a rate Λ 1 due to birth or immigration; there is a natural per capita human mortality µ 1 . Similarly, new mosquitoes are recruited into the susceptible compartment at a rate Λ 2 , and there is a natural per capita mosquito mortality µ 2 . We disregard the human disease-induced mortality because it is low, and doing so allows us to compute endemic equilibria analytically.</ns0:p><ns0:p>Susceptible humans who are bitten by infectious mosquitoes become exposed. The force of infection f 1 , which is the rate at which susceptible individuals move to the exposed class, depends on the density of susceptible humans S/N (i.e., the probability that an infectious mosquito bites a susceptible individual), the number of infectious mosquitoes Z, and the mosquito-to-human transmission coefficient β 1 . We assume that mosquitoes have a consistent average number of encounters with humans over a given time span, and that repellent usage directly decreases the force of infection by deterring biting upon encounter.</ns0:p><ns0:p>Similarly, susceptible mosquitoes who bite infectious humans become exposed. The force of infection f 2 , which is the rate at which susceptible mosquitoes move to the exposed compartment, depends on the the density of infectious humans (I + I a )/N (i.e., the probability that a mosquito bites an infectious individual), the number of susceptible mosquitoes X, and the human-to-mosquito transmission coefficient</ns0:p><ns0:formula xml:id='formula_0'>β 2 .</ns0:formula><ns0:p>Exposed humans become infectious (after a latency period) at a rate λ 1 ; a proportion φ of infectious individuals develop symptoms of the disease. Exposed mosquitoes become infectious at a rate λ 2 . Infectious humans (both symptomatic and asymptomatic) recover at a rate γ and acquire immunity from future infections. The lifespan of a mosquito is too short to recover; an infectious mosquito remains as such until it dies.</ns0:p><ns0:p>Figure <ns0:ref type='figure' target='#fig_0'>1</ns0:ref> shows a diagram for the chikungunya transmission model on Reunion Island. The parameters of the epidemiological model are summarized in table 1. The table includes the baseline value of the mosquito-to-human transmission parameter, denoted by β 0 1 . Later this parameter will be affected by an intervention measure (insect repellent), and hence it will become a function of the level of insect repellent usage, given by ( <ns0:ref type='formula' target='#formula_12'>7</ns0:ref>). The dynamics of the compartment model in figure <ns0:ref type='figure' target='#fig_0'>1</ns0:ref> is described by the following system of differential equations:</ns0:p><ns0:formula xml:id='formula_1'>dS dt = Λ 1 − β 1 SZ N − µ 1 S, dE dt = β 1 SZ N − λ 1 E − µ 1 E, dI dt = φ λ 1 E − γI − µ 1 I, dI a dt = (1 − φ )λ 1 E − γI a − µ 1 I a , dR dt = γ(I + I a ) − µ 1 R, dX dt = Λ 2 − β 2 X(I + I a ) N − µ 2 X, dY dt = β 2 X(I + I a ) N − λ 2 Y − µ 2 Y,<ns0:label>and</ns0:label></ns0:formula><ns0:formula xml:id='formula_2'>dZ dt = λ 2 Y − µ 2 Z.</ns0:formula><ns0:p>(1)</ns0:p><ns0:p>The disease-free equilibrium (DFE) of this system is given by</ns0:p><ns0:formula xml:id='formula_3'>S 0 , E 0 , I 0 , I 0 a , R 0 , X 0 ,Y 0 , Z 0 = Λ 1 µ 1 , 0, 0, 0, 0, Λ 2 µ 2 , 0, 0 .<ns0:label>(2)</ns0:label></ns0:formula><ns0:p>To compute the basic reproduction number R 0 , we use the next-generation matrix approach (van den <ns0:ref type='bibr' target='#b43'>Driessche and Watmough, 2002)</ns0:ref>. To simplify this computation, we temporarily combined the symptomatic and asymptomatic infectious compartments into one infectious compartment I + I a : individuals in both I and I a compartments have identical contributions to the dynamics of chikungunya. We order the compartments that contribute to new infections as follows: E, I + I a , Y , and Z. Then the vector of the rates of appearance of new infections in these four compartments F and the vector of the rates of transfer Proportion of hosts that develop symptoms 0.97 <ns0:ref type='bibr' target='#b47'>Yakob and Clements (2013)</ns0:ref> of existing infections between these four compartments V are given by</ns0:p><ns0:formula xml:id='formula_4'>F =     β 1 SZ N 0 β 2 X(I+I a ) N 0     and V =     λ 1 E + µ 1 E −λ 1 E + γ(I + I a ) + µ 1 (I + I a ) λ 2 Y + µ 2 Y −λ 2 Y + µ 2 Z     .</ns0:formula><ns0:p>(3)</ns0:p><ns0:p>The matrices F and V are the Jacobians of F and V respectively, evaluated at DFE; they are given by</ns0:p><ns0:formula xml:id='formula_5'>F =     0 0 0 β 1 0 0 0 0 0 Λ 2 β 2 µ 1 Λ 1 µ 2 0 0 0 0 0 0     and V =     λ 1 + µ 1 0 0 0 −λ 1 γ + µ 1 0 0 0 0 λ 2 + µ 2 0 0 0 −λ 2 µ 2     .<ns0:label>(4)</ns0:label></ns0:formula><ns0:p>The basic reproduction number is the spectral radius of the matrix FV −1 ; it is given by</ns0:p><ns0:formula xml:id='formula_6'>R 0 = 1 µ 2 Λ 2 β 1 β 2 µ 1 λ 1 λ 2 Λ 1 (γ + µ 1 )(λ 1 + µ 1 )(λ 2 + µ 2 ) . (<ns0:label>5</ns0:label></ns0:formula><ns0:formula xml:id='formula_7'>)</ns0:formula><ns0:p>If R 0 > 1, then the system converges to the endemic equilibrium (EE) given by</ns0:p><ns0:formula xml:id='formula_8'>S * = Λ 1 − (λ 1 + µ 1 )E * µ 1 , E * = Λ 1 Λ 2 β 1 β 2 µ 1 λ 1 λ 2 − Λ 2 1 µ 2 2 (λ 1 + µ 1 )(λ 2 + µ 2 )(γ + µ 1 ) Λ 2 β 1 β 2 µ 1 λ 1 λ 2 (λ 1 + µ 1 ) + Λ 1 β 2 µ 1 µ 2 λ 1 (λ 1 + µ 1 )(λ 2 + µ 2 ) , I * = φ λ 1 E * γ + µ 1 , I * a = (1 − φ )λ 1 E * γ + µ 1 , R * = γλ 1 E * µ 1 (γ + µ 1 )</ns0:formula><ns0:p>,</ns0:p><ns0:formula xml:id='formula_9'>X * = Λ 1 Λ 2 (γ + µ 1 ) β 2 µ 1 λ 1 E * + Λ 1 µ 2 (γ + µ 1 ) , Y * = Λ 2 β 2 µ 1 λ 1 E * (λ 2 + µ 2 )(β 2 µ 1 λ 1 E * + Λ 1 µ 2 (γ + µ 1 ))</ns0:formula><ns0:p>, and</ns0:p><ns0:formula xml:id='formula_10'>Z * = Λ 2 β 2 µ 1 λ 1 λ 2 E * µ 2 (λ 2 + µ 2 )(β 2 µ 1 λ 1 E * + Λ 1 µ 2 (γ + µ 1 )) . (<ns0:label>6</ns0:label></ns0:formula><ns0:formula xml:id='formula_11'>)</ns0:formula><ns0:p>In the game-theoretic models constructed in the next section, we will be assuming that the system has reached an endemic equilibrium. In particular, we will use the values from (6) for relevant compartment sizes.</ns0:p></ns0:div>
<ns0:div><ns0:head>5/17</ns0:head><ns0:p>PeerJ reviewing PDF | (2019:08:40674:2:0:NEW 12 May 2020)</ns0:p><ns0:p>Manuscript to be reviewed</ns0:p></ns0:div>
<ns0:div><ns0:head>RESULTS</ns0:head><ns0:p>We consider two intervention measures to fight the chikungunya outbreak on Reunion Island: (1) using insect repellent to prevent mosquito bites; and (2) emigrating to the neighboring Mauritius Island.</ns0:p></ns0:div>
<ns0:div><ns0:head>Optimal levels of voluntary insect repellent usage</ns0:head><ns0:p>We adopt a modeling approach of <ns0:ref type='bibr' target='#b4'>Bauch and Earn (2004)</ns0:ref>; <ns0:ref type='bibr' target='#b18'>Dorsett et al. (2016)</ns0:ref>. The strategy of an individual is the proportion of the day r ∈ [0, 1] the individual is protected from mosquito bites; the protection is granted by insect repellent. We assume that the repellent provides complete protection from mosquito bites while it is active. Since mosquitoes cannot bites humans while they are protected by the insect repellent, the mosquito-to-human transmission coefficient β 1 becomes a function of r. If no protection is used (r = 0), then β 1 (0) is at its base value β 0 1 (given in table <ns0:ref type='table' target='#tab_0'>1</ns0:ref>). If humans are protected at all times (r = 1), then mosquitoes cannot bite these humans at all, and hence they cannot infect humans: β 1 (1) = 0. We therefore assume that the mosquito-to-human transmission coefficient is a linear function of r given by The basic reproduction number is at its maximum value-given by ( <ns0:ref type='formula' target='#formula_6'>5</ns0:ref>)-when no insect repellent is used by susceptible individuals (r = 0), and it becomes zero if the population employs complete protection from mosquito bites (r = 1). The threshold for disease eradication (R 0 = 1) is achieved at the herd immunity protection level r HI .</ns0:p><ns0:formula xml:id='formula_12'>β 1 = β 0 1 (1 − r).<ns0:label>(7</ns0:label></ns0:formula><ns0:p>We define the utility function (expected payoff) of a susceptible individual using strategy r in a population that adopted strategy r pop as</ns0:p><ns0:formula xml:id='formula_13'>E(r, r pop ) = −π(r, r pop )C i − rC p ,<ns0:label>(8)</ns0:label></ns0:formula><ns0:p>where C i is the cost of infection, C p is the cost of complete protection through insect repellent, and π(r, r pop ) is the probability of infection. The latter depends on the individual's strategy r because it determines how often mosquitoes may bite the individual, and on the population strategy r pop because it affects the prevalence of the disease (e.g., the number of infected mosquitoes). The outcome of a game does not change if the utility function is scaled, so we divide the right-hand side of (8) by C i to obtain</ns0:p><ns0:formula xml:id='formula_14'>E(r, r pop ) = −π(r, r pop ) − rC,<ns0:label>(9)</ns0:label></ns0:formula><ns0:p>where C = C p /C i is the cost of complete protection relative to the cost of infection.</ns0:p></ns0:div>
<ns0:div><ns0:head>6/17</ns0:head><ns0:p>PeerJ reviewing PDF | (2019:08:40674:2:0:NEW 12 May 2020)</ns0:p><ns0:p>Manuscript to be reviewed</ns0:p><ns0:p>We next compute the probability of getting infected and becoming symptomatic as the transition probability from the susceptible compartment S to the symptomatic infectious compartment I. This probability is the product of the probability that a susceptible individual becomes exposed f 1 /(µ 1 + f 1 ) multiplied by the probability that an exposed individual becomes symptomatically infected φ λ 1 /(µ</ns0:p><ns0:formula xml:id='formula_15'>1 + φ λ 1 + (1 − φ )λ 1 ) = φ λ 1 /(µ 1 + λ 1 ): π(r, r pop ) = f 1 (r, r pop ) µ 1 + f 1 (r, r pop ) φ λ 1 µ 1 + λ 1 ,<ns0:label>(10)</ns0:label></ns0:formula><ns0:p>where for β 1 into (6). In particular, Z * and N * do not depend on the individual protection level r.</ns0:p><ns0:formula xml:id='formula_16'>f 1 (r, r pop ) = β 0 1 (1 − r) Z * N *<ns0:label>(</ns0:label></ns0:formula><ns0:p>To find the Nash equilibrium population protection level, we attempt to maximize the utility function (9) of a focal individual. Observe that</ns0:p><ns0:formula xml:id='formula_17'>f 1 (r, r pop ) = (1 − r) f 1 (0, r pop ),<ns0:label>(12)</ns0:label></ns0:formula><ns0:p>and hence</ns0:p><ns0:formula xml:id='formula_18'>∂ ∂ r f 1 (r, r pop ) = − f 1 (0, r pop ).<ns0:label>(13)</ns0:label></ns0:formula><ns0:p>It follows that</ns0:p><ns0:formula xml:id='formula_19'>∂ 2 ∂ r 2 E(r, r pop ) = 2µ 1 f 1 (0, r pop ) 2 (µ 1 + f 1 (r, r pop )) 3 > 0. (<ns0:label>14</ns0:label></ns0:formula><ns0:formula xml:id='formula_20'>)</ns0:formula><ns0:p>Consequently, the utility function is a convex function of r, and thus it attains its maximum value at one of the endpoints: r = 0 or r = 1.</ns0:p><ns0:p>This conclusion can be interpreted as follows. If the population repellent usage is sufficiently high, then the probability of getting infected is very low. A focal individual would rather bypass the potentially costly preventive measure and face the low morbidity risk instead. Hence individuals may improve their payoff by deviating from the population strategy (they should stop using repellent). On the other hand, if the population repellent usage is low, then the probability of getting infected is too high, and a focal individual should prefer to pay the cost of complete protection rather than face the high morbidity risk.</ns0:p><ns0:p>In this case, individuals may also improve their payoff by deviating from the population strategy (they should use repellent 100% of the time).</ns0:p><ns0:p>So, if the population repellent usage is too high, then individuals would do better if they stop using repellent, and hence the population repellent usage will decrease. Conversely, if the population repellent usage is too low, then individuals would do better if they start using repellent 100% of the time, and hence the population repellent usage will increase. If the population repellent usage is 'just right' (Nash equilibrium), then individuals cannot improve their payoffs by using repellent either less frequently or more frequently. We note that there is a presumption of the population-wide adoption of treatment rates in our model that is common to ESS-modeling; however, in situations such as ( <ns0:ref type='formula' target='#formula_19'>14</ns0:ref>), there is a potential implication that the population should in fact separate into distinct groups with different adoption rates.</ns0:p><ns0:p>As it goes beyond the framework discussed here, we leave that investigation for future research.</ns0:p><ns0:p>The Nash equilibrium protection level of the population r NE is thus a solution to the equation</ns0:p><ns0:formula xml:id='formula_21'>E(0, r NE ) = E(1, r NE )<ns0:label>(15)</ns0:label></ns0:formula><ns0:p>or</ns0:p><ns0:formula xml:id='formula_22'>f 1 (0, r NE ) µ 1 + f 1 (0, r NE ) φ λ 1 µ 1 + λ 1 = C. (<ns0:label>16</ns0:label></ns0:formula><ns0:formula xml:id='formula_23'>)</ns0:formula></ns0:div>
<ns0:div><ns0:head>7/17</ns0:head><ns0:p>PeerJ reviewing PDF | (2019:08:40674:2:0:NEW 12 May 2020)</ns0:p><ns0:p>Manuscript to be reviewed</ns0:p><ns0:p>The graph of the optimal (Nash equilibrium) repellent usage as a function of the relative cost of protection C is shown in figure <ns0:ref type='figure'>3a</ns0:ref>. The optimal repellent usage r NE reaches the herd immunity r HI level only when the cost of the protective measure relative to the cost of chikungunya infection is negligible (i.e., zero mathematically). The optimal repellent usage remains very close to the herd immunity level for a range of values of the relative cost C, and then drops off sharply. Once the relative cost of protection becomes too large (C max ), then everyone stops using insect repellent because its high cost forces individuals to prefer to risk the cost of infection.</ns0:p></ns0:div>
<ns0:div><ns0:head>Figure 3. (a)</ns0:head><ns0:p>The graph of the optimal level of population repellent usage r NE as a function of the relative cost of protection. The optimal repellent usage reaches the herd immunity level only when C = 0. Everyone stops using repellent if its relative cost is too high: larger than the threshold value C max . (b) The graph of the basic reproduction number computed at the optimal population repellent usage level r NE as a function of the relative cost of protection. When C = 0, the optimal protection level is equal to the herd immunity threshold, so R 0 = 1. When the relative cost of protection exceeds the threshold value C max , the population stops using repellent, and the basic reproduction number reaches its maximum value.</ns0:p></ns0:div>
<ns0:div><ns0:head>Optimal levels of mandatory insect repellent usage</ns0:head><ns0:p>We now consider a scenario where an organization (e.g., the government) enforces the use of insect repellent in the population to fight chikungunya. The organization must balance the cost of prevention of the disease in the population and the cost of treatment of symptomatically infected individuals. On the one hand, every individual who utilizes insect repellent 100% of the time results in a cost C p (same as the cost of voluntary complete protection). On the other hand, every symptomatically infected individual results in a cost C i .</ns0:p><ns0:p>The goal of the mandating organization is to find the repellent usage level for the population r pop ∈ [0, 1] so that the expected payoff (negative of the total cost) Manuscript to be reviewed </ns0:p><ns0:formula xml:id='formula_24'>E(r pop ) = −C i I * − r pop C p S *<ns0:label>(</ns0:label></ns0:formula></ns0:div>
<ns0:div><ns0:head>Optimal levels of voluntary emigration</ns0:head><ns0:p>We are going to operate under the assumption that the chikungunya outbreak on Reunion Island did not affect the neighboring island of Mauritius Island, located 140 miles to the north-east of Reunion. Susceptible individuals-and only susceptible individuals, perhaps identified through a screening or quarantine procedure-may elect to emigrate from Reunion to Mauritius to protect themselves from the outbreak. To model the emigration as a potential personal protection measure against chikungunya, we allow residents of Reunion Island to leave the susceptible compartment of the epidemiological model at an emigration rate ω. This modification to (1) replaces the first equation describing the change in the susceptible population with</ns0:p><ns0:formula xml:id='formula_25'>dS dt = Λ 1 − β 1 SZ N − µ 1 S − ωS. (<ns0:label>19</ns0:label></ns0:formula><ns0:formula xml:id='formula_26'>)</ns0:formula><ns0:p>The DFE of the modified system is given by</ns0:p><ns0:formula xml:id='formula_27'>(S 0 , E 0 , I 0 , I 0 a , R 0 , X 0 ,Y 0 , Z 0 ) = Λ 1 µ 1 + ω , 0, 0, 0, 0, Λ 2 µ 2 , 0, 0 ,<ns0:label>(20)</ns0:label></ns0:formula><ns0:p>and the corresponding basic reproduction number of the disease is</ns0:p><ns0:formula xml:id='formula_28'>R 0 = 1 µ 2 Λ 2 β 1 β 2 λ 1 λ 2 (µ 1 + ω) Λ 1 (γ + µ 1 )(λ 1 + µ 1 )(λ 2 + µ 2 ) . (<ns0:label>21</ns0:label></ns0:formula><ns0:formula xml:id='formula_29'>)</ns0:formula><ns0:p>The graph of the basic reproduction number as a function of the emigration rate ω is shown in figure <ns0:ref type='figure' target='#fig_6'>5</ns0:ref>. It is an increasing function of ω, and hence emigrating susceptible individuals paradoxically make it worse for the remaining susceptible population. When the fresh blood supply is reduced due to emigration, susceptible mosquitoes are more likely to prey upon infectious humans, increasing the disease prevalence in the vector population. Consequently, the remaining susceptible human population is at an increased risk of contracting the disease from a mosquito bite. It follows that, while being potentially beneficial to specific individuals, voluntary emigration may result in a tragedy-of-the-commons effect for the remaining islanders.</ns0:p><ns0:p>To further investigate the effect of voluntary emigration on the chikungunya epidemic and whether (selfish) susceptible individuals should emigrate, we compute the EE values of all compartments in the </ns0:p><ns0:formula xml:id='formula_30'>N * = Λ 1 µ 1 + ω(λ 1 + µ 1 )E * µ 1 (µ 1 + ω) , S * = Λ 1 − (λ 1 + µ 1 )E * µ 1 + ω , I * = φ λ 1 E * γ + µ 1 , I * a = (1 − φ )λ 1 E * γ + µ 1 , R * = γλ 1 E * µ 1 (γ + µ 1 )</ns0:formula><ns0:p>,</ns0:p><ns0:formula xml:id='formula_31'>X * = Λ 2 (γ + µ 1 )(Λ 1 µ 1 + ω(λ 1 + µ 1 )E * d + µ 2 (γ + µ 1 )(Λ 1 µ 1 + ω(λ 1 + µ 1 )E * ) , Y * = Λ 2 β 2 µ 1 λ 1 (µ 1 + ω)E * (λ 2 + µ 2 )[d + Λ 1 µ 1 µ 2 (γ + µ 1 ) + µ 2 ω(λ 1 + µ 1 )(γ + µ 1 )E * ]</ns0:formula><ns0:p>, and</ns0:p><ns0:formula xml:id='formula_32'>Z * = Λ 2 β 2 µ 1 λ 1 λ 2 (µ 1 + ω)E * µ 2 (λ 2 + µ 2 )[d + Λ 1 µ 1 µ 2 (γ + µ 1 ) + µ 2 ω(λ 1 + µ 1 )(γ + µ 1 )E * ] ,<ns0:label>(22)</ns0:label></ns0:formula><ns0:p>where</ns0:p><ns0:formula xml:id='formula_33'>d = β 2 λ 1 µ 1 (µ 1 + ω)E * ,<ns0:label>(23)</ns0:label></ns0:formula><ns0:p>and E * is the solution to the quadratic equation</ns0:p><ns0:formula xml:id='formula_34'>aE 2 + bE + c = 0<ns0:label>(24)</ns0:label></ns0:formula><ns0:p>with coefficients</ns0:p><ns0:formula xml:id='formula_35'>a = −µ 2 ω(λ 1 + µ 1 ) 2 (λ 2 + µ 2 )[β 2 µ 1 λ 1 (µ 1 + ω) + µ 2 ω(λ 1 + µ 1 )(γ + µ 1 )], b = −Λ 2 β 1 β 2 µ 2 1 λ 1 λ 2 (λ 1 + µ 1 )(µ 1 + ω) − 2Λ 1 µ 1 µ 2 2 ω(λ 1 + µ 1 ) 2 (λ 2 + µ 2 )(γ + µ 1 ) − Λ 1 β 2 µ 2 1 µ 2 λ 1 (λ 1 + µ 1 )(λ 2 + µ 2 )(µ 1 + ω), and c = Λ 1 Λ 2 β 1 β 2 µ 2 1 λ 1 λ 2 (µ 2 + ω) − Λ 2 1 µ 2 1 µ 2 2 (λ 1 + µ 1 )(λ 2 + µ 2 )(γ + µ 1 ).<ns0:label>(25)</ns0:label></ns0:formula><ns0:p>The biologically meaningful root of this equation is given by</ns0:p><ns0:formula xml:id='formula_36'>E * = −b − √ b 2 − 4ac 2a . (<ns0:label>26</ns0:label></ns0:formula><ns0:formula xml:id='formula_37'>)</ns0:formula><ns0:p>Figure <ns0:ref type='figure' target='#fig_7'>6</ns0:ref> shows the graphs of the number and proportion of symptomatically infectious individuals in the population as functions of the emigration rate ω. As more individuals leave the island, the overall population level declines, and hence there are fewer infected individuals. However, the infection spreads faster among the remaining inhabitants, resulting in a greater proportion of infected individuals in the population. The proportion of symptomatically infectious individuals grows with the migration rate, but it asymptotically approaches the value lim ω→∞ We next consider a game-theoretic model of individual migration decisions. Suppose that the population adopted the emigration rate ω pop . A focal susceptible individual is presented with a choice to either migrate or not migrate. Each of the two strategic choices carries a corresponding payoff: E m for migrate and E nm for not migrate, given by</ns0:p><ns0:formula xml:id='formula_38'>I * N * = φ µ 1 λ 1 (γ + µ 1 )(λ 1 + µ 1 ) . (<ns0:label>27</ns0:label></ns0:formula><ns0:formula xml:id='formula_39'>E m (ω pop ) = −C b − ω pop C s ,<ns0:label>and</ns0:label></ns0:formula><ns0:formula xml:id='formula_40'>E nm (ω pop ) = −π(ω pop )C i ,<ns0:label>(28)</ns0:label></ns0:formula><ns0:p>where C b is the base (fixed) cost of migration, C s is the scaling cost of migration, C i is the cost of the (symptomatic) chikungunya infection, and π(ω pop ) is the probability of getting infected given the population emigration rate ω pop . We assume that the cost of emigration is an increasing function of the migration rate because of the limited immigration potential of Mauritius: the more individuals migrate to Mauritius, the harder it becomes to find housing and jobs. For simplicity, we model the increasing emigration cost as a linear function of the migration rate. The probability of getting infected and incurring the cost of a symptomatic chikungunya infection if remaining on Reunion Island is the transition probability from the susceptible class S to the symptomatically infectious class I:</ns0:p><ns0:formula xml:id='formula_41'>π(ω pop ) = f 1 (ω pop ) µ 1 + f 1 (ω pop ) φ λ 1 µ 1 + λ 1 . (<ns0:label>29</ns0:label></ns0:formula><ns0:formula xml:id='formula_42'>)</ns0:formula><ns0:p>This probability is an increasing function of the emigration rate because each of the remaining susceptible 221 individuals faces a higher risk of infection (cf. figure <ns0:ref type='figure' target='#fig_6'>5</ns0:ref>); the graph is shown in figure <ns0:ref type='figure' target='#fig_8'>7</ns0:ref>.</ns0:p></ns0:div>
<ns0:div><ns0:head>222</ns0:head><ns0:p>To find conditions when a focal susceptible individual should emigrate to Mauritius or remain on Reunion, we scale the payoffs in equation ( <ns0:ref type='formula' target='#formula_40'>28</ns0:ref>) by 1/C i and obtain E m = − Cb − ω pop Cs , and</ns0:p><ns0:formula xml:id='formula_43'>E nm = −π(ω pop ),<ns0:label>(30)</ns0:label></ns0:formula><ns0:p>where Cb and Cs are relative base and scaling costs of emigration, respectively. A susceptible individual to emigrate as long as the relative base cost of emigration Cb is sufficiently small (figure <ns0:ref type='figure' target='#fig_9'>8a</ns0:ref>). On the other hand, as the relative scaling cost of emigration Cs grows, the individual's decision to emigrate starts to depend on the emigration decisions of other individuals (figure <ns0:ref type='figure' target='#fig_9'>8b-c</ns0:ref>), until it becomes unprofitable to emigrate regardless of the relative base cost of emigration if the emigration rate is too high (figure <ns0:ref type='figure' target='#fig_9'>8d</ns0:ref>).</ns0:p></ns0:div>
<ns0:div><ns0:head>Optimal levels of mandatory emigration</ns0:head><ns0:p>Finally, we consider the potential impacts on the chikungunya epidemic on Reunion Island of coordinated emigration efforts. A mandating organization attempts to minimize overall costs, which are comprised of the cost of treatment of symptomatically infected individuals and the relocation costs of emigrating individuals. To estimate the number of emigrating susceptible individuals, we consider the difference between the total population size at equilibrium without emigration (N * = Λ 1 /µ 1 ) and the total population size at equilibrium given the population migration rate ω (this expression is given in the first equation of ( <ns0:ref type='formula' target='#formula_32'>22</ns0:ref>)); we denote this difference by N * ω .</ns0:p><ns0:p>The payoff of the emigration policy with migration rate ω is given by</ns0:p><ns0:formula xml:id='formula_44'>E(ω) = −I * − Cm N * ω ,<ns0:label>(31)</ns0:label></ns0:formula><ns0:p>where Cm = C m /C i is the cost of migration relative to the cost of infection. The graphs of this function for several values of Cm are shown in figure <ns0:ref type='figure' target='#fig_10'>9</ns0:ref>. There are three qualitatively different outcomes:</ns0:p><ns0:p>1. For very low relative migration cost ( Cm ≤ 0.00022), higher migration rates result in smallest overall costs; however, the near-optimal costs are quickly achieved by small values of migration rate (ω = 0.01)-see figure <ns0:ref type='figure' target='#fig_10'>9a</ns0:ref>.</ns0:p><ns0:p>2. There is a small interval of the relative migration cost values (0.00023 ≤ Cm ≤ 0.00025) where the optimal cost is achieved in the interior for very small values of the migration rate (ω < 0.002)-see figures 9b and 9c.</ns0:p><ns0:p>3. For all sufficiently large values of the relative migration cost ( Cm ≥ 0.00026), it is best not to allow individuals to emigrate from the island-see figure <ns0:ref type='figure' target='#fig_10'>9d</ns0:ref>.</ns0:p><ns0:p>In practice, however, the cost of emigration (such as relocation from Reunion to Mauritius) is usually comparable to or higher than the cost of the symptomatic chikungunya infection. Therefore, the scenario shown in figure <ns0:ref type='figure' target='#fig_10'>9d</ns0:ref> is the most realistic one: it is best not to allow susceptible individuals to leave the island during the outbreak.</ns0:p><ns0:p>12/17 </ns0:p></ns0:div>
<ns0:div><ns0:head>DISCUSSION</ns0:head><ns0:p>We investigated potential implications of both voluntary and mandatory intervention measures to fight the chikungunya outbreak on Reunion Island. Susceptible individuals may either prevent the infection by using insect repellent and hence reduce the frequency of mosquito bites, or leave Reunion Island and emigrate to neighboring Mauritius. We adopted a version of a previous epidemiological model of the chikungunya transmission on Reunion Island by <ns0:ref type='bibr' target='#b47'>Yakob and Clements (2013)</ns0:ref>. The epidemiological model informed the payoff functions in the game-theoretic models of individual and centralized decisions on the level of adoption of the protective measures. We found that the two protocols resulted in qualitatively different predictions concerning optimal allocations, with the latter measure creating an additional hazard for non-participants.</ns0:p><ns0:p>Voluntary participation in the two intervention measures produced opposite population-level effects.</ns0:p><ns0:p>The more susceptible individuals spray themselves with insect repellent, the less likely the infectious mosquitoes generate new human infections before they die. Consequently, higher adoption levels of insect repellent usage in the population resulted in lower basic reproduction number values for the disease.</ns0:p><ns0:p>Individuals using repellent provide (near) herd-immunity-effect benefits to the entire population. In contrast, if susceptible individuals vacated the island, then susceptible mosquitoes were more likely to bite infectious humans as a percentage of the remaining population, thus increasing the disease prevalence among mosquitoes. The remaining susceptible individuals subsequently faced an increased risk of contracting the infection from a mosquito bite. Increased migration levels resulted in drastically elevated basic reproduction number values. Thus, the impact of voluntary emigration is similar to the tragedyof-the-commons effect: while being potentially beneficial to specific individuals, it hurts the remaining islanders.</ns0:p><ns0:p>The mandated repellent usage protocol resulted in the same outcome as the voluntary (i.e., selfishly rational) compliance scenario if the cost of the preventive measure relative to the cost of the disease was too high: it was best to bypass the repellent usage altogether. But if the relative cost of protection was sufficiently low, so that repellent usage was warranted, then the two scenarios effected different outcomes.</ns0:p><ns0:p>In the voluntary compliance case, the population repellent usage fell short-albeit not by much-of the herd immunity threshold. In the mandated protocol case, reaching the herd immunity usage level and thus eradicating the disease was most effective.</ns0:p><ns0:p>That voluntary adoption of preventative measures against an infectious disease falls short of the herd immunity threshold has also been observed in other studies <ns0:ref type='bibr' target='#b22'>(Geoffard and Philipson, 1997;</ns0:ref><ns0:ref type='bibr' target='#b4'>Bauch and Earn, 2004;</ns0:ref><ns0:ref type='bibr' target='#b18'>Dorsett et al., 2016)</ns0:ref>. Yet looking at a mandated repellent usage scenario revealed that a mandatory protocol might have eliminated the epidemic if the relative cost of the preventive measure was sufficiently low.</ns0:p><ns0:p>Mandatory emigration from Reunion Island demonstrated that this preventive measure made sense for the public benefit only when the cost of relocation was significantly lower than the public cost of infection. Since this mathematical assumption is not likely to hold in practice, the model predicted that it was best to avoid migration of susceptible individuals from the island. The potentially high cost of relocating susceptible individuals away from the epidemic was not compensated by the minimal decrease in the number of infected individuals.</ns0:p><ns0:p>The qualitative differences in optimal behavior under the two alternative treatment protocols invite further examination of our model's behavior and assumptions. Both evacuation/emigration of the human populace and the use of repellent reduce the pool of potential blood hosts for the mosquito population; however, they produce contrasting effects on the force of infection. A base assumption in the model is that each insect has a consistent average number of encounters with humans over a given time span.</ns0:p><ns0:p>Repellent usage directly decreases the force of infection by deterring biting upon encounter-it is this feature of 'wasted' encounters that permits the development of herd immunity. In contrast, reduction in the size of the standing human population elevates the force of infection by increasing the number of encounters an individual human experiences. Secondarily, this results in increased prevalence of the disease in the vector-population as their blood hosts are more likely to be infected. We hypothesize that distinct protocol results depend upon the presence of (1) a distinct vector population; (2) an assumption of constant predation encounters for vectors; (3) the proportional allocation of encounters across humans;</ns0:p><ns0:p>(4) an inability of vectors to pre-judge encounters and thereby shift towards more palatable hosts; and (5) a secondary food source to support constant recruitment of new vectors. An experimental study may be warranted to verify the validity of our assumptions. We also propose a followup study to this paper that focuses specifically on the dynamic analysis of the force of infection as these assumptions are introduced or removed.</ns0:p></ns0:div>
<ns0:div><ns0:head>CONCLUSIONS</ns0:head><ns0:p>There are several additional directions in which our model can be improved. First, we are assuming that individuals possess complete information regarding the prevalence of the disease and the costs of protection relative to infection. But individuals rarely have access to the exact disease prevalence data, and hence they may only guess the relevant numbers. Second, the cost of intervention (such as using insect repellent) and the cost of the disease must be estimated individually. These costs include both direct costs such as paying for repellent or medical treatment, and indirect costs such as potentially harmful side effects of the chemicals in repellent or morbidity risks of the infection. Additionally, different individuals may have various opinions about the risk of using repellent or getting infected with chikungunya virus.</ns0:p><ns0:p>Building these uncertainties into the model should allow a broader outlook at different strategies to combat such outbreaks.</ns0:p><ns0:p>Moreover, our model assumes that the population has reached an equilibrium with respect to the disease dynamics. But reaching this equilibrium usually occurs on a different timescale compared to individual preventive actions. For example, individuals could be more likely to participate in preventive efforts when the epidemic is at its peak rather then when the disease reached the endemic state. A dynamic model where susceptible individuals inform their preventive decisions on the current state of the prevalence of the disease which, in turn, affects the dynamics of the disease transmission, should present a more realistic analysis of selfish individual decisions to prevent the infection.</ns0:p></ns0:div><ns0:figure xml:id='fig_0'><ns0:head>Figure 1 .</ns0:head><ns0:label>1</ns0:label><ns0:figDesc>Figure 1. The compartment model of chikungunya virus transmission on Reunion Island. The human population is divided into five compartments: susceptible (S), exposed (E), symptomatic infectious (I), asymptomatic infectious (I a ), and recovered (R). The mosquito population is divided into three compartments: susceptible (X), exposed (Y ), and infectious (Z). The forces of infection on human and mosquito populations, f 1 and f 2 , respectively, are population frequency-dependent functions of the state variables.</ns0:figDesc><ns0:graphic coords='5,203.77,63.78,289.51,191.53' type='bitmap' /></ns0:figure>
<ns0:figure xml:id='fig_1'><ns0:head /><ns0:label /><ns0:figDesc>)If all susceptible humans in the population adopt the same strategy r pop , then the basic reproduction number becomes a function of r pop by substituting the expression (7) for β 1 into (5). The graph of the basic reproduction number as a function of the population strategy r pop is shown in figure2. The herd immunity protection level r HI is the population protection level that reaches the threshold R 0 = 1 for disease eradication.</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_2'><ns0:head>Figure 2 .</ns0:head><ns0:label>2</ns0:label><ns0:figDesc>Figure2. The graph of the basic reproduction number as a function of the population protection level r pop . The basic reproduction number is at its maximum value-given by (5)-when no insect repellent is used by susceptible individuals (r = 0), and it becomes zero if the population employs complete protection from mosquito bites (r = 1). The threshold for disease eradication (R 0 = 1) is achieved at the herd immunity protection level r HI .</ns0:figDesc><ns0:graphic coords='7,267.87,341.70,161.28,142.03' type='bitmap' /></ns0:figure>
<ns0:figure xml:id='fig_3'><ns0:head /><ns0:label /><ns0:figDesc>11) is the force of infection, which depends on the individual protection level r and on the population protection level r pop . The individual protection level r determines the rate at which mosquitoes bite the individual β 0 1 (1 − r). The population protection level r pop affects the prevalence of the disease in the population; it determines the size of the compartment Z * via the substitution of the expression β 0 1 (1 − r pop )</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_4'><ns0:head /><ns0:label /><ns0:figDesc>17) is maximal. (Note that the equilibrium values I * and S * depend on r pop .) Here we analyze the case where the mandating organization addresses mosquito-to-human transmission by advising susceptible individuals to spray themselves with insect repellent. One may also consider an alternative scenario where the infectious individuals are using insect repellent to reduce the human-to-mosquito transmission. As before, we scale the payoff function and consider E(r pop ) = −I * − r pop CS * , (18) where C = C p /C i is the relative cost of protection. The graphs of this function for different values of C are shown in figure 4. There are two possible outcomes: (1) the susceptible individuals should adopt the repellent usage level equal to that of the herd immunity threshold r HI , leading to the eradication of the disease; or (2) no insect repellent should be used, and it is more cost-effective to treat symptomatically infected individuals only. The first outcome occurs for sufficiently low values of C (less than 0.00024), and the second outcome occurs for greater values of C (greater than 0.00024); the threshold value of C separating the two outcomes was found numerically.</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_5'><ns0:head>Figure 4 .</ns0:head><ns0:label>4</ns0:label><ns0:figDesc>Figure 4. The graphs of the expected payoff function E given by the equation (18). (a) C = 0.0001, (b) C = 0.01. The graphs show two qualitatively different outcomes. If C < 0.00024 then mandating repellent usage necessary to reach the herd immunity threshold is most effective. If C > 0.00024 then no insect repellent should be used, and all efforts should be devoted to treating infected individuals.</ns0:figDesc><ns0:graphic coords='10,162.41,63.78,372.21,149.72' type='bitmap' /></ns0:figure>
<ns0:figure xml:id='fig_6'><ns0:head>Figure 5 .</ns0:head><ns0:label>5</ns0:label><ns0:figDesc>Figure 5. The graph of R 0 as a function of ω. If susceptible individuals emigrate from Reunion Island, then the remaining inhabitants face an increased spread of the disease.</ns0:figDesc><ns0:graphic coords='11,267.87,63.78,161.30,139.51' type='bitmap' /></ns0:figure>
<ns0:figure xml:id='fig_7'><ns0:head>)Figure 6 .</ns0:head><ns0:label>6</ns0:label><ns0:figDesc>Figure 6. The graphs of the (a) number and (b) proportion of symptomatically infectious individuals in the population as functions of the emigration rate ω. Increased migration levels result in fewer infectious individuals overall but a greater proportion of infectious individuals in the population.</ns0:figDesc><ns0:graphic coords='12,162.41,187.58,372.20,146.65' type='bitmap' /></ns0:figure>
<ns0:figure xml:id='fig_8'><ns0:head>Figure 7 .</ns0:head><ns0:label>7</ns0:label><ns0:figDesc>Figure 7. The probability of symptomatic chikungunya infection for a susceptible individual on Reunion Island is increasing with the emigration rate ω.</ns0:figDesc><ns0:graphic coords='13,267.87,63.78,161.29,134.97' type='bitmap' /></ns0:figure>
<ns0:figure xml:id='fig_9'><ns0:head>Figure 8 .</ns0:head><ns0:label>8</ns0:label><ns0:figDesc>Figure 8. The regions in the ( Cb , ω)-parameter space showing whether a focal susceptible individual should emigrate to Mauritius or remain on Reunion. Color code: white-emigrate, gray-stay. (a) Cs = 0, (b) Cs = 0.1, (c) Cs = 1, (d) Cs = 10.</ns0:figDesc><ns0:graphic coords='14,162.41,63.78,372.22,295.08' type='bitmap' /></ns0:figure>
<ns0:figure xml:id='fig_10'><ns0:head>Figure 9 .</ns0:head><ns0:label>9</ns0:label><ns0:figDesc>Figure 9. The overall cost of the mandated emigration policy as a function of the migration rate ω. (a) Cm = 0.00022, (b) Cm = 0.00023, (c) Cm = 0.00024, (d) Cm = 0.00026.</ns0:figDesc><ns0:graphic coords='15,162.41,63.78,372.22,292.17' type='bitmap' /></ns0:figure>
<ns0:figure type='table' xml:id='tab_0'><ns0:head>Table 1 .</ns0:head><ns0:label>1</ns0:label><ns0:figDesc>Summary of the model parameters</ns0:figDesc><ns0:table><ns0:row><ns0:cell cols='2'>Symbol Meaning</ns0:cell><ns0:cell>Value</ns0:cell><ns0:cell>Source</ns0:cell></ns0:row><ns0:row><ns0:cell>β 0 1 β 2</ns0:cell><ns0:cell>Mosquito-to-human transmission Human-to-mosquito transmission</ns0:cell><ns0:cell>0.37 0.37</ns0:cell><ns0:cell>Dumont et al. (2008) Dumont et al. (2008)</ns0:cell></ns0:row><ns0:row><ns0:cell>γ</ns0:cell><ns0:cell>Human recovery rate</ns0:cell><ns0:cell>0.14</ns0:cell><ns0:cell>Yakob and Clements (2013)</ns0:cell></ns0:row><ns0:row><ns0:cell>Λ 1</ns0:cell><ns0:cell>Human birth rate</ns0:cell><ns0:cell>3.58</ns0:cell><ns0:cell>Assumed</ns0:cell></ns0:row><ns0:row><ns0:cell>Λ 2 λ 1</ns0:cell><ns0:cell>Mosquito birth rate Rate of humans becoming infectious</ns0:cell><ns0:cell>4.76 × 10 3 0.5</ns0:cell><ns0:cell>Assumed Yakob and Clements (2013)</ns0:cell></ns0:row><ns0:row><ns0:cell>λ 2</ns0:cell><ns0:cell>Rate of mosquitoes becoming infectious</ns0:cell><ns0:cell>0.5</ns0:cell><ns0:cell>Yakob and Clements (2013)</ns0:cell></ns0:row><ns0:row><ns0:cell>µ 1 µ 2</ns0:cell><ns0:cell>Human natural death rate Mosquito natural death rate</ns0:cell><ns0:cell cols='2'>3.58 × 10 −5 Assumed 0.05 Yakob and Clements (2013)</ns0:cell></ns0:row><ns0:row><ns0:cell>φ</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row></ns0:table></ns0:figure>
<ns0:note place='foot'>PeerJ reviewing PDF | (2019:08:40674:2:0:NEW 12 May 2020)</ns0:note>
<ns0:note place='foot' n='3'>/17 PeerJ reviewing PDF | (2019:08:40674:2:0:NEW 12 May 2020)</ns0:note>
<ns0:note place='foot' n='4'>/17 PeerJ reviewing PDF | (2019:08:40674:2:0:NEW 12 May 2020)Manuscript to be reviewed</ns0:note>
<ns0:note place='foot' n='9'>/17 PeerJ reviewing PDF | (2019:08:40674:2:0:NEW 12 May 2020)</ns0:note>
<ns0:note place='foot' n='10'>/17 PeerJ reviewing PDF | (2019:08:40674:2:0:NEW 12 May 2020)</ns0:note>
<ns0:note place='foot' n='223'>should emigrate when the relative cost of doing so is less than the probability of getting infected:</ns0:note>
<ns0:note place='foot' n='13'>/17 PeerJ reviewing PDF | (2019:08:40674:2:0:NEW 12 May 2020)</ns0:note>
<ns0:note place='foot' n='14'>/17 PeerJ reviewing PDF | (2019:08:40674:2:0:NEW 12 May 2020)</ns0:note>
<ns0:note place='foot' n='17'>/17 PeerJ reviewing PDF | (2019:08:40674:2:0:NEW 12 May 2020)Manuscript to be reviewed</ns0:note>
</ns0:body>
" | "We would like to thank the reviewers for taking the time to again carefully read the paper
and provide constructive comments and suggestions. We applied the PeerJ LaTeX class
style to the manuscript in this minor revision, and we used this revision opportunity to
update the bibliography to include several recent works.
Response to the Reviewer 1
Although we believe that the previous presentations of our theoretical/modeling findings
were well within the norm of the literature, we recognize that some readers such as the
reviewer may have been confused that we were offering direct policy recommendations
rather than providing insight into emerging dynamic properties of a model. To the extent
that we can allay the reviewer's concern, we note that we have made the following
changes to our manuscript:
1. The sentence “An underlying assumption about the conservation of human-vector
encounter rates in mosquito biting behavior informs our conclusions and may
warrant additional experimental verification” has been appended to the abstract
(lines 31–32).
2. The last sentence of the Introduction “We find that the latter protocol has a
paradoxically worsening of outcomes for the non-participating population” has
been changed to “We find that the latter protocol has a paradoxically worsening of
outcomes for the non-participating population potentially as a consequence of the
non-responsiveness in the mosquitoes feeding behavior to decreases in the human
population relative to other blood sources” (lines 105–107).
3. We added the sentence “We assume that mosquitoes have a consistent average
number of encounters with humans over a given time span, and that repellent
usage directly decreases the force of infection by deterring biting upon encounter”
to the third paragraph of the Methods, which discusses the force of infection in the
epidemiological model (lines 132–134).
4. We added the sentence “An experimental study may be warranted to verify the
validity of our assumptions” to the last paragraph of the Discussion (lines
308–309).
We hope that these changes help clear any potential confusion regarding our conclusions,
and that the manuscript is acceptable for publication in PeerJ in its current form.
" | Here is a paper. Please give your review comments after reading it. |
9,797 | "<ns0:abstract xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:p>Study Objective. To assess the adequacy of clinical information with reference to the Ottawa Ankle Rules in X-ray referrals for adults with traumatic ankle injury in the emergency department of a South Australian tertiary hospital and report upon referring trends between emergency department clinicians. Methods. A retrospective clinical audit of adult ankle X-ray referrals in the emergency department was conducted. Eligible referrals were screened for their adherence to the OAR, patient details, clinical history and referrer. A logistic regression was used to determine the influence of these factors on the likelihood of being referred for X-rays despite not meeting the OAR criteria. Sensitivity, specificity, positive and negative likelihood ratios and their associated confidence intervals were calculated to assess the diagnostic accuracy of the OAR for those referred. Results.</ns0:p><ns0:p>Out of the 262 eligible referrals, 163 were deemed to have met the criteria for the OAR.</ns0:p><ns0:p>Physiotherapists showed the highest OAR compliance of 77.3% and were the most accurate in their use of the rules, with a sensitivity of 0.86. Medical officers, registrars and interns were 2.5 x more likely to still refer a patient for X-ray if they did not meet the OAR criteria, compared to physiotherapists as the baseline. Patient age, duration of injury etc.</ns0:p><ns0:p>were not significantly associated with likelihood of referral (even when they did not meet OAR criteria). The overall sensitivity, specificity, positive and negative likelihood ratios of the OAR were 0.59 (95% CI: 0.47 -0.71), 0.37 (95% CI: 0.30 -0.44), 0.93 (95% CI: 0.76 -1.16) and 1.10 (95% CI: 0.82 -1.48) respectively. Conclusion. The results of this audit demonstrated poor sensitivity and moderate compliance by referrers with the rule.</ns0:p><ns0:p>Reasonable evidence exists for the implementation of individual and/or institutional-based change strategies to improve clinician compliance and accuracy with use of the OAR.</ns0:p></ns0:div>
</ns0:abstract>
<ns0:body xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:head>Background</ns0:head><ns0:p>Ankle and foot injuries are the most frequently presenting musculoskeletal injuries in Australian emergency departments (ED) <ns0:ref type='bibr' target='#b16'>(Strudwick et al. 2018)</ns0:ref>. Over 46,000 presentations occurred in South Australia in the year 2017-18 (Australian Institute of Health and Welfare 2018). Acute ankle trauma is often a result of inversion injury, commonly causing a sprain or disruption of ligaments. Ankle fractures, however, are more likely observed with blunt ankle trauma, such as those associated with sporting injuries or motor vehicle related accidents <ns0:ref type='bibr' target='#b10'>(Goergen et al. 2015)</ns0:ref>. Acute ankle injuries are commonly diagnosed by subjective patient history, objective physical examination and/or using radiographic imaging <ns0:ref type='bibr' target='#b10'>(Goergen et al. 2015)</ns0:ref>.</ns0:p><ns0:p>Radiographic imaging is one of the most routinely used assessment methods for ankle trauma <ns0:ref type='bibr' target='#b2'>(Beckenkamp et al. 2017)</ns0:ref>. Despite the benefits of this modality, the continuous referral of patients for imaging of ankle trauma often leads to increased waiting times in the emergency department (ED), contributes to rising healthcare costs and unnecessarily exposes patients to ionising radiation <ns0:ref type='bibr' target='#b2'>(Beckenkamp et al. 2017)</ns0:ref>.. Consequently, it is imperative that the management of ankle trauma within EDs is optimised to facilitate best management, minimise costs and improve the quality of care provided to patients <ns0:ref type='bibr' target='#b16'>(Strudwick et al. 2018)</ns0:ref>.</ns0:p><ns0:p>The Ottawa Ankle Rules (OAR) (Figure <ns0:ref type='figure'>1</ns0:ref>) are part of a clinical decision-making tool to help clinicians accurately rule out ankle fractures, potentially precluding the need for diagnostic X-ray imaging <ns0:ref type='bibr' target='#b1'>(Bachmann et al. 2003)</ns0:ref>. The OAR were introduced by <ns0:ref type='bibr' target='#b15'>Stiell et al. (1992)</ns0:ref> and have proven to be a highly accurate tool, with good sensitivity and demonstrated ability to reduce imaging requests and waiting times in the ED <ns0:ref type='bibr' target='#b6'>(Cheng, Varma & Smith 2016;</ns0:ref><ns0:ref type='bibr' target='#b16'>Strudwick et al. 2018</ns0:ref>).</ns0:p><ns0:p>The OAR have been widely applied in many countries <ns0:ref type='bibr' target='#b9'>(Daþ, Temiz & Çevik 2016)</ns0:ref> and there have been active dissemination and education strategies developed to encourage clinicians to incorporate them into practice <ns0:ref type='bibr' target='#b5'>(Cameron & Naylor 1999)</ns0:ref>. However, <ns0:ref type='bibr' target='#b5'>Cameron and Naylor (1999)</ns0:ref> report that OAR have not been universally adopted into practice. Potentially due to the convenience of referring patients with ankle trauma for imaging, or due to practitioner concerns around litigation <ns0:ref type='bibr' target='#b14'>(Pires et al. 2014)</ns0:ref>. Evidence suggests that in order to influence changes in clinical behaviour, implementing guidelines on a local level is paramount <ns0:ref type='bibr' target='#b5'>(Cameron & Naylor 1999)</ns0:ref>.</ns0:p><ns0:p>A cross-sectional study evaluated the international adoption of the OAR and reported significant differences in the use of the rules by physicians in five countries. More than 80% of physicians in the United Kingdom and Canada that were aware of the OAR, reported using them frequently. However, the usage was far lower for physicians in the United States (31%), France (31%) and Spain (9%) <ns0:ref type='bibr' target='#b11'>(Graham et al. 2001)</ns0:ref>.</ns0:p><ns0:p>Evidence of uptake in the Australian context is limited. An Adelaide-based validation study found the rules were correctly applied by both junior and senior Emergency Department (ED) physicians; 47 of 54 physicians (87%) correctly interpreted the requirement for radiography in 327 patients (97.3%) <ns0:ref type='bibr' target='#b4'>(Broomhead & Stuart 2003)</ns0:ref>. In a separate study, analysis of the use of the OAR by nurse practitioners, triage nurses and other medical staff identified a gap between evidence and practice; with reasons for ordering radiographs including an obligation to the patient, streamlining patient flow through the ED and wanting to avoid patients 're-presenting' with the injury <ns0:ref type='bibr' target='#b3'>(Bessen et al. 2009</ns0:ref>). = Furthermore, a retrospective review of a major metropolitan ED in Victoria showed positive results, with a 87.9% compliance rate with the OAR <ns0:ref type='bibr' target='#b6'>(Cheng, Varma & Smith 2016)</ns0:ref>.</ns0:p><ns0:p>To the best of the authors' knowledge, however, the adequacy of clinical information on adult ankle radiograph requests with reference to the OAR has not been recently investigated in a South Australian context. Therefore, the aims of this study are to:</ns0:p><ns0:formula xml:id='formula_0'>(i)</ns0:formula><ns0:p>assess the current usage of the OAR in ruling out ankle fractures in a major South Australian metropolitan tertiary care emergency department, (ii) evaluate the current concordance rate of scoring with positive findings on radiography, and (iii) report upon referring trends between professions, including consultants, registrars/medical officers/interns, physiotherapists and nurse practitioners.</ns0:p><ns0:p>These research findings can provide the first step into further research on the awareness, dissemination and uptake of the OAR in Australia.</ns0:p></ns0:div>
<ns0:div><ns0:head>Materials & Methods</ns0:head><ns0:p>A retrospective clinical audit was performed on X-ray referrals for adult patients presenting with acute ankle injuries to the emergency department of Flinders Medical Centre (South Australia). Approval was granted by the University of South Australia Human Research Ethics Committee (Approval number: 202798) and waived by the Human Research Ethics Committee of Flinders Medical Centre. A confidentiality report was signed by the data collector and all patient information was de-identified. The requirement for obtaining informed consent was waived Referrals for imaging were excluded from the study if participants were under 18 years of age, returning for follow-up imaging, presenting with a multi-trauma, had an injury more than two weeks old, had cognitive impairment (including intoxication), had an incomplete examination, had an inflammatory, neurological or musculoskeletal condition that impeded ankle joint function, the referral gave no fracture subscription (i.e. query foreign body, prosthesis position), or requested imaging for a known dislocation or post-reduction of a dislocation. This is consistent with previous literature <ns0:ref type='bibr'>(Daþ, Temiz & Çevik 2016, p.362;</ns0:ref><ns0:ref type='bibr'>Pijnenburg et al. 2002, p.601)</ns0:ref>.</ns0:p><ns0:p>The hospital's Picture Archiving and Communications (PACS) system was reviewed in January 2020 to extract relevant ankle X-ray referrals from March 2019 to January 2020. The referrals were screened for their adherence to the OAR and patient details. Data extracted included patients' date of birth, clinical history and the type of referring health professional (i.e. consultant, registrar/medical officer (MO)/intern, physiotherapist or nurse). A referral was deemed to have met the criteria for the OAR if it (i) indicated pain in the malleolar zone and (ii) indicated pain or bone tenderness of the posterior distal tibia/medial malleolus tip or the posterior distal fibula/lateral malleolus tip or an inability to weight bear for 4 steps both immediately and in the ED <ns0:ref type='bibr' target='#b17'>(Tiemstra 2012)</ns0:ref>. All data was entered into Excel (Microsoft Corporation 2016). The radiologist's report was also screened for the presence/absence of an acute fracture.</ns0:p><ns0:p>Descriptive and inferential statistics were evaluated using SPSS software (IBM SPSS Statistics for Windows, Version 26). Descriptive analyses were conducted for patient demographics (i.e. mean age, duration of injury, etc.). Sensitivity, specificity, positive and negative likelihood ratios were calculated to determine the diagnostic accuracy of the OAR <ns0:ref type='bibr' target='#b12'>(Lowry 2004</ns0:ref>) on referrals received. A logistic regression was also performed to assess the impact of other potential decision-making factors when referring for X-rays in patients who did not meet the OAR criteria.</ns0:p><ns0:p>Outcomes were considered statistically significant if p < 0.05.</ns0:p></ns0:div>
<ns0:div><ns0:head>Results</ns0:head><ns0:p>The search yielded 750 referrals. Of the 750 referrals screened, 488 were excluded (Figure <ns0:ref type='figure' target='#fig_0'>2</ns0:ref>). The characteristics of the 262 eligible referrals are summarised in Table <ns0:ref type='table' target='#tab_0'>1</ns0:ref>.</ns0:p><ns0:p>Participants' mean age was 38 (SD 13.8) and the ages ranged from 18-97 years. Males and females were equally represented. Over three-quarters of referrals for imaging did not specify the duration of injury and rolling or twisting injuries were the most commonly recorded mechanism of injury (42.4%), with inversion/eversion injuries also frequently reported (32.4%). Registrars, MOs and interns were grouped into one category due to an inability to differentiate between them on referrals. As such, they were the largest referring profession (47.3%), followed by nurses (30.9%).</ns0:p><ns0:p>Out of the 262 eligible referrals for imaging, 163 met the criteria for the OAR (OAR +), while 99 did not (OAR-). This suggests that 38% (99 of 262) referrals for imaging may not have been necessary. Table <ns0:ref type='table' target='#tab_2'>2</ns0:ref> summarises the characteristics of the two groups.</ns0:p><ns0:p>Registrars, MOs and interns were the most frequently referring profession, while nurses constituted the second-highest reporting profession in both groups. Physiotherapists were the most compliant profession in their use of the OAR criteria (77.3%). Nurses and consultants showed 65.4% and 61.5% compliance respectively, while registrars/MOs/interns showed the least compliance at 54.8%.</ns0:p><ns0:p>A logistic regression was performed to assess the impact of clinical factors on referring for Xrays in patients who did not meet the OAR criteria. The model contained several independent variables (duration of injury, mechanism of injury, referring profession and participant age). The full model containing all predictors was statistically significant, χ 2 (9, N = 262) = 16.9, p = 0.05, indicating the model was able to distinguish between OAR+ and OAR-referrals. As shown in Table <ns0:ref type='table'>3</ns0:ref>, the only statistically significant independent variable that contributed to the model was registrars, MO's and interns, recording an odds ratio of 2.48 (p = 0.026). This indicates that this group of professions was 2.5 x more likely to still refer a patient for X-ray if they did not meet the criteria, compared to physiotherapists as the baseline.</ns0:p><ns0:p>The number of referrals for imaging that resulted in identified ankle fractures are summarised in Table <ns0:ref type='table'>4</ns0:ref>.</ns0:p><ns0:p>Of the 163 OAR+ referrals, 44 (27.0%) resulted in a fracture or potential fracture as per the radiologist's report, while 119 (73.0%) reported no fracture. Of the 99 referrals that were OAR-, 30 (30.3 %) reported a fracture/potential fracture, while 69 (69.7%) did not report a fracture.</ns0:p><ns0:p>Overall, the sensitivity and specificity of the OAR were 0.59 (95% CI: 0.47 -0.71) and 0.37 (95% CI: 0.30 -0.44) respectively. The positive (LR+) and negative (LR-) likelihood ratios were 0.93 (95% CI: 0.76 -1.16) and 1.10 (95% CI: 0.82 -1.48) respectively.</ns0:p><ns0:p>The referrals were aggregated according to the referring clinician to assess the clinician-specific accuracy of the OAR. The sensitivity, specificity, LR+ and LR-of the OAR when applied by each emergency clinician were calculated along with 95% confidence intervals, as were true positives (TP), false positives (FP), true negatives (TN) and false negatives (FN). This information is displayed in Table <ns0:ref type='table' target='#tab_4'>5</ns0:ref>.</ns0:p><ns0:p>Physiotherapists were the most accurately reporting profession (sensitivity 0.86), followed by consultants (sensitivity 0.75).</ns0:p></ns0:div>
<ns0:div><ns0:head>Discussion</ns0:head><ns0:p>The OAR have been validated on an international scale and are regarded as a highly sensitive clinical decision tool with the capacity to reduce the number of unnecessary ankle radiographs ordered <ns0:ref type='bibr' target='#b2'>(Beckenkamp et al. 2017)</ns0:ref>. While the application of the OAR by medical doctors in an emergency setting has been previously validated, <ns0:ref type='bibr' target='#b2'>Beckenkamp et al. (2017)</ns0:ref> highlight the importance of the accurate application of the rules by triage nurses and physiotherapists, as they play an increasingly important role in the outflow of patients within EDs. The uptake of the OAR and application of the rules by different health professionals has been sparsely validated in the Australian context.</ns0:p><ns0:p>This clinical audit assessed the current use of the OAR in a major South Australian tertiary hospital. Based on our findings, physiotherapists showed the highest OAR compliance of 77.3% (n=34) and were the most accurate in their use of the rules, with the highest sensitivity of 0.86. Consultants, although not the most compliant profession (61.5%), displayed a reasonably high sensitivity of 0.75. They also referred the least number of patients (n = 13), while registrars/MOs/interns referred the largest proportion of patients (n= 124). We hypothesise that this may be due to the distribution of highly complex cases (e.g. motor vehicle accidents) to consultants, while junior doctors are responsible for the triage of a wider variety of less complex cases (e.g. rolling, twisting injuries). This may have biased our results. Nurses and registrars/MOs/interns were the least accurate in their use of the rules, with a sensitivity of 0.68 and 0.43 respectively, which may be due to the inexperience of junior physicians interpreting and applying the OAR.</ns0:p><ns0:p>The most recent prospective validation of the OAR in a South Australian ED, conducted by <ns0:ref type='bibr' target='#b4'>Broomhead and Stuart (2003)</ns0:ref>, reported a sensitivity of 100% (95% CI: 77-100) and a specificity of 15.8% (95% CI: 11-21) for ankle fractures. Given our investigation was a retrospective and not a prospective study, the sensitivity and specificity of the rules could not be calculated based on the follow-up of participants that did not receive X-rays. A bias may exist among referrals that were deemed non-compliant (i.e. did not meet the criteria for the OAR), as there is the possibility that clinicians used the OAR during triage and whilst x-rays were not indicated still made a request based on other clinical reasoning concerns. Previous research suggest that an inability to weight-bear is the most important factor influencing referral <ns0:ref type='bibr' target='#b14'>(Pires et al. 2014)</ns0:ref>. We found that 'bone tenderness at the posterior distal fibula or tip of lateral malleolus' was the most common OAR item reported across referrals (44.1%), however, we did not analyse the factors independently.</ns0:p><ns0:p>Despite the large numbers of referrals deemed not compliant to the OAR, our findings did not suggest other possible clinical indicators such as patient age, nor duration or mechanism of injury affected this. The only influencing factor for referral when independent of OAR as an indicator was the referring profession. Results of the logistic regression found that medical officers, registrars and interns were 2.5 x more likely to still refer a patient for X-ray if they did not meet the OAR criteria, compared to physiotherapists as the baseline. As discussed by <ns0:ref type='bibr' target='#b14'>Pires et al. (2014)</ns0:ref>, this may be due to the convenience of requesting imaging for ankle trauma and/or the fear of litigation. In these instances, the indicator for X-rays is unclear and decreases the accuracy with which we can report on the sensitivity and specificity of the rules. A retrospective study, however, decreases the influence of the Hawthorne effect around application of the OAR. The influence of subjective examiner perception on the referral for X-rays should also be noted.</ns0:p><ns0:p>For example, the subjective examination of pain can vary between examiners and differences in clinical skills and experience may impact the perception of fracture occurrence <ns0:ref type='bibr' target='#b14'>(Pires et al. 2014)</ns0:ref>. It is therefore likely that differences in examiner perception during the clinical examination influenced the referral for X-rays.</ns0:p><ns0:p>Within the limits of this study, this audit provides a good summary of the use of the OAR by different emergency clinicians. It provides a starting point for potential further study into the reasons for or against the use of the OAR, particularly amongst Registrars/MO's/Intern, as well as the diagnostic accuracy of the OAR within a South Australian context. Our study did not involve teaching the correct use and interpretation of the OAR and hence, solely evaluated their individual ability to correctly apply the rules at baseline to patients presenting with ankle trauma.</ns0:p><ns0:p>To improve accuracy and compliance rates, a thorough investigation into the current knowledge and application of the OAR is recommended. Prospective validation of the OAR could follow a small-scale change strategy similar to a study conducted by <ns0:ref type='bibr' target='#b3'>Bessen et al. (2009)</ns0:ref>; this involved educating clinicians in the use of the OAR and a specific ankle radiography request form, each designed to target barriers in the use of the OAR at an individual and institutional level. As expected, a significant change in practice was noted, with nurses demonstrating the greatest uptake in the OAR <ns0:ref type='bibr' target='#b3'>(Bessen et al. 2009)</ns0:ref>. <ns0:ref type='bibr' target='#b3'>Bessen et al. (2009)</ns0:ref> also trialled the concept of 'gatekeeping' among ED radiographers, who were empowered to reject request forms that demonstrated non-compliance with the OAR. These are important considerations in improving the compliance and hence accurate use of the OAR within the ED.</ns0:p></ns0:div>
<ns0:div><ns0:head>Conclusions</ns0:head><ns0:p>In conclusion, although the OAR is internationally regarded as highly accurate clinical decision tool, its local uptake has been varied. The results of this audit demonstrate moderate compliance and poor sensitivity of the rule. Despite limitations in the reporting of the diagnostic accuracy of the rule, this audit demonstrates reasonable evidence that individual and/or institutional-based change strategies are warranted to improve clinician compliance in the use of the OAR in the local tertiary care emergency department setting. Doing so will improve the implementation of the rule and reduce the frequency of radiography requests to optimise patient flow through the emergency department.</ns0:p></ns0:div>
<ns0:div><ns0:head>Figure 1</ns0:head><ns0:p>The Ottowa Ankle Rules</ns0:p><ns0:p>The Ottowa Ankle Rules for X-ray referral</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:07:51585:1:1:NEW 18 Sep 2020)</ns0:p><ns0:p>Manuscript to be reviewed Manuscript to be reviewed Manuscript to be reviewed </ns0:p></ns0:div><ns0:figure xml:id='fig_0'><ns0:head>Figure 2</ns0:head><ns0:label>2</ns0:label><ns0:figDesc>Figure 2</ns0:figDesc></ns0:figure>
<ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='10,42.52,204.37,525.00,265.50' type='bitmap' /></ns0:figure>
<ns0:figure type='table' xml:id='tab_0'><ns0:head>Table 1 (on next page)</ns0:head><ns0:label>1</ns0:label><ns0:figDesc /><ns0:table><ns0:row><ns0:cell>Referred patient characteristics</ns0:cell></ns0:row><ns0:row><ns0:cell>A table outlining numbers and percentage for referral characteristics</ns0:cell></ns0:row></ns0:table><ns0:note>PeerJ reviewing PDF | (2020:07:51585:1:1:NEW 18 Sep 2020)Manuscript to be reviewed 1</ns0:note></ns0:figure>
<ns0:figure type='table' xml:id='tab_1'><ns0:head>Table 1 :</ns0:head><ns0:label>1</ns0:label><ns0:figDesc>Characteristics of the of 262 eligible patients with ankle trauma</ns0:figDesc><ns0:table><ns0:row><ns0:cell>Characteristics</ns0:cell><ns0:cell>Number</ns0:cell><ns0:cell>%</ns0:cell></ns0:row><ns0:row><ns0:cell>Mean age in years (SD)</ns0:cell><ns0:cell>38 (± 18.3)</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>Gender</ns0:cell><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>-Male</ns0:cell><ns0:cell>132</ns0:cell><ns0:cell>50.4</ns0:cell></ns0:row><ns0:row><ns0:cell>-Female</ns0:cell><ns0:cell>130</ns0:cell><ns0:cell>49.6</ns0:cell></ns0:row><ns0:row><ns0:cell>Duration of Injury</ns0:cell><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>-None specified</ns0:cell><ns0:cell>200</ns0:cell><ns0:cell>76.3</ns0:cell></ns0:row><ns0:row><ns0:cell>-< 24 hours</ns0:cell><ns0:cell>24</ns0:cell><ns0:cell>9.2</ns0:cell></ns0:row><ns0:row><ns0:cell>-1-7 days</ns0:cell><ns0:cell>34</ns0:cell><ns0:cell>13.0</ns0:cell></ns0:row><ns0:row><ns0:cell>-7-14 days</ns0:cell><ns0:cell>4</ns0:cell><ns0:cell>1.5</ns0:cell></ns0:row><ns0:row><ns0:cell>Mechanism of injury (as stated on</ns0:cell><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>referral)</ns0:cell><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>-Fall</ns0:cell><ns0:cell>40</ns0:cell><ns0:cell>15.3</ns0:cell></ns0:row><ns0:row><ns0:cell>-Motor vehicle accident (MVA)</ns0:cell><ns0:cell>3</ns0:cell><ns0:cell>1.1</ns0:cell></ns0:row><ns0:row><ns0:cell>-Sporting injury</ns0:cell><ns0:cell>23</ns0:cell><ns0:cell>8.8</ns0:cell></ns0:row><ns0:row><ns0:cell>-Inversion/eversion injury</ns0:cell><ns0:cell>85</ns0:cell><ns0:cell>32.4</ns0:cell></ns0:row><ns0:row><ns0:cell>-Other (i.e. rolling, twisting)</ns0:cell><ns0:cell>111</ns0:cell><ns0:cell>42.4</ns0:cell></ns0:row><ns0:row><ns0:cell>Referring profession</ns0:cell><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>-Nurse</ns0:cell><ns0:cell>81</ns0:cell><ns0:cell>30.9</ns0:cell></ns0:row><ns0:row><ns0:cell>-Consultant</ns0:cell><ns0:cell>13</ns0:cell><ns0:cell>5.0</ns0:cell></ns0:row><ns0:row><ns0:cell>-Registrar/MO/Intern</ns0:cell><ns0:cell>124</ns0:cell><ns0:cell>47.3</ns0:cell></ns0:row><ns0:row><ns0:cell>-Physiotherapist</ns0:cell><ns0:cell>44</ns0:cell><ns0:cell>16.8</ns0:cell></ns0:row></ns0:table><ns0:note>2PeerJ reviewing PDF | (2020:07:51585:1:1:NEW 18 Sep 2020)</ns0:note></ns0:figure>
<ns0:figure type='table' xml:id='tab_2'><ns0:head>Table 2 (on next page)</ns0:head><ns0:label>2</ns0:label><ns0:figDesc>Characteristics of patients meeting or not meeting OAR criteriaNumbers and percentages of characteristics for referrals meeting or not meeting OAR criteria</ns0:figDesc><ns0:table /><ns0:note>PeerJ reviewing PDF | (2020:07:51585:1:1:NEW 18 Sep 2020)Manuscript to be reviewed 1</ns0:note></ns0:figure>
<ns0:figure type='table' xml:id='tab_3'><ns0:head>Table 2 :</ns0:head><ns0:label>2</ns0:label><ns0:figDesc>Characteristics of OAR positive (+) and negative (-) referrals</ns0:figDesc><ns0:table><ns0:row><ns0:cell>OAR +</ns0:cell><ns0:cell>OAR -</ns0:cell></ns0:row></ns0:table></ns0:figure>
<ns0:figure type='table' xml:id='tab_4'><ns0:head>Table 5 :</ns0:head><ns0:label>5</ns0:label><ns0:figDesc>Diagnostic accuracy of OAR when applied by different health professionals</ns0:figDesc><ns0:table><ns0:row><ns0:cell /><ns0:cell>Nurse</ns0:cell><ns0:cell>Consultant</ns0:cell><ns0:cell>Registrar/MO/Intern</ns0:cell><ns0:cell>Physiotherapist</ns0:cell></ns0:row><ns0:row><ns0:cell>No. of</ns0:cell><ns0:cell>81</ns0:cell><ns0:cell>13</ns0:cell><ns0:cell>124</ns0:cell><ns0:cell>44</ns0:cell></ns0:row><ns0:row><ns0:cell>patients</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>TP</ns0:cell><ns0:cell>13</ns0:cell><ns0:cell>3</ns0:cell><ns0:cell>16</ns0:cell><ns0:cell>12</ns0:cell></ns0:row><ns0:row><ns0:cell>FP</ns0:cell><ns0:cell>40</ns0:cell><ns0:cell>5</ns0:cell><ns0:cell>52</ns0:cell><ns0:cell>22</ns0:cell></ns0:row><ns0:row><ns0:cell>TN</ns0:cell><ns0:cell>22</ns0:cell><ns0:cell>4</ns0:cell><ns0:cell>35</ns0:cell><ns0:cell>8</ns0:cell></ns0:row><ns0:row><ns0:cell>FN</ns0:cell><ns0:cell>6</ns0:cell><ns0:cell>1</ns0:cell><ns0:cell>21</ns0:cell><ns0:cell>2</ns0:cell></ns0:row><ns0:row><ns0:cell cols='3'>Sensitivity 0.68 (0.43 -0.86) 0.75 (0.22 -0.99)</ns0:cell><ns0:cell>0.43 (0.28 -0.60)</ns0:cell><ns0:cell>0.86 (0.56 -0.97)</ns0:cell></ns0:row><ns0:row><ns0:cell cols='3'>Specificity 0.35 (0.24 -0.49) 0.44 (0.15 -0.77)</ns0:cell><ns0:cell>0.40 (0.30 -0.51)</ns0:cell><ns0:cell>0.27 (0.13 -0.46)</ns0:cell></ns0:row><ns0:row><ns0:cell>LR+</ns0:cell><ns0:cell cols='2'>1.06 (0.74 -1.52) 1.35 (0.60 -3.04)</ns0:cell><ns0:cell>0.72 (0.48 -1.09)</ns0:cell><ns0:cell>1.17 (0.86 -1.58)</ns0:cell></ns0:row><ns0:row><ns0:cell>LR-</ns0:cell><ns0:cell cols='2'>0.89 (0.43 -1.82) 0.56 (0.078 -4.03)</ns0:cell><ns0:cell>1.41 (1.02 -1.93)</ns0:cell><ns0:cell>0.54 (0.12 -2.31)</ns0:cell></ns0:row></ns0:table></ns0:figure>
<ns0:note place='foot'>PeerJ reviewing PDF | (2020:07:51585:1:1:NEW 18 Sep 2020)</ns0:note>
</ns0:body>
" | "
Dear Editor,
We thank the reviewers for their comments on the manuscript and have edited the manuscript to address these comments.
We believe the manuscript is now suitable for publication in PeerJ.
Kind regards,
Yolanda Gomes
3rd year Medical Imaging Student, University of South Australia
On behalf of all authors.
Reviewer 1 (Eugen Divjak)
Basic reporting
The article presents the clinical problem, methods used and results of the study clearly. Professional English is used throughout the text. Authors provided abundant data regarding the background with corresponding references. Article is well-structured with appropriate use of figures and tables. Raw data was available for the review.
Results of the study are clearly presented and discussed in context of previous knowledge and the goal of the study.
Experimental design
Clinical problem of the article is well known in scientific community, but lack of data regarding local, Southern Australian region, can justify the research. It is important for both quality control on a local level as well as for comparison of results on a global scale to gather and publish information even regarding previously researched clinical problems. Experimental design is of a sufficient research integrity and reproducible.
Validity of the findings
The Authors admit that this is a novel research on a local level only, clearly underlying that there are some weaknesses of the study. Results seem statistically sound and are interesting to a medical professional. Discussion is well-written summarizing the results and possible impacts of the study.
Comments for the Author
The article offers a new insight to an old problem, offering possible reasons why Ottawa Ankle Rules application in Emergency department setting differs between medical professionals who refer patients for radiography. While not assessing the accuracy of the test itself, but rather in a context of different medical professions, this is still important information for scientific community.
Reviewer 2 (Robinson Pires)
Basic reporting
The manuscript is clear and well written. References, figure, and tables are adequate. Results, although not entirely new, are relevant.
Experimental design
The study is in accordance with the scope of the journal. Research question and hypothesis were adequately addressed, following all the required issues in terms of ethical standard. Methods were thoroughly described.
Validity of the findings
Although the findings are not entirely new, the results are interesting and worthy of disclosure. The manuscript results are adequately presented and compared with previous similar studies.
Comments for the Author
Dear authors, I appreciate the opportunity to review this interesting manuscript.
The topic is relevant in terms of cost-effectiveness regarding ankle trauma.
The article is well written and the methodology is adequate and thoroughly presented.
Ethical standards were also adequately addressed.
However, I have two minor suggestions aiming to improve the manuscript quality:
1-Background section is excessively long. Make it more straightforward and compare the previously published studies with your findings in the discussion section.
2- Discuss about the relationship between Ottawa Ankle Rules and subjective examinador perception to evaluate radiograph necessity following ankle sprain. Furthermore, some discussion should be presented in terms of the most important item of OAR to predict the presence of a foot/ankle fracture.
1-Paragraphs 1 and 2 have been edited to make them more concise. Paragraph 6 has been shortened to briefly summarise the limited OAR uptake in Australia and help justify the need for further research at a local level.
2-The potential effects of subjective examiner perception have been briefly summarised in the discussion section (lines 214-219).
We have briefly commented on the most frequently occurring OAR item across referrals (lines 200-204) but did not analyse the factors independently to report on the most important OAR item predicting the presence of an ankle fracture.
Editor comments (John Ringo)
Reviewer #2 wants minor revisions. Let's consider them in turn. First, the length of the Background section. From my point of view, all the paragraphs of this section are pretty brief as is, except for #6 (lines 80-93). Maybe you could shorten this one, and maybe even scour the whole section to eliminate needless words. Second, the requested discussion of the relationship between OAR and 'subjective examindor [maybe examiner?] perception'. Please consider this point, but keep remarks about the subject brief. Third, discuss 'the most important item of OAR'. If you can think of a useful remark to make, please add that, though I must say that the OAR seems to me (a non-clinician, naive to this whole business) very simple and brief altogether.
Revisions addressed above.
Revisions addressed above.
" | Here is a paper. Please give your review comments after reading it. |
9,798 | "<ns0:abstract xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:p>Background. Gene tree discordance is common in phylogenetic analyses. Many phylogenetic studies have excluded non-coding regions of the plastome without evaluating their impact on tree topology. In general, plastid loci have often been treated as a single unit, and tree discordance among these loci has seldom been examined. Using samples of Laureae (Lauraceae) plastomes, we explored plastome variation among the tribe, examined the influence of non-coding regions on tree topology, and quantified intra-plastome conflict.</ns0:p><ns0:p>Results. We found that the plastomes of Laureae have low inter-specific variation and are highly similar in structure, size, and gene content. Laureae was divided into three groups, subclades I, II and III. The inclusion of non-coding regions changed the phylogenetic relationship among the three subclades. Topologies based on coding and non-coding regions were largely congruent except for the relationship among subclades I, II and III. By measuring the distribution of phylogenetic signal across loci that supported different topologies, we found that nine loci (two coding regions, two introns and five intergenic spacers) played a critical role at the contentious node.</ns0:p></ns0:div>
<ns0:div><ns0:head>Conclusions.</ns0:head><ns0:p>Our results suggest that subclade III and subclade II are successively sister to subclade I. Conflicting phylogenetic signals exist between coding and non-coding regions of Laureae plastomes. Our study highlights the importance of evaluating the influence of non-coding regions on tree topology and emphasizes the necessity of examining discordance among different plastid loci in phylogenetic studies.</ns0:p></ns0:div>
</ns0:abstract>
<ns0:body xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:head>Introduction</ns0:head><ns0:p>Gene tree discordance is relatively common in phylogenomic studies. The conflicts can be caused by biological factors like incomplete lineage sorting (ILS), hybridization, horizontal gene transfer, gene loss, and gene duplication <ns0:ref type='bibr' target='#b43'>(Maddison, 1997;</ns0:ref><ns0:ref type='bibr' target='#b74'>Sun et al., 2015;</ns0:ref><ns0:ref type='bibr' target='#b17'>Gonçalves et al., 2019;</ns0:ref><ns0:ref type='bibr' target='#b60'>Sato et al., 2019)</ns0:ref>. Most relevant studies have focused on incongruent tree topologies among different genomic compartments <ns0:ref type='bibr' target='#b74'>(Sun et al., 2015;</ns0:ref><ns0:ref type='bibr' target='#b95'>Zhao et al., 2016;</ns0:ref><ns0:ref type='bibr' target='#b80'>Walker et al., 2019)</ns0:ref> because these genes have evolved independently and their gene tree topologies have been influenced by biological processes. By contrast, relatively few studies have focused on tree conflicts among plastid genes (e.g., <ns0:ref type='bibr' target='#b15'>Foster, Henwood & Ho, 2018;</ns0:ref><ns0:ref type='bibr' target='#b17'>Gonçalves et al., 2019;</ns0:ref><ns0:ref type='bibr' target='#b80'>Walker et al., 2019;</ns0:ref><ns0:ref type='bibr' target='#b94'>Zhang et al., 2020)</ns0:ref>. Usually, plastomes are considered to be uniparentally inherited and to have evolved as a single unit, free from such biological sources of conflict <ns0:ref type='bibr' target='#b2'>(Birky, 1995;</ns0:ref><ns0:ref type='bibr' target='#b82'>Wicke et al., 2011)</ns0:ref>. However, the branched and linear structure of plastid DNA, which arose from recombination-dependent replication, is indicative of recombination <ns0:ref type='bibr' target='#b49'>(Oldenburg & Bendich, 2016;</ns0:ref><ns0:ref type='bibr' target='#b57'>Ruhlman et al., 2017)</ns0:ref>. In addition, biparental inheritance and heteroplasmy (e.g., the presence of different plastomes within an individual or a cell) have been reported in seed plants <ns0:ref type='bibr' target='#b75'>(Szmidt, Aldén & Hällgren, 1987;</ns0:ref><ns0:ref type='bibr' target='#b25'>Johnson & Palmer, 1989;</ns0:ref><ns0:ref type='bibr' target='#b55'>Reboud & Zeyl, 1994;</ns0:ref><ns0:ref type='bibr' target='#b4'>Carbonell-Caballero et al., 2015)</ns0:ref>. Heteroplasmy may, in rare cases, give rise to heteroplasmic recombination, which has been invoked to explain gene tree discordance <ns0:ref type='bibr' target='#b44'>(Marshall, Newton & Ritland, 2001;</ns0:ref><ns0:ref type='bibr' target='#b73'>Sullivan et al., 2017;</ns0:ref><ns0:ref type='bibr' target='#b59'>Sancho et al., 2018)</ns0:ref>. In addition to recombination events, the transfer of genes among plastid, mitochondrial and nuclear genomes; positive selection; tree length (gene evolutionary rate); and GC content may also generate PeerJ reviewing PDF | (2020:05:49532:1:1:NEW 30 Jul 2020)</ns0:p><ns0:p>Manuscript to be reviewed phylogenomic conflict (e.g., <ns0:ref type='bibr' target='#b72'>Stegemann et al., 2003;</ns0:ref><ns0:ref type='bibr' target='#b66'>Smith, 2014;</ns0:ref><ns0:ref type='bibr' target='#b86'>Wysocki et al., 2015;</ns0:ref><ns0:ref type='bibr' target='#b53'>Piot et al., 2018;</ns0:ref><ns0:ref type='bibr' target='#b58'>Saarela et al., 2018;</ns0:ref><ns0:ref type='bibr' target='#b15'>Foster, Henwood & Ho, 2018)</ns0:ref>. Aside from biological factors, non-biological factors (e.g., outlier genes, uninformative loci, and gaps) may cause conflict as well. For example, <ns0:ref type='bibr' target='#b10'>Duvall, Burke & Clark (2020)</ns0:ref> found that alternative topologies arose from alignment gaps. Given that most studies assume no conflict and treat the plastome as a single unit, taking biological and non-biological factors into consideration and quantifying the extent of conflict among different plastid loci is of great importance <ns0:ref type='bibr' target='#b85'>(Wolfe & Randle, 2004)</ns0:ref>.</ns0:p><ns0:p>Owing to the rapid development of next-generation sequencing (NGS), more plastomes are becoming available at a reasonable cost, driving advances in phylogenomics and promoting a more comprehensive understanding of plant evolution <ns0:ref type='bibr' target='#b32'>(Li et al., 2019)</ns0:ref>. Phylogenetic relationships among Lauraceae <ns0:ref type='bibr' target='#b68'>(Song et al., 2017)</ns0:ref>, as well as many other groups (e.g., <ns0:ref type='bibr' target='#b12'>Eserman et al., 2014;</ns0:ref><ns0:ref type='bibr' target='#b1'>Barrett et al., 2016)</ns0:ref>, have been well resolved using plastome data. In phylogenomic studies of plastomes <ns0:ref type='bibr' target='#b18'>(Guo et al., 2017;</ns0:ref><ns0:ref type='bibr' target='#b17'>Gonçalves et al., 2019;</ns0:ref><ns0:ref type='bibr' target='#b90'>Xu et al., 2019;</ns0:ref><ns0:ref type='bibr' target='#b32'>Li et al., 2019)</ns0:ref>, plastome coding genes have generally been used, and non-coding regions have been excluded.</ns0:p><ns0:p>Only a few studies have noted the potential impact of non-coding regions on tree topology. <ns0:ref type='bibr' target='#b51'>Parks, Cronn & Liston (2009)</ns0:ref> revealed that the phylogenetic position of Pinus albicaulis Engelm. based on complete plastomes differed from that based on exon sequences. A similar situation also occurred for phylogenetic relationships within Rubiaceae <ns0:ref type='bibr' target='#b84'>(Wikström, Bremer & Rydin, 2020)</ns0:ref>, suggesting that there were conflicting phylogenetic signals between coding-and non-coding regions. Because tree topology is the foundation of comparative studies that infer biogeographic history, phylogenetic diversity and other evolutionary patterns <ns0:ref type='bibr'>(Walker et al.,</ns0:ref> and in evergreen broadleaf forests <ns0:ref type='bibr' target='#b81'>(Wang et al., 2009)</ns0:ref>, and Laurus nobilis L. has been used in remedies for centuries <ns0:ref type='bibr' target='#b46'>(Nayak et al., 2006)</ns0:ref>.</ns0:p><ns0:p>Although Laureae is monophyletic, generic delimitation within this tribe remains unclear <ns0:ref type='bibr' target='#b29'>(Kostermans, 1957;</ns0:ref><ns0:ref type='bibr' target='#b23'>Hutchinson, 1964;</ns0:ref><ns0:ref type='bibr' target='#b37'>Li et al., 2008b)</ns0:ref>. Adenodaphne, endemic to New Caledonia, is closely related to Litsea <ns0:ref type='bibr' target='#b5'>(Chanderbali, van der Werff & Renner, 2001)</ns0:ref>. However, morphological confusion still exists between this genus and Litsea, meaning that their distinctiveness and the monophyly of Adenodaphne require further study <ns0:ref type='bibr' target='#b5'>(Chanderbali, van der Werff & Renner, 2001)</ns0:ref>. Actinodaphne is polyphyletic and closely related to the monophyletic genus Neolitsea <ns0:ref type='bibr' target='#b38'>(Li et al., 2007;</ns0:ref><ns0:ref type='bibr' target='#b37'>Li et al., 2008b;</ns0:ref><ns0:ref type='bibr' target='#b13'>Fijridiyanto & Murakami, 2009a</ns0:ref><ns0:ref type='bibr' target='#b14'>, 2009b)</ns0:ref>.</ns0:p><ns0:p>Although <ns0:ref type='bibr' target='#b13'>Fijridiyanto & Murakami (2009a</ns0:ref><ns0:ref type='bibr' target='#b14'>, 2009b)</ns0:ref> argued that Actinodaphne was monophyletic, the species of Actinodaphne sampled in their analyses were totally different from those sampled in <ns0:ref type='bibr' target='#b38'>Li et al. (2007)</ns0:ref> and <ns0:ref type='bibr' target='#b37'>Li et al. (2008b)</ns0:ref>. Furthermore, Lindera and Litsea have been shown to be polyphyletic, with Dodecadenia, Iteadaphne, Laurus, Parasassafras and Sinosassafras nested within them <ns0:ref type='bibr' target='#b36'>(Li et al., 2004;</ns0:ref><ns0:ref type='bibr' target='#b37'>Li et al., 2008b)</ns0:ref>. <ns0:ref type='bibr' target='#b41'>Liu et al. (2017)</ns0:ref> used three plastid barcode loci combined with the internal transcribed spacer (ITS) region for species identification and found that the Laureae tree was polytomic. Despite these efforts, phylogenetic relationships among and within these genera have been poorly resolved based on molecular markers like the ITS, the external transcribed spacer (ETS), matK, trnL-F and trnH-psbA.</ns0:p><ns0:p>Compared with these molecular markers, complete plastomes have better performance at the species level within Laureae, although generic delimitation still remains unclear due to limited taxon sampling <ns0:ref type='bibr' target='#b96'>(Zhao et al., 2018;</ns0:ref><ns0:ref type='bibr' target='#b67'>Song et al., 2019;</ns0:ref><ns0:ref type='bibr' target='#b76'>Tian, Ye & Song, 2019)</ns0:ref>.</ns0:p><ns0:p>Genomic DNA was extracted from silica-gel-dried leaf tissue using the cetyl trimethyl ammonium bromide (CTAB) method <ns0:ref type='bibr' target='#b9'>(Doyle & Doyle, 1987)</ns0:ref>. The yields of genomic DNA extracts were quantified by fluorometric quantification on a Qubit instrument (Invitrogen, Carlsbad, California, USA) using the dsDNA HS kit, and the DNA size distribution was assessed visually on a 1% agarose gel. DNA libraries with an average insert size of 270 bp were prepared by the Beijing Genomics Institute (BGI, Shenzhen, China). Paired-end reads of 2 × 151 bp were generated on the Illumina X ten sequencing system (Illumina Inc.).</ns0:p></ns0:div>
<ns0:div><ns0:head>Plastid genome assembly, annotation and comparison</ns0:head><ns0:p>Low-quality reads and adaptors were removed using Trimmomatric v0.36 <ns0:ref type='bibr' target='#b3'>(Bolger, Lohse & Usadel, 2014)</ns0:ref>, generating approximately 3 Gb of high-quality clean reads per sample. The clean reads were analyzed for quality control with FastQC <ns0:ref type='bibr' target='#b0'>(Andrews, 2010)</ns0:ref> and then used to assemble plastomes with NOVOPlasty v2.7.2 <ns0:ref type='bibr' target='#b8'>(Dierckxsens, Mardulyn & Smits, 2016)</ns0:ref>. To guarantee assembly quality, clean reads were mapped to the assembled plastid genomes using the Burrows-Wheeler Aligner (BWA 0.7.17-r1188 <ns0:ref type='bibr' target='#b34'>(Li & Durbin, 2010)</ns0:ref>) and samtools 1.9 <ns0:ref type='bibr' target='#b35'>(Li et al., 2009)</ns0:ref>, and were visually checked in Geneious Prime 2019.1.</ns0:p><ns0:p>Plastome annotation was performed using the program GeSeq -Annotation of Organellar Genomes <ns0:ref type='bibr' target='#b77'>(Tillich et al., 2017)</ns0:ref>. Start and stop codons were inspected and manually adjusted in Geneious Prime when necessary. Plastomes were submitted to GenBank (MN274947, MN428456-MN428466). Maps of all 12 plastomes were drawn using the OrganellarGenomeDRAW tool (OGDRAW) <ns0:ref type='bibr' target='#b42'>(Lohse et al., 2013)</ns0:ref>. A summary of the newly PeerJ reviewing PDF | (2020:05:49532:1:1:NEW 30 Jul 2020)</ns0:p><ns0:p>Manuscript to be reviewed sequenced plastomes is presented in Table <ns0:ref type='table'>2</ns0:ref>.</ns0:p><ns0:p>To illustrate interspecific sequence variation within Laureae, plastomes of A. obovata, I. caudata, Laurus nobilis (KY085912), Lindera erythrocarpa, Litsea acutivena, N. pallens and Parasassafras confertiflorum (Meisn.) D. G. Long (MH729378) were aligned using MAFFT <ns0:ref type='bibr' target='#b26'>(Katoh & Standley, 2013)</ns0:ref> with default settings. Sequence identity was plotted with the mVISTA program using the LAGAN mode <ns0:ref type='bibr' target='#b16'>(Frazer et al., 2004)</ns0:ref>, with Lindera glauca (Siebold et Zucc.)</ns0:p><ns0:p>Bl. (MF188124) as a reference.</ns0:p></ns0:div>
<ns0:div><ns0:head>Phylogenetic reconstruction and tests for selection</ns0:head><ns0:p>To evaluate potential conflicts, phylogenetic trees were constructed using maximum likelihood (ML) methods based on six datasets: (1) complete plastome (cp), (2) coding regions (CDS), (3) non-coding regions (non-CDS), (4) large single copy region (LSC), (5) small single copy region (SSC), and (6) one inverted repeat region (IR).</ns0:p><ns0:p>Sequences were aligned using MAFFT with default settings and manually edited with BioEdit v7.2.5 <ns0:ref type='bibr' target='#b19'>(Hall, 1999)</ns0:ref> when necessary. The best-fitting DNA substitution models for the six unpartitioned datasets were selected using ModelTest-NG <ns0:ref type='bibr' target='#b7'>(Darriba et al., 2020)</ns0:ref> under the corrected Akaike Information Critierion (AICc). The aligned sequences and selected DNA substitution models were used for ML analyses, and ML trees were constructed using RAxML-NG <ns0:ref type='bibr' target='#b30'>(Kozlov et al., 2019)</ns0:ref>. We also implemented a partitioning strategy on two datasets, the cp with one IR region removed (cp-reduced) and CDS (configuration details shown in Supplemental File 1). The best partition schemes were inferred with PartitionFinder 2 <ns0:ref type='bibr' target='#b31'>(Lanfear et al., 2016)</ns0:ref>, <ns0:ref type='table'>PDF | (2020:05:49532:1:1:NEW 30 Jul 2020)</ns0:ref> Manuscript to be reviewed and the best partition schemes and models for each partition were used for ML analyses in RAxML-NG.</ns0:p><ns0:formula xml:id='formula_0'>PeerJ reviewing</ns0:formula><ns0:p>Because gaps can affect tree topology <ns0:ref type='bibr' target='#b10'>(Duvall, Burke & Clark, 2020)</ns0:ref>, we also performed the following analysis based on the cp dataset. 'Mask Alignment' in Geneious Prime was used to strip the gaps from the MAFFT alignment, with the threshold set to 0 (no gaps), 2%, 10%, 20%, 50% or 75%. The resulting alignments were used to infer ML trees in RAxML-NG.</ns0:p><ns0:p>Positive selection on plastid coding genes has the potential to bias phylogenies (e.g., <ns0:ref type='bibr' target='#b53'>Piot et al., 2018;</ns0:ref><ns0:ref type='bibr' target='#b58'>Saarela et al., 2018)</ns0:ref>, and we therefore performed natural selection tests using CODEML in PAML 4.9j <ns0:ref type='bibr' target='#b91'>(Yang, 2007)</ns0:ref>. Coding genes were extracted and aligned in Geneious Prime using MAFFT, stop codons were removed manually, and the aligned sequences were converted to paml format. To statistically test for positive selection, we compared the performance of two branch models (M0 and M2) for each gene. Three foreground branches were labeled on the unpartitioned CDS ML tree. Likelihood ratio tests (LRT) were performed using pchisq function in R 3.6.2 (R Core Team, 2018).</ns0:p></ns0:div>
<ns0:div><ns0:head>Node support investigation and tree topology tests</ns0:head><ns0:p>Because gene contents were not identical among Cryptocaryeae, Cassytha, Caryodaphnopsis, Neocinnamomum and other clades, the following analyses were performed using a dataset from which six plastomes had been removed (Beilschmiedia pauciflora H. W. </ns0:p></ns0:div>
<ns0:div><ns0:head>Manuscript to be reviewed</ns0:head><ns0:p>We extracted all loci (coding regions, introns, tRNA, rRNA and intergenic spacers) using a python script <ns0:ref type='bibr' target='#b24'>(Jin, 2019)</ns0:ref> and aligned them using MAFFT with default settings. These alignments were used to infer gene trees by rapid bootstrap analyses (option -f a) in RAxML <ns0:ref type='bibr' target='#b71'>(Stamatakis, 2014)</ns0:ref> with the GTRGAMMA model. The number of bootstrap replicates was set to 1000, as <ns0:ref type='bibr' target='#b65'>Simmons & Kessenich (2019)</ns0:ref> have suggested that fewer replicates may be insufficient to find the optimal gene tree topology. The best-scoring ML trees were used to estimate the species tree with local posterior probability (LPP) <ns0:ref type='bibr' target='#b61'>(Sayyari & Mirarab, 2016)</ns0:ref> in ASTRAL III <ns0:ref type='bibr' target='#b93'>(Zhang et al., 2018)</ns0:ref>.</ns0:p><ns0:p>We performed constrained maximum likelihood analyses in IQ-TREE <ns0:ref type='bibr' target='#b47'>(Nguyen et al., 2014)</ns0:ref> to obtain the ML trees that supported different topologies. To understand which loci supported the alternative topologies, we calculated site-wise log-likelihood values for each topology in RAxML using option '-f G'. After obtaining site-wise lnL differences, we converted site-wise differences to locus-wise lnL differences (ΔlnL) in R 3.6.2. The lnL differences were plotted against each locus using ggplot2 <ns0:ref type='bibr' target='#b83'>(Wickham, 2016)</ns0:ref>. It has been suggested that loci with an absolute ΔlnL > 2 are statistically significant <ns0:ref type='bibr' target='#b11'>(Edwards, 1984)</ns0:ref>. Therefore, we conducted separate ML analyses on datasets from which these loci (absolute ΔlnL > 2) had been removed to test whether small subsets of sequence matrices determined tree topology <ns0:ref type='bibr' target='#b62'>(Shen, Hittinger & Rokas, 2017)</ns0:ref>.</ns0:p><ns0:p>The Kishino-Hasegawa test (KH test) <ns0:ref type='bibr' target='#b27'>(Kishino & Hasegawa, 1989)</ns0:ref>, Shimodaira-Hasegawa test (SH test) <ns0:ref type='bibr' target='#b64'>(Shimodaira & Hasegawa, 1999)</ns0:ref> and Approximately-Unbiased test (AU test) <ns0:ref type='bibr' target='#b63'>(Shimodaira, 2002)</ns0:ref> were used in IQ-TREE to assess the statistical significance of incongruence PeerJ reviewing PDF | (2020:05:49532:1:1:NEW 30 Jul 2020)</ns0:p><ns0:p>Manuscript to be reviewed based on complete plastomes (including only one copy of the IR regions). We specified 10,000 RELL (resampling of estimated log-likelihoods) replicates for the topological tests.</ns0:p></ns0:div>
<ns0:div><ns0:head>Results</ns0:head></ns0:div>
<ns0:div><ns0:head>Plastome features of Laureae</ns0:head><ns0:p>The sizes of the 12 newly generated Laureae plastid genomes ranged from 152,132 bp (Litsea szemaois) to 152,916 bp (Lindera erythrocarpa) (Table <ns0:ref type='table'>2</ns0:ref>), similar to previously published Laureae plastomes (152,211-153,011 bp, Table <ns0:ref type='table'>S1</ns0:ref>). All had a typical quadripartite structure and were assembled into a single, circular and double-stranded DNA sequence (Fig. <ns0:ref type='figure' target='#fig_1'>1</ns0:ref>). The length of the LSC, SSC and IR regions ranged from 93,119 bp (Litsea szemaois) to 93,921 bp (Lindera erythrocarpa), 18,796 bp (N. pallens) to 18,936 bp (Litsea mollis), and 20,057 bp (A. obovata) to 20,144 bp (I. caudata), respectively, with little variation in size (Table <ns0:ref type='table'>2</ns0:ref>). The overall GC contents ranged from 39.1% to 39.2%. GC content was unequally distributed within the plastomes; it was highest in IR regions (44.4-44.5%), moderate in LSC regions (37.9-38.1%), and lowest in SSC regions (33.8-34.0%, Table <ns0:ref type='table'>2</ns0:ref>).</ns0:p><ns0:p>The 12 newly sequenced plastomes contained 112 single-copy genes: 78 protein-coding genes, 30 tRNA genes, and 4 rRNA genes (Table <ns0:ref type='table'>2 and Table S2</ns0:ref>). Sixteen genes had one intron, and two genes had two introns. There were 13 duplicated genes in the IR regions (Table <ns0:ref type='table'>S2</ns0:ref>), and rps12, ycf1, and ycf2 were partly duplicated in the IR regions (Fig. <ns0:ref type='figure' target='#fig_1'>1</ns0:ref>).</ns0:p></ns0:div>
<ns0:div><ns0:head>Phylogenetic reconstruction and positive selection tests</ns0:head><ns0:p>The GTR+I+G4 model was selected for the six unpartitioned datasets (cp, CDS, non-CDS, LSC, Manuscript to be reviewed subclade I based on the unpartitioned CDS dataset (Fig. <ns0:ref type='figure' target='#fig_1'>S1</ns0:ref>). Both topologies were strongly supported.</ns0:p><ns0:p>The sister relationship of subclades I and II was confirmed in the ML tree based on partitioned plastomes (one IR removed, cp-reduced dataset; Fig. <ns0:ref type='figure'>S7</ns0:ref>), and subclade II was sister to subclade III in the ML tree based on the partitioned CDS dataset (Fig. <ns0:ref type='figure'>S8</ns0:ref>), indicating that partitioning did not affect our tree topology.</ns0:p><ns0:p>The sister relationship of subclades I and II (BS values ranging from 80% to 92%) was consistently revealed even as the percentage of gaps increased (Table <ns0:ref type='table'>S3</ns0:ref>), indicating that gaps had no impact on our tree topology.</ns0:p><ns0:p>LRT showed that the dN/dS ratios of labeled lineages (subclades I, II and III) were not significantly different from background (p > 0.05), of which dN/dS ratios were less than one (Table <ns0:ref type='table'>S4</ns0:ref>), suggesting that there was no positive selection on the plastid genes.</ns0:p></ns0:div>
<ns0:div><ns0:head>Investigating incongruent nodes and differences in tree topology</ns0:head><ns0:p>The tree topology inferred from ASTRAL III (Fig. <ns0:ref type='figure'>3</ns0:ref>) was largely congruent with that of the ML trees (Figs. 2 and S1-S4), except that the former showed a sister relationship of subclade I and subclade III. We performed constrained maximum likelihood analyses in IQ-TREE and obtained three suboptimal ML trees that supported the subclade II-subclade I (called T1 hereafter), subclade II-subclade III (T2) and subclade I-subclade III (T3) affinities. We extracted 243 loci and assessed how each locus supported one of the three topologies by examining the gene-wise log-likelihoods (Fig. <ns0:ref type='figure' target='#fig_2'>4</ns0:ref>). T1 was strongly supported by six loci (rpoC1 intron, trnG-trnfM, ndhA PeerJ reviewing PDF | (2020:05:49532:1:1:NEW 30 Jul 2020)</ns0:p><ns0:p>Manuscript to be reviewed intron, psaJ-rpl33, rpl2-rpl23 and petN-psbM; absolute ΔlnL > 2); T2 was strongly supported by three loci (psaB, trnS-ycf3 and ycf2; absolute ΔlnL > 2); and T3 was moderately supported by one locus (clpP intron1; absolute ΔlnL > 1 and < 2) (Table <ns0:ref type='table'>S5</ns0:ref>). The sum of absolute ΔlnL of T1 was higher than that of T2 and T3 (Fig. <ns0:ref type='figure' target='#fig_2'>4</ns0:ref>), suggesting that our data support the topology of T1 rather than T2 or T3. After the removal of six loci (rpoC1 intron, trnG-trnfM, ndhA intron, psaJ-rpl33, rpl2-rpl23 and petN-psbM), a sister relationship of subclade II and subclade III was revealed (Fig. <ns0:ref type='figure'>S9</ns0:ref>). After the removal of three loci (psaB, trnS-ycf3, and ycf2), subclade II was sister to subclade I (Fig. <ns0:ref type='figure' target='#fig_1'>S10</ns0:ref>). These results underscore the decisive role played by small subsets of loci in phylogenetic inference.</ns0:p><ns0:p>The topological tests showed that T2 did not differ significantly from T1 (p > 0.05, Table <ns0:ref type='table'>S6</ns0:ref>).</ns0:p><ns0:p>T3 was statistically rejected by the KH and AU tests (p < 0.05) but not by the Shimodaira-Hasegawa (SH) test (p = 0.0505). That T3 was rejected according to the KH and AU tests suggests that the sister relationship between subclades I and III may be misleading.</ns0:p></ns0:div>
<ns0:div><ns0:head>Discussion</ns0:head></ns0:div>
<ns0:div><ns0:head>Plastome features</ns0:head><ns0:p>It has been noted that most plastid genomes of land plants and algae range from 120 to 160 kilobase pairs (kb) in length <ns0:ref type='bibr' target='#b50'>(Palmer, 1985)</ns0:ref>. In this study, the plastid genome sizes of 12 species from five Laureae genera ranged from 152,132 bp to 152,916 bp, indicating that plastid genome size was conserved within Laureae. GC content was highest in the IR region rather than in the single copy regions, owing to the presence of a ribosomal RNA gene cluster in the IR region, PeerJ reviewing PDF | (2020:05:49532:1:1:NEW 30 Jul 2020)</ns0:p><ns0:p>Manuscript to be reviewed consistent with a previous study <ns0:ref type='bibr' target='#b21'>(Huotari & Korpelainen, 2012)</ns0:ref>. GC contents of the IR, LSC and SSC regions of the newly sequenced plastomes were identical to those of nine Lindera species studied earlier <ns0:ref type='bibr' target='#b96'>(Zhao et al., 2018)</ns0:ref>. In contrast to the gene losses recognized in several Lauraceae lineages <ns0:ref type='bibr' target='#b68'>(Song et al., 2017)</ns0:ref>, our analysis revealed that gene content among Laureae was highly conserved. <ns0:ref type='bibr' target='#b68'>Song et al. (2017)</ns0:ref> suggested that plastome contraction in Lauraceae was largely driven by fragment loss events in the IR regions. In our study, we found no gene loss among Laureae plastomes.</ns0:p></ns0:div>
<ns0:div><ns0:head>Phylogenetic relationships within Laureae</ns0:head><ns0:p>Previous phylogenetic studies <ns0:ref type='bibr' target='#b68'>(Song et al., 2017;</ns0:ref><ns0:ref type='bibr' target='#b96'>Zhao et al., 2018)</ns0:ref> based on complete plastomes suggested that Laureae was sister to Cinnamomeae and that together they were sister to Perseeae.</ns0:p><ns0:p>The same phylogenetic relationships among these groups were recognized in our study (Figs. <ns0:ref type='figure'>2 and 3</ns0:ref>). In previous work, Actinodaphne and Neolitsea were resolved as monophyletic groups based on matK, ITS and rpb2 <ns0:ref type='bibr' target='#b13'>(Fijridiyanto & Murakami, 2009a</ns0:ref><ns0:ref type='bibr' target='#b14'>, 2009b)</ns0:ref>, but Actinodaphne was not a monophyletic group based on complete plastid genomes <ns0:ref type='bibr' target='#b67'>(Song et al., 2019)</ns0:ref>. In this study, the non-monophyletic status of Actinodaphne was supported. The conclusion of Actinodaphne monophyly may have been caused by sampling bias in previous studies <ns0:ref type='bibr' target='#b14'>(Fijridiyanto & Murakami, 2009b</ns0:ref><ns0:ref type='bibr' target='#b13'>, 2009a)</ns0:ref>. The monophyly of Neolitsea can be neither rejected nor supported in the present study. Actinodaphne cupularis (Hemsl.) Gamble was grouped with Neolitsea oblongifolia Merr. et Chun, N. pallens and N. chui Merr. with low bootstrap support (54%; Fig. <ns0:ref type='figure'>2</ns0:ref>), and sampling of Neolitsea and related genera was limited. Lindera and Litsea were PeerJ reviewing PDF | (2020:05:49532:1:1:NEW 30 Jul 2020)</ns0:p><ns0:p>Manuscript to be reviewed polyphyletic in our analysis, consistent with previous studies <ns0:ref type='bibr' target='#b37'>(Li et al., 2008b;</ns0:ref><ns0:ref type='bibr' target='#b14'>Fijridiyanto & Murakami, 2009b)</ns0:ref>. The phylogenetic position of P. confertiflorum was unresolved based on ETS and ITS <ns0:ref type='bibr' target='#b37'>(Li et al., 2008b)</ns0:ref>, and the ambiguity of its position still remains, despite the integration of complete plastid genomes in our analysis and a previous study <ns0:ref type='bibr' target='#b39'>(Liao, Ye & Song, 2018)</ns0:ref>.</ns0:p><ns0:p>Subclade III was sister to subclade I and II in our study, consistent with previous analyses <ns0:ref type='bibr' target='#b96'>(Zhao et al., 2018;</ns0:ref><ns0:ref type='bibr' target='#b67'>Song et al., 2019;</ns0:ref><ns0:ref type='bibr' target='#b76'>Tian, Ye & Song, 2019)</ns0:ref>. The three Lindera species in subclade I share common morphological traits, such as alternate and pinninerved leaves, a persistent involucre, vegetative terminal buds in inflorescences and 3-merous flowers <ns0:ref type='bibr' target='#b33'>(Li et al., 2008a)</ns0:ref>. However, these characters also occur in several members of the other two subclades (e.g., Lindera benzoin (L.) Bl. and Laurus nobilis), perhaps resulting from convergent and/or parallel evolution <ns0:ref type='bibr' target='#b37'>(Li et al., 2008b)</ns0:ref>. These traits are not good indicators for delimiting the three subclades of Laureae. In subclade III, the trinerved or triplinerved species of Lindera (Lindera aggregata, L. chunii, L. fragrans, L. limprichtii, L. pulcherrima, L. supracostata, L. thomsonii and L. thomsonii var. vernayana) formed a well-supported clade in both our study and that of <ns0:ref type='bibr' target='#b76'>Tian, Ye & Song (2019)</ns0:ref>. However, triplinerved leaves also exist in most species of Neolitsea <ns0:ref type='bibr' target='#b37'>(Li et al., 2008b;</ns0:ref><ns0:ref type='bibr' target='#b33'>Li et al., 2008a)</ns0:ref>. Therefore, traditional morphological traits are of limited use in taxon delimitation, even within subclades of Laureae. Given the limited samples and data in our analyses, more sampling and DNA sequences are needed to further elucidate the relationships within Laureae.</ns0:p></ns0:div>
<ns0:div><ns0:head>Phylogenetic incongruence in the plastome</ns0:head><ns0:p>PeerJ reviewing PDF | (2020:05:49532:1:1:NEW 30 Jul 2020)</ns0:p><ns0:p>Manuscript to be reviewed Although many studies have treated plastid protein-coding genes or the complete plastome as a single unit (e.g., <ns0:ref type='bibr' target='#b67'>Song et al., 2019;</ns0:ref><ns0:ref type='bibr' target='#b76'>Tian, Ye & Song, 2019)</ns0:ref>, potential conflicts among sequence types (i.e., coding vs. non-coding regions) have been reported in several studies. By comparing phylogenies based on complete plastomes and coding regions <ns0:ref type='bibr' target='#b92'>(Yu et al., 2017)</ns0:ref>, it was inferred that non-coding regions did not significantly influence the tree topology of Theaceae. By contrast, non-coding regions had an impact on the phylogenetic relationships of several tribes in Rubiaceae <ns0:ref type='bibr' target='#b84'>(Wikström, Bremer & Rydin, 2020)</ns0:ref> and subtribes in Poaceae <ns0:ref type='bibr' target='#b58'>(Saarela et al., 2018)</ns0:ref>. A conflicting signal between coding and non-coding regions was also reported in Leguminosae <ns0:ref type='bibr' target='#b94'>(Zhang et al., 2020)</ns0:ref>. In this study, inclusion of non-coding regions altered tree topology in the tribe Laureae, suggesting the existence of a conflicting signal between coding and non-coding regions. Non-coding regions are often discarded for being uninformative, or for being misleading due to saturation at deep time scales <ns0:ref type='bibr' target='#b15'>(Foster, Henwood & Ho, 2018)</ns0:ref>. In our study, tree topologies based on coding and non-coding regions were largely congruent, except for the relationships among the three subclades (Figs. S1-S2), indicating that non-coding regions are as informative as coding regions in Laureae. Thus, it is imperative to evaluate the influence of noncoding regions on tree topology rather than treating the whole plastome as a single unit or simply excluding non-coding regions from phylogenetic analysis.</ns0:p><ns0:p>To accommodate the conflicts among different plastid regions, a species tree was inferred through summary coalescent analysis. It has been suggested that the coalescent method is more robust than the concatenation method when the level of ILS is high <ns0:ref type='bibr' target='#b40'>(Liu, Xi & Davis, 2014;</ns0:ref><ns0:ref type='bibr' target='#b45'>Mirarab, Bayzid & Warnow, 2014)</ns0:ref>. High ILS tends to occur when the time interval between PeerJ reviewing PDF | (2020:05:49532:1:1:NEW 30 Jul 2020)</ns0:p><ns0:p>Manuscript to be reviewed consecutive speciation events is short <ns0:ref type='bibr' target='#b74'>(Sun et al., 2015;</ns0:ref><ns0:ref type='bibr' target='#b60'>Sato et al., 2019)</ns0:ref>, and the core Lauraceae group (Perseeae, Cinnamomeae and Laureae) is thought to have undergone a rapid radiation <ns0:ref type='bibr' target='#b5'>(Chanderbali, van der Werff & Renner, 2001;</ns0:ref><ns0:ref type='bibr' target='#b55'>Rohwer & Rudolph, 2005;</ns0:ref><ns0:ref type='bibr' target='#b48'>Nie, Wen & Sun, 2007)</ns0:ref>. We therefore chose to implement the coalescent method. Nonetheless, it should be noted that, with this method, short and uninformative loci may lead to problematic gene trees and therefore result in a less accurate species tree <ns0:ref type='bibr' target='#b88'>(Xi, Liu & Davis, 2015;</ns0:ref><ns0:ref type='bibr' target='#b69'>Springer & Gatesy, 2016)</ns0:ref>.</ns0:p><ns0:p>In our study, only nine of 243 loci <ns0:ref type='bibr'>(rpoC1 intron, trnG-trnfM, ndhA intron, psaJ-rpl33, rpl2-rpl23, petN-psbM, psaB, trnS-ycf3, and ycf2</ns0:ref>) had a strong phylogenetic signal at the contentious node. The other 234 loci with weak phylogenetic signals may have resulted in gene trees with uncertainties and led to inaccurate topology at this node.</ns0:p><ns0:p>Exploration of the factors that underlie conflicts in phylogenetic signals is of great importance-but it is also challenging. Previous studies have examined whether biological and non-biological factors contribute to such conflicts (e.g., <ns0:ref type='bibr' target='#b10'>Duvall, Burke & Clark, 2020;</ns0:ref><ns0:ref type='bibr' target='#b94'>Zhang et al., 2020)</ns0:ref>. For example, gaps have been found to cause alternate, but conflicting topologies in Poaceae <ns0:ref type='bibr' target='#b10'>(Duvall, Burke & Clark, 2020)</ns0:ref>. However, the inclusion of alignment gaps did not alter our tree topology (Table <ns0:ref type='table'>S3</ns0:ref>). Although previous studies indicated that partitioning improves phylogenetic inference <ns0:ref type='bibr' target='#b89'>(Xi et al., 2012)</ns0:ref>, ML tree topologies based on partitioned and unpartitioned datasets did not differ significantly in our study. It has been suggested that plastid genes under positive selection may bias phylogenies (e.g., <ns0:ref type='bibr' target='#b53'>Piot et al., 2018;</ns0:ref><ns0:ref type='bibr' target='#b58'>Saarela et al., 2018)</ns0:ref>.</ns0:p><ns0:p>Although psaB and ycf2 were shown to influence topology, neither gene evolved under positive selection, suggesting that natural selection is not the cause of the conflict. In this study, the low PeerJ reviewing PDF | (2020:05:49532:1:1:NEW 30 Jul 2020)</ns0:p><ns0:p>Manuscript to be reviewed support values and short branch lengths of the estimated species tree (Fig. <ns0:ref type='figure'>3</ns0:ref>) suggested that each locus had a significantly incongruent topology and may indicate the existence of ILS. High levels of ILS are thought to yield similar numbers of loci supporting alternative topologies <ns0:ref type='bibr' target='#b22'>(Huson et al., 2005)</ns0:ref>. In our study, the numbers of loci supporting each topology were different (six for T1, three for T2, and zero for T3 after exclusion of loci with absolute ΔlnL ≤ 2), suggesting that ILS may not be the primary cause of the discordance among loci. Another plausible explanation for the conflict is heteroplasmic recombination, which can occur in species with biparental plastome inheritance <ns0:ref type='bibr' target='#b80'>(Walker et al., 2019)</ns0:ref>. Although heteroplasmic recombination has been reported with clear evidence in Brachypodium and Picea <ns0:ref type='bibr' target='#b73'>(Sullivan et al., 2017;</ns0:ref><ns0:ref type='bibr' target='#b59'>Sancho et al., 2018)</ns0:ref>, to our knowledge it has never been documented in Lauraceae. Based on the data reported here, it is too early to draw a firm conclusion about the causes of the conflict in phylogenetic signals. Although fully resolved phylogenies may still remain elusive based on different genomic compartments (i.e., nuclear, mitochondrial and plastid), phylogenomic studies that incorporate these compartments can provide new insights into tree discordance and its underlying causes <ns0:ref type='bibr' target='#b28'>(Koenen et al., 2020)</ns0:ref>. Therefore, more genetic information (e.g., nuclear genes) will be required to solve this problem in future work.</ns0:p></ns0:div>
<ns0:div><ns0:head>Conclusion</ns0:head><ns0:p>In summary, this study revealed that Laureae plastomes are conserved in structure, size and gene content. A conflicting phylogenetic signal was detected between coding and non-coding regions, suggesting that the plastid genome should not be treated as a single unit. ML trees based on PeerJ reviewing PDF | (2020:05:49532:1:1:NEW 30 Jul 2020)</ns0:p><ns0:p>Manuscript to be reviewed coding and non-coding regions were largely congruent except at the contentious node, indicating that coding regions are as informative as non-coding regions and that the influence of non-coding regions on tree inference should be evaluated. We also found that small subsets of plastome loci determined the topology at specific nodes, consistent with the results of a previous study <ns0:ref type='bibr' target='#b62'>(Shen, Hittinger & Rokas, 2017)</ns0:ref>. Through quantification and analysis of intra-plastome conflicts, the sister relationship of subclade I (including Lindera communis, L. glauca and L. nacusua) and II (including Laurus azorica, L. nobilis, Lindera megaphylla, Litsea acutivena, L. glutinosa, L. monopetala and L. pungens) was supported by our study. Biological factors may contribute to the conflicts among plastid loci; however, more information is needed to determine the underlying mechanism(s). </ns0:p></ns0:div><ns0:figure xml:id='fig_0'><ns0:head /><ns0:label /><ns0:figDesc>Figs. 2 and S1-S4), except for the ML tree based on the IR region (71%, Fig.S5). This result was</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_1'><ns0:head>Figure 1</ns0:head><ns0:label>1</ns0:label><ns0:figDesc>Figure 1</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_2'><ns0:head>Figure 4</ns0:head><ns0:label>4</ns0:label><ns0:figDesc>Figure 4</ns0:figDesc></ns0:figure>
<ns0:note place='foot'>PeerJ reviewing PDF | (2020:05:49532:1:1:NEW 30 Jul 2020)Manuscript to be reviewed</ns0:note>
</ns0:body>
" | "Dear editor
We thank the reviewers that their comments are very helpful to improve our manuscript. We have already revised our manuscript to address their concerns.
In particular, we have implemented a partitioning strategy to our datasets and found that partitioning did not affect topology inference. We have carried out positive selection test and found no impact of selection on our tree topology. In addition to corrections required by the reviewers, we also followed guidance of PeerJ and revised several places to improve our manuscript.
According to International Code of Botanical Nomenclature, when the plant Latin name was firstly mentioned in the text, we added the author of name.
We believe that the revised manuscript is now ready for publication in PeerJ.
Tian-Wen Xiao
South China Botanical Garden, University of Chinese Academy of Sciences
On behalf of all authors.
Reviewer 1 (Anonymous):
Line 42 The authors fail to include one major biological factor that has the potential to introduce artifacts to phylogenies. This is natural selection, especially positive selection, of protein coding sequences. Selection is well-documented for plastid genes and has the potential to bias phylogenies (e.g., see Piot et al., 2018; Saarela et al. 2018).
Response: We agree with the reviewer that positive selection is an important biological factor which may have impact on tree topology, thus we have added this biological factor in “Introduction” section (lines 58-61; please note that all line numbers in this rebuttal letter refer to lines of the “Revised Manuscript with tracked changes”).
Lines 52-53 The phrase: “…biparental inheritance and heteroplasmy…has been reported…” should be “…biparental inheritance and heteroplasmy…have been reported…”
Response: We agree with the reviewer and have changed “has” to “have” (line 53).
Lines 72-73, 330-332 A recent paper, in which the impact of non-coding partitions in the plastomes of Poaceae were investigated, is Saarela et al. (2018). The authors should cite this relevant paper and compare their methods and results with those of Saarela et al. (2018).
Response: We agree that the study of Saarela et al. (2018) is relevant and important. Both our and their studies showed that non-coding regions had impact on tree topology and could lead to conflicts, thus we have cited their study (lines 58-61). However, Saarela et al. (2018) mainly focused on gapped sites and positively selected sites, they did neither explore which loci led to conflicts between ML trees based on datasets with or without non-coding regions nor discuss why these loci caused the conflicts. Thus, we only mentioned results of Saarela et al. (2018) about the effect of non-coding partitions in “Discussion” section (line 360).
Lines 99-100 The phrase: “…Neolitsea sericea is a constructive species …” is unclear and should be reworded. Perhaps the intended meaning was that N. sericea is a source of lumber used for construction.
Response: Corrected. “…Neolitsea sericea is a dominant species …” (line 100).
Line 124 The phrase: “…we now report 12 new plastomes…” should be “…we now report 12 newly sequenced plastomes…”
Response: Corrected. We have changed “report 12 new plastomes” to “report 12 newly sequenced plastomes” (line 124).
Lines 232-233 This sentence regarding model selection may be more appropriate in the “Methods” section of the manuscript.
Response: We agree that sentence regarding model selection should be in “Method” section, however, we already did like this (lines 178-180). The selected models are results of model selection and we think it is properly placed (see lines 246-247).
Line 362 The phrase: “…cause alternate but conflicting topologies…” should be “…cause alternate, but conflicting topologies…” (insert comma).
Response: We agree and have inserted a comma (line 388).
Lines 273-275 When small subsets of genes have a “decisive role” in the inference of phylogenies (line 274), the possibility exists that selection in one or two genes (in this case, ycf2 and psaB) may bias the result creating false phylogenetic signal. The authors should acknowledge this possibility.
Response: As positive selection has the potential to bias phylogenies, we performed selection tests using two branch models (M0 vs. M2) in PAML(Yang, 2007). Our results suggested that psaB and ycf2 did not evolve under positive selection (Supplemental Table S4), therefore, selection is not the cause of conflict in our study (lines 191-198, 278-280, 392-395).
The original release date for the paper by Duvall et al. was 2019. However, the paper wasn’t officially published in Botanical Journal of the Linnean Society until 2020. The date of the citations (lines 62, 180, 361, 363) and the reference (line 456) should probably be changed to 2020.
Response: Corrected (lines 62-63, 187, 387, 389)
One factor that is overlooked in this otherwise thorough study is the possibility of selection bias. Selection, and the potential impact of positive selection artifacts on phylogenetic results, would ideally be measured for the protein coding sequences, but at least should be discussed.
Response: We conducted selection tests for plastid coding genes and found no evidence for positive selection (lines 191-198, 278-280). We also discussed the influence of selection on tree topology in “Discussion” section (lines 392-395).
Reviewer 2 (Anonymous):
The figure captions could be more complete by describing the phylogenetic trees in full. The ML trees should mention the program that they were inferred with. The ASTRAL tree should mention that ASTRAL the tree was inferred using a multispecies coalescent approach in ASTRAL.
Response: We agree and have mentioned the RAxML-NG program in the legends of ML trees, and the “multispecies coalescent approach” in legend of figure 3.
Given that many trees were inferred in the study, the reason for presenting those two particular trees in the main text should be justified.
Response: Though there were many ML trees in our analysis, we decided to present figure 2 in the main text after we finished all analysis and confirmed this tree topology was the right one. As for figure 3, multispecies coalescent approach is a widely used method and always compared to the concatenation method, we therefore thought it was reasonable to present this tree (Fig. 3) in main text as well. We decided not to explain this in our manuscript for concision.
The authors point out that: “By contrast, relatively few studies have focused on tree conflicts among plastid genes”, and “Aside from biological factors, non-biological factors (e.g., outlier genes, uninformative loci, and gaps) may cause conflict as well.” These are good points, and the authors have cited several of the few recent studies that address conflict among plastid genes. However, an additional, relevant citation would be: Foster et al. (2018) Plastome sequences and exploration of tree-space help to resolve the phylogeny of riceflowers (Thymelaeaceae: Pimelea), Molecular Phylogenetics and Evolution, 127: 156-167. In Foster et al. (2018), a topology-clustering approach was used, finding discordance among chloroplast gene trees. The discordance was then investigated with respect to strength of selection (dN/dS), tree length, and GC content. These are useful biological reasons for discordance among chloroplast genes that should be mentioned in the present study.
Response: Corrected. We have cited the relevant paper and include positive selection, tree length (gene evolutionary rate) and GC content in biological factors that can cause conflicts (lines 58-61).
I found the following sentence a little confusing: “The sum of absolute ΔlnL of T1 was higher than that of T2 and T3 (Fig. 4), suggesting that our data support the topology of T1 rather than T2 or T3.” If it could be rephrased to better explain the rationale behind justifying T1 as better based on likelihood ratios, that would be beneficial.
Response: In short, the larger the log likelihood value, the better the topology is. If we name sites in plastomes that support T1 than T2 site1, site2, site3, then ΔlnL(T1-T2) = lnL(site1_T1)-lnL(site1_T2)+lnL(site2_T1)-lnL(site2_T2)+lnL(site3_T1)-lnL(site3_T2); if we name sites in plastomes that support T2 than T1 site4, site5, site6, then ΔlnL(T2-T1) = lnL(site4_T2)-lnL(site4_T1)+lnL(site5_T2)-lnL(site5_T1)+lnL(site6_T2)-lnL(site6_T1). ΔlnL(T1-T2) > ΔlnL(T2-T1) means the plastome data support T1 than T2, that is to say, T1 is more likely to be the true topology. We decided not to add this explanation in our manuscript for concision.
Finally, it should be mentioned in the discussion that, at deep timescales, non-coding genes are often discarded for being uninformative, or for being misleading due to saturation.
Response: We agree and have mentioned “Non-coding regions are often discarded for being uninformative, or for being misleading due to saturation at deep time scales (Foster, Henwood & Ho, 2018)” in the “Discussion” section (lines 364-365).
The authors describe six main, initial data sets for phylogenetic analysis: “(1) complete plastome (cp), (2) coding regions (CDS), (3) non-coding regions (non-CDS), (4) large single copy region (LSC), (5) small single copy region (SSC), and (6) inverted repeat region (IR)”. As the name of the IR implies, it contains two copies of its constituent genes. I’m assuming that only one copy of each gene was used for phylogenetic analysis, but this should be made explicit in the manuscript.
Response: We agree with the reviewer and have rephrased the sentence: “(6) one inverted repeat region” (line 176).
All sequences for each data set were aligned, then analyzed using RAxML-NG to estimate trees. A key part of multilocus phylogenetic analysis is partitioning of the data, but there is no mention of partitioning in the manuscript. For example, the CDS data set could/should have been partitioned by gene and by codon position, with appropriate nucleotide substitution models applied to each partition.
Was any partitioning carried out? If yes, please report it in the manuscript. If not, why not? Given the impact of adequate partitioning on topology inference, it’s an important omission.
Response: In the previous manuscript, we did not use the partitioning strategy. However, we agree with the reviewer that our datasets should be partitioned. Because large partition number could lead to overparameterization (Lanfear et al., 2012), we decided to obtain the best partition scheme with PartitionFinder 2 (Lanfear et al., 2016). Each coding gene was divided into three blocks based on codon position (see Supplemental File 1). To reduce computation cost, tRNA + rRNA, introns, and intergenic spacers were treated as three independent data blocks in the configuration file of PartitionFinder 2 (see Supplemental File 1). ML tree topology based on partitioned plastomes (one IR removed, cp-reduced) was congruent with topology based on unpartitioned complete plastomes (cp) (Figure S7 vs. Figure 2), and tree topology based on partitioned coding regions (CDS) was congruent with unpartitioned CDS (Figure S8 vs. Figure S1), suggesting no impact of partitioning on topology inference. See lines 182-186, 271-274, 390-392.
Reference:
Foster CSP, Henwood MJ, Ho SYW. 2018. Plastome sequences and exploration of tree-space help to resolve the phylogeny of riceflowers (Thymelaeaceae: Pimelea). Molecular Phylogenetics and Evolution 127:156-167 DOI: 10.1016/j.ympev.2018.05.018.
Lanfear R, Calcott B, Ho SYW, Guindon S. 2012. PartitionFinder: Combined Selection of Partitioning Schemes and Substitution Models for Phylogenetic Analyses. Molecular Biology and Evolution 29:1695-1701 DOI: 10.1093/molbev/mss020.
Lanfear R, Frandsen PB, Wright AM, Senfeld T, Calcott B. 2016. PartitionFinder 2: New methods for selecting partitioned models of evolution for molecular and morphological phylogenetic analyses. Molecular Biology and Evolution 34:772-773 DOI: 10.1093/molbev/msw260.
Saarela JM, Burke SV, Wysocki WP, Barrett MD, Clark LG, Craine JM, Peterson PM, Soreng RJ, Vorontsova MS, Duvall MR. 2018. A 250 plastome phylogeny of the grass family (Poaceae): topological support under different data partitions. PeerJ 6:e4299 DOI: 10.7717/peerj.4299.
Yang Z. 2007. PAML 4: Phylogenetic Analysis by Maximum Likelihood. Molecular Biology and Evolution 24:1586-1591 DOI: 10.1093/molbev/msm088.
" | Here is a paper. Please give your review comments after reading it. |
9,799 | "<ns0:abstract xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:p>Background. Gene tree discordance is common in phylogenetic analyses. Many phylogenetic studies have excluded non-coding regions of the plastome without evaluating their impact on tree topology. In general, plastid loci have often been treated as a single unit, and tree discordance among these loci has seldom been examined. Using samples of Laureae (Lauraceae) plastomes, we explored plastome variation among the tribe, examined the influence of non-coding regions on tree topology, and quantified intra-plastome conflict.</ns0:p><ns0:p>Results. We found that the plastomes of Laureae have low inter-specific variation and are highly similar in structure, size, and gene content. Laureae was divided into three groups, subclades I, II and III. The inclusion of non-coding regions changed the phylogenetic relationship among the three subclades. Topologies based on coding and non-coding regions were largely congruent except for the relationship among subclades I, II and III. By measuring the distribution of phylogenetic signal across loci that supported different topologies, we found that nine loci (two coding regions, two introns and five intergenic spacers) played a critical role at the contentious node.</ns0:p></ns0:div>
<ns0:div><ns0:head>Conclusions.</ns0:head><ns0:p>Our results suggest that subclade III and subclade II are successively sister to subclade I. Conflicting phylogenetic signals exist between coding and non-coding regions of Laureae plastomes. Our study highlights the importance of evaluating the influence of non-coding regions on tree topology and emphasizes the necessity of examining discordance among different plastid loci in phylogenetic studies.</ns0:p></ns0:div>
</ns0:abstract>
<ns0:body xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:head>Introduction</ns0:head><ns0:p>Gene tree discordance is relatively common in phylogenomic studies. The conflicts can be caused by biological factors like incomplete lineage sorting (ILS), hybridization, horizontal gene transfer, gene loss, and gene duplication <ns0:ref type='bibr' target='#b44'>(Maddison, 1997;</ns0:ref><ns0:ref type='bibr' target='#b74'>Sun et al., 2015;</ns0:ref><ns0:ref type='bibr' target='#b18'>Gonçalves et al., 2019;</ns0:ref><ns0:ref type='bibr' target='#b61'>Sato et al., 2019)</ns0:ref>. Most relevant studies have focused on incongruent tree topologies among different genomic compartments <ns0:ref type='bibr' target='#b74'>(Sun et al., 2015;</ns0:ref><ns0:ref type='bibr' target='#b96'>Zhao et al., 2016;</ns0:ref><ns0:ref type='bibr' target='#b80'>Walker et al., 2019)</ns0:ref> because these genes have evolved independently and their gene tree topologies have been influenced by biological processes. By contrast, relatively few studies have focused on tree conflicts among plastid genes (e.g., <ns0:ref type='bibr' target='#b15'>Foster, Henwood & Ho, 2018;</ns0:ref><ns0:ref type='bibr' target='#b18'>Gonçalves et al., 2019;</ns0:ref><ns0:ref type='bibr' target='#b80'>Walker et al., 2019;</ns0:ref><ns0:ref type='bibr' target='#b95'>Zhang et al., 2020)</ns0:ref>. Usually, plastomes are considered to be uniparentally inherited and to have evolved as a single unit, free from such biological sources of conflict <ns0:ref type='bibr' target='#b2'>(Birky, 1995;</ns0:ref><ns0:ref type='bibr' target='#b82'>Wicke et al., 2011)</ns0:ref>. However, the branched and linear structure of plastid DNA, which arose from recombination-dependent replication, is indicative of recombination <ns0:ref type='bibr' target='#b50'>(Oldenburg & Bendich, 2016;</ns0:ref><ns0:ref type='bibr' target='#b58'>Ruhlman et al., 2017)</ns0:ref>. In addition, biparental inheritance and heteroplasmy (e.g., the presence of different plastomes within an individual or a cell) have been reported in seed plants <ns0:ref type='bibr' target='#b75'>(Szmidt, Aldén & Hällgren, 1987;</ns0:ref><ns0:ref type='bibr' target='#b25'>Johnson & Palmer, 1989;</ns0:ref><ns0:ref type='bibr' target='#b56'>Reboud & Zeyl, 1994;</ns0:ref><ns0:ref type='bibr' target='#b4'>Carbonell-Caballero et al., 2015)</ns0:ref>. Heteroplasmy may, in rare cases, give rise to heteroplasmic recombination, which has been invoked to explain gene tree discordance <ns0:ref type='bibr' target='#b45'>(Marshall, Newton & Ritland, 2001;</ns0:ref><ns0:ref type='bibr' target='#b73'>Sullivan et al., 2017;</ns0:ref><ns0:ref type='bibr' target='#b60'>Sancho et al., 2018)</ns0:ref>. In addition to recombination events, the transfer of genes among plastid, mitochondrial and nuclear genomes; positive selection; tree length (a proxy for evolutionary rate); and GC content may also generate Manuscript to be reviewed phylogenomic conflict (e.g., <ns0:ref type='bibr' target='#b72'>Stegemann et al., 2003;</ns0:ref><ns0:ref type='bibr' target='#b67'>Smith, 2014;</ns0:ref><ns0:ref type='bibr' target='#b86'>Wysocki et al., 2015;</ns0:ref><ns0:ref type='bibr' target='#b54'>Piot et al., 2018;</ns0:ref><ns0:ref type='bibr' target='#b59'>Saarela et al., 2018;</ns0:ref><ns0:ref type='bibr' target='#b15'>Foster, Henwood & Ho, 2018)</ns0:ref>. Aside from biological factors, non-biological factors (e.g., outlier genes, uninformative loci, and gaps) may cause conflict as well. For example, <ns0:ref type='bibr' target='#b10'>Duvall, Burke & Clark (2020)</ns0:ref> found that alternative topologies arose from alignment gaps. Given that most studies assume no conflict and treat the plastome as a single unit, taking biological and non-biological factors into consideration and quantifying the extent of conflict among different plastid loci is of great importance <ns0:ref type='bibr' target='#b85'>(Wolfe & Randle, 2004)</ns0:ref>.</ns0:p><ns0:p>Owing to the rapid development of next-generation sequencing (NGS), more plastomes are becoming available at a reasonable cost, driving advances in phylogenomics and promoting a more comprehensive understanding of plant evolution <ns0:ref type='bibr' target='#b32'>(Li et al., 2019)</ns0:ref>. Phylogenetic relationships among Lauraceae <ns0:ref type='bibr' target='#b69'>(Song et al., 2017)</ns0:ref>, as well as many other groups (e.g., <ns0:ref type='bibr' target='#b12'>Eserman et al., 2014;</ns0:ref><ns0:ref type='bibr' target='#b1'>Barrett et al., 2016)</ns0:ref>, have been well resolved using plastome data. In phylogenomic studies of plastomes <ns0:ref type='bibr' target='#b19'>(Guo et al., 2017;</ns0:ref><ns0:ref type='bibr' target='#b18'>Gonçalves et al., 2019;</ns0:ref><ns0:ref type='bibr' target='#b91'>Xu et al., 2019;</ns0:ref><ns0:ref type='bibr' target='#b32'>Li et al., 2019)</ns0:ref>, plastome coding genes have generally been used, and non-coding regions have been excluded.</ns0:p><ns0:p>Only a few studies have noted the potential impact of non-coding regions on tree topology. <ns0:ref type='bibr' target='#b52'>Parks, Cronn & Liston (2009)</ns0:ref> revealed that the phylogenetic position of Pinus albicaulis Engelm. based on complete plastomes differed from that based on exon sequences. A similar situation also occurred for phylogenetic relationships within Rubiaceae <ns0:ref type='bibr' target='#b84'>(Wikström, Bremer & Rydin, 2020)</ns0:ref>, suggesting that there were conflicting phylogenetic signals between coding-and non-coding regions. Because tree topology is the foundation of comparative studies that infer biogeographic history, phylogenetic diversity and other evolutionary patterns <ns0:ref type='bibr'>(Walker et al.,</ns0:ref> and in evergreen broadleaf forests <ns0:ref type='bibr' target='#b81'>(Wang et al., 2009)</ns0:ref>, and Laurus nobilis L. has been used in remedies for centuries <ns0:ref type='bibr' target='#b47'>(Nayak et al., 2006)</ns0:ref>.</ns0:p><ns0:p>Although Laureae is monophyletic, generic delimitation within this tribe remains unclear <ns0:ref type='bibr' target='#b29'>(Kostermans, 1957;</ns0:ref><ns0:ref type='bibr' target='#b23'>Hutchinson, 1964;</ns0:ref><ns0:ref type='bibr' target='#b38'>Li et al., 2008b)</ns0:ref>. Adenodaphne, endemic to New Caledonia, is closely related to Litsea <ns0:ref type='bibr' target='#b5'>(Chanderbali, van der Werff & Renner, 2001)</ns0:ref>. However, morphological confusion still exists between this genus and Litsea, meaning that their distinctiveness and the monophyly of Adenodaphne require further study <ns0:ref type='bibr' target='#b5'>(Chanderbali, van der Werff & Renner, 2001)</ns0:ref>. Actinodaphne is polyphyletic and closely related to the monophyletic genus Neolitsea <ns0:ref type='bibr' target='#b39'>(Li et al., 2007;</ns0:ref><ns0:ref type='bibr' target='#b38'>Li et al., 2008b;</ns0:ref><ns0:ref type='bibr' target='#b13'>Fijridiyanto & Murakami, 2009a</ns0:ref><ns0:ref type='bibr' target='#b14'>, 2009b)</ns0:ref>.</ns0:p><ns0:p>Although <ns0:ref type='bibr' target='#b13'>Fijridiyanto & Murakami (2009a</ns0:ref><ns0:ref type='bibr' target='#b14'>, 2009b)</ns0:ref> argued that Actinodaphne was monophyletic, the species of Actinodaphne sampled in their analyses were totally different from those sampled in <ns0:ref type='bibr' target='#b39'>Li et al. (2007)</ns0:ref> and <ns0:ref type='bibr' target='#b38'>Li et al. (2008b)</ns0:ref>. Furthermore, Lindera and Litsea have been shown to be polyphyletic, with Dodecadenia, Iteadaphne, Laurus, Parasassafras and Sinosassafras nested within them <ns0:ref type='bibr' target='#b37'>(Li et al., 2004;</ns0:ref><ns0:ref type='bibr' target='#b38'>Li et al., 2008b)</ns0:ref>. <ns0:ref type='bibr' target='#b42'>Liu et al. (2017)</ns0:ref> used three plastid barcode loci combined with the internal transcribed spacer (ITS) region for species identification and found that the Laureae tree was polytomic. Despite these efforts, phylogenetic relationships among and within these genera have been poorly resolved based on molecular markers like the ITS, the external transcribed spacer (ETS), matK, trnL-F and trnH-psbA.</ns0:p><ns0:p>Compared with these molecular markers, complete plastomes have better performance at the species level within Laureae, although generic delimitation still remains unclear due to limited taxon sampling <ns0:ref type='bibr' target='#b97'>(Zhao et al., 2018;</ns0:ref><ns0:ref type='bibr' target='#b68'>Song et al., 2019;</ns0:ref><ns0:ref type='bibr' target='#b76'>Tian, Ye & Song, 2019)</ns0:ref>. Manuscript to be reviewed Thirty-five plastomes representing 28 species and six genera of Laureae have been published (Table <ns0:ref type='table'>S1</ns0:ref>). Compared with the vast diversity of Laureae, the published plastome data for this group are relatively limited. Hence, we now report 12 newly sequenced plastomes (Table <ns0:ref type='table'>1</ns0:ref>) and combine them with existing plastomes to address three primary goals: (1) reinvestigation of phylogenetic relationships within Laureae; (2) examination of conflict between coding and noncoding regions; and (3) quantification of conflicts among different plastid loci. were collected and identified by the authors (Table <ns0:ref type='table'>1</ns0:ref>). Voucher specimens were deposited in the herbarium of the South China Botanical Garden (IBSC) at the Chinese Academy of Sciences. No specific permissions were required for the relevant locations and activities. Including the plastomes downloaded from GenBank and the Lauraceae Chloroplast Genome Database (LCGDB, https://lcgdb.wordpress.com) (Table <ns0:ref type='table'>S1</ns0:ref>), this study included 47 Laureae plastomes, representing seven genera and all subclades identified by <ns0:ref type='bibr' target='#b68'>Song et al. (2019)</ns0:ref>. Twelve plastomes from other tribes were also downloaded (Table <ns0:ref type='table'>S1</ns0:ref>).</ns0:p></ns0:div>
<ns0:div><ns0:head>Methods</ns0:head><ns0:p>PeerJ reviewing PDF | (2020:05:49532:2:0:NEW 3 Sep 2020)</ns0:p><ns0:p>Manuscript to be reviewed Genomic DNA was extracted from silica-gel-dried leaf tissue using the cetyl trimethyl ammonium bromide (CTAB) method <ns0:ref type='bibr' target='#b9'>(Doyle & Doyle, 1987)</ns0:ref>. The yields of genomic DNA extracts were quantified by fluorometric quantification on a Qubit instrument (Invitrogen, Carlsbad, California, USA) using the dsDNA HS kit, and the DNA size distribution was assessed visually on a 1% agarose gel. DNA libraries with an average insert size of 270 bp were prepared by the Beijing Genomics Institute (BGI, Shenzhen, China). Paired-end reads of 2 × 151 bp were generated on the Illumina X ten sequencing system (Illumina Inc.).</ns0:p></ns0:div>
<ns0:div><ns0:head>Plastid genome assembly, annotation and comparison</ns0:head><ns0:p>Low-quality reads and adaptors were removed using Trimmomatric v0.36 <ns0:ref type='bibr' target='#b3'>(Bolger, Lohse & Usadel, 2014)</ns0:ref>, generating approximately 3 Gb of high-quality clean reads per sample. The clean reads were analyzed for quality control with FastQC <ns0:ref type='bibr' target='#b0'>(Andrews, 2010)</ns0:ref> and then used to assemble plastomes with NOVOPlasty v2.7.2 <ns0:ref type='bibr' target='#b8'>(Dierckxsens, Mardulyn & Smits, 2016)</ns0:ref>. To guarantee assembly quality, clean reads were mapped to the assembled plastid genomes using the Burrows-Wheeler Aligner (BWA 0.7.17-r1188 <ns0:ref type='bibr' target='#b34'>(Li & Durbin, 2010)</ns0:ref>) and samtools 1.9 <ns0:ref type='bibr' target='#b36'>(Li et al., 2009)</ns0:ref>, and were visually checked in Geneious Prime 2019.1.</ns0:p></ns0:div>
<ns0:div><ns0:head>Plastome annotation was performed using the program GeSeq -Annotation of Organellar</ns0:head><ns0:p>Genomes <ns0:ref type='bibr' target='#b77'>(Tillich et al., 2017)</ns0:ref>. Start and stop codons were inspected and manually adjusted in Geneious Prime when necessary. Plastomes were submitted to GenBank (MN274947, MN428456-MN428466). Maps of all 12 plastomes were drawn using the OrganellarGenomeDRAW tool (OGDRAW) <ns0:ref type='bibr' target='#b43'>(Lohse et al., 2013)</ns0:ref>. A summary of the newly PeerJ reviewing <ns0:ref type='table'>PDF | (2020:05:49532:2:0:NEW 3 Sep 2020)</ns0:ref> Manuscript to be reviewed sequenced plastomes is presented in Table <ns0:ref type='table'>2</ns0:ref>.</ns0:p><ns0:p>To illustrate interspecific sequence variation within Laureae, plastomes of A. obovata, I. caudata, Laurus nobilis (KY085912), Lindera erythrocarpa, Litsea acutivena, N. pallens and Parasassafras confertiflorum (Meisn.) D. G. Long (MH729378) were aligned using MAFFT <ns0:ref type='bibr' target='#b26'>(Katoh & Standley, 2013)</ns0:ref> with default settings. Sequence identity was plotted with the mVISTA program using the LAGAN mode <ns0:ref type='bibr' target='#b16'>(Frazer et al., 2004)</ns0:ref>, with Lindera glauca (Siebold et Zucc.)</ns0:p><ns0:p>Bl. (MF188124) as a reference.</ns0:p></ns0:div>
<ns0:div><ns0:head>Phylogenetic reconstruction and tests for selection</ns0:head><ns0:p>To evaluate potential conflicts, phylogenetic trees were constructed using maximum likelihood (ML) methods based on six datasets: (1) complete plastome (CP), (2) coding regions (CDS), (3) non-coding regions (non-CDS), (4) large single copy region (LSC), (5) small single copy region (SSC), and (6) one inverted repeat region (IR).</ns0:p><ns0:p>Sequences were aligned using MAFFT with default settings and manually edited with BioEdit v7.2.5 <ns0:ref type='bibr' target='#b20'>(Hall, 1999)</ns0:ref> when necessary. The best-fitting DNA substitution models for the six unpartitioned datasets were selected using ModelTest-NG <ns0:ref type='bibr' target='#b7'>(Darriba et al., 2020)</ns0:ref> under the corrected Akaike Information Critierion (AICc). The aligned sequences and selected DNA substitution models were used for ML analyses, and ML trees were constructed using RAxML-NG <ns0:ref type='bibr' target='#b30'>(Kozlov et al., 2019)</ns0:ref>. We also implemented a partitioning strategy on two datasets, the CP with one IR region removed (CP-reduced) and CDS (configuration details shown in Supplemental File 1). The optimal partitioning schemes for each dataset were inferred with PeerJ reviewing <ns0:ref type='table'>PDF | (2020:05:49532:2:0:NEW 3 Sep 2020)</ns0:ref> Manuscript to be reviewed PartitionFinder 2 <ns0:ref type='bibr' target='#b31'>(Lanfear et al., 2016)</ns0:ref>, and the optimal partitioning schemes, and nucleotide substitution models for each partition were used for ML analyses in RAxML-NG.</ns0:p><ns0:p>Because gaps can affect tree topology <ns0:ref type='bibr' target='#b10'>(Duvall, Burke & Clark, 2020)</ns0:ref>, we also performed the following analysis based on the CP dataset. 'Mask Alignment' in Geneious Prime was used to strip the gaps from the MAFFT alignment, with the threshold set to 0 (no gaps), 2%, 10%, 20%, 50% or 75%. The resulting alignments were used to infer ML trees in RAxML-NG.</ns0:p><ns0:p>Positive selection on plastid coding genes has the potential to bias phylogenies (e.g., <ns0:ref type='bibr' target='#b54'>Piot et al., 2018;</ns0:ref><ns0:ref type='bibr' target='#b59'>Saarela et al., 2018)</ns0:ref>, and we therefore performed positive selection tests using CODEML in PAML 4.9j <ns0:ref type='bibr' target='#b92'>(Yang, 2007)</ns0:ref>. Coding genes were extracted and aligned in Geneious Prime using MAFFT, stop codons were removed manually, and the aligned sequences were converted to PAML format. Because site models allow dN/dS ratio to vary among different sites, we implemented M0, M1a, M2a, M3, M7 and M8 to identify positively selected sites. Likelihood ratio tests (LRTs) were performed using pchisq function in R 3.6.2 (R Core Team, 2018) to test if there was significant difference between models (M0 vs M3, M2a vs M1a, M8 vs M7). We manually deleted positively selected sites when LTRs was significant (M2a vs M1a and/or M8 vs M7 with p value less than 0.05). Coding gene alignments with positively selected sites removed were concatenated (CDS-reduced dataset), and used for ML tree inference to examine whether positive selection can bias phylogeny or not.</ns0:p></ns0:div>
<ns0:div><ns0:head>Node support investigation and tree topology tests</ns0:head><ns0:p>Because gene contents were not identical among Cryptocaryeae, Cassytha, Caryodaphnopsis, We extracted all loci (coding regions, introns, tRNA, rRNA and intergenic spacers) using a python script <ns0:ref type='bibr' target='#b24'>(Jin, 2019)</ns0:ref> and aligned them using MAFFT with default settings. These alignments were used to infer gene trees by rapid bootstrap analyses (option -f a) in RAxML <ns0:ref type='bibr' target='#b71'>(Stamatakis, 2014)</ns0:ref> with the GTRGAMMA model. The number of bootstrap replicates was set to 1000, as <ns0:ref type='bibr' target='#b66'>Simmons & Kessenich (2019)</ns0:ref> have suggested that fewer replicates may be insufficient to find the optimal gene tree topology. The best-scoring ML trees were used to estimate the species tree with local posterior probability (LPP) <ns0:ref type='bibr' target='#b62'>(Sayyari & Mirarab, 2016)</ns0:ref> in ASTRAL III <ns0:ref type='bibr' target='#b94'>(Zhang et al., 2018)</ns0:ref>.</ns0:p><ns0:p>We performed constrained maximum likelihood analyses in IQ-TREE <ns0:ref type='bibr' target='#b48'>(Nguyen et al., 2014)</ns0:ref> to obtain the ML trees that supported different topologies. To understand which loci supported the alternative topologies, we calculated site-wise log-likelihood values for each topology in RAxML using option '-f G'. After obtaining site-wise lnL differences, we converted site-wise differences to locus-wise lnL differences (ΔlnL) in R 3.6.2. The lnL differences were plotted against each locus using ggplot2 <ns0:ref type='bibr' target='#b83'>(Wickham, 2016)</ns0:ref>. It has been suggested that loci with an absolute ΔlnL > 2 are statistically significant <ns0:ref type='bibr' target='#b11'>(Edwards, 1984)</ns0:ref>. Therefore, we conducted separate ML analyses on datasets from which these loci (absolute ΔlnL > 2) had been removed to test whether small subsets of sequence matrices determined tree topology <ns0:ref type='bibr'>(Shen, Hittinger & Rokas,</ns0:ref> PeerJ reviewing PDF | (2020:05:49532:2:0:NEW 3 Sep 2020)</ns0:p><ns0:p>Manuscript to be reviewed</ns0:p></ns0:div>
<ns0:div><ns0:head>2017).</ns0:head><ns0:p>The Kishino-Hasegawa test (KH test) <ns0:ref type='bibr' target='#b27'>(Kishino & Hasegawa, 1989)</ns0:ref>, Shimodaira-Hasegawa test (SH test) <ns0:ref type='bibr' target='#b65'>(Shimodaira & Hasegawa, 1999)</ns0:ref> and Approximately-Unbiased test (AU test) <ns0:ref type='bibr' target='#b64'>(Shimodaira, 2002)</ns0:ref> were used in IQ-TREE to assess the statistical significance of incongruence based on complete plastomes (including only one copy of the IR regions). We specified 10,000 RELL (resampling of estimated log-likelihoods) replicates for the topological tests.</ns0:p></ns0:div>
<ns0:div><ns0:head>Results</ns0:head></ns0:div>
<ns0:div><ns0:head>Plastome features of Laureae</ns0:head><ns0:p>The sizes of the 12 newly generated Laureae plastid genomes ranged from 152,132 bp (Litsea szemaois) to 152,916 bp (Lindera erythrocarpa) (Table <ns0:ref type='table'>2</ns0:ref>), similar to previously published Laureae plastomes (152,211-153,011 bp, Table <ns0:ref type='table'>S1</ns0:ref>). All had a typical quadripartite structure and were assembled into a single, circular and double-stranded DNA sequence (Fig. <ns0:ref type='figure' target='#fig_11'>1</ns0:ref>). The length of the LSC, SSC and IR regions ranged from 93,119 bp (Litsea szemaois) to 93,921 bp (Lindera erythrocarpa), 18,796 bp (N. pallens) to 18,936 bp (Litsea mollis), and 20,057 bp (A. obovata) to 20,144 bp (I. caudata), respectively, with little variation in size (Table <ns0:ref type='table'>2</ns0:ref>). The overall GC contents ranged from 39.1% to 39.2%. GC content was unequally distributed within the plastomes; it was highest in IR regions (44.4-44.5%), moderate in LSC regions (37.9-38.1%), and lowest in SSC regions (33.8-34.0%, Table <ns0:ref type='table'>2</ns0:ref>).</ns0:p><ns0:p>The 12 newly sequenced plastomes contained 112 single-copy genes: 78 protein-coding genes, 30 tRNA genes, and 4 rRNA genes (Table <ns0:ref type='table'>2 and Table S2</ns0:ref>). Sixteen genes had one intron, Manuscript to be reviewed and two genes had two introns. There were 13 duplicated genes in the IR regions (Table <ns0:ref type='table'>S2</ns0:ref>), and rps12, ycf1, and ycf2 were partly duplicated in the IR regions (Fig. <ns0:ref type='figure' target='#fig_11'>1</ns0:ref>).</ns0:p></ns0:div>
<ns0:div><ns0:head>Phylogenetic reconstruction and positive selection tests</ns0:head><ns0:p>The GTR+I+G4 model was selected for the six unpartitioned datasets (CP, CDS, non-CDS, LSC, SSC and IR). Perseeae was sister to Cinnamomeae and Laureae (Figs. <ns0:ref type='figure' target='#fig_11'>2 and S1-S5</ns0:ref>). All the ML trees indicated the monophyly of Laureae with high bootstrap (BS) support values (99-100%, Figs. 2 and S1-S4), except for the ML tree based on the IR region (71%, Fig. <ns0:ref type='figure'>S5</ns0:ref>). This result was caused by the low variability of the IR region (Fig. <ns0:ref type='figure'>S6</ns0:ref>). In the five ML trees (Figs. Subclade II was sister to subclade I based on four unpartitioned datasets (CP, non-CDS, LSC, SSC; Figs. 2 and S2-S4, respectively). However, subclade II was sister to subclade III rather than subclade I based on the unpartitioned CDS dataset (Fig. <ns0:ref type='figure' target='#fig_11'>S1</ns0:ref>). Both topologies were strongly supported. The sister relationship of subclades I and II was supported in the ML tree based on partitioned plastomes (one IR removed, CP-reduced dataset; Fig. <ns0:ref type='figure'>S7</ns0:ref>), and subclade II was sister to subclade III in the ML tree based on the partitioned CDS dataset (Fig. <ns0:ref type='figure'>S8</ns0:ref>), indicating that our results were robust to different partitioning schemes.</ns0:p><ns0:p>The sister relationship of subclades I and II (BS values ranging from 80% to 92%) was consistently revealed even as the percentage of gaps increased (Table <ns0:ref type='table'>S3</ns0:ref>), indicating that gaps had no impact on our tree topology. Positively selected sites were detected in 27 coding genes (Table <ns0:ref type='table'>S4</ns0:ref>). The ML tree based on CDS-reduced dataset supported a sister relationship of subclades II and III (Fig. <ns0:ref type='figure'>S9</ns0:ref>), consistent with ML trees based on CDS dataset (Figs. <ns0:ref type='figure' target='#fig_11'>S1 and S8</ns0:ref>), suggesting that positive selection did not affect the relationship of the three subclades.</ns0:p></ns0:div>
<ns0:div><ns0:head>Investigating incongruent nodes and differences in tree topology</ns0:head><ns0:p>The tree topology inferred from ASTRAL III (Fig. <ns0:ref type='figure'>3</ns0:ref>) was largely congruent with that of the ML trees (Figs. 2 and S1-S4), except that the former showed a sister relationship of subclade I and subclade III. We performed constrained maximum likelihood analyses in IQ-TREE and obtained Manuscript to be reviewed three suboptimal ML trees that supported the subclade II-subclade I (called T1 hereafter), subclade II-subclade III (T2) and subclade I-subclade III (T3) affinities. We extracted 243 loci and assessed how each locus supported one of the three topologies by examining the gene-wise log-likelihoods (Fig. <ns0:ref type='figure' target='#fig_12'>4</ns0:ref>). T1 was strongly supported by six loci (rpoC1 intron, trnG-trnfM, ndhA intron, psaJ-rpl33, rpl2-rpl23 and petN-psbM; absolute ΔlnL > 2); T2 was strongly supported by three loci (psaB, trnS-ycf3 and ycf2; absolute ΔlnL > 2); and T3 was moderately supported by one locus (clpP intron1; absolute ΔlnL > 1 and < 2) (Table <ns0:ref type='table'>S5</ns0:ref>). The sum of absolute ΔlnL of T1 was higher than that of T2 and T3 (Fig. <ns0:ref type='figure' target='#fig_12'>4</ns0:ref>), suggesting that our data support the topology of T1 rather than T2 or T3. After the removal of six loci (rpoC1 intron, trnG-trnfM, ndhA intron, psaJ-rpl33, rpl2-rpl23 and petN-psbM), a sister relationship of subclade II and subclade III was revealed (Fig. <ns0:ref type='figure' target='#fig_11'>S10</ns0:ref>). After the removal of three loci (psaB, trnS-ycf3, and ycf2), subclade II was sister to subclade I (Fig. <ns0:ref type='figure' target='#fig_11'>S11</ns0:ref>). These results underscore the decisive role played by small subsets of loci in phylogenetic inference.</ns0:p><ns0:p>The topological tests showed that T2 did not differ significantly from T1 (p > 0.05, Table <ns0:ref type='table'>S6</ns0:ref>).</ns0:p><ns0:p>T3 was statistically rejected by the KH and AU tests (p < 0.05) but not by the Shimodaira-Hasegawa (SH) test (p = 0.0505). That T3 was rejected according to the KH and AU tests suggests that the sister relationship between subclades I and III may be misleading.</ns0:p></ns0:div>
<ns0:div><ns0:head>Discussion</ns0:head></ns0:div>
<ns0:div><ns0:head>Plastome features</ns0:head><ns0:p>It has been noted that most plastid genomes of land plants and algae range from 120 to 160 Manuscript to be reviewed kilobase pairs (kb) in length <ns0:ref type='bibr' target='#b51'>(Palmer, 1985)</ns0:ref>. In this study, the plastid genome sizes of 12 species from five Laureae genera ranged from 152,132 bp to 152,916 bp, indicating that plastid genome size was conserved within Laureae. GC content was highest in the IR region rather than in the single copy regions, owing to the presence of a ribosomal RNA gene cluster in the IR region, consistent with a previous study <ns0:ref type='bibr' target='#b21'>(Huotari & Korpelainen, 2012)</ns0:ref>. GC contents of the IR, LSC and SSC regions of the newly sequenced plastomes were identical to those of nine Lindera species studied earlier <ns0:ref type='bibr' target='#b97'>(Zhao et al., 2018)</ns0:ref>. In contrast to the gene losses recognized in several Lauraceae lineages <ns0:ref type='bibr' target='#b69'>(Song et al., 2017)</ns0:ref>, our analysis revealed that gene content among Laureae was highly conserved. <ns0:ref type='bibr' target='#b69'>Song et al. (2017)</ns0:ref> suggested that plastome contraction in Lauraceae was largely driven by fragment loss events in the IR regions. In our study, we found no gene loss among Laureae plastomes.</ns0:p></ns0:div>
<ns0:div><ns0:head>Phylogenetic relationships within Laureae</ns0:head><ns0:p>Previous phylogenetic studies <ns0:ref type='bibr' target='#b69'>(Song et al., 2017;</ns0:ref><ns0:ref type='bibr' target='#b97'>Zhao et al., 2018)</ns0:ref> based on complete plastomes suggested that Laureae was sister to Cinnamomeae and that together they were sister to Perseeae.</ns0:p><ns0:p>The same phylogenetic relationships among these groups were recognized in our study (Figs. <ns0:ref type='figure'>2 and 3</ns0:ref>). In previous work, Actinodaphne and Neolitsea were resolved as monophyletic groups based on matK, ITS and rpb2 <ns0:ref type='bibr' target='#b13'>(Fijridiyanto & Murakami, 2009a</ns0:ref><ns0:ref type='bibr' target='#b14'>, 2009b)</ns0:ref>, but Actinodaphne was not a monophyletic group based on complete plastid genomes <ns0:ref type='bibr' target='#b68'>(Song et al., 2019)</ns0:ref>. In this study, the non-monophyletic status of Actinodaphne was supported. The conclusion of Actinodaphne monophyly may have been caused by sampling bias in previous studies <ns0:ref type='bibr'>(Fijridiyanto &</ns0:ref> PeerJ reviewing PDF | (2020:05:49532:2:0:NEW 3 Sep 2020)</ns0:p><ns0:p>Manuscript to be reviewed <ns0:ref type='bibr' target='#b14'>Murakami, 2009b</ns0:ref><ns0:ref type='bibr' target='#b13'>Murakami, , 2009a))</ns0:ref>. The monophyly of Neolitsea can be neither rejected nor supported in the present study. Actinodaphne cupularis (Hemsl.) Gamble was grouped with Neolitsea oblongifolia Merr. et Chun, N. pallens and N. chui Merr. with low bootstrap support (54%; Fig. <ns0:ref type='figure'>2</ns0:ref>), and sampling of Neolitsea and related genera was limited. Lindera and Litsea were polyphyletic in our analysis, consistent with previous studies <ns0:ref type='bibr' target='#b38'>(Li et al., 2008b;</ns0:ref><ns0:ref type='bibr' target='#b14'>Fijridiyanto & Murakami, 2009b)</ns0:ref>. The phylogenetic position of P. confertiflorum was unresolved based on ETS and ITS <ns0:ref type='bibr' target='#b38'>(Li et al., 2008b)</ns0:ref>, and the ambiguity of its position still remains, despite the integration of complete plastid genomes in our analysis and a previous study <ns0:ref type='bibr' target='#b40'>(Liao, Ye & Song, 2018)</ns0:ref>.</ns0:p><ns0:p>Subclade III was sister to subclade I and II in our study, consistent with previous analyses <ns0:ref type='bibr' target='#b97'>(Zhao et al., 2018;</ns0:ref><ns0:ref type='bibr' target='#b68'>Song et al., 2019;</ns0:ref><ns0:ref type='bibr' target='#b76'>Tian, Ye & Song, 2019)</ns0:ref>. The three Lindera species in subclade I share common morphological traits, such as alternate and pinninerved leaves, a persistent involucre, vegetative terminal buds in inflorescences and 3-merous flowers <ns0:ref type='bibr' target='#b33'>(Li et al., 2008a)</ns0:ref>. However, these characters also occur in several members of the other two subclades (e.g., Lindera benzoin (L.) Bl. and Laurus nobilis), perhaps resulting from convergent and/or parallel evolution <ns0:ref type='bibr' target='#b38'>(Li et al., 2008b)</ns0:ref>. These traits are not good indicators for delimiting the three subclades of Laureae. In subclade III, the trinerved or triplinerved species of Lindera (Lindera aggregata, L. chunii, L. fragrans, L. limprichtii, L. pulcherrima, L. supracostata, L. thomsonii and L. thomsonii var. vernayana) formed a well-supported clade in both our study and that of <ns0:ref type='bibr' target='#b76'>Tian, Ye & Song (2019)</ns0:ref>. However, triplinerved leaves also exist in most species of Neolitsea <ns0:ref type='bibr' target='#b38'>(Li et al., 2008b;</ns0:ref><ns0:ref type='bibr' target='#b33'>Li et al., 2008a)</ns0:ref>. Therefore, traditional morphological traits are of limited use in taxon delimitation, even within subclades of Laureae. Given the limited samples and data in our Manuscript to be reviewed analyses, more sampling and DNA sequences are needed to further elucidate the relationships within Laureae.</ns0:p></ns0:div>
<ns0:div><ns0:head>Phylogenetic incongruence in the plastome</ns0:head><ns0:p>Although many studies have treated plastid protein-coding genes or the complete plastome as a single unit (e.g., <ns0:ref type='bibr' target='#b68'>Song et al., 2019;</ns0:ref><ns0:ref type='bibr' target='#b76'>Tian, Ye & Song, 2019)</ns0:ref>, potential conflicts among sequence types (i.e., coding vs. non-coding regions) have been reported in several studies. By comparing phylogenies based on complete plastomes and coding regions <ns0:ref type='bibr' target='#b93'>(Yu et al., 2017)</ns0:ref>, it was inferred that non-coding regions did not significantly influence the tree topology of Theaceae. By contrast, non-coding regions had an impact on the phylogenetic relationships of several tribes in Rubiaceae <ns0:ref type='bibr' target='#b84'>(Wikström, Bremer & Rydin, 2020)</ns0:ref> and subtribes in Poaceae <ns0:ref type='bibr' target='#b59'>(Saarela et al., 2018)</ns0:ref>. A conflicting signal between coding and non-coding regions was also reported in Leguminosae <ns0:ref type='bibr' target='#b95'>(Zhang et al., 2020)</ns0:ref>. In this study, inclusion of non-coding regions altered tree topology in the tribe Laureae, suggesting the existence of a conflicting signal between coding and non-coding regions. Non-coding regions are often discarded for being uninformative, or for being misleading due to saturation at deep time scales <ns0:ref type='bibr' target='#b15'>(Foster, Henwood & Ho, 2018)</ns0:ref>. In our study, tree topologies based on coding and non-coding regions were largely congruent, except for the relationships among the three subclades (Figs. <ns0:ref type='figure' target='#fig_11'>S1-S2</ns0:ref>), indicating that non-coding regions are as informative as coding regions in Laureae. Thus, it is imperative to evaluate the influence of noncoding regions on tree topology rather than treating the whole plastome as a single unit or simply excluding non-coding regions from phylogenetic analysis. </ns0:p></ns0:div>
<ns0:div><ns0:head>Manuscript to be reviewed</ns0:head><ns0:p>To accommodate the conflicts among different plastid regions, a species tree was inferred through summary coalescent analysis. It has been suggested that the coalescent method is more robust than the concatenation method when the level of ILS is high <ns0:ref type='bibr' target='#b41'>(Liu, Xi & Davis, 2014;</ns0:ref><ns0:ref type='bibr' target='#b46'>Mirarab, Bayzid & Warnow, 2014)</ns0:ref>. High ILS tends to occur when the time interval between consecutive speciation events is short <ns0:ref type='bibr' target='#b74'>(Sun et al., 2015;</ns0:ref><ns0:ref type='bibr' target='#b61'>Sato et al., 2019)</ns0:ref>, and the core Lauraceae group (Perseeae, Cinnamomeae and Laureae) is thought to have undergone a rapid radiation <ns0:ref type='bibr' target='#b5'>(Chanderbali, van der Werff & Renner, 2001;</ns0:ref><ns0:ref type='bibr' target='#b56'>Rohwer & Rudolph, 2005;</ns0:ref><ns0:ref type='bibr' target='#b49'>Nie, Wen & Sun, 2007)</ns0:ref>. We therefore chose to implement the coalescent method. Nonetheless, it should be noted that, with this method, short and uninformative loci may lead to problematic gene trees and therefore result in a less accurate species tree <ns0:ref type='bibr' target='#b89'>(Xi, Liu & Davis, 2015;</ns0:ref><ns0:ref type='bibr' target='#b70'>Springer & Gatesy, 2016)</ns0:ref>.</ns0:p><ns0:p>In our study, only nine of 243 loci (rpoC1 intron, trnG-trnfM, ndhA intron, psaJ-rpl33, rpl2-rpl23, petN-psbM, psaB, trnS-ycf3, and ycf2) had a strong phylogenetic signal at the contentious node. The other 234 loci with weak phylogenetic signals may have resulted in gene trees with uncertainties and led to inaccurate topology at this node.</ns0:p><ns0:p>Exploration of the factors that underlie conflicts in phylogenetic signals is of great importance-but it is also challenging. Previous studies have examined whether biological and non-biological factors contribute to such conflicts (e.g., <ns0:ref type='bibr' target='#b10'>Duvall, Burke & Clark, 2020;</ns0:ref><ns0:ref type='bibr' target='#b95'>Zhang et al., 2020)</ns0:ref>. For example, gaps have been found to cause alternate, but conflicting topologies in Poaceae <ns0:ref type='bibr' target='#b10'>(Duvall, Burke & Clark, 2020)</ns0:ref>. However, the inclusion of alignment gaps did not alter our tree topology (Table <ns0:ref type='table'>S3</ns0:ref>). Although previous studies indicated that partitioning improves phylogenetic inference <ns0:ref type='bibr' target='#b90'>(Xi et al., 2012)</ns0:ref>, ML tree topologies based on partitioned and Manuscript to be reviewed unpartitioned datasets did not differ significantly in our study. It has been suggested that plastid genes under positive selection may bias phylogenies (e.g., <ns0:ref type='bibr' target='#b54'>Piot et al., 2018;</ns0:ref><ns0:ref type='bibr' target='#b59'>Saarela et al., 2018)</ns0:ref>, however, we found that the relationship among subclades I, II and III was not affected by positively selected sites, suggesting that positive selection was not the cause of tree conflicts. In this study, the low support values and short branch lengths of the estimated species tree (Fig. <ns0:ref type='figure'>3</ns0:ref>) suggested that each locus had a significantly incongruent topology and may indicate the existence of ILS. High levels of ILS are thought to yield similar numbers of loci supporting alternative topologies <ns0:ref type='bibr' target='#b22'>(Huson et al., 2005)</ns0:ref>. In our study, the numbers of loci supporting each topology were different (six for T1, three for T2, and zero for T3 after exclusion of loci with absolute ΔlnL ≤ 2), suggesting that ILS may not be the primary cause of the discordance among loci. Another plausible explanation for the conflict is heteroplasmic recombination, which can occur in species with biparental plastome inheritance <ns0:ref type='bibr' target='#b80'>(Walker et al., 2019)</ns0:ref>. Although heteroplasmic recombination has been reported with clear evidence in Brachypodium and Picea <ns0:ref type='bibr' target='#b73'>(Sullivan et al., 2017;</ns0:ref><ns0:ref type='bibr' target='#b60'>Sancho et al., 2018)</ns0:ref>, to our knowledge it has never been documented in Lauraceae. Based on the data reported here, it is too early to draw a firm conclusion about the causes of the conflict in phylogenetic signals. Although fully resolved phylogenies may still remain elusive based on different genomic compartments (i.e., nuclear, mitochondrial and plastid), phylogenomic studies that incorporate these compartments can provide new insights into tree discordance and its underlying causes <ns0:ref type='bibr' target='#b28'>(Koenen et al., 2020)</ns0:ref>. Therefore, more genetic information (e.g., nuclear genes) will be required to solve this problem in future work. Manuscript to be reviewed</ns0:p></ns0:div>
<ns0:div><ns0:head>Conclusion</ns0:head><ns0:p>In summary, this study revealed that Laureae plastomes are conserved in structure, size and gene content. A conflicting phylogenetic signal was detected between coding and non-coding regions, suggesting that the plastid genome should not be treated as a single unit. ML trees based on coding and non-coding regions were largely congruent except at the contentious node, indicating that coding regions are as informative as non-coding regions and that the influence of non-coding regions on tree inference should be evaluated. We also found that small subsets of plastome loci determined the topology at specific nodes, consistent with the results of a previous study <ns0:ref type='bibr' target='#b63'>(Shen, Hittinger & Rokas, 2017)</ns0:ref>. Through quantification and analysis of intra-plastome conflicts, the sister relationship of subclade I (including Lindera communis, L. glauca and L. nacusua) and II (including Laurus azorica, L. nobilis, Lindera megaphylla, Litsea acutivena, L. glutinosa, L. monopetala and L. pungens) was supported by our study. Biological factors may contribute to the conflicts among plastid loci; however, more information is needed to determine the underlying mechanism(s). </ns0:p></ns0:div><ns0:figure xml:id='fig_0'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:05:49532:2:0:NEW 3 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_1'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:05:49532:2:0:NEW 3 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_2'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:05:49532:2:0:NEW 3 Sep 2020) Manuscript to be reviewed Neocinnamomum and other clades, the following analyses were performed using a dataset from which six plastomes had been removed (Beilschmiedia pauciflora H. W. Li, Caryodaphnopsis malipoensis Bing Liu et Y. Yang, Cassytha filiformis L., Cryptocarya chinensis (Hance) Hemsl. and Eusideroxylon zwageri Teijsm. et Binn.).</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_3'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:05:49532:2:0:NEW 3 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_4'><ns0:head /><ns0:label /><ns0:figDesc>2 and S1-S4), Laureae was divided into three groups. Subclade I included Lindera communis Hemsl., L. glauca and L. nacusua (D. Don) Merr.; subclade II included Laurus azorica (Seub.) Franco, L. nobilis, Lindera megaphylla Hemsl., Litsea acutivena, L. glutinosa, L. monopetala and L. pungens; and subclade III included the other Laureae species used in the study. In subclade I, Lindera glauca was sister to L. communis and L. nacusua. In subclade II, Laurus was sister to Litsea acutivena, L. glutinosa and Lindera megaphylla, and the position of Litsea pungens was unstable (Figs. 2 and S1-S4). Litsea monopetala (LAU00063) was embedded within three samples of Litsea glutinosa in subclade II, highlighting the necessity of re-identification for L. monopetala (LAU00063). Topologies within subclade III based on different datasets were largely congruent (Figs. 2 and S1-S4). In subclade III, samples of Litsea, together with Lindera obtusiloba Bl., were monophyletic. Lindera erythrocarpa, L. latifolia Hook. f., L. metcalfiana Allen and L. robusta (Allen) Tsui were monophyletic as well. Lindera aggregata, L. chunii PeerJ reviewing PDF | (2020:05:49532:2:0:NEW 3 Sep 2020) Manuscript to be reviewed Merr., L. fragrans Oliv., L. limprichtii H. Winkl., L. pulcherrima (Wall.) Benth., L. supracostata Lec., L. thomsonii Allen and L. thomsonii var. vernayana (Allen) H.P. Tsui formed a wellsupported clade. Neolitsea was closer to Actinodaphne than to other Laureae species.</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_5'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:05:49532:2:0:NEW 3 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_6'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:05:49532:2:0:NEW 3 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_7'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:05:49532:2:0:NEW 3 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_8'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:05:49532:2:0:NEW 3 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_9'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:05:49532:2:0:NEW 3 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_10'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:05:49532:2:0:NEW 3 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_11'><ns0:head>Figure 1</ns0:head><ns0:label>1</ns0:label><ns0:figDesc>Figure 1</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_12'><ns0:head>Figure 4</ns0:head><ns0:label>4</ns0:label><ns0:figDesc>Figure 4</ns0:figDesc></ns0:figure>
<ns0:figure type='table' xml:id='tab_0'><ns0:head>Plant materials, DNA extraction and genome sequencing Materials</ns0:head><ns0:label /><ns0:figDesc>from 12 species in five genera (Actinodaphne obovata (Nees) Bl., Iteadaphne caudata (Nees) H. W. Li, Lindera erythrocarpa Makino, Litsea acutivena Hay., L. elongata (Wall. ex</ns0:figDesc><ns0:table><ns0:row><ns0:cell>Nees) Benth. et Hook. f., L. glutinosa (Lour.) C. B. Rob., L. dilleniifolia P. Y. Pai et P. H.</ns0:cell></ns0:row><ns0:row><ns0:cell>Huang, L. mollis Hemsl., L. monopetala (Roxb.) Pers., L. pungens Hemsl., L. szemaois (H. Liu)</ns0:cell></ns0:row><ns0:row><ns0:cell>J. Li et H.W. Li, and Neolitsea pallens (D. Don) Momiy. et H. Hara) (tribe Laureae, Lauraceae)</ns0:cell></ns0:row></ns0:table></ns0:figure>
<ns0:note place='foot' n='3'>PeerJ reviewing PDF | (2020:05:49532:2:0:NEW 3 Sep 2020)Manuscript to be reviewed</ns0:note>
</ns0:body>
" | "Dear editor
We thank the reviewers that their splendid comments have helped to improve our manuscript. We have already revised our manuscript to address their concerns.
In particular, we used the site models to perform positive selection tests. We identified the positively selected sites, manually deleted them and used the remaining CDS to infer a ML tree. We compared this tree to ML trees based on complete CDS dataset and found no impact of positive selection on our tree topology.
We believe that the revised manuscript is now ready for publication in PeerJ.
Tian-Wen Xiao
South China Botanical Garden, University of Chinese Academy of Sciences
On behalf of all authors.
Reviewer 1 (Anonymous):
Comments for the Author
In this revision of manuscript #49532 by Xiao and coworkers, the suggestions on the previous draft of the manuscript have been satisfactorily addressed.
No further revisions are needed.
Reviewer 2 (Anonymous):
59: change “gene evolutionary rate” to “a proxy for evolutionary rate:
Response: Agreed. We have changed “gene evolutionary rate” to “a proxy for evolutionary rate” (line 60).
186: personal preference, but change “cp” to “CP” here (and henceforth). This would also require changing “cp-reduced” to “CP-reduced”. I feel capitalizing the data set name helps to delimit the abbreviation in the text, rather than having it look like a typo.
Response: Agreed. All “cp” and “cp-reduced” has been changed to “CP” and “CP-reduced” respectively in our manuscript.
194–195: change to “the CP dataset”
Response: Agreed. See line 190.
196: Change to “the optimal partitioning schemes for each dataset”.
Response: Agreed. See line 185.
197: change “the best partition schemes and models for each partition “to“ the optimal partitioning schemes, and nucleotide substitution models for each partition,”
Response: Agreed. See lines 186-187.
212: change “paml” to “PAML”
Response: Agreed. We have changed “paml” to “PAML” (line 197).
293: change “confirmed” to “supported”
Response: Agreed. See line 278.
295–296: change “indicating that partitioning did not affect our tree topology.” to “indicating that our results were robust to different partitioning schemes”.
Response: Agreed. See lines 280-281.
293–302: instead of having three one-sentence paragraphs, merge all into one paragraph. The following sentence also needs to be reworded for clarity: “LRT showed that the dN/dS ratios of labeled lineages (subclades I, II and III) were not significantly different from background (p > 0.05), of which dN/dS ratios were less than one (Table S4), suggesting that there was no positive selection on the plastid genes.” I suggest something like:
“Additionally, a likelihood ratio test showed that the dN/dS ratios of labeled lineages (subclades I, II and III) were not significantly different from background (p > 0.05), of which dN/dS ratios were less than one (Table S4), suggesting that there was no positive selection on the plastid genes.”
212–215: The explanation of the tests for selection is very confusing: “To statistically test for positive selection, we compared the performance of two branch models (M0 and M2) for each gene. Three foreground branches were labeled on the unpartitioned CDS ML tree. Likelihood ratio tests (LRT) were performed using pchisq function in R 3.6.2 (R Core Team, 2018).”
In a study like this one, it would be useful to test whether there is evidence for selection in any of the protein-coding genes. Such an approach might show that (e.g.) most protein-coding genes are evolving under purifying selection, but a handful are evolving under positive selection. Most plastome studies take this approach. However, the authors in the present study chose to test foreground vs background lineages using branch models in CODEML. This approach is useful to determine whether particular lineage(s) are evolving under different regimes of selection, and can be interesting when there is an a priori explanation of why particular lineages might be expected to be evolving differently from others. To include the selection analyses in the manuscript, the authors need to either (a) justify why CODEML branch models were used, and what particular lineages were selected as foreground lineages and why (in the methods, not just results); or (b) use site models in CODEML to test for global signals of selection in genes based on multiple sequence alignments, rather than comparing particular lineages to a background.
I think the discrepancy in finding no genes under positive selection in the present study, compared to previous studies, might be because of the choice of tests conducted in CODEML.
Response: We agree with the reviewer that the branch models we chose for positive selection is inappropriate. We therefore followed the reviewer’s suggestion and chose the site models (M0, M1a, M2a, M3, M7, M8) in CODEML to identify positively selected codon sites. To examine the effect of positive selection on tree topology, we deleted these codon sites, used the remaining CDS, namely CDS-reduced dataset, to infer a ML tree. We compare this tree to ML trees based on CDS dataset and found that the relationship of subclades I, II and III were identical, suggesting that positive selection is not the cause of tree conflicts (lines 197-206, 284-290, 401-406).
" | Here is a paper. Please give your review comments after reading it. |
9,800 | "<ns0:abstract xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:p>Impatiens capensis (jewelweed, orange balsam) is an annual plant native to eastern North America that is currently spreading across Europe. In Poland, due to its rapid spread in the secondary range and high competitiveness in relation to native species, it is considered a locally invasive species.</ns0:p><ns0:p>Numerous studies have been devoted to the morphology, ecology, biology and genetics of this species. However, a review of available literature showed scarcity of data on seed size and a complete lack of information describing the morphological variation of I. capensis seeds.</ns0:p></ns0:div>
</ns0:abstract>
<ns0:body xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:p>The current work presents a detailed description of the morphology of the seeds of I. capensis using SEM for the first time. Our research shows also two types of cells on the ultrastructure of seeds, between the ribs and on the ribs, not yet described in the literature. The paper gives a new data about the seed morphology of I. capensis growing in different habitat conditions as well as the information about the area -the circuit that has not been reported before. The relationship between seeds variability and various environmental factors was shown.</ns0:p></ns0:div>
<ns0:div><ns0:head>Introduction</ns0:head><ns0:p>The Impatiens are the most species-rich genus of the family Balsaminaceae, with ca. 1000 species distributed primarily in the Old World tropics and subtropics <ns0:ref type='bibr' target='#b38'>(Grey-Wilson 1980;</ns0:ref><ns0:ref type='bibr' target='#b31'>Fischer 2004;</ns0:ref><ns0:ref type='bibr' target='#b118'>Yu et al. 2015)</ns0:ref>.</ns0:p><ns0:p>For many years the Impatiens taxa has been a subject of numerous studies regarding its distribution <ns0:ref type='bibr' target='#b44'>(Iljanić et al. 1994;</ns0:ref><ns0:ref type='bibr' target='#b116'>Xiang et al. 2011;</ns0:ref><ns0:ref type='bibr' target='#b123'>Zhou et al. 2019)</ns0:ref>, ecology <ns0:ref type='bibr' target='#b2'>(Abrahamson, Hershey 1977;</ns0:ref><ns0:ref type='bibr' target='#b113'>Winsor 1983;</ns0:ref><ns0:ref type='bibr' target='#b12'>Bell et al. 1991;</ns0:ref><ns0:ref type='bibr' target='#b18'>Boyer et al. 2016)</ns0:ref>, physiology <ns0:ref type='bibr' target='#b15'>(Bhattacharya et al. 1976;</ns0:ref><ns0:ref type='bibr' target='#b66'>Nanda, Kumar 1983;</ns0:ref><ns0:ref type='bibr' target='#b100'>Tooke et al. 2005)</ns0:ref>, biochemistry <ns0:ref type='bibr' target='#b92'>(Sreelakshmi et al. 2018;</ns0:ref><ns0:ref type='bibr' target='#b95'>Szewczyk 2018)</ns0:ref>, biology <ns0:ref type='bibr' target='#b109'>(Waller 1980;</ns0:ref><ns0:ref type='bibr' target='#b45'>Jacquemart et al. 2015)</ns0:ref>, pollination <ns0:ref type='bibr' target='#b91'>(Sreekala 2016;</ns0:ref><ns0:ref type='bibr' target='#b1'>Abrahamczyk et al. 2017)</ns0:ref>, morphology <ns0:ref type='bibr' target='#b86'>(Shimizu 1982;</ns0:ref><ns0:ref type='bibr' target='#b7'>Akiyama, Ohba 2000;</ns0:ref><ns0:ref type='bibr' target='#b0'>Abid et al. 2011;</ns0:ref><ns0:ref type='bibr' target='#b48'>Janssens et al. 2018)</ns0:ref>, systematics <ns0:ref type='bibr' target='#b22'>(Chen et al. 2007b;</ns0:ref><ns0:ref type='bibr' target='#b14'>Bhaskar 2012;</ns0:ref><ns0:ref type='bibr' target='#b36'>Gogoi et al. 2018;</ns0:ref><ns0:ref type='bibr' target='#b77'>Ruchisansakun et al. 2018)</ns0:ref>, phylogeny and evolution <ns0:ref type='bibr' target='#b46'>(Janssens et al. 2007;</ns0:ref><ns0:ref type='bibr' target='#b106'>Utami, Ardiyani 2015;</ns0:ref><ns0:ref type='bibr' target='#b76'>Ruchisansakun et al. 2015;</ns0:ref><ns0:ref type='bibr' target='#b118'>Yu et al. 2015;</ns0:ref><ns0:ref type='bibr' target='#b72'>Rahelivololona et al. 2018)</ns0:ref>, and others (see <ns0:ref type='bibr'>Adamowski 2016 onwards)</ns0:ref>.</ns0:p><ns0:p>Despite the plethora of publications on various attributes of Impatiens species, our knowledge about this genus is still incomplete. It is because taxonomically Impatiens is one of the most difficult groups to classify and still remains a strong challenge due to the enormous species richness and their extraordinary morphological variability <ns0:ref type='bibr'>(Hooker 1904</ns0:ref><ns0:ref type='bibr'>(Hooker -1906;;</ns0:ref><ns0:ref type='bibr' target='#b38'>Grey-Wilson 1980;</ns0:ref><ns0:ref type='bibr' target='#b89'>Song et al. 2005;</ns0:ref><ns0:ref type='bibr' target='#b36'>Gogoi et al. 2018)</ns0:ref>.</ns0:p><ns0:p>Balsams are extremely variable in morphology, from several centimetres high annual plants with one flower to four metres high subshrubs <ns0:ref type='bibr' target='#b38'>(Grey-Wilson 1980;</ns0:ref><ns0:ref type='bibr' target='#b14'>Bhaskar 2012;</ns0:ref><ns0:ref type='bibr' target='#b77'>Ruchisansakun et al. 2018)</ns0:ref>. Plants from this genus generally prefer humid habitats as montane forests, banks of water courses and vicinity of waterfalls, but some, often rhizomatous or tuberous could thrive in drier situations. Only in Western Ghats grow balsams from Scapigerae section, often epiphytic, having all leaves in rosette <ns0:ref type='bibr' target='#b14'>(Bhaskar 2012)</ns0:ref>. On the other hand, members of Trimorphopetalum section, having invariably spurless flowers, inhabit only Madagascar <ns0:ref type='bibr' target='#b72'>(Rahelivololona et al. 2018)</ns0:ref>.</ns0:p><ns0:p>Majority of balsam species grow in hardly accessible mountainous areas and have delicate flowers with complicated morphology <ns0:ref type='bibr' target='#b14'>(Bhaskar 2012;</ns0:ref><ns0:ref type='bibr' target='#b117'>Yu 2012;</ns0:ref><ns0:ref type='bibr' target='#b36'>Gogoi et al. 2018;</ns0:ref><ns0:ref type='bibr' /> PeerJ reviewing PDF | (2020:02:45912:1:2:NEW 14 Apr 2020)</ns0:p><ns0:p>Manuscript to be reviewed <ns0:ref type='bibr' target='#b72'>Rahelivololona et al. 2018;</ns0:ref><ns0:ref type='bibr' target='#b77'>Ruchisansakun et al. 2018)</ns0:ref>. Herbarium specimens of balsams are difficult to prepare, due to succulent nature of stems. Specimens need special preparations such as floral dissections <ns0:ref type='bibr' target='#b38'>(Grey-Wilson 1980;</ns0:ref><ns0:ref type='bibr'>Shui et al. 2011</ns0:ref>) and extensive field notes, otherwise are of limited value. Flower colors quickly fade and position of particular flower parts is often impossible to recreate from traditionally prepared specimens. One of taxonomically important features within the genus Impatiens is related to the morphology of seeds . The first information on the diversity of the seed coat in the genus Impatiens was reported by <ns0:ref type='bibr' target='#b42'>Hooker and Thomson (1859)</ns0:ref> and <ns0:ref type='bibr' target='#b112'>Warburg and Reiche (1895)</ns0:ref>. Other works were concerned mostly with the shape and size of seeds rather than details of their surface ornamentation <ns0:ref type='bibr' target='#b84'>(Shimizu 1977)</ns0:ref>.</ns0:p><ns0:p>The development of new imaging methods enabled observation and studies of small-sized structures. With the use of the scanning electron microscopy technique (SEM), a detailed morphological analysis of Impatiens seeds, including seed coat micromorphology, has become possible <ns0:ref type='bibr' target='#b54'>(Lu, Chen 1991;</ns0:ref><ns0:ref type='bibr' target='#b89'>Song et al. 2005;</ns0:ref><ns0:ref type='bibr' target='#b107'>Utami, Shimizu 2005;</ns0:ref><ns0:ref type='bibr' target='#b21'>Chen et al. 2007a;</ns0:ref><ns0:ref type='bibr' target='#b122'>Zhang et al. 2016)</ns0:ref>. The seed morphology of Impatiens and seed characters were found very significant in solving problems of classification as well as in phylogenetic considerations <ns0:ref type='bibr' target='#b114'>(Wunderlich 1967;</ns0:ref><ns0:ref type='bibr' target='#b19'>Brisson, Peterson 1976;</ns0:ref><ns0:ref type='bibr' target='#b85'>Shimizu 1979;</ns0:ref><ns0:ref type='bibr' target='#b89'>Song et al. 2005;</ns0:ref><ns0:ref type='bibr' target='#b107'>Utami, Shimizu 2005</ns0:ref><ns0:ref type='bibr' target='#b118'>, Yu et al. 2015)</ns0:ref>, and provided useful taxonomic characters correlated with flower and pollen morphology.</ns0:p><ns0:p>Despite an increasing number of publications on the surface of Impatiens seeds based on the use of the SEM technique (e.g. <ns0:ref type='bibr' target='#b85'>Shimizu 1979;</ns0:ref><ns0:ref type='bibr' target='#b54'>Lu, Chen 1991;</ns0:ref><ns0:ref type='bibr' target='#b107'>Utami, Shimizu 2005;</ns0:ref><ns0:ref type='bibr' target='#b21'>Chen et al. 2007a;</ns0:ref><ns0:ref type='bibr' target='#b49'>Jin et al. 2008;</ns0:ref><ns0:ref type='bibr' target='#b20'>Cai et al. 2013;</ns0:ref><ns0:ref type='bibr' target='#b122'>Zhang et al. 2016;</ns0:ref><ns0:ref type='bibr'>Xia et al. 2019 and others)</ns0:ref>, there is still a lack of information on the seed micromorphology for majority of species. In fact, the seed morphology of the genus Impatiens has been observed only for nearly 170 species (Maciejewska-Rutkowska, Janczak 2016).</ns0:p><ns0:p>Despite several major molecular studies using novel imaging methods (e.g. <ns0:ref type='bibr' target='#b120'>Yuan et al. 2004;</ns0:ref><ns0:ref type='bibr' target='#b76'>Ruchisansakun et al. 2015;</ns0:ref><ns0:ref type='bibr' target='#b118'>Yu et al. 2015;</ns0:ref><ns0:ref type='bibr' target='#b106'>Utami, Ardiyani 2015</ns0:ref><ns0:ref type='bibr'>, Shajita et al. 2016;</ns0:ref><ns0:ref type='bibr' target='#b72'>Rahelivololona et al. 2018</ns0:ref>) taxonomy of Impatiens species is far from being resolved. Still only about one fifth of all species were investigated.</ns0:p><ns0:p>One of the morphologically undescribed species is Impatiens capensis (jewelweed, orange balsam), an annual plant native to eastern North America <ns0:ref type='bibr' target='#b61'>(Meusel et al. 1978)</ns0:ref>, that is currently spreading across Europe. Its native distribution reaches from Florida Panhandle to New Manuscript to be reviewed moist forests <ns0:ref type='bibr' target='#b35'>(Gleason, Cronquist 1991;</ns0:ref><ns0:ref type='bibr' target='#b105'>USDA 2019)</ns0:ref>. In the early 19th century, the species was introduced to Great Britain, and a few decades later also to France and other European countries, as an ornamental plant. It subsequently escaped from cultivation and spread, colonizing especially waterways <ns0:ref type='bibr' target='#b3'>(Adamowski 2008;</ns0:ref><ns0:ref type='bibr' target='#b96'>Tabak, von Wettberg 2008)</ns0:ref>. The species was seen for the first time as a naturalized plant in Great Britain in 1822 <ns0:ref type='bibr' target='#b69'>(Perring, Walters 1962)</ns0:ref> and in France in 1898 <ns0:ref type='bibr' target='#b32'>(Fournier 1961)</ns0:ref>.</ns0:p><ns0:p>Today I. capensis is considered as naturalized in several European countries <ns0:ref type='bibr' target='#b59'>(Matthews et al. 2015)</ns0:ref>, including Poland, where the species is locally established and invasive, due to its rapid spread in the secondary range and high competitiveness in relation to native species, even perennials <ns0:ref type='bibr' target='#b99'>(Tokarska-Guzik et al. 2012)</ns0:ref>. In Poland, it was found for the first time in 1987 <ns0:ref type='bibr' target='#b68'>(Pawlaczyk, Adamowski 1991)</ns0:ref> and it is spreading in the Western Pomerania region <ns0:ref type='bibr' target='#b70'>(Popiela et al. 2015;</ns0:ref><ns0:ref type='bibr'>M Myśliwy, 2017, personal observations)</ns0:ref>. The species occurs in the area of the Szczecin Lagoon and enters alder forests, willow shrubs, rushes and riparian tall herb fringe communities <ns0:ref type='bibr' target='#b68'>(Pawlaczyk, Adamowski 1991;</ns0:ref><ns0:ref type='bibr' target='#b101'>Torbé 2000;</ns0:ref><ns0:ref type='bibr' target='#b65'>Myśliwy et al. 2009;</ns0:ref><ns0:ref type='bibr'>M Myśliwy, 2014, personal observations)</ns0:ref>. It also appears in moist anthropogenic habitats, e.g. along roadside ditches (M Myśliwy, 2017, personal observations).</ns0:p><ns0:p>I. capensis is an annual plant growing 0.5-1.5 m or more in height. The stems are sparsely branched, glabrous, succulent, pale green to pale reddish-green and slightly translucent. The leaves are alternate, to 10 cm long and 5 cm wide, ovate or elliptical and crenate, similar to native Impatiens noli-tangere L. but with 5-12(-14) teeth on each side. The flowers are 2.5-3.0 cm long, orange with darker patches in most common f. capensis. The lower sepal forms a light orange nectar spur, 5-9 mm long, which is bent through 180° to lie parallel with the sepal-sac <ns0:ref type='bibr' target='#b124'>(Zika 2006)</ns0:ref>. Besides color, it differs from predominantly Eurasiatic I. noli-tangere in the lower sepal more rapidly constricted into the spur and the position of the spur <ns0:ref type='bibr' target='#b125'>(Zika 2009)</ns0:ref>.</ns0:p><ns0:p>The fruit is a five-valved capsule, 2.0-2.5 cm long and 0.3-0.5 cm wide, which uses explosive dehiscence to eject seeds <ns0:ref type='bibr' target='#b62'>(Moore 1968;</ns0:ref><ns0:ref type='bibr' target='#b35'>Gleason, Cronquist 1991;</ns0:ref><ns0:ref type='bibr' target='#b25'>Day et al. 2012)</ns0:ref>.</ns0:p><ns0:p>The seeds are oval or lanceolate, with four strong ribs, the apex and bottom narrowed, 5-5.6 x 2.7-3.1 mm <ns0:ref type='bibr' target='#b17'>(Bojňanský, Fargašová 2007)</ns0:ref>. The weight ranges from 6.4 to 26.9 mg <ns0:ref type='bibr' target='#b88'>(Simpson et al. 1985)</ns0:ref>. <ns0:ref type='bibr' target='#b80'>Schemske (1978)</ns0:ref> gives 11.5 mg for cleistogamous seeds and 13.3 mg for chasmogamous ones, and <ns0:ref type='bibr' target='#b110'>Waller (1982)</ns0:ref> 10.6 mg. The seed surface is wrinkly or rough, lusterless, dark-brown, with some roundish and paler spots <ns0:ref type='bibr' target='#b17'>(Bojňanský, Fargašová 2007)</ns0:ref>.</ns0:p><ns0:p>PeerJ reviewing <ns0:ref type='table'>PDF | (2020:02:45912:1:2:NEW 14 Apr 2020)</ns0:ref> Manuscript to be reviewed Numerous studies (over several hundred -see Adamowski 2016 onward and the literature cited therein) have been devoted to the morphology, ecology, biology and genetics of this species (e.g. <ns0:ref type='bibr' target='#b8'>Antlfinger 1989;</ns0:ref><ns0:ref type='bibr' target='#b9'>Argyres, Schmitt 1991;</ns0:ref><ns0:ref type='bibr' target='#b81'>Schmitt 1993;</ns0:ref><ns0:ref type='bibr' target='#b82'>Schmitt et al. 1985;</ns0:ref><ns0:ref type='bibr' target='#b26'>Donohue, Schmitt 1999;</ns0:ref><ns0:ref type='bibr' target='#b27'>Donohue et al. 2000;</ns0:ref><ns0:ref type='bibr' target='#b124'>Zika 2006</ns0:ref><ns0:ref type='bibr' target='#b125'>, Zika 2009;</ns0:ref><ns0:ref type='bibr' target='#b96'>Tabak, von Wettberg 2008;</ns0:ref><ns0:ref type='bibr' target='#b25'>Day et al. 2012</ns0:ref>). However, a review of available literature showed scarcity of data on seed size and a complete lack of information describing the morphological variation of I. capensis seeds. The laconic information about the size and weight of seeds of this species came from works by <ns0:ref type='bibr' target='#b80'>Schemske (1978)</ns0:ref>, <ns0:ref type='bibr' target='#b110'>Waller (1982)</ns0:ref>, <ns0:ref type='bibr' target='#b88'>Simpson et al. (1985)</ns0:ref>, <ns0:ref type='bibr' target='#b17'>Bojňanský and Fargašová (2007)</ns0:ref>.</ns0:p><ns0:p>Simultaneously, the micromorphology of I. capensis seeds has not been extensively examined yet.</ns0:p><ns0:p>The current work presents a detailed description of the morphology and their potential morphological variability of the seeds of I. capensis growing in various habitat conditions using SEM for the first time. The aims of our work were to: i) characterize the morphological traits of I. capensis seeds; ii) analyze the variability of the examined morphological traits of seeds in the studied populations and compare it with the available literature; iii) analyze the effect of habitat conditions on quantitative and qualitative traits of seeds; and iv) characterize the ultrastructure of seed coat.</ns0:p></ns0:div>
<ns0:div><ns0:head>Materials and methods</ns0:head></ns0:div>
<ns0:div><ns0:head>Study sites</ns0:head><ns0:p>We collected the seeds within one year at the turn of August and September 2018 (to eliminate seasonal variability) from eight populations of I. capensis in Poland. The research covered the entire Polish range of this species and various habitats, from natural (alder carrs, hydrophilous tall herbs along rivers, near water seeps and on the banks of the Szczecin Lagoon) to anthropogenic (tall herbs along roadside ditches, transformed forests along artificial canals) (Table <ns0:ref type='table'>1</ns0:ref>, Fig. <ns0:ref type='figure'>1</ns0:ref>). The studied populations were also subject to different lighting conditions, which were scored using a 3-point scale: plants which grew in willow forests and alder carrs were strongly shaded (3), while those from tall herb communities received full sun for almost the entire day (2) or the entire day (1). As the height of I. capensis specimens, the location of capsules within the plant (main stem vs. branches) and their derivation from flowers of various types (cleistogamous vs. chasmogamous) may influence seeds weight <ns0:ref type='bibr' target='#b110'>(Waller 1982)</ns0:ref>. The seeds PeerJ reviewing PDF | (2020:02:45912:1:2:NEW 14 Apr 2020)</ns0:p><ns0:p>Manuscript to be reviewed for our study were collected always from the main stems of 8-10 plants of similar (average for the population) height and from capsules derived from chasmogamous flowers to minimize the biases. Species nomenclature was adopted after Euro+Med PlantBase (Euro+Med PlantBase).</ns0:p></ns0:div>
<ns0:div><ns0:head>Biometric and SEM analysis</ns0:head><ns0:p>We analyzed from 24 to 30 seeds from each population. We measured four quantified seed traits: seed length (SL); seed width (SW); seed circuit (SC); and seed area (SA). The seeds were measured as previously described in <ns0:ref type='bibr' target='#b73'>Rewicz et al. (2017)</ns0:ref>. In order to describe the seed mass, we used 15 seeds from each population. The weighing of the seeds was performed with the use of Ohaus PA 21 weight. We used Roundness according to the formula: R=4 x area/π [Major axis]^2 defined by <ns0:ref type='bibr' target='#b30'>Ferreira and Wayne (2010)</ns0:ref>.</ns0:p><ns0:p>The seeds were sputter-coated with 4nm thick layer of gold. The SEM work was performed with the Phenom Pro X Scanning Electron Microscope at the Department of Invertebrate Zoology and Hydrobiology, University of Lodz, Poland. The 3D models of the seed surface were generated using dedicated software 3D Roughness Reconstruction for Phenom.</ns0:p><ns0:p>Pictures of seeds obtained using SEM were analyzed as previously described in <ns0:ref type='bibr' target='#b73'>Rewicz et al. (2017)</ns0:ref>. The seed shape terminology and types of seed surfaces were adopted from: Murley (1951), <ns0:ref type='bibr' target='#b89'>Song et al. (2005)</ns0:ref> and <ns0:ref type='bibr' target='#b17'>Bojňanský and Fargašová (2007)</ns0:ref>.</ns0:p></ns0:div>
<ns0:div><ns0:head>Soil properties</ns0:head><ns0:p>In order to characterize habitat conditions at each locality, five soil samples cores (0-20 cm depth) were collected and then mixed together into one composite sample. Soil samples were dried at room temperature, and then rub through a sieve to remove fractions larger than 1 mm.</ns0:p><ns0:p>The following physico-chemical soil parameters were determined <ns0:ref type='bibr' target='#b11'>(Bednarek et al. 2011</ns0:ref>): organic matter content defined as the loss on ignition (LOI) -soil samples annealed at 550ºC (%); grain composition (the content of sand, silt, clay) -Bouyoucos's sedimentation method, modified by Casagrande and Prószyński; granulometric categories according to the PSSS <ns0:ref type='bibr'>(2009)</ns0:ref> classification; soil reaction (pH) -the potentiometric method, in 1-M solution of KCl; soil calcium carbonate (CaCO 3 ) content (%) -the Scheibler's method; organic carbon (C org ) content (%), and total nitrogen (N tot ) content (%) were determined by an elemental analyzer CHNS/O FlashSmart (Thermo Scientific), and the C/N ratio; the content of available forms of soil PeerJ reviewing PDF | (2020:02:45912:1:2:NEW 14 Apr 2020)</ns0:p><ns0:p>Manuscript to be reviewed nutrients (mg/100 g soil): calcium (Ca) and sodium (Na) were determined spectrophotometrically (Ca -AAS and Na -EAS) on ICE3000; potassium (K) and phosphorus (P) -were measured according to the Egner-Riehm method, level of magnesium (Mg) -was measured using the Schachtschabel's method; soil moisture content, hand-felt assessed directly in the field using a 4-point scale: (1) dry, (2) fresh, (3) moist, (4) wet, as it was first described in <ns0:ref type='bibr' target='#b64'>Myśliwy (2019)</ns0:ref>.</ns0:p></ns0:div>
<ns0:div><ns0:head>Data analysis</ns0:head><ns0:p>The five following basic characteristic traits were calculated: arithmetic average (x), minimum and maximum values (min/max), coefficient of variation (CV) and standard deviation (SD). The distribution of the data was not normal; statistical analysis was based on the Kruskal-Wallis test (for p ≤0.05), which is a non-parametric alternative to ANOVA <ns0:ref type='bibr' target='#b121'>(Zar 1984)</ns0:ref>.</ns0:p><ns0:p>Correlation between pairs of morphological characters was evaluated using Spearman's correlation coefficient and the values were adopted after Meissner (2010), (correlation; less than 0.20 -very poor, 0.21-0.39 -weak, 0.40-0.69 -moderate, 0.70-0.89 -strong and above 0.89very strong).</ns0:p><ns0:p>The cluster analysis based on the nearest neighbor method was performed using the matrix on the population's mean values. As the dataset required a linear response model <ns0:ref type='bibr' target='#b50'>(Jongman et al. 1995)</ns0:ref>, the Redundancy Analysis (RDA) was used to relate the variability of morphological traits of seeds to environmental variables. The variables C org and N tot were excluded from the RDA as they were strongly correlated with organic matter content (LOI). The </ns0:p></ns0:div>
<ns0:div><ns0:head>Monte</ns0:head></ns0:div>
<ns0:div><ns0:head>Results</ns0:head></ns0:div>
<ns0:div><ns0:head>Biometric analysis</ns0:head><ns0:p>The seeds from the G (Police) population is characterized by the biggest seeds. This population characterized by the highest average values of the following traits: seed length (SL)</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:02:45912:1:2:NEW 14 Apr 2020)</ns0:p><ns0:p>Manuscript to be reviewed Comparatively high values of analyzed traits were observed in the E population (Święta). In the B (Lubin) population was observed the shortest and narrowest seeds (respectively: SL (3.88 mm), SW (2.03 mm), SC (9.55 mm) and SA (5.79 mm 2 ) (Table <ns0:ref type='table'>2</ns0:ref>).</ns0:p><ns0:p>The minimum values of analyzed traits were recorded in the B (Lubin) population (SL -3.16 mm, SW -1.12 mm, SC -7.27 mm, SA -2.43 mm 2 ). The maximum values of length (5.74 mm), circuit (14.59 mm) and area (13.54 mm 2 ) were recorded in the G population (Police).</ns0:p><ns0:p>A very strong Spearman's correlation (r = 0.94) was observed between the seed area and circuit (Table <ns0:ref type='table'>3</ns0:ref>). The most variable features were seed area (CV = 21.76%) and width (CV = 15.35%). The variation of seed traits ranged insignificantly from 6.19 % (H population) to 10.28% (B) -SL; from 7.30% (D) to 20.57% (B) -SW; from 6.22% (H) to 11.42% (B) -SC; and from 13.88% (D) to 26.06% (B) -SA, respectively.</ns0:p><ns0:p>The G (Police -11.42 mg) and E (Święta -9.82 mg) populations are characterized by the heaviest seeds. The lightest seeds were observed in the following populations: B (Lubin -6.52 mg) and H (Trzebieradz -6.92 mg) (Fig. <ns0:ref type='figure'>2</ns0:ref>, Table <ns0:ref type='table'>2</ns0:ref>).</ns0:p><ns0:p>The Kruskal-Wallis test showed that the I. capensis populations differ significantly in each of the analyzed traits. The conducted post hoc test (DunnTest) showed that the populations from: Police (G), followed by Czarnocin (D), Święta (E) and Trzebieradz (H) showed the greatest variation in terms of studied traits among all the populations (Table <ns0:ref type='table'>4</ns0:ref>).</ns0:p><ns0:p>The similarity analysis made using Euclidean's distances showed two main clusters (Fig. <ns0:ref type='figure'>3</ns0:ref>). The first cluster includes six populations of I. capensis (A-D, F, H), all derived from natural habitats, while the other cluster groups two populations (E, G) from anthropogenic habitats, where examined plants were the highest (Table <ns0:ref type='table'>1</ns0:ref>). According to the dendrogram (Fig. <ns0:ref type='figure'>3</ns0:ref>), populations C and F are the closest to each other; both were associated with river valleys (the Dziwna and Oder Rivers, respectively) and close to the river bed, hence under the influence of flooding. The D and A populations were growing in tall herb communities on the banks of the Szczecin Lagoon. The most distinct populations in the first cluster (H and B) were also found on the banks of the Szczecin Lagoon, but they had the lowest average height and differed in habitat conditions from the other populations of this cluster (Table <ns0:ref type='table'>1</ns0:ref>).</ns0:p><ns0:p>The morphological parameters of seed shape -roundness -showed statistically significant differences between the populations (p < 0.05). For roundness, the highest value was PeerJ reviewing PDF | (2020:02:45912:1:2:NEW 14 Apr 2020)</ns0:p><ns0:p>Manuscript to be reviewed recorded at D -Czarnocin (0.58) (tall herbs on the bank of the Szczecin Lagoon) and the lowest -at H -Trzebieradz (0.47) (alder carr) (Table <ns0:ref type='table'>5</ns0:ref>).</ns0:p></ns0:div>
<ns0:div><ns0:head>Biometric variability of seeds and its relationship with environmental conditions</ns0:head><ns0:p>All environmental variables included in the RDA explained 35.6% of the total variation.</ns0:p><ns0:p>The results of step-wise forward selection of variables indicated that five variables: anthropogenic disturbances, carbonates (CaCO 3 ), loose sand presence, potassium (K), and soil moisture content were statistically significant and differentiated the studied populations of I. capensis (Fig. <ns0:ref type='figure' target='#fig_2'>4</ns0:ref>). Along the gradient represented by Axis I, the highest correlation between the sample position and environmental variables (the so-called inter-set correlation) was typical of anthropogenic disturbances and CaCO 3 , followed by the degree of shading and soil Ca, while the C/N ratio was most closely correlated with Axis II, followed by soil contents of P, Na, K, and organic soil.</ns0:p><ns0:p>The location of population H (Trzebieradz) in the ordination space (the upper part of the RDA diagram) was associated with the highest C/N ratio, the highest soil moisture and shading, as well as with the lowest soil pH and the lowest soil contents of CaCO 3 , Ca, P and K (Fig. <ns0:ref type='figure' target='#fig_2'>4</ns0:ref>, Table <ns0:ref type='table'>6</ns0:ref>). The H population was dominated by short specimens (Table <ns0:ref type='table'>1</ns0:ref>), with relatively light seeds and average values of biometric traits (Table <ns0:ref type='table'>2</ns0:ref>). In contrast, populations E (Święta) and G (Police), located in the right-hand side of the RDA diagram, were also related to a relatively high C/N ratio, but unlike the previous population, they were associated with a low level of soil moisture as well as the highest anthropogenic disturbances (Fig. <ns0:ref type='figure' target='#fig_2'>4</ns0:ref>), and consisted of the tallest specimens (Table <ns0:ref type='table'>1</ns0:ref>), with the heaviest and largest seeds (Table <ns0:ref type='table'>2</ns0:ref>). Population D (Czarnocin) occupied the bottom part of the diagram and was distinct in its organic soil, with the highest content of organic matter (LOI), P, K, Mg, and Na contents in the soil, as well as the lowest C/N ratio (Table <ns0:ref type='table'>6</ns0:ref>). The lowest values of seeds biometric traits were typical of population B (Lubin), located in the left-hand part of the RDA diagram (Fig. <ns0:ref type='figure' target='#fig_2'>4</ns0:ref>, Table <ns0:ref type='table'>2</ns0:ref>), and associated with high soil pH and the highest contents of soil carbonates and calcium, as well as a low level of soil moisture (Table <ns0:ref type='table'>6</ns0:ref>). The other populations (A, Podgrodzie; C, Unin; F, Szczecin-Zdroje) were also on the left side of the diagram, but closer to the center (Fig. <ns0:ref type='figure' target='#fig_2'>4</ns0:ref>). Neither their seeds biometric traits nor habitat conditions were distinct (Table <ns0:ref type='table'>2, 6</ns0:ref>).</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:02:45912:1:2:NEW 14 Apr 2020)</ns0:p></ns0:div>
<ns0:div><ns0:head>Manuscript to be reviewed</ns0:head><ns0:p>The structure of seed surface Seeds from all studied populations did not differ in their ultrastructure. Seeds had a round shape, four strong, clear ribs, the apex and bottom narrowed. Each rib was built of rows of 4-5 cells and had a darker color than the surface of the seeds between them. The surface was rough, lusterless and dark-brown, without roundish and paler spots on the surface (Fig. <ns0:ref type='figure'>5</ns0:ref>). We found two types of cells with different ornamentation on the surface of the seed. The cells of the seed surface between the ribs were elongated and irregular, with cuticle grains visible on anticyclic walls at higher magnification. There were rows of 4-7 round cells with reticular ultrastructure located near the ribs.</ns0:p></ns0:div>
<ns0:div><ns0:head>Discussion</ns0:head><ns0:p>The genus Impatiens has the most diverse and elaborately sculptured seed coat. Although seed coat morphology alone does not provide universally applicable key characters for identification, it can be as helpful as many other characters used in taxonomy (e.g. <ns0:ref type='bibr' target='#b89'>Song et al. 2005;</ns0:ref><ns0:ref type='bibr' target='#b107'>Utami, Shimizu 2005;</ns0:ref><ns0:ref type='bibr' target='#b21'>Chen et al. 2007a;</ns0:ref><ns0:ref type='bibr' target='#b49'>Jin et al. 2008)</ns0:ref>. Earlier works focused on seed dimensions rarely described the ultrastructure of seeds <ns0:ref type='bibr' target='#b85'>(Shimizu 1979;</ns0:ref><ns0:ref type='bibr' target='#b54'>Lu, Chen 1991)</ns0:ref>, whereas the sculpture on seed coats offers a set of characters which could be used to identify the species and in combination with other data, e.g. pollen morphology, could provide crucial evidence for the taxonomy of the genus <ns0:ref type='bibr' target='#b54'>(Lu, Chen 1991;</ns0:ref><ns0:ref type='bibr' target='#b89'>Song et al. 2005;</ns0:ref><ns0:ref type='bibr' target='#b20'>Cai et al. 2013)</ns0:ref>. Unfortunately, till now seed morphology has been observed only for a small group from Balsaminaceae species, so the use of some morphological traits of seeds in taxonomy and classification of species in this group remains an open issue.</ns0:p><ns0:p>The paper provides new data about the seed morphology and seed coat micromorphology of I. capensis, as well as the information about the area -the circuit that has not been reported before. Moreover, our studies have shown maximum length values (5.74 mm) and seed widths (3.21 mm) beyond the available literature data. <ns0:ref type='bibr' target='#b17'>Bojňanský and Fargašová (2007)</ns0:ref> reported seeds of I. capensis 5-5.6 (length) x 2.7-3.1 (widths) mm. <ns0:ref type='bibr' target='#b17'>Bojňanský and Fargašová (2007)</ns0:ref> reported a lot of roundish spots on the surface of seeds (Fig. <ns0:ref type='figure'>6</ns0:ref>). We have not confirmed this trait on the surface of seeds in any of the studied populations. Our studies show also two types of cells on the ultrastructure of seeds, between the ribs and on the ribs, not yet described in the literature (Fig. <ns0:ref type='figure'>5</ns0:ref>). Such detailed seed coat PeerJ reviewing PDF | (2020:02:45912:1:2:NEW 14 Apr 2020)</ns0:p><ns0:p>Manuscript to be reviewed ornamentation was described for the first time. The observed differences may result from different geographical origin of the examined seeds. Seeds of I. capensis in our study came from populations from spontaneous localities while <ns0:ref type='bibr' target='#b17'>Bojňanský and Fargašová (2007)</ns0:ref> described seeds from cultivation of unknown origin. As it seems climatic conditions have limited influence on investigated parameters of seeds, due to small area of secondary distribution of I. capensis in Poland <ns0:ref type='bibr' target='#b5'>(Adamowski et al. 2018</ns0:ref>; Fig. <ns0:ref type='figure'>1</ns0:ref>) and short time of residence -little over 30 years. At this moment, due to insufficient sampling, including the lack of data from the native range of orange jewelweed, it is not possible to elucidate the overall variation in seed coat micromorphology and its implementation for taxonomy of I. capensis.</ns0:p><ns0:p>The analysis of SEM micrographs of I. noli-tangere seeds, closely related to I. capensis <ns0:ref type='bibr' target='#b118'>(Yu et al. 2015)</ns0:ref>, shows that seed coats of this species vary significantly depending on the geographical origin of the seeds <ns0:ref type='bibr' target='#b107'>(Utami, Shimizu 2005;</ns0:ref><ns0:ref type='bibr'>Chen et al. 2007 a;</ns0:ref><ns0:ref type='bibr' target='#b49'>Jin et al. 2008</ns0:ref>). On the other hand, comparison of the seed micromorphology of I. capensis did not show similarity to seed coat ornamentation of the aforementioned I. noli-tangere <ns0:ref type='bibr' target='#b89'>(Song et al. 2005;</ns0:ref><ns0:ref type='bibr' target='#b107'>Utami, Shimizu 2005;</ns0:ref><ns0:ref type='bibr' target='#b21'>Chen et al. 2007a;</ns0:ref><ns0:ref type='bibr' target='#b49'>Jin et al. 2008)</ns0:ref>. Despite the fact that these species are closely related and may be confused <ns0:ref type='bibr' target='#b125'>(Zika 2009;</ns0:ref><ns0:ref type='bibr' target='#b118'>Yu et al. 2015)</ns0:ref>, their seeds clearly differ morphologically. It proves that new data presented here may be useful in the identification of both species. In turn, there is no information about seed morphology of I. pallida, which is sympatric and synchronic species to I. capensis <ns0:ref type='bibr' target='#b78'>(Rust 1977)</ns0:ref>, which makes this subject even more difficult. It is, therefore, necessary to conduct further detailed observations on more samples and an overall evaluation of seed coat micromorphology not only in I. capensis. This phenomenon should be basis for further comparisons and studies due to the fact that the seed ultrastructure is considered as a constant feature within a taxonomic unit <ns0:ref type='bibr' target='#b93'>(Stace 1992)</ns0:ref> and as the morphological studies show, the seeds shape and size are highly diverse at genera and species level <ns0:ref type='bibr' target='#b102'>(Ullah et al. 2019a</ns0:ref><ns0:ref type='bibr' target='#b104'>(Ullah et al. , 2019b;;</ns0:ref><ns0:ref type='bibr' target='#b40'>Hadidchi et al. 2020)</ns0:ref>. Nowadays, with the present development of SEM techniques, the morphological characters of seeds and their ultrastructural characteristics are very useful for identification and taxonomic delimitation of various angiosperms groups, e.g. Brasssicaceae <ns0:ref type='bibr' target='#b97'>(Tantaway et al. 2004)</ns0:ref>, Caryophyllaceae <ns0:ref type='bibr' target='#b102'>(Ullah et al. 2019a</ns0:ref><ns0:ref type='bibr' target='#b104'>(Ullah et al. , 2019b))</ns0:ref>, Poaceae <ns0:ref type='bibr' target='#b58'>(Martín-Gómez et al. 2019b)</ns0:ref>, Ranunculaceae <ns0:ref type='bibr' target='#b24'>(Constantinidis et al. 2001;</ns0:ref><ns0:ref type='bibr' target='#b73'>Rewicz et al. 2017;</ns0:ref><ns0:ref type='bibr' target='#b57'>Martín-Gómez et al. 2019a;</ns0:ref><ns0:ref type='bibr' target='#b40'>Hadidchi et al. 2020)</ns0:ref>, Rosaceae <ns0:ref type='bibr' target='#b10'>(Ballian, Mujagić-Pašić 2013)</ns0:ref>, Onagraceae <ns0:ref type='bibr' target='#b6'>(Akbari, Azizian 2006)</ns0:ref>, Orchidaceae <ns0:ref type='bibr' target='#b33'>(Gamarra et al. 2007</ns0:ref><ns0:ref type='bibr' target='#b34'>(Gamarra et al. , 2010;;</ns0:ref><ns0:ref type='bibr'>Rewicz et al.</ns0:ref> PeerJ reviewing PDF | (2020:02:45912:1:2:NEW 14 Apr 2020)</ns0:p><ns0:p>Manuscript to be reviewed 2016). Data concerning the macro-and microstructure of seeds not only have been used as an important tool for solving various taxonomic problems within Impatiens genus but also provide results useful for determining the impact of various environmental factors on the phenotypic variability of species <ns0:ref type='bibr' target='#b12'>(Bell et al. 1991;</ns0:ref><ns0:ref type='bibr' target='#b9'>Argyres, Schmitt 1991;</ns0:ref><ns0:ref type='bibr' target='#b81'>Schmitt 1993;</ns0:ref><ns0:ref type='bibr' target='#b23'>Chmura et al. 2013;</ns0:ref><ns0:ref type='bibr' target='#b55'>Maciejewska-Rutkowska, Janczak 2016)</ns0:ref>. The understanding of environmentally induced variation in an individual plant phenotype is crucial for predicting population responses to environmental changes. This issue is particularly important in relation to invasive species which occupy a wide range of habitats in the invaded range <ns0:ref type='bibr' target='#b75'>(Richards et al. 2006)</ns0:ref>. Impatiens capensis is recognized as invasive species in Poland. Therefore, it is suspected that this species, while adapting to occupy new territories and compete with native species, developed specific adaptations, contributing to its success in the new environments. <ns0:ref type='bibr' target='#b88'>Simpson et al. (1985)</ns0:ref> have shown that I. capensis vegetative and reproductive growth parameters reflect site differences.</ns0:p><ns0:p>Our results confirm those findings. The size difference between Polish I. capensis populations is probably a consequence of ecological conditions of habitats. Populations G (Police) and E (Święta), occurring in the most disturbed anthropogenic habitats, have the highest stems and the largest and heaviest seeds as a result of growth in favorable environmental conditions (neutral or slightly acidic soil with a relatively high C/N ratio). In turn, population B (Lubin) with the lowest average plant height and the smallest and lightest seeds was associated with high soil pH, and the highest contents of soil carbonates and calcium. Light availability <ns0:ref type='bibr' target='#b88'>(Simpson et al. 1985)</ns0:ref> as well as soil moisture and pH <ns0:ref type='bibr' target='#b109'>(Waller 1980)</ns0:ref> were reported to affect growth patterns of I. capensis.</ns0:p><ns0:p>According to <ns0:ref type='bibr' target='#b79'>Skálová et al. (2013)</ns0:ref>, the stem height of I. capensis showed no marked response to deep shade under high soil moisture. Only the combined effect of a low level of moisture and deep shade resulted in decreased stem height. Our studies showed that five environmental variables were statistically significant and differentiated studied populations: anthropogenic disturbances (which may serve as a proxy for habitat fertility), carbonates (CaCO 3 ), loose sand presence, potassium (K), and soil moisture (Fig. <ns0:ref type='figure' target='#fig_2'>4</ns0:ref>).</ns0:p><ns0:p>According to <ns0:ref type='bibr' target='#b87'>Silvertown (1989)</ns0:ref>, correlation between seed size and place where plant is growing is an adaptative feature. Bigger seeds occur in habitats with stable environmental conditions, where seedlings may grow slowly. Small seeds are generally produced by fast growing plants with short life cycle. Bigger seeds has tendency to producing bigger seedlings, which have better chance to survive than smaller seedlings (produced by smaller seeds). Bigger PeerJ reviewing PDF | (2020:02:45912:1:2:NEW 14 Apr 2020)</ns0:p><ns0:p>Manuscript to be reviewed seedlings are also better adapted to natural threats like i.e.: deep shadow, drought, physical damage and occurrence of competitive plants <ns0:ref type='bibr'>(Leishmann et al. 2000)</ns0:ref>.</ns0:p></ns0:div>
<ns0:div><ns0:head>Conclusions</ns0:head><ns0:p>Understanding the variability of various traits, including taxonomically significant ones, is not only important to determine the limits of species variability, but in the case of invasive species, such research provides information whether a given feature is actually stable or is susceptible to environmental changes. Moreover, seed size is among key functional traits for the investigation of adaptive phenotypic plasticity <ns0:ref type='bibr' target='#b67'>(Nicotra et al. 2010)</ns0:ref>. Species with a greater adaptive plasticity may have a better chance of persisting in their environments under future climate and land-use changes. Such studies contribute to the knowledge of plants response to the environment and may help with predicting species distribution changes <ns0:ref type='bibr' target='#b67'>(Nicotra et al. 2010;</ns0:ref><ns0:ref type='bibr' target='#b37'>Gratani 2014)</ns0:ref>. It is particularly important in the case of I. capensis in Poland, where the anthropogenic range of this species is still limited, and some actions should be undertaken in the near future to eliminate it completely <ns0:ref type='bibr' target='#b5'>(Adamowski et al. 2018)</ns0:ref>.</ns0:p><ns0:p>We should note that the majority of the species of Impatiens still remain to be studied.</ns0:p><ns0:p>Further examination of more species may reveal further morphological diversity and seed coat types. Studies on the developmental variation of seed coat sculpture, which may provide insights into a better understanding of the evolutionary relationships among different types of sculpture, are also urgently needed <ns0:ref type='bibr' target='#b89'>(Song et al. 2005)</ns0:ref> </ns0:p></ns0:div><ns0:figure xml:id='fig_0'><ns0:head>Foundland</ns0:head><ns0:label /><ns0:figDesc>and from Texas to Saskatchewan, where it grows along watercourses, in bogs and PeerJ reviewing PDF | (2020:02:45912:1:2:NEW 14 Apr 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_1'><ns0:head /><ns0:label /><ns0:figDesc>Carlo permutation test with the forward selection of environmental variables was applied to determine the importance and statistical significance of variables in explaining the variability in seeds. The software package Canoco v.4.5 (ter Braak, Šmilauer 2002), MVSP 3.2 (Kovach 2010) and STATISTICA PL. ver. 13.1 (Stat-Soft Inc. 2011), and were used for all analyses (van Emden 2008; Lepš, Šmilauer 2003).</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_2'><ns0:head>( 4 .</ns0:head><ns0:label>4</ns0:label><ns0:figDesc>60 mm), width (SW) (2.71 mm), circuit (SC) (11.65 mm) and area (SA) (9.26 mm 2 ).</ns0:figDesc></ns0:figure>
<ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='45,42.52,199.12,525.00,162.75' type='bitmap' /></ns0:figure>
<ns0:figure type='table' xml:id='tab_0'><ns0:head /><ns0:label /><ns0:figDesc>.Dendrogram of similarities of populations of Impatiens capensis Meerb. in Poland, obtained by the nearest neighbor method.</ns0:figDesc><ns0:table><ns0:row><ns0:cell cols='2'>1 1 Figure 3</ns0:cell><ns0:cell>A</ns0:cell><ns0:cell>Length</ns0:cell><ns0:cell>B</ns0:cell><ns0:cell>C Width</ns0:cell><ns0:cell>D</ns0:cell><ns0:cell>E Circuit</ns0:cell><ns0:cell>F</ns0:cell><ns0:cell>Area</ns0:cell><ns0:cell>G</ns0:cell><ns0:cell>H</ns0:cell><ns0:cell>x</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell cols='3'>Weight (mg) 7.66 Length 1.00</ns0:cell><ns0:cell>6.52</ns0:cell><ns0:cell>8.16 0.47</ns0:cell><ns0:cell>7.82</ns0:cell><ns0:cell>9.82 0.85</ns0:cell><ns0:cell cols='2'>8.62 0.72</ns0:cell><ns0:cell>11.42</ns0:cell><ns0:cell>6.92</ns0:cell><ns0:cell>8.37</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>SL (mm) Width</ns0:cell><ns0:cell cols='2'>4.05</ns0:cell><ns0:cell>3.88</ns0:cell><ns0:cell>4.23 1.00</ns0:cell><ns0:cell>4.11</ns0:cell><ns0:cell>4.46 0.73</ns0:cell><ns0:cell cols='2'>4.17 0.84</ns0:cell><ns0:cell>4.60</ns0:cell><ns0:cell>4.41</ns0:cell><ns0:cell>4.24</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>Min-max Circuit</ns0:cell><ns0:cell cols='2'>3.50-</ns0:cell><ns0:cell>3.16-</ns0:cell><ns0:cell>3.68-</ns0:cell><ns0:cell>3.59-</ns0:cell><ns0:cell>3.85-1.00</ns0:cell><ns0:cell cols='2'>3.43-0.94</ns0:cell><ns0:cell>3.88-</ns0:cell><ns0:cell>3.59-</ns0:cell><ns0:cell>3.16-</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>Area</ns0:cell><ns0:cell cols='2'>4.64</ns0:cell><ns0:cell>4.48</ns0:cell><ns0:cell>4.70</ns0:cell><ns0:cell>4.75</ns0:cell><ns0:cell>5.26</ns0:cell><ns0:cell cols='2'>4.73 1.00</ns0:cell><ns0:cell>5.74</ns0:cell><ns0:cell>4.82</ns0:cell><ns0:cell>5.74</ns0:cell></ns0:row><ns0:row><ns0:cell>2</ns0:cell><ns0:cell>SD</ns0:cell><ns0:cell cols='2'>0.26</ns0:cell><ns0:cell>0.40</ns0:cell><ns0:cell>0.29</ns0:cell><ns0:cell>0.30</ns0:cell><ns0:cell>0.32</ns0:cell><ns0:cell>0.40</ns0:cell><ns0:cell /><ns0:cell>0.40</ns0:cell><ns0:cell>0.27</ns0:cell><ns0:cell>0.40</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>CV</ns0:cell><ns0:cell cols='2'>6.50</ns0:cell><ns0:cell>10.28</ns0:cell><ns0:cell>6.95</ns0:cell><ns0:cell>7.35</ns0:cell><ns0:cell>7.21</ns0:cell><ns0:cell>9.71</ns0:cell><ns0:cell /><ns0:cell>8.75</ns0:cell><ns0:cell>6.19</ns0:cell><ns0:cell>9.34</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>SW (mm)</ns0:cell><ns0:cell cols='2'>2.23</ns0:cell><ns0:cell>2.03</ns0:cell><ns0:cell>2.36</ns0:cell><ns0:cell>2.56</ns0:cell><ns0:cell>2.60</ns0:cell><ns0:cell>2.40</ns0:cell><ns0:cell /><ns0:cell>2.71</ns0:cell><ns0:cell>2.23</ns0:cell><ns0:cell>2.39</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>Min-max</ns0:cell><ns0:cell cols='2'>1.53-</ns0:cell><ns0:cell>1.12-</ns0:cell><ns0:cell>1.78-</ns0:cell><ns0:cell>2.14-</ns0:cell><ns0:cell>2.19-</ns0:cell><ns0:cell cols='2'>1.88-</ns0:cell><ns0:cell>2.15-</ns0:cell><ns0:cell>1.71-</ns0:cell><ns0:cell>1.12-</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell cols='2'>2.82</ns0:cell><ns0:cell>2.61</ns0:cell><ns0:cell>3.00</ns0:cell><ns0:cell>2.94</ns0:cell><ns0:cell>3.33</ns0:cell><ns0:cell>2.99</ns0:cell><ns0:cell /><ns0:cell>3.21</ns0:cell><ns0:cell>2.93</ns0:cell><ns0:cell>3.33</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>SD</ns0:cell><ns0:cell cols='2'>0.28</ns0:cell><ns0:cell>0.42</ns0:cell><ns0:cell>0.33</ns0:cell><ns0:cell>0.19</ns0:cell><ns0:cell>0.31</ns0:cell><ns0:cell>0.32</ns0:cell><ns0:cell /><ns0:cell>0.30</ns0:cell><ns0:cell>0.30</ns0:cell><ns0:cell>0.37</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>CV</ns0:cell><ns0:cell cols='2'>12.50</ns0:cell><ns0:cell>20.57</ns0:cell><ns0:cell>13.98</ns0:cell><ns0:cell>7.30</ns0:cell><ns0:cell>12.02</ns0:cell><ns0:cell cols='2'>13.19</ns0:cell><ns0:cell>11.04</ns0:cell><ns0:cell>13.51</ns0:cell><ns0:cell>15.35</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>SC (mm)</ns0:cell><ns0:cell cols='2'>10.00</ns0:cell><ns0:cell>9.55</ns0:cell><ns0:cell>10.51</ns0:cell><ns0:cell>10.61</ns0:cell><ns0:cell>11.21</ns0:cell><ns0:cell cols='2'>10.49</ns0:cell><ns0:cell>11.65</ns0:cell><ns0:cell>10.75</ns0:cell><ns0:cell>10.61</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>Min-max</ns0:cell><ns0:cell cols='2'>8.63-</ns0:cell><ns0:cell>7.27-</ns0:cell><ns0:cell>9.13-</ns0:cell><ns0:cell>9.35-</ns0:cell><ns0:cell>9.77-</ns0:cell><ns0:cell cols='2'>8.61-</ns0:cell><ns0:cell>10.17-</ns0:cell><ns0:cell>8.9-</ns0:cell><ns0:cell>7.27-</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell cols='2'>11.94</ns0:cell><ns0:cell>10.98</ns0:cell><ns0:cell>11.70</ns0:cell><ns0:cell>12.70</ns0:cell><ns0:cell>13.27</ns0:cell><ns0:cell cols='2'>12.10</ns0:cell><ns0:cell>14.59</ns0:cell><ns0:cell>12.20</ns0:cell><ns0:cell>14.59</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>SD</ns0:cell><ns0:cell cols='2'>0.77</ns0:cell><ns0:cell>1.09</ns0:cell><ns0:cell>0.67</ns0:cell><ns0:cell>0.69</ns0:cell><ns0:cell>0.89</ns0:cell><ns0:cell>0.89</ns0:cell><ns0:cell /><ns0:cell>1.04</ns0:cell><ns0:cell>0.67</ns0:cell><ns0:cell>1.02</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>CV</ns0:cell><ns0:cell cols='2'>7.65</ns0:cell><ns0:cell>11.42</ns0:cell><ns0:cell>6.39</ns0:cell><ns0:cell>6.50</ns0:cell><ns0:cell>7.90</ns0:cell><ns0:cell>8.53</ns0:cell><ns0:cell /><ns0:cell>8.97</ns0:cell><ns0:cell>6.22</ns0:cell><ns0:cell>9.64</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>SA (mm 2 )</ns0:cell><ns0:cell cols='2'>6.52</ns0:cell><ns0:cell>5.79</ns0:cell><ns0:cell>7.15</ns0:cell><ns0:cell>7.71</ns0:cell><ns0:cell>8.42</ns0:cell><ns0:cell>6.74</ns0:cell><ns0:cell /><ns0:cell>9.26</ns0:cell><ns0:cell>7.25</ns0:cell><ns0:cell>7.44</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>Min-max</ns0:cell><ns0:cell cols='2'>4.72-</ns0:cell><ns0:cell>2.43-</ns0:cell><ns0:cell>5.10-</ns0:cell><ns0:cell>5.83-</ns0:cell><ns0:cell>6.15-</ns0:cell><ns0:cell cols='2'>4.82-</ns0:cell><ns0:cell>6.74-</ns0:cell><ns0:cell>5.27-</ns0:cell><ns0:cell>2.43-</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell cols='2'>9.53</ns0:cell><ns0:cell>7.69</ns0:cell><ns0:cell>9.10</ns0:cell><ns0:cell>1.23</ns0:cell><ns0:cell>12.16</ns0:cell><ns0:cell>1.21</ns0:cell><ns0:cell /><ns0:cell>13.54</ns0:cell><ns0:cell>9.61</ns0:cell><ns0:cell>13.54</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>SD</ns0:cell><ns0:cell cols='2'>1.14</ns0:cell><ns0:cell>1.51</ns0:cell><ns0:cell>1.21</ns0:cell><ns0:cell>1.07</ns0:cell><ns0:cell>1.48</ns0:cell><ns0:cell>1.27</ns0:cell><ns0:cell /><ns0:cell>1.63</ns0:cell><ns0:cell>1.08</ns0:cell><ns0:cell>1.62</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>CV</ns0:cell><ns0:cell cols='2'>17.54</ns0:cell><ns0:cell>26.06</ns0:cell><ns0:cell>16.92</ns0:cell><ns0:cell>13.88</ns0:cell><ns0:cell>17.63</ns0:cell><ns0:cell cols='2'>17.58</ns0:cell><ns0:cell>17.65</ns0:cell><ns0:cell>14.96</ns0:cell><ns0:cell>21.76</ns0:cell></ns0:row></ns0:table></ns0:figure>
</ns0:body>
" | "Łódź, 27.03.2020
Dear Editors and Reviewers,
Encouraged by your letter, we would like to re-submit the manuscript entitled: Seed morphology and sculpture of invasive Impatiens capensis Meerb. from different habitats by Agnieszka Rewicz, Monika Myśliwy, Wojciech Adamowski, Marek Podlasiński, Anna Bomanowska. We are very much thankful to the Reviewers for their deep and thorough reviews. We have revised our paper according to their useful suggestions and comments.
All corrections indicated on the manuscript by the Reviewers and by the Editor have been carefully implemented in the text; we provided also version of manuscript with track changes mode.
We hope that you will find this revised and improved version of the manuscript suitable for publishing in the PeerJ.
Thank you once again for considering the publication of our manuscript in PeerJ.
On behalf of the authors’ team.
Yours sincerely,
Agnieszka Rewicz
Response to Reviewer #1 Comments
Reviewer: Kamuran Aktas
Basic reporting
Rewicz and colleagues performed seed morphology and sculpture of Impatiens capensis Meerb. from different habitats. The authors organized the manuscript quite well, with a good introduction, that reports information about family and the genus which Impatiens taxa belonging to, and important historical citations. Moreover, tables and figures are clear and readable.
Experimental design
Material and methods are briefly enough. Research question and aim of study well defined.
Validity of the findings
Results and discussion are adequately structured to clearly describe their findings. Conclusions are well stated.
We are very grateful for this comment.
Response to Reviewer #2 Comments
Basic reporting
Comments to author In the present research work, the authors did work on the species Impatiens capensis seed morphology (SEM), their ultrastructure, and the relationship and variability of various environmental factors have been discussed. The paper is according to the scope of the journal, but I have some suggestions and comments on the present manuscript. The introduction is to the point but need one paragraphs to explain the morphological and molecular difficulties of the genus Impatiens and specifically focus on the specie I. capensis.
Thank you very much for comment. We added this information to the sections Introduction (line 56 – 70; 91-94).
Experimental design
In methodology some paragraphs are more looking background then the methodology. The subtitle “The Species” should be the part of introduction. Methodology is up to the point and well explained the study. This section is well explained the methodology section of the article despite with the results.
Changed: we moved chapter titled: The Species to the Introduction section (line 97-130). We also rephrased and developed this section
Validity of the findings
The results should be written in the past tense, change the present tense to the past tense. The results are written concise, and well explained with relation with methodology. Discussion is well explained and provide some basic information for the study of the species. I would like to suggest comparing your study with the recently published articles of other groups of plants i.e. 1) 10.1002/jemt.23393 2) 10.1002/jemt.23167 Some articles missing volume number and pages number, provide page and volume number in the revised manuscript.
Thank you very much for comment. We corrected the past tense in the Result section (line 268; 285;289;290;303;333;334;338). We discussed our results with the suggested articles (line 392-400).
Comments for the author
The manuscript is well written, but i would like to suggest some major comments on the present manuscript. After that i will be agree to accept the manuscript.
Response to Reviewer #3 Comments
Reviewer 3
Basic reporting
This manuscript is clear and easy to read. The background is also enough to follow and understand the story.
Experimental design
Good experimental design. The methods are clear.
Validity of the findings
For me, the manuscript is a bit too simple for PeerJ. It is not very interesting to me. Many publications report on seed morphology of Impatiens.
Thank you very much for comment.
Editor's Decision
Two of the reviewers suggested improvements in your article to be considered again in PeerJ. The most important issue is that based on previous palynological analyses in Impatiens, in other members of Convolvulaceae and even in other angiosperms hypotheses should be proposed for instance on the variation of the attributes analyzed. Climate or orther ecological variables in addition to soil should be considered. Please pay attention to suggestions of these two reviewers included below and in the attached file.
Dear Editor,
All corrections indicated on the manuscript by the Reviewers have been implemented in the text. We considered also our results with climate in discussion section (line 371-374). In our opinion, the occurrence of the species in the study area is too short (about 33 years) to be able to analyze the results in the context of climate change.
Editor suggestion about discussion previous palynological analyses and comparison with other members of Convolvulaceae is not clear as we didn’t analyzed such topic in our manuscript. We added to the Discussion section information about the taxonomic significance of seeds in other plant families (line 392-400)
" | Here is a paper. Please give your review comments after reading it. |
9,801 | "<ns0:abstract xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:p>Impatiens capensis is an annual plant native to eastern North America that is currently spreading across Europe. In Poland, due to its rapid spread in the secondary range and high competitiveness in relation to native species, it is considered a locally invasive species. The microstructure of seeds is an important tool for solving various taxonomic problems and also provides results useful for determining the impact of various environmental factors on the phenotypic variability of species. This issue is particularly important in relation to invasive species which occupy a wide range of habitats in the invaded range. Unfortunately, we generally found a shortage of information about seed size and absence of descriptions of the ultrastructure of I. capensis seeds in the analyzed literature. The paper presents new data about the seed morphology of I. capensis growing in different habitat conditions in the secondary range of species. Studied populations differed significantly in each of the analyzed traits (seed length, width, circuit, area, roundness, and mass). Our findings showed that anthropogenic disturbances in habitats and some soil parameters (presence of carbonates, potassium, loose sand and moisture) were statistically significant and differentiated seed sizes and morphology in the studied populations of I. capensis. Moreover, our studies showed maximum length values (5.74 mm) and seed widths (3.21 mm) beyond the available literature data. The presented paper provides also a detailed description of the ultrastructure of the seed coat of I capensis using SEM for the first time. Our research indicates two types of epidermal cells on the seeds, between the ribs (elongated with straight anticlinal walls, slightly concave outer periclinal walls and micropapillate secondary sculpture on the edges anticyclic walls), and on the ribs (isodiametric cells with straight anticlinal walls and concave outer periclinal walls). Unlike the size and weight of seeds, the coat ornamentation have turned out to be a permanent feature within the secondary range of I. capensis.</ns0:p></ns0:div>
</ns0:abstract>
<ns0:body xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:head>Introduction</ns0:head><ns0:p>The Impatiens are the most species-rich genus of the family Balsaminaceae, with ca. 1000 species distributed primarily in the Old World tropics and subtropics <ns0:ref type='bibr' target='#b38'>(Grey-Wilson 1980;</ns0:ref><ns0:ref type='bibr' target='#b114'>Yu et al. 2015)</ns0:ref>.</ns0:p><ns0:p>For many years the Impatiens species has been a subject of numerous studies regarding its distribution <ns0:ref type='bibr' target='#b118'>(Zhou et al. 2019)</ns0:ref>, ecology <ns0:ref type='bibr' target='#b1'>(Abrahamson, Hershey 1977;</ns0:ref><ns0:ref type='bibr' target='#b17'>Boyer et al. 2016)</ns0:ref>, physiology <ns0:ref type='bibr' target='#b67'>(Nanda, Kumar 1983;</ns0:ref><ns0:ref type='bibr' target='#b98'>Tooke et al. 2005)</ns0:ref>, biochemistry <ns0:ref type='bibr' target='#b90'>(Sreelakshmi et al. 2018)</ns0:ref>, biology <ns0:ref type='bibr' target='#b43'>(Jacquemart et al. 2015)</ns0:ref>, pollination <ns0:ref type='bibr' target='#b0'>(Abrahamczyk et al. 2017)</ns0:ref>, morphology <ns0:ref type='bibr' target='#b6'>(Akiyama, Ohba 2000;</ns0:ref><ns0:ref type='bibr' target='#b46'>Janssens et al. 2018)</ns0:ref>, systematics <ns0:ref type='bibr'>(Chen et al. 2007a, b;</ns0:ref><ns0:ref type='bibr' target='#b37'>Gogoi et al. 2018)</ns0:ref>, phylogeny and evolution <ns0:ref type='bibr' target='#b44'>(Janssens et al. 2007;</ns0:ref><ns0:ref type='bibr' target='#b76'>Ruchisansakun et al. 2015)</ns0:ref>, and others (see <ns0:ref type='bibr'>Adamowski 2016 onwards)</ns0:ref>. Despite the plethora of publications on various attributes of Impatiens species, our knowledge about this genus is still incomplete. It is because taxonomically Impatiens is one of the most difficult groups to classify and still remains a strong challenge due to the enormous species richness and their extraordinary morphological variability, from several centimeters high annual plants with one flower to four-meter high subshrubs <ns0:ref type='bibr'>(Hooker 1904</ns0:ref><ns0:ref type='bibr'>(Hooker -1906;;</ns0:ref><ns0:ref type='bibr' target='#b38'>Grey-Wilson 1980;</ns0:ref><ns0:ref type='bibr' target='#b37'>Gogoi et al. 2018;</ns0:ref><ns0:ref type='bibr' target='#b77'>Ruchisansakun et al. 2018)</ns0:ref>.</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:02:45912:2:0:NEW 13 Jul 2020)</ns0:p><ns0:p>Manuscript to be reviewed Majority of balsam species grow in hardly accessible mountainous areas and have delicate flowers with complicated morphology <ns0:ref type='bibr' target='#b15'>(Bhaskar 2012;</ns0:ref><ns0:ref type='bibr' target='#b111'>Yu 2012;</ns0:ref><ns0:ref type='bibr' target='#b71'>Rahelivololona et al. 2018)</ns0:ref>. Herbarium specimens of balsams are difficult to prepare due to succulent nature of stems.</ns0:p><ns0:p>Specimens need special preparations such as floral dissections <ns0:ref type='bibr' target='#b85'>(Shui et al. 2011</ns0:ref>) and extensive field notes, otherwise are of limited value. Flower colors quickly fade and the position of particular flower parts is often impossible to recreate from traditionally prepared specimens. One of taxonomically important features within the genus Impatiens is related to the morphology of seeds. The first information on the diversity of the seed coat in the genus Impatiens was reported by <ns0:ref type='bibr' target='#b42'>Hooker and Thomson (1859)</ns0:ref> and <ns0:ref type='bibr' target='#b107'>Warburg and Reiche (1895)</ns0:ref>. Other works were concerned mostly with the shape and size of seeds rather than details of their surface ornamentation <ns0:ref type='bibr' target='#b82'>(Shimizu 1977)</ns0:ref>.</ns0:p><ns0:p>The development of new imaging methods enabled observation and studies of small-sized structures. With the use of the scanning electron microscopy technique (SEM), a detailed analysis of seed coat micromorphology of Impatiens seeds has become possible <ns0:ref type='bibr' target='#b89'>(Song et al. 2005;</ns0:ref><ns0:ref type='bibr' target='#b19'>Chen et al. 2007a;</ns0:ref><ns0:ref type='bibr' target='#b117'>Zhang et al. 2016)</ns0:ref>. Earlier works focused on seed dimensions rarely described the ultrastructure of seeds <ns0:ref type='bibr' target='#b83'>(Shimizu 1979;</ns0:ref><ns0:ref type='bibr' target='#b52'>Lu, Chen 1991)</ns0:ref>, whereas the sculpture on seed coats offers a set of characters which could be used to identify the species, and in combination with other morphological data, could provide crucial evidence for the taxonomy of the genus <ns0:ref type='bibr' target='#b52'>(Lu, Chen 1991;</ns0:ref><ns0:ref type='bibr' target='#b89'>Song et al. 2005;</ns0:ref><ns0:ref type='bibr' target='#b101'>Utami, Shimizu 2005;</ns0:ref><ns0:ref type='bibr' target='#b18'>Cai et al. 2013;</ns0:ref><ns0:ref type='bibr' target='#b114'>Yu et al. 2015)</ns0:ref>.</ns0:p><ns0:p>The seed morphology of Impatiens not only has been used as an important tool for solving various taxonomic problems within the genus Impatiens but also provides results useful for determining the impact of various environmental factors on the phenotypic variability of balsam species <ns0:ref type='bibr' target='#b9'>(Argyres, Schmitt 1991;</ns0:ref><ns0:ref type='bibr' target='#b80'>Schmitt 1993;</ns0:ref><ns0:ref type='bibr' target='#b53'>Maciejewska-Rutkowska, Janczak 2016)</ns0:ref>.</ns0:p><ns0:p>The understanding of environmentally induced variation in an individual plant phenotype is crucial for predicting population responses to environmental changes. This issue is particularly important in relation to invasive species which occupy a wide range of habitats in the invaded range <ns0:ref type='bibr' target='#b75'>(Richards et al. 2006)</ns0:ref>.</ns0:p><ns0:p>Despite an increasing number of publications on the surface of Impatiens seeds based on the use of the SEM technique (e.g. <ns0:ref type='bibr' target='#b83'>Shimizu 1979;</ns0:ref><ns0:ref type='bibr' target='#b112'>Yu et al. 2007;</ns0:ref><ns0:ref type='bibr' target='#b85'>Shui et al. 2011;</ns0:ref><ns0:ref type='bibr'>Xia et al. 2019 and others)</ns0:ref>, there is still a lack of information on the seed micromorphology for majority of species. In fact, despite major studies using novel imaging methods (e.g. <ns0:ref type='bibr' target='#b115'>Yuan et al. 2004;</ns0:ref><ns0:ref type='bibr' target='#b76'>Ruchisansakun et al. 2015;</ns0:ref><ns0:ref type='bibr' target='#b71'>Rahelivololona et al. 2018)</ns0:ref>, the seed morphology of the genus Impatiens is far from being resolved. As yet, it has been investigated only for nearly 170 species, which is about one fifth of all known balsams (Maciejewska-Rutkowska, Janczak 2016).</ns0:p><ns0:p>One of the species with morphologically undescribed seeds is Impatiens capensis (jewelweed, orange balsam), an annual plant native to eastern North America <ns0:ref type='bibr' target='#b58'>(Meusel et al. 1978)</ns0:ref>, which is currently spreading across Europe. Today I. capensis is considered as naturalized in several European countries <ns0:ref type='bibr' target='#b56'>(Matthews et al. 2015)</ns0:ref>, including Poland, where the species is locally established and invasive due to its rapid spread in the secondary range and high competitiveness in relation to native species, even perennials <ns0:ref type='bibr' target='#b96'>(Tokarska-Guzik et al. 2012)</ns0:ref>. In Poland, it was found for the first time in 1987 <ns0:ref type='bibr' target='#b68'>(Pawlaczyk, Adamowski 1991)</ns0:ref>, and it is spreading in the Western Pomerania region <ns0:ref type='bibr' target='#b69'>(Popiela et al. 2015;</ns0:ref><ns0:ref type='bibr'>M Myśliwy, 2017, personal observations)</ns0:ref>.</ns0:p><ns0:p>The species occurs in the area of the Szczecin Lagoon and enters alder forests, willow shrubs, rushes and riparian tall herb fringe communities <ns0:ref type='bibr' target='#b68'>(Pawlaczyk, Adamowski 1991;</ns0:ref><ns0:ref type='bibr' target='#b65'>Myśliwy et al. 2009;</ns0:ref><ns0:ref type='bibr'>M Myśliwy, 2014, personal observations)</ns0:ref>. It also appears in moist anthropogenic habitats, e.g. along roadside ditches (M Myśliwy, 2017, personal observations).</ns0:p><ns0:p>I. capensis is an annual plant growing 0.5-1.5 m or more in height. The flowers are 2.5-3.0 cm long, orange with darker patches in most common f. capensis. The lower sepal forms a light orange nectar spur, 5-9 mm long, which is bent through 180° to lie parallel with the sepalsac <ns0:ref type='bibr' target='#b119'>(Zika 2006)</ns0:ref>. Besides color, it differs from predominantly Eurasiatic I. noli-tangere in the lower sepal more rapidly constricted into the spur and the position of the spur <ns0:ref type='bibr' target='#b120'>(Zika 2009</ns0:ref>). The fruit is a five-valved capsule, 2.0-2.5 cm long and 0.3-0.5 cm wide, which uses explosive dehiscence to eject seeds <ns0:ref type='bibr' target='#b63'>(Moore 1968;</ns0:ref><ns0:ref type='bibr' target='#b36'>Gleason, Cronquist 1991;</ns0:ref><ns0:ref type='bibr' target='#b26'>Day et al. 2012)</ns0:ref>. The seeds are laterally compressed, prolate spheroids, with four strong ribs 5-5.6 x 2.7-3.1 mm <ns0:ref type='bibr' target='#b16'>(Bojňanský, Fargašová 2007)</ns0:ref>. The weight ranges from 6.4 to 26.9 mg <ns0:ref type='bibr' target='#b87'>(Simpson et al. 1985)</ns0:ref>. <ns0:ref type='bibr' target='#b79'>Schemske (1978)</ns0:ref> gives 11.5 mg for cleistogamous seeds and 13.3 mg for chasmogamous ones, and <ns0:ref type='bibr' target='#b105'>Waller (1982)</ns0:ref> 10.6 mg. The seed surface is wrinkly or rough, lusterless, dark-brown, with some roundish and paler spots <ns0:ref type='bibr' target='#b16'>(Bojňanský, Fargašová 2007)</ns0:ref>.</ns0:p><ns0:p>Numerous studies (over several hundred -see Adamowski 2016 onward and the literature cited therein) have been devoted to the ecology, biology and genetics of this species (e.g. <ns0:ref type='bibr' target='#b8'>Antlfinger 1989;</ns0:ref><ns0:ref type='bibr' target='#b81'>Schmitt et al. 1985;</ns0:ref><ns0:ref type='bibr' target='#b28'>Donohue, Schmitt 1999;</ns0:ref><ns0:ref type='bibr' target='#b29'>Donohue et al. 2000;</ns0:ref><ns0:ref type='bibr' target='#b120'>Zika 2009;</ns0:ref><ns0:ref type='bibr' target='#b93'>Tabak, von Wettberg 2008;</ns0:ref><ns0:ref type='bibr' target='#b26'>Day et al. 2012</ns0:ref>). However, a review of the available literature showed scarcity of data on seed size and a complete lack of information describing the morphological variation and the seed coat of I. capensis seeds <ns0:ref type='bibr' target='#b79'>(Schemske 1978;</ns0:ref><ns0:ref type='bibr' target='#b105'>Waller 1982;</ns0:ref><ns0:ref type='bibr' target='#b87'>Simpson et al. 1985;</ns0:ref><ns0:ref type='bibr' target='#b16'>Bojňanský, Fargašová 2007)</ns0:ref>.</ns0:p><ns0:p>The aims of our work were to characterize the micro-morphological traits and ultrastructural data of I. capensis seeds growing in various habitat conditions and their potential morphological variability. Anthropogenic changes in habitats were expected to be an important factor in shaping seed variation in populations.</ns0:p></ns0:div>
<ns0:div><ns0:head>Materials and methods</ns0:head></ns0:div>
<ns0:div><ns0:head>Study sites</ns0:head><ns0:p>We collected the seeds within one year at the turn of August and September 2018 (to eliminate seasonal variability) from eight populations of I. capensis in Poland. The research covered the entire Polish range of this species and various habitats, from natural (alder carrs, hydrophilous tall herbs along rivers, near water seeps, and on the banks of the Szczecin Lagoon) to anthropogenic (tall herbs along roadside ditches, transformed forests along artificial canals) (Table <ns0:ref type='table'>1</ns0:ref>, Fig. <ns0:ref type='figure'>1</ns0:ref>). The studied populations were also subject to different lighting conditions, which were scored using a 3-point scale: plants which grew in willow forests and the understory of alder carrs were strongly shaded (3), while those from tall herb communities were partly shaded by solitary trees (2) or received full sun (1). As the height of I. capensis specimens, the location of capsules within the plant (main stem vs. branches) and their derivation from flowers of various types (cleistogamous vs. chasmogamous) may influence seeds weight <ns0:ref type='bibr' target='#b105'>(Waller 1982)</ns0:ref>, the seeds for our study were collected always from the main stems of 8-10 plants of similar (average for the population) height and from capsules derived from chasmogamous flowers to minimize the biases. Species nomenclature was adopted after Euro+Med PlantBase (Euro+Med PlantBase).</ns0:p></ns0:div>
<ns0:div><ns0:head>Biometric and SEM analysis</ns0:head><ns0:p>We analyzed from 24 to 30 mature seeds from each population for biometric analysis. We measured four quantified seed traits: seed length (SL); seed width (SW); seed circuit (SC); and seed area (SA). The seeds were measured as previously described in <ns0:ref type='bibr' target='#b73'>Rewicz et al. (2017)</ns0:ref>. In order to describe the seed mass, we used 15 seeds from each population. The weighing of the seeds was performed with the use of Ohaus PA 21 weight. We used Roundness according to the formula: R=4 x area/π [Major axis]^2 defined by <ns0:ref type='bibr' target='#b31'>Ferreira and Wayne (2010)</ns0:ref>.</ns0:p><ns0:p>We used eight seeds from each population for SEM analyses. The seeds were air-dried and sputter-coated with a 4nm thick layer of gold (Leica EM ACE200). The SEM work was performed with the Phenom Pro X Scanning Electron Microscope at the Department of Invertebrate Zoology and Hydrobiology, University of Lodz, Poland. The 3D models of the seed surface were generated using dedicated software 3D Roughness Reconstruction for Phenom. SEM micrographs were analyzed as previously described in <ns0:ref type='bibr' target='#b73'>Rewicz et al. (2017)</ns0:ref>. The seed shape terminology and types of seed surfaces were adopted from: <ns0:ref type='bibr' target='#b11'>Barthlott (1981)</ns0:ref> </ns0:p></ns0:div>
<ns0:div><ns0:head>Soil properties</ns0:head><ns0:p>In order to characterize habitat conditions at each locality, five soil samples cores (0-20 cm depth) were collected and then mixed together into one composite sample. Soil samples were dried at room temperature, and then rub through a sieve to remove fractions larger than 1 mm.</ns0:p><ns0:p>The following physico-chemical soil parameters were determined <ns0:ref type='bibr' target='#b12'>(Bednarek et al. 2011)</ns0:ref>, as it was first described in <ns0:ref type='bibr' target='#b64'>Myśliwy (2019)</ns0:ref>: organic matter content defined as the loss on ignition (LOI) -soil samples annealed at 550ºC (%); grain composition (the content of sand, silt, clay) -Bouyoucos's sedimentation method, modified by Casagrande and Prószyński; granulometric categories according to the PSSS (2009) classification; soil reaction (pH) -the potentiometric method, in 1-M solution of KCl; soil calcium carbonate (CaCO 3 ) content (%) -the Scheibler's method; organic carbon (C org ) content (%), and total nitrogen (N tot ) content (%) were determined by an elemental analyzer CHNS/O FlashSmart (Thermo Scientific), and the C/N ratio; the content of available forms of soil nutrients (mg/100 g soil): calcium (Ca) and sodium (Na) were determined spectrophotometrically (Ca -AAS and Na -EAS) on ICE3000; potassium (K) and phosphorus (P) -were measured according to the Egner-Riehm method, level of magnesium (Mg) -was measured using Schachtschabel's method; soil moisture content, hand-felt assessed directly in the field using a 4-point scale recommended by the Soil Science Society of Poland (2017): (1) dry (the soil crumbles and turns to dust, it is neither cool nor moist to touch; it darkens visibly after wetting), ( <ns0:ref type='formula'>2</ns0:ref>) fresh (the soil feels cool, but no moisture is felt; darkens after wetting), (3) moist (the soil moistens fingers and tissue paper, but water does not leak when squeezed; clayey, loamy and some dusty soils are plastic; does not darken after wetting), ( <ns0:ref type='formula'>4</ns0:ref>) wet (water leaks from the soil when squeezed aggregates, soil smears).</ns0:p></ns0:div>
<ns0:div><ns0:head>Data analysis</ns0:head><ns0:p>The five following basic characteristic traits were calculated: arithmetic average (x), minimum and maximum values (min/max), coefficient of variation (CV) and standard deviation (SD). The distribution of the data was not normal; statistical analysis was based on the Kruskal-Wallis test (for p ≤0.05), which is a non-parametric alternative to ANOVA <ns0:ref type='bibr' target='#b116'>(Zar 1984)</ns0:ref>.</ns0:p><ns0:p>Correlation between pairs of morphological characters was evaluated using Spearman's correlation coefficient and the values were adopted after Meissner (2010), (correlation; less than 0.20 -very poor, 0.21-0.39 -weak, 0.40-0.69 -moderate, 0.70-0.89 -strong and above 0.89very strong).</ns0:p><ns0:p>The cluster analysis based on the nearest neighbor method was performed using the matrix on the population's mean values. As the dataset required a linear response model <ns0:ref type='bibr' target='#b49'>(Jongman et al. 1995)</ns0:ref>, the Redundancy Analysis (RDA) was used to relate the variability of morphological traits of seeds to environmental variables. The variables C org and N tot were excluded from the RDA as they were strongly correlated with organic matter content (LOI). The Monte Carlo permutation test with the forward selection of environmental variables was applied to determine the importance and statistical significance of variables in explaining the variability in seeds. The software package Canoco v.4.5 (ter Braak, Šmilauer 2002), MVSP 3.2 <ns0:ref type='bibr' target='#b50'>(Kovach 2010</ns0:ref>) and STATISTICA <ns0:ref type='bibr'>PL. ver. 13.1 (Stat-Soft Inc. 2011)</ns0:ref> were used for all analyses (van Emden 2008; <ns0:ref type='bibr' target='#b51'>Lepš, Šmilauer 2003)</ns0:ref>.</ns0:p></ns0:div>
<ns0:div><ns0:head>Results</ns0:head></ns0:div>
<ns0:div><ns0:head>Biometric analysis</ns0:head><ns0:p>The seeds from the G (Police) population were characterized by the biggest seeds with the average values of seed length (SL) 4.60 mm, width (SW) 2.71 mm, circuit (SC) 11.65 mm and area (SA) 9.26 mm 2 . Comparatively high values of analyzed traits were observed in the E population (Święta). In the B (Lubin) population, the shortest (mean SL 3.88 mm) and narrowest seeds (mean SW 2.03 mm) were observed (Table <ns0:ref type='table'>2</ns0:ref>).</ns0:p><ns0:p>The minimum values of analyzed traits were also recorded in the B (Lubin) population (SL -3.16 mm, SW -1.12 mm, SC -7.27 mm, SA -2.43 mm 2 ). The maximum values of length (5.74 mm), circuit (14.59 mm) and area (13.54 mm 2 ) were recorded in the G population (Police).</ns0:p><ns0:p>A very strong Spearman correlation (r = 0.94) was observed between the seed area and circuit (Table <ns0:ref type='table'>3</ns0:ref>). The most variable features were the seed area (CV = 21.76 %) and width (CV = 15.35 %). The variation of seed traits ranged insignificantly from 6.19 % (H population) to 10.28 % (B) -SL; from 7.30 % (D) to 20.57 % (B) -SW; from 6.22 % (H) to 11.42 % (B) -SC; and from 13.88 % (D) to 26.06 % (B) -SA, respectively.</ns0:p><ns0:p>The G (Police -11.42 mg) and E (Święta -9.82 mg) populations are characterized by the heaviest seeds. The lightest seeds were observed in the following populations: B (Lubin -6.52 mg) and H (Trzebieradz -6.92 mg) (Fig. <ns0:ref type='figure'>2</ns0:ref>, Table <ns0:ref type='table'>2</ns0:ref>).</ns0:p><ns0:p>The Kruskal-Wallis test showed that the I. capensis populations differed significantly in each of the analyzed traits. The conducted post hoc test (DunnTest) showed that the populations from: Police (G), followed by Czarnocin (D), Święta (E) and Trzebieradz (H) showed the greatest variation in terms of studied traits among all the populations (Table <ns0:ref type='table'>4</ns0:ref>).</ns0:p><ns0:p>The similarity analysis made using Euclidean's distances showed two main clusters (Fig. <ns0:ref type='figure'>3</ns0:ref>). The first cluster includes six populations of I. capensis (A-D, F, H), all derived from natural habitats, while the other cluster groups two populations (E, G) from anthropogenic habitats, where examined plants were the highest (Table <ns0:ref type='table'>1</ns0:ref>). According to the dendrogram (Fig. <ns0:ref type='figure'>3</ns0:ref>), populations C and F are the closest to each other; both were associated with river valleys (the Dziwna and Oder Rivers, respectively) and close to the river bed, hence under the influence of flooding. The D and A populations were growing in tall herb communities on the banks of the Szczecin Lagoon. The most distinct populations in the first cluster (H and B) were also found on the banks of the Szczecin Lagoon, but they had the lowest average height and differed in habitat conditions from the other populations of this cluster (Table <ns0:ref type='table'>1</ns0:ref>).</ns0:p><ns0:p>The morphological parameters of seed shape -roundness -showed statistically significant differences between the populations (p < 0.05). For roundness, the highest value was recorded at D -Czarnocin (0.58) (tall herbs on the bank of the Szczecin Lagoon) and the lowest -at H -Trzebieradz (0.47) (alder carr) (Table <ns0:ref type='table'>5</ns0:ref>).</ns0:p></ns0:div>
<ns0:div><ns0:head>Biometric variability of seeds and its relationship with environmental conditions</ns0:head><ns0:p>PeerJ reviewing PDF | (2020:02:45912:2:0:NEW 13 Jul 2020)</ns0:p></ns0:div>
<ns0:div><ns0:head>Manuscript to be reviewed</ns0:head><ns0:p>All environmental variables included in the RDA explained 35.6 % of the total variation.</ns0:p><ns0:p>The results of step-wise forward selection of variables indicated that five variables: anthropogenic disturbances (Anthrop), carbonates (CaCO 3 ), loose sand presence (LoSa), potassium (K), and soil moisture content (Moist) were statistically significant and differentiated the studied populations of I. capensis (Fig. <ns0:ref type='figure' target='#fig_1'>4</ns0:ref>). Along the gradient represented by Axis I, the highest correlation between the sample position and environmental variables (the so-called interset correlation) was typical of anthropogenic disturbances and CaCO 3 , followed by the degree of shading and soil Ca, while the C/N ratio was most closely correlated with Axis II, followed by soil contents of P, Na, K, and organic soil.</ns0:p><ns0:p>The location of population H (Trzebieradz) in the ordination space (the upper part of the RDA diagram) was associated with the highest C/N ratio, the highest soil moisture and shading, as well as with the lowest soil pH and the lowest soil contents of CaCO 3 , Ca, P and K (Fig. <ns0:ref type='figure' target='#fig_1'>4</ns0:ref>, Table <ns0:ref type='table'>6</ns0:ref>). At the same time, the H population was dominated by short specimens (Table <ns0:ref type='table'>1</ns0:ref>), with one of the lightest seeds and average values of biometric traits (Table <ns0:ref type='table'>2</ns0:ref>). In contrast, populations characterized by the longest, widest and heaviest seeds -E (Święta) and G (Police), located in the right-hand side of the RDA diagram, were also related to a relatively high C/N ratio, but unlike the previous population, they were associated with a low level of soil moisture as well as the highest anthropogenic disturbances (Fig. <ns0:ref type='figure' target='#fig_1'>4</ns0:ref>), and consisted of the tallest specimens (Table <ns0:ref type='table'>1</ns0:ref>).</ns0:p><ns0:p>Population D (Czarnocin) occupied the bottom part of the diagram and was distinct in its organic soil, with the highest content of organic matter (LOI), P, K, Mg, and Na contents in the soil, as well as the lowest C/N ratio (Table <ns0:ref type='table'>6</ns0:ref>). The lowest values of seed biometric traits were typical of population B (Lubin), located in the left-hand part of the RDA diagram (Fig. <ns0:ref type='figure' target='#fig_1'>4</ns0:ref>, Table <ns0:ref type='table'>2</ns0:ref>), and associated with high soil pH and the highest contents of soil carbonates and calcium, as well as a low level of soil moisture (Table <ns0:ref type='table'>6</ns0:ref>). The other populations (A, Podgrodzie; C, Unin; F, Szczecin-Zdroje) were also on the left side of the diagram, but closer to the center (Fig. <ns0:ref type='figure' target='#fig_1'>4</ns0:ref>). Neither their seed biometric traits nor habitat conditions were distinct (Table <ns0:ref type='table'>2, 6</ns0:ref>).</ns0:p></ns0:div>
<ns0:div><ns0:head>The structure of seed surface</ns0:head><ns0:p>Seeds of I. capensis had a round shape, a lusterless, rough, and dark-brown surface, without roundish and paler spots (Fig. <ns0:ref type='figure'>5</ns0:ref>). The seeds had four strong, clear ribs, the apex and bottom narrowed. Each rib was built of rows of 4-5 cells and had a darker color than the surface of the seeds between them (Fig. <ns0:ref type='figure'>5 H</ns0:ref>). The seed coat is composed of two types of epidermal cells (Fig. <ns0:ref type='figure'>5 E, H</ns0:ref>) creating a net-like pattern. The cells of the seed surface between the ribs were: elongated with straight anticlinal walls (Fig. <ns0:ref type='figure'>5 E</ns0:ref>), raised cell boundaries between the cells (Fig. <ns0:ref type='figure'>5</ns0:ref> G), slightly concave outer periclinal walls (Fig. <ns0:ref type='figure'>5 F, G</ns0:ref>) and a micropapillate secondary sculpture on the edges of anticyclic walls (Fig. <ns0:ref type='figure'>5 F</ns0:ref>). Near the ribs, there were rows of 4-7 isodiametric cells (Fig. <ns0:ref type='figure'>5 I, K</ns0:ref>) with straight anticlinal walls (Fig. <ns0:ref type='figure'>5 L</ns0:ref>), with raised cell boundaries (Fig. <ns0:ref type='figure'>5 M</ns0:ref>) and concave outer periclinal walls. Seeds from all studied populations did not differ in their ultrastructure (Fig. <ns0:ref type='figure'>5 A -D</ns0:ref>).</ns0:p></ns0:div>
<ns0:div><ns0:head>Discussion</ns0:head><ns0:p>Nowadays, with the present development of SEM techniques, the morphological characters of seeds and their ultrastructural characteristics are very useful for identification and taxonomic delimitation of various angiosperms groups, e.g. Brasssicaceae <ns0:ref type='bibr' target='#b94'>(Tantaway et al. 2004</ns0:ref>), Caryophyllaceae <ns0:ref type='bibr' target='#b99'>(Ullah et al. 2019a</ns0:ref><ns0:ref type='bibr' target='#b100'>(Ullah et al. , 2019b))</ns0:ref>, Poaceae (Martín-Gómez et al. 2019b), Cyperaceae <ns0:ref type='bibr' target='#b109'>(Więcław et al. 2017)</ns0:ref>, Ranunculaceae <ns0:ref type='bibr' target='#b25'>(Constantinidis et al. 2001;</ns0:ref><ns0:ref type='bibr' target='#b73'>Rewicz et al. 2017;</ns0:ref><ns0:ref type='bibr' target='#b54'>Martín-Gómez et al. 2019a;</ns0:ref><ns0:ref type='bibr' target='#b40'>Hadidchi et al. 2020)</ns0:ref>, Rosaceae <ns0:ref type='bibr' target='#b10'>(Ballian, Mujagić-Pašić 2013)</ns0:ref>, Onagraceae <ns0:ref type='bibr' target='#b5'>(Akbari, Azizian 2006)</ns0:ref>, and Orchidaceae <ns0:ref type='bibr' target='#b32'>(Gamarra et al. 2007</ns0:ref><ns0:ref type='bibr' target='#b33'>(Gamarra et al. , 2010;;</ns0:ref><ns0:ref type='bibr' target='#b74'>Rewicz et al. 2016)</ns0:ref>. Although seed morphology alone does not provide universally applicable key characters for species identification, it can be as helpful as many other characters used in taxonomy.</ns0:p><ns0:p>The Balsaminaceae species have a diverse and elaborately sculptured seed coat.</ns0:p><ns0:p>Unfortunately, till now seed morphology has been observed only for a small group from the Impatiens species, so the use of morphological traits of seeds in taxonomy and classification of species in this group is limited (e.g. <ns0:ref type='bibr' target='#b89'>Song et al. 2005;</ns0:ref><ns0:ref type='bibr' target='#b101'>Utami, Shimizu 2005;</ns0:ref><ns0:ref type='bibr' target='#b19'>Chen et al. 2007a;</ns0:ref><ns0:ref type='bibr' target='#b112'>Yu et al. 2007;</ns0:ref><ns0:ref type='bibr' target='#b47'>Jin et al. 2008;</ns0:ref><ns0:ref type='bibr' target='#b85'>Shui et al. 2011)</ns0:ref>.</ns0:p><ns0:p>The paper has provided new data about the seed morphology and seed coat sculpture of I. capensis, as well as new information about the area -the circuit that has not been reported before. Moreover, our studies have shown maximum length values (5.74 mm) and seed widths (3.21 mm) beyond the available literature data. <ns0:ref type='bibr' target='#b16'>Bojňanský and Fargašová (2007)</ns0:ref> reported seeds of I. capensis 5-5.6 (length) x 2.7-3.1 (widths) mm. Our studies have shown two types of cells on the ultrastructure of seeds, between the ribs and on the ribs, not described in the literature yet (Fig. <ns0:ref type='figure'>5</ns0:ref>). The occurrence of several types of epidermal cells on seeds of the genus Impatiens was previously noted, e.g. in Impatiens aconitoides <ns0:ref type='bibr' target='#b85'>Shui et al. (2011)</ns0:ref> reported three types of epidermal cells. We also have not confirmed in any of the studied populations roundish spots on the surface of seeds (Fig. <ns0:ref type='figure'>6</ns0:ref>) reported by <ns0:ref type='bibr' target='#b16'>Bojňanský and Fargašová (2007)</ns0:ref>. The observed differences may result from a different geographical origin of the examined seeds. Seeds of I. capensis in our study came from populations from spontaneous localities with various habitat conditions, while <ns0:ref type='bibr' target='#b16'>Bojňanský and Fargašová (2007)</ns0:ref> described seeds from cultivation of unknown origin.</ns0:p><ns0:p>The analysis of SEM micrographs of I. noli-tangere seeds closely related to I. capensis <ns0:ref type='bibr' target='#b114'>(Yu et al. 2015)</ns0:ref> has shown that seed coats of this species vary significantly depending on the geographical origin of the seeds <ns0:ref type='bibr' target='#b101'>(Utami, Shimizu 2005;</ns0:ref><ns0:ref type='bibr'>Chen et al. 2007 a;</ns0:ref><ns0:ref type='bibr' target='#b47'>Jin et al. 2008</ns0:ref>). On the other hand, the comparison of the seed micromorphology of I. capensis has not shown similarity to seed coat ornamentation of the aforementioned I. noli-tangere (with narrow and ellipsoid seeds, fine reticulate subtype, testa cells with reticulate thickened outer walls; <ns0:ref type='bibr' target='#b89'>Song et al. 2005;</ns0:ref><ns0:ref type='bibr' target='#b101'>Utami, Shimizu 2005;</ns0:ref><ns0:ref type='bibr' target='#b19'>Chen et al. 2007a;</ns0:ref><ns0:ref type='bibr' target='#b47'>Jin et al. 2008)</ns0:ref>. Despite the fact that both species are closely related and may be confused <ns0:ref type='bibr' target='#b120'>(Zika 2009;</ns0:ref><ns0:ref type='bibr' target='#b114'>Yu et al. 2015)</ns0:ref>, their seeds clearly differ morphologically. It proves that new data presented here may be useful in the identification of these species. In turn, there is no information about the seed morphology of I. pallida, which is sympatric and synchronic species to I. capensis <ns0:ref type='bibr' target='#b78'>(Rust 1977)</ns0:ref>, which makes this subject even more difficult. To elucidate the overall variation in seed coat micromorphology and to implement this feature to taxonomy of I. capensis, more samplings have to be conducted, including the native range of orange jewelweed, as well as other closely related species. This phenomenon should be the basis for further comparisons and studies due to the fact that the seed ultrastructure is considered as a constant feature within a taxonomic unit <ns0:ref type='bibr' target='#b91'>(Stace 1992)</ns0:ref> and, as morphological studies show, seed shape and size are highly diverse at the genera and species level <ns0:ref type='bibr' target='#b112'>(Yu et al. 2007;</ns0:ref><ns0:ref type='bibr' target='#b47'>Jin et al. 2008;</ns0:ref><ns0:ref type='bibr' target='#b85'>Shui et al. 2011;</ns0:ref><ns0:ref type='bibr' target='#b99'>Ullah et al. 2019a</ns0:ref><ns0:ref type='bibr' target='#b100'>Ullah et al. , 2019b;;</ns0:ref><ns0:ref type='bibr' target='#b40'>Hadidchi et al. 2020</ns0:ref>). Both statements have been proven for I. capensis in Poland.</ns0:p><ns0:p>Data concerning the size, shape and structure of seeds not only have been used as an important tool for solving various taxonomic problems within the genus Impatiens but also provide results useful for determining the impact of various environmental factors on the PeerJ reviewing PDF | (2020:02:45912:2:0:NEW 13 Jul 2020)</ns0:p><ns0:p>Manuscript to be reviewed phenotypic variability of these species <ns0:ref type='bibr' target='#b13'>(Bell et al. 1991;</ns0:ref><ns0:ref type='bibr' target='#b9'>Argyres, Schmitt 1991;</ns0:ref><ns0:ref type='bibr' target='#b80'>Schmitt 1993;</ns0:ref><ns0:ref type='bibr' target='#b21'>Chmura et al. 2013;</ns0:ref><ns0:ref type='bibr' target='#b53'>Maciejewska-Rutkowska, Janczak 2016)</ns0:ref>.</ns0:p><ns0:p>Environmental heterogeneity is indicated as a major factor driving morphological changes <ns0:ref type='bibr' target='#b66'>(Nakazato et al. 2008)</ns0:ref>. Seeds are sensitive to changes in biotic and abiotic conditions <ns0:ref type='bibr' target='#b59'>(Moles et al. 2005)</ns0:ref>. According to <ns0:ref type='bibr' target='#b86'>Silvertown (1989)</ns0:ref>, the correlation between seed size and the place where plant is growing is an adaptative feature. Bigger seeds occur in habitats with stable environmental conditions, where seedlings may grow slowly. Small seeds are generally produced by plants with a short life cycle, growing mainly in disturbed habitats.</ns0:p><ns0:p>Orange balsam is known for colonizing a wide range of habitats <ns0:ref type='bibr' target='#b79'>(Schemske 1978;</ns0:ref><ns0:ref type='bibr' target='#b104'>Waller 1980)</ns0:ref>. Moreover, <ns0:ref type='bibr' target='#b87'>Simpson et al. (1985)</ns0:ref> have shown that I. capensis vegetative and reproductive growth parameters reflect habitat differences. Light availability <ns0:ref type='bibr' target='#b87'>(Simpson et al. 1985)</ns0:ref> as well as soil moisture and pH <ns0:ref type='bibr' target='#b104'>(Waller 1980)</ns0:ref> have been reported to affect growth patterns of I. capensis.</ns0:p><ns0:p>Our studies indicate that five environmental variables were statistically significant and differentiated studied populations in terms of seed size and weight: anthropogenic disturbances (which may serve as a proxy for habitat fertility), carbonates (CaCO 3 ), loose sand presence, potassium (K), and soil moisture (Fig. <ns0:ref type='figure' target='#fig_1'>4</ns0:ref>). Populations G (Police) and E (Święta), occurring in the most disturbed anthropogenic habitats (artificial canal and roadside), have the heaviest seeds as a result of growth in favorable environmental conditions (neutral or slightly acidic soil with a relatively high C/N ratio). In turn, population B (Lubin) with the smallest and lightest seeds was associated with high soil pH, and the highest contents of soil carbonates and calcium. At the same time <ns0:ref type='bibr' target='#b105'>Waller (1982)</ns0:ref> proved that higher nodes of I. capensis individuals tended to produce heavier seeds. In <ns0:ref type='bibr' target='#b105'>Waller's (1982)</ns0:ref> opinion, the position effect probably leads to a greater mean seed size for higher plants. Our results are consistent with both studies, as the largest and heaviest seeds were obtained from populations G (Police) and E (Święta), formed by the highest plants, growing in large numbers and densities.</ns0:p><ns0:p>Another important factor shaping a diverse array of plant traits, including morphological features, is climate <ns0:ref type='bibr' target='#b66'>(Nakazato et al. 2008;</ns0:ref><ns0:ref type='bibr' target='#b24'>Colautti, Barrett 2013;</ns0:ref><ns0:ref type='bibr' target='#b102'>van Boheemen et al. 2019)</ns0:ref>, and temperature and precipitation gradients are the main climatic factors driving the adaptive diversification of species <ns0:ref type='bibr' target='#b66'>(Nakazato et al. 2008)</ns0:ref>. As it seems, climatic conditions have had a limited influence on investigated parameters of seeds till now, due to a small area of secondary distribution of I. capensis in Poland <ns0:ref type='bibr' target='#b3'>(Adamowski et al. 2018;</ns0:ref><ns0:ref type='bibr' /> Manuscript to be reviewed -little over 30 years. Although this investigated plant has only a few localities and in a relatively small area in Poland, it should be noted that rapid expansion across environmental gradients has been reported for several plants introduced to a new area and species could evolve very quickly under the effect of environmental factors <ns0:ref type='bibr' target='#b27'>(Dlugosh, Parker 2008;</ns0:ref><ns0:ref type='bibr'>Colautti, Barret 2013;</ns0:ref><ns0:ref type='bibr' target='#b60'>Molina-Montenegro et al. 2018;</ns0:ref><ns0:ref type='bibr' target='#b102'>van Boheemen et al. 2019)</ns0:ref>.</ns0:p><ns0:p>Phenotypic plasticity has been seen as the primary mechanism enabling aliens to colonize new, environmentally diverse areas <ns0:ref type='bibr' target='#b75'>(Richards et al. 2006;</ns0:ref><ns0:ref type='bibr' target='#b62'>Molina-Montenegro et al. 2010)</ns0:ref>.</ns0:p><ns0:p>However, the latest research indicates that alien plants can evolve quickly in recently occupied areas, so bothphenotypic plasticity and evolution of reproductive featurescould be a relevant factor for successful invasions <ns0:ref type='bibr' target='#b35'>(Geng et al. 2007;</ns0:ref><ns0:ref type='bibr' target='#b60'>Molina-Montenegro et al. 2018</ns0:ref>). An evolutionary explanation for plant invasiveness implies that seed and fruit traits are crucial for invasive plants since they are related to dispersal strategies and mechanisms to cope with environmental stress. Some works indicate that native and invasive populations employ different strategies for growth and reproduction <ns0:ref type='bibr' target='#b23'>(Chun et al. 2007;</ns0:ref><ns0:ref type='bibr' target='#b62'>Molina-Montenegro et al. 2010</ns0:ref><ns0:ref type='bibr'>, 2018)</ns0:ref>. Results by <ns0:ref type='bibr' target='#b60'>Molina-Montenegro et al. (2018)</ns0:ref> suggest that some seed traits of invasive plant species with rapid adaptive capacity can evolve in order to maximize their establishment in new environments and can be heritable.</ns0:p><ns0:p>Due to scarcity of data we could not point out the presence of morphological differentiation between native and invasive populations of I. capensis, and it is impossible to conclude whether the seed traits evolve. However, we should be aware that I. capensis is recognized as an invasive species in Poland. Therefore, it is suspected that this species, while adapting to occupy new territories and competing with native species, developed specific adaptations, contributing to its success in the new environments.</ns0:p></ns0:div>
<ns0:div><ns0:head>Conclusions</ns0:head><ns0:p>The paper provides new data about the seed morphology and seed coat sculpture of I. capensis, which may be useful in the taxonomic identification of this species among other closely related species. Such detailed seed coat ornamentation was described for the first time.</ns0:p><ns0:p>Studies on the developmental variation of seed coat sculpture especially with species closely related with I. capensis, which may provide insights into a better understanding of the evolutionary relationships among different types of sculpture, are also urgently needed.</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:02:45912:2:0:NEW 13 Jul 2020)</ns0:p><ns0:p>Manuscript to be reviewed</ns0:p><ns0:p>Our data have provided new information on the plasticity of seeds of the tested species.</ns0:p><ns0:p>Moreover, there are few papers about the problem of phenotypic variability within the Impatiens genus. Data on the morphology of seeds also provide results useful for determining the impact of various environmental factors on the morphological traits of species. Such research provides information whether a given feature is actually stable or is susceptible to environmental changes.</ns0:p><ns0:p>Our results confirm that some habitat variables, especially anthropogenic disturbances and selected soil properties, have proven to be important factors in shaping I. capensis seeds morphological variation. In turn, the seed coat sculpture has turned out to be a constant feature within the secondary range of this species in Poland. . Explanations: *Statistically significant variables;Shadow, degree of shading. For codes of populations and soil properties, see Table <ns0:ref type='table'>1 and 6</ns0:ref>, respectively</ns0:p></ns0:div><ns0:figure xml:id='fig_0'><ns0:head /><ns0:label /><ns0:figDesc>Fig. 1) and short time of residence PeerJ reviewing PDF | (2020:02:45912:2:0:NEW 13 Jul 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_1'><ns0:head>Figure 4</ns0:head><ns0:label>4</ns0:label><ns0:figDesc>Figure 4</ns0:figDesc></ns0:figure>
<ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='37,42.52,250.12,525.00,371.25' type='bitmap' /></ns0:figure>
<ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='38,42.52,255.37,525.00,371.25' type='bitmap' /></ns0:figure>
<ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='45,42.52,199.12,525.00,162.75' type='bitmap' /></ns0:figure>
<ns0:figure type='table' xml:id='tab_0'><ns0:head /><ns0:label /><ns0:figDesc>Dendrogram of similarities of populations of Impatiens capensis Meerb. in Poland, obtained by the nearest neighbor method.</ns0:figDesc><ns0:table><ns0:row><ns0:cell cols='5'>Code SL (mm) Locality 4.05 Width Weight (mg) 7.66 Length 1.00 A Length Latitude 3.88 6.52 B Figure 3 1 1</ns0:cell><ns0:cell>Longitude 4.23 1.00 8.16 0.47 C Width</ns0:cell><ns0:cell cols='2'>Habitat all herbs on 4.11 4.46 0.73 7.82 9.82 0.85 D E Circuit</ns0:cell><ns0:cell>Average plants height 4.17 0.84 [cm] 8.62 0.72 F Area</ns0:cell><ns0:cell cols='2'>29 4.60 analyzed seeds 11.42 Number of 4.41 6.92 Population size 20-30 4.24 individuals) 8.37 (mature G H x</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell cols='3'>A Min-max Podgrodzie 3.50-Circuit 4.64 Area</ns0:cell><ns0:cell cols='4'>53.740222°14.306667°t the bank of 3.16-3.68-3.59-3.85-1.00 Szczecin Lagoon 4.48 4.70 4.75 5.26</ns0:cell><ns0:cell>3.43-0.94 130 4.73 1.00</ns0:cell><ns0:cell>3.88-5.74</ns0:cell><ns0:cell>3.59-4.82</ns0:cell><ns0:cell>3.16-5.74</ns0:cell></ns0:row><ns0:row><ns0:cell>2</ns0:cell><ns0:cell>B SD</ns0:cell><ns0:cell>Lubin</ns0:cell><ns0:cell>0.26</ns0:cell><ns0:cell cols='4'>53.865056°14.426778°t all herbs and grasses near 0.40 0.29 0.30 0.32</ns0:cell><ns0:cell>50 0.40</ns0:cell><ns0:cell>24 0.40</ns0:cell><ns0:cell>not less than 20 0.27 0.40</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>CV</ns0:cell><ns0:cell /><ns0:cell>6.50</ns0:cell><ns0:cell>10.28</ns0:cell><ns0:cell>6.95</ns0:cell><ns0:cell cols='2'>water seeps 7.35 7.21</ns0:cell><ns0:cell>9.71</ns0:cell><ns0:cell>8.75</ns0:cell><ns0:cell>6.19</ns0:cell><ns0:cell>9.34</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell cols='2'>C SW (mm) Unin</ns0:cell><ns0:cell>2.23</ns0:cell><ns0:cell cols='4'>53.894806°14.634444°t all herbs along the 2.03 2.36 2.56 2.60</ns0:cell><ns0:cell>120 2.40</ns0:cell><ns0:cell>27 2.71</ns0:cell><ns0:cell>2.23</ns0:cell><ns0:cell>20-30 2.39</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell cols='2'>Min-max</ns0:cell><ns0:cell>1.53-</ns0:cell><ns0:cell>1.12-</ns0:cell><ns0:cell>1.78-</ns0:cell><ns0:cell cols='2'>river all herbs on 2.14-2.19-</ns0:cell><ns0:cell>1.88-</ns0:cell><ns0:cell>30 2.15-</ns0:cell><ns0:cell>1.71-</ns0:cell><ns0:cell>over 50 1.12-</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>D SD</ns0:cell><ns0:cell cols='2'>Czarnocin 2.82 0.28</ns0:cell><ns0:cell cols='4'>53.722306°14.549167°t the bank of 2.61 3.00 2.94 3.33 Szczecin Lagoon 0.42 0.33 0.19 0.31</ns0:cell><ns0:cell>2.99 130 0.32</ns0:cell><ns0:cell>3.21 0.30</ns0:cell><ns0:cell>2.93 0.30</ns0:cell><ns0:cell>3.33 0.37</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell cols='2'>E CV SC (mm) Święta</ns0:cell><ns0:cell>12.50 10.00</ns0:cell><ns0:cell cols='4'>53.559861°14.659083°t all herbs 20.57 13.98 7.30 12.02 along roadside 9.55 10.51 10.61 11.21</ns0:cell><ns0:cell>13.19 165 10.49</ns0:cell><ns0:cell>27 11.04 11.65</ns0:cell><ns0:cell>13.51 10.75</ns0:cell><ns0:cell>over 100 15.35 10.61</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell cols='3'>F Min-max Szczecin-8.63-Zdroje 11.94</ns0:cell><ns0:cell cols='4'>ditch all herbs 9.35-9.77-53.382861°14.614944°t 7.27-9.13-along the river 10.98 11.70 12.70 13.27</ns0:cell><ns0:cell>8.61-120 12.10</ns0:cell><ns0:cell>29 10.17-14.59</ns0:cell><ns0:cell>8.9-12.20</ns0:cell><ns0:cell>7.27-over 50 14.59</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>SD</ns0:cell><ns0:cell /><ns0:cell>0.77</ns0:cell><ns0:cell>1.09</ns0:cell><ns0:cell>0.67</ns0:cell><ns0:cell cols='2'>illow forest 0.69 0.89</ns0:cell><ns0:cell>0.89</ns0:cell><ns0:cell>28 1.04</ns0:cell><ns0:cell>much more than 0.67 1.02</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>G CV</ns0:cell><ns0:cell>Police</ns0:cell><ns0:cell>7.65</ns0:cell><ns0:cell cols='3'>53.573194°14.581472°w along artificial 11.42 6.39 6.50</ns0:cell><ns0:cell>7.90</ns0:cell><ns0:cell>145 8.53</ns0:cell><ns0:cell>8.97</ns0:cell><ns0:cell>6.22</ns0:cell><ns0:cell>100 9.64</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell cols='7'>canal Trzebieradz 53.675417°14.441444°alder carr SA (mm 2 ) H 6.52 5.79 7.15 7.71 8.42 Min-max 4.72-2.43-5.10-5.83-6.15-</ns0:cell><ns0:cell>6.74 70 4.82-</ns0:cell><ns0:cell>30 9.26 6.74-</ns0:cell><ns0:cell>7.25 much more than 7.44 100 5.27-2.43-</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell>9.53</ns0:cell><ns0:cell>7.69</ns0:cell><ns0:cell>9.10</ns0:cell><ns0:cell>1.23</ns0:cell><ns0:cell>12.16</ns0:cell><ns0:cell>1.21</ns0:cell><ns0:cell>13.54</ns0:cell><ns0:cell>9.61</ns0:cell><ns0:cell>13.54</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>SD</ns0:cell><ns0:cell /><ns0:cell>1.14</ns0:cell><ns0:cell>1.51</ns0:cell><ns0:cell>1.21</ns0:cell><ns0:cell>1.07</ns0:cell><ns0:cell>1.48</ns0:cell><ns0:cell>1.27</ns0:cell><ns0:cell>1.63</ns0:cell><ns0:cell>1.08</ns0:cell><ns0:cell>1.62</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>CV</ns0:cell><ns0:cell /><ns0:cell>17.54</ns0:cell><ns0:cell>26.06</ns0:cell><ns0:cell>16.92</ns0:cell><ns0:cell>13.88</ns0:cell><ns0:cell>17.63</ns0:cell><ns0:cell>17.58</ns0:cell><ns0:cell>17.65</ns0:cell><ns0:cell>14.96</ns0:cell><ns0:cell>21.76</ns0:cell></ns0:row></ns0:table></ns0:figure>
</ns0:body>
" | "
Łódź, 13.07.2020
Dear Editors and Reviewers,
Encouraged by your letter, we would like to re-submit the manuscript entitled: Seed morphology and sculpture of invasive Impatiens capensis Meerb. from different habitats by Agnieszka Rewicz, Monika Myśliwy, Wojciech Adamowski, Marek Podlasiński, Anna Bomanowska
We are very much thankful to the Reviewers for their deep and thorough reviews. We have revised our paper according to their useful suggestions and comments. All specific comments are addressed in the revised version of the manuscript; we provided also version of manuscript with track changes mode. We hope that you will find this revised and improved version of the manuscript suitable for publishing in the PeerJ
Thank you once again for considering the publication of our manuscript in PeerJ.
Yours sincerely,
On behalf of all authors,
Agnieszka Rewicz
Reviewer #2 (Remarks to the Author)
Basic reporting
I read the revised article and I am agree to accept the manuscript in current form. This manuscript is clear to understand and easy for the readers. The manuscript is have enough data to understand the scientific method of the article.
Experimental design
The experimental design is enough to understand the article.
Validity of the findings
The findings are well written and understandable.
Comments for the Author
I would like to accept the manuscript in current form.
Thank you for your comments.
Reviewer #4 (Remarks to the Author)
Basic reporting
The manuscript aims to investigate the seed morphology and sculpture of invasive species Impatiens capensis collecting from different locations. It also aims to show the effect of environmental variables on the seed features of the species. The text is well written but lacks clarity and sharpness in the discussion conclusion.
Discussion and conclusions were rewrited.
Experimental design
The methodology and statistical analyses are care, and interesting data are provided for the single species studied.
Thank you for your comments.
Validity of the findings
I don’t know the editor’s and journal policy, but in my opinion, this is very simple with a very narrow scope for the readers of PeerJ. As per the literature cited in the manuscript, the seed morphology of Impatiens capensis and related species is already published, and in my knowledge environmental factors including soil, characteristics have little but obvious effects on the seed morphological traits. They found no variation in ultrastructure including shape and structure of rib. The variation in seed size and mass is superficial might depend on the collection time, maturity, condition of the mother plants, and some other factors. The comparative study including some other related species would have been interesting though.
In our opinion, basic research is very important and necessary for more detailed research. As we wrote in the introduction, research on the ecology and biology of I. capensis are known, but so far only four papers concerned about seeds morphology descriptions of this species. Available literature does not described such an important feature as the ultrastructure of this seed. We do not agree that our work is insignificant. The genus Impatiens currently has a description of the ultrastructure and in some cases seed measurements only for about 200 (c.a. 170) species out of 1000. Moreover, these works were based on material from single populations without other analyzes, e.g. habitats. In our work, we analyzed seeds from different habitats and covered the whole Polish range of this species. Our results bring information not only about the ultrastructure of the seed but about the stability of this structure and biometric traits of seeds. We probably did not manage to eliminate the influence of some factors on the size and weight of seeds, but we certainly eliminated the impact of collection time (seeds were collected in the same growing season, within thirteen days), degree of maturity (only ripe seeds were collected), and the state of mother plants (seeds were collected from undamaged individuals, of average height for a given population, always from the main stem). We agree that the differences in seeds are small but our results bring new data on the phenotypic plasticity of the species. We are also convinced that our results may be of interest to a wide audience, including taxonomists, ecologists, biological invasion researchers.
Thank you for your suggestions on expanding our research. The current article is the first in a series of articles on the taxonomic differences of three Impatiens spp..
Comments for the Author
Specific comments are presented below, to guide the authors in the revision.
Abstract:
The abstract lacks key results and conclusions of the study.
The abstract has been refined, the most important results have been added.
Line 26: Our research shows also two types of cells on the ultrastructure of seeds, between the ribs and on the ribs, not yet described in the literature. I think this is a normal feature in the seed sculpture.
Line 27: The paper gives new data about the seed morphology of I. capensis growing in different habitat conditions as well as the information about the area – the circuit that has not been reported before. For me, this is a very general statement.
The occurrence of cells of various shapes, numbers, arrangements, or sizes is very often a characteristic feature for species, genus, family. Different seed ultrastructure allows one to distinguish species in many plants. A lot of articles concern that it is a very important feature. The authors enclose two figures Fig 1 and Fig 2. they come from articles (New data on seed coat micromorphology of some Impatiens spp. from Northeast India Agnieszka Rewicz, Wojciech Adamowski, Souravjyoti Borah, Rajib Gogoi - currently in review). Please, see that the seed structure, cell shape are different (please see especially Fig 1 B, Fig 2 F).
Fig 1. Seed general view and coat micromorphology of analyzed balsams from Northeast India: A – B I. anjawensis, C – D I. decipiens, E – F I. drepanophora
Fig 2. Seed general view and coat micromorphology of analyzed balsams from Northeast India: A – B I. haridasanii, C – D I. jurpia, E – F I. nyimana
To describe the surface of the seed, a SEM image should be taken. However, to state unequivocally that this feature is permanent, i.e. taxonomically significant, i.e. 'good', we should test material from different habitats or locations. Not every species is subjected to such detailed research. Our research describes the surface of the seeds for the first time and also proven that this feature is a taxonomically significant feature within this species.
Same for line 29: The relationship between seed variability and various environmental factors was shown.
Changed.
Introduction:
The introduction is very long. It covers the unnecessary history of the genus and species should be specific on the problems and objectives of the study. There are lots of unnecessary citations as well.
Thank you very much for comment. We deleted the history of the species (line 95-102) and some citations from the Introduction section. The introduction has been shortened and the content focused on research problems undertaken in the work.
Materials and methods:
Line 149: anthropogenic? How about cultivated or human used land?
We considered for anthropogenic habitats this habitats which has been disturbed by human or which was created by human impact like an arable fields (after The Biology of Disturbed Habitats, Lawrence R. Walker, 2011).
Line 152-153: the difference between almost the entire day and the entire day is confusing? What are the criteria for the condition?
Thank you for this comment. This statement has been clarified.
Line: 167: how do you measure the thickness of the coating layer? Is it instrument-specific? I don’t think it is necessary to mention here.
Covering structures with gold before taking SEM images is standard practice. The thickness of this layer is important because for different structures and species it is different. The thickness of the gold layer should be selected depending on the ultrastructure tested, if the layer is too thick we can not see the tested surfaces. The layer thickness is automatically measured by the spraying device. We used (Leica EM ACE200). This information we added to the text (line 172).
Results: Tables 3 and 4 can be removed as the statistical data are presented in the text.
In our opinion, these tables are necessary for understanding the text, if deleted, the reader will not have a full picture of our analysis. Only some data are cited in the text, for the 2-3 most distinct populations.
Discussion:
Discussion lacks clarity and sharpness. In several parts, the writing is also confusing.
The discussion has been refined and the language has been improved by native speaker.
Line 304-306: the seed measurement is very much comparable. The difference margin is small so it is very difficult to say the larger seed than previous reports.
We agree that the differences in literature data and our results are small, but they prove that there are such differences, moreover, our research extends the range of phenotypic plasticity of the examined features. Each data on the value of the analyzed features is important in understanding the morphology of the species and its potential for phenotypic variation.
Line 320-327: Writing is very confusing and vague. The discussion and comparison with I. nolitangere are superficial. The author needs to specify how the surface sculptures of both species differ.
We added the information about I. noli-tangere ultrastructure (line 333) and rephrased this sentences.
Line 355-342: Not related to the discussion of this study. Please delete or you can adjust in the introduction part.
We deleted the part of this sentence.
Line 346-367: The author discussed about the phenotypic character like height which is I think is beyond the scope of this study. Authors need to discuss how the seed morphology of I. capensis varies in different populations and what is the most influential environmental factor?
The height of studied species stems is not quite beyond the scope of this study. As we explained in Materials and Methods section, the height of I. capensis specimens may influence seeds weight (Waller 1982). That is why the information of average height of studied population was provided in Table 1 and then used in discussion of obtained results, among other discussed factors.
Line 368-374: I think this part is not related to the result of this study or not linked properly with the result of this study.
We deleted this fragment.
Conclusion: The whole part is just a general statement. I think this is not the conclusion of this study.
Changed.
Reviewer #5 (Remarks to the Author)
Basic reporting
I appreciate the general structure of the paper. The subject is properly introduced and the background knowledge about the studied species is rich. However, I have several points to report.
In the abstract is explained the kind of data revealed by the study but I would expect that the authors explicitly detail what are the main results obtained in one or two sentences (regarding relation to environmental variables and seed morphology variations between populations). Moreover, I am surprised that the emphasis is put on the methodology used (SEM) but less on the questions addressed, and would suggest to better balance the abstract towards hypotheses and results.
The abstract has been refined, the most important results have been added.
In the manuscript, I am surprised to find very limited details about molecular differentiations between Impatiens species, although the taxonomy of this group of species is discussed.
I understand the motivations of the authors to focus on Impatiens capensis, due to its invasive nature, however I am frustrated that the study only report data about seed morphology restricted to this species without including measurement and comparison with closely related species at least, or populations from the native area to make more relevant the study.
My main concern is about the very limited geographical range within which the seeds have been collected. The prospected area cover only approximately 20 x 60 km, and I am concerned about the representativeness of the data (trait mean and variations) at a larger scale (country, and European range) for the species considered. To address the requirements of the journal, I would suggest including to the study seeds from more populations, either from different part of the country (to make more relevant the effect of environmental factors on seed traits) or from different ranges (to confirm your observations apply to native population in the same extend).
The current article is the first article in the series on which our team is working. We wanted to focus on Polish populations in the context of different habitats. Our research covered the entire Polish range of this species, so in fact it is country-scale work.
Currently, we have collected material from the natural range of I. capensis (by the first author who is currently staying in Canada). Material outside Poland is also collected. The next article will be about seeds variability between the natural and invasive range. Due to the fact that it is not always possible to collect soil, we decided to elaborate on this problem in detail in this article. In the research, we used all available populations within Poland. Our goal was to describe the ultrastructure and morphology of seeds and verify that it will be stable within different habitats.
In addition, we are working on a molecular-morphological problem within these three closely related species (I. capensis, I. pallida, I noli-tangere).
Experimental design
The statistical analyses performed are appropriated, well conducted and explained. Figures are appropriate and well detailed (maybe check Figure 4, Ordination diagram, it seems that the asterisks are missing on the statistically significant variables, and maybe verify the caption of this figure as I do not understand what the “Anthrop” refers to, no such variable in the diagram).
Thank you very much for this comment. By mistake we have included another version of the Figure 4, without the “Anthrop” (anthropogenic disturbances) variable. Now the correct version of the chart has been submitted and the caption carefully checked.
Indeed, as evolution of life history traits (reproductive characters) is being more and more supported as an evolutionary explanation for plant invasiveness, this aspect is important to consider, at least to discuss. Although the authors acknowledge that plant is naturalized for about 30 years in Poland, rapid evolution as been reported for several plants and can occur very quickly under the effect of climatic factors, and reproductive characters being central to invasiveness, I would expect this to be verified (Beheemen et al., 2019, doi: 10.1111/nph.15564; Molina-Montenegro et al., 2018, doi: 10.3389/fpls).
Thank you for this comment, we supplemented the discussion with evolutionary explanation for plant invasiveness
I am surprised to see that moisture content has been hand-felt assessed and wonder how precise this method is. This variable being a fundamental resource for plant development its variations may be important to consider.
In our work, we used a four-point scale of current soil moisture: dry soil (the soil crumbles and dust, it is neither cool nor moist to touch; after wetting it darkens visibly), fresh soil (the soil feels cool, but no moisture is felt; darkens after wetting), moist soil (the soil moistens fingers and tissue paper, but water does not leak when squeezed; clayey, loamy and some dusty soils are plastic; does not darken after wetting), and wet soil (water leaks from the soil when squeezed aggregates, soil smears). Such a description is widely used in field research, recommended, among others by Soil Science Society of Poland in the Fieldguide for soil description (2017) and sufficient for the assumed purposes of our work. It was used e.g. in the cited monograph by Myśliwy (2019). The relevant details were added to the Methods section.
Validity of the findings
The results regarding inter populations variations of seed traits with environmental factors is interesting and should be emphasized. Morphological characterization seems less relevant because no comparison with closely related species is directly made in the study. For someone intending to address taxonomical differentiation issues between Impatiens sp. it would require to perform proper comparisons (e.g. through morphometric measurements on the different species concerned).
The text of our article has been better balanced towards our hypotheses and results and inter-population variation of seed traits with environmental factors have been emphasized.
Comments for the Author
In my opinion, this study brings significant and relevant results although it may miss an hindsight with the inclusion of populations from different locations.
Thank you for your comment, we are currently working on comparing the three most similar species of Impatiens in terms of molecular and carpology.
Reviewer #6 (Remarks to the Author)
Basic reporting
Authors of the reviewed manuscript examined Impatiens capensis, an annual plant native to eastern North America which spreads across Europe and in Poland it is considered a locally invasive species. The species is an interesting and worth studying taxon, especially according to seeds which most probably play an important role in a rapid spread of that invasive species into secondary localities. Although the important differentiation in seed size and weight between examined populations was proved, the studies concerning seed biology and seedling recruitment in different environmental conditions were not undertaken. With such additional studies the work would be much more complete and would allow to evaluate the importance of seed size differentiation in spreading strategy of that invasive species. Anyway, the above mentioned problem might be an interesting aim of the future studies on I. capensis.
The actual aims of the reviewed work were as follows: the first aim was to examine seed morphology based on the standard biometric measurements; the next aim was the examination of the seed coat ultrastructure with the use of SEM; the third aim was the evaluation of the potential variability of seed morphology between individuals growing in different populations and in different habitat conditions.
Thank you very much for the comment. We currently have I. capensis material from natural (Canadian) and invasive range. Studies concerning seed biology and seedling recruitment will be the subject of another article.
Abstract:
It is not fully in agreement with the obtained results. The Authors stated in lines: 24-25 that “The current work presents a detailed description of the morphology of the seeds of I. capensis using SEM for the first time”. In my opinion such really complete description is not given in the manuscript– for the details concerning several lacks please see some of my comments given to the paragraphs Results and Discussion. Additionally, to make the Abstract more concise and clear I would suggest to unit all new and all the most important findings and underline together their taxonomic significance and/or the novelty to science.
Abstract was rewrited.
Introduction
The Introduction is to the point, however it is sometimes too detailed in my opinion, like e.g. the information concerning the botanical description of the stem and leaves which are not the subject of the study.
Thank you for this comments. We decided to delete this sentences (line 110 – 113)
Please make the text more concise and do not describe the obvious facts: e.g. lines:74-75; also devoid the repetitions: e.g. lines: 51-52 and 53; 130-132 and lines: 120-123 on the previous page; I also suggest to use proper descriptions for the 3-dimensional objects which the seeds are (instead of those suitable for the 2-dimensional ones): e.g. line 120: “seed are oval to lanceolate with …. “– instead I suggest the following description: seeds were narrowly ovate and tapering to a point at the apex; you may also use the description: seeds were laterally compressed, prolate spheroids, with four strong ribs.
Changed.
There are also some inconsistences in the text e.g. in line: 93 it is written that …” One of the morphologically undescribed species is Impatiens capensis” while later in line: 125-126 the text is as follows …” Numerous studies (over several hundred – see Adamowski 2016 onward and the literature cited therein) have been devoted to the morphology, ecology, biology and genetics of this species…” – maybe the Authors meant the whole genus Impatiens, not the examined species? But it is not clean and should be corrected.
Changed.
The aim of the study described in the last paragraph of the Introduction is awkward.
Changed.
The English language should be improved in my opinion; it should be checked by the native speaker.
The text was double checked by native speaker – once before sending the manuscript to the Editor, and the second time after review.
Results and Discussion:
I have listed below just few examples/observations (not all of them …) to be corrected:
- lines: 221-222: the sentences: “The seeds from the G (Police) population is characterized by the biggest seeds” and “This population characterized by the highest average values of the following traits:…” are not clear;.
Changed.
- Lines: 219-220: “In the B (Lubin) population was observed the shortest and narrowest seeds (respectively:…)” again the English language correction is necessary
Changed.
- The paragraph : The structure of seed surface must be rewritten according to the given below suggestions; in the present form it is not clear and the interesting results are not fully and properly described;
- The finding concerning the clear difference between seed morphology of the examined species and the closely related I. noli-tangere is important, but it would be also interesting to compare the observed type of the seed sculpture with the findings concerning other Impatiens species described by different authors (many of them were cited by the Authors, but I also recommend to see the additional following articles:
1. Three New Species of Impatiens L. from China and Vietnam: Preparation of Flowers and Morphology of Pollen and Seeds Author(s) :Yu-Min Shui, Steven Janssens, Su-Hua Huang, Wen-Hong Chen, and Zhi-Guo Yang Source: Systematic Botany, 36(2):428-439. 2011 and
2. 45 (5): 708–712(2007) doi:10.1360/aps06037 Acta Phytotaxonomica Sinica Impatiens macrovexilla var. yaoshanensis S. X. Yu, Y. L. Chen & H. N. Qin, a new variety of Balsaminaceae from Guangxi, China. 1,2YU Sheng-Xiang 1CHEN Yi-Lin 1QIN Hai-Ning*
Thank you for this comments. Added to discussion.
But to do such comparison it will be necessary to improve the present description of the seed coat micromorphology with several missing elements. Based on the observations of the structure and the microornamentation pattern of the epidermal cells of the testa, the following characters of the primary/secondary sculpture should be described or corrected: cellular/microornametation pattern should be defined; the cell outline was not clearly described; specific characters of the anticlinal walls should be described (degree of their elevation and thy type of their boundary) - by the way, from which literature comes the description anticyclic walls?; specific characters of the outer periclinal walls should be described (flat, convex, concave etc.); the secondary sculpture (if present) should also be described (e.g. cuticle striations, perhaps the presence of some protrusions etc ….).
Changed. Seed shape terminology and types of seed surfaces were adopted after Song et al 2005.
The present descriptions of SEM images (figure captions) are not clear enough, they should be more informative, and the arrows pointing particular details should be described.
Added.
Experimental design
It is an original primary research which fits to the scope of the journal.
However the methodology needs some additional information:
- How big were the examined populations (How many individuals were growing in them approx.? Were there only few, several or more individuals, e.g. not less than 50?)
The size of studied populations has been provided in Table 1.
- Did you do any preparations of the seeds before examining them in SEM; were they anyhow dried? Was the seed dehydration necessary? If yes, how was it done?
Part of seeds was air-dried before taking SEM pictures, and part (test was dehydrated). SEM photos showed no differences in the way the material was prepared. Thank you for the comment - this information has been added to materials and methods (line 214).
- From how many individuals in each population were the seeds collected?
The seeds for our study were collected from 8-10 plants from each population, which was described in the first part of Materials and Methods. We used eight seeds from each population for SEM analyzes. We added this information to methods (line 214)
- line 171: instead “pictures of seed obtained using SEM’ you may write: SEM images of seeds or SEM micrographs etc. …
Changed.
Validity of the findings
Concerning the variety of the examined plant material the performed studies are not too extensive, but the additional aspects concerning the relationships between the examined seed morphological characters and different environmental conditions gives the reviewed work an additional value.
The results are important and interesting. I suggest to complete and discuss the results of seed SEM analysis according to the given comments and to prepare the revised version of the manuscript. While preparing the description of the seed micromorphology and testa sculpture I would advice to follow the terminology of Barthlott W. (1981; 1984).
Changed.
Comments for the Author
I would like to suggest Authors some minor revision of the manuscript, and encourage them to take under consideration my comments.
Thank you very much for your comments, we followed all your suggestions.
Editor (Remarks to the Author)
I appreciate that you considered previous suggestions by reviewers. However, as indicated below several reviewers suggested a number of issues that have not been yet resolved. I agree with them, for instance the Abstract does not present clearly the objective of the paper. The decision to study pollen characters to understand an invasive plant is not well presented. On the other hand no hypothesis is stated. One of the reviewers expressed concerns on selection of populations and a strong justification is needed for considering only these populations in a small area. Also figures and captions need changes. Please read and examine every issue raised by all reviewers belos.
Dear Editor,
Thank you for sending reviewers suggestions. Our answers and comments are given below.
Here we would like to clarify some misunderstanding. Our article concerned the morphology and ultrastructure of seed of orange balsam, and we have not dealt with pollen of this species in any paragraph of our work. Seed and pollen are two different plant structures.
" | Here is a paper. Please give your review comments after reading it. |
9,802 | "<ns0:abstract xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:p>Impatiens capensis is an annual plant native to eastern North America that is currently spreading across Europe. In Poland, due to this plant's rapid spread in the secondary range and high competitiveness in relation to native species, it is considered a locally invasive species. The microstructure of seeds is an important tool for solving various taxonomic problems and also provides data useful for determining the impact of various environmental factors on the phenotypic variability of species. This issue is particularly important regarding invasive species which occupy a wide range of habitats in the invaded range. There are few reports on seed size and thus far no descriptions of the seed ultrastructure of I. capensis in the analyzed literature. We present new data on the seed morphology of I. capensis growing in different habitats and conditions in the secondary range of the species. The studied populations differed significantly in each of the investigated traits (seed length, width, circumference, area, roundness, and mass). Our findings showed that anthropogenic disturbances in habitats and some soil parameters (presence of carbonates, potassium, loose sand, and moisture) were statistically significant with various seed sizes and morphology in the studied populations of I. capensis. Moreover, our studies showed maximum seed length (5.74 mm) and width (3.21 mm) exceeding those values given in the available literature. For the first time, we also provide a detailed SEM study of the ultrastructure of the seed coat of I. capensis. There are two types of epidermal cells on the seeds: a) between the ribs (elongated with straight anticlinal walls, slightly concave outer periclinal walls, and micropapillate secondary sculpture on the edges with anticyclic walls), and b) on the ribs (isodiametric cells with straight anticlinal walls and concave outer periclinal walls). Unlike the variability of size and weight of seeds, the coat ornamentation has turned out to be a steady feature within the studied secondary range of I. capensis.</ns0:p></ns0:div>
</ns0:abstract>
<ns0:body xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:head>Introduction</ns0:head><ns0:p>The genus Impatiens is the most species-rich within the family Balsaminaceae, with ca. 1000 species distributed primarily in the Old World tropics and subtropics <ns0:ref type='bibr' target='#b37'>(Grey-Wilson 1980;</ns0:ref><ns0:ref type='bibr' target='#b114'>Yu et al. 2015)</ns0:ref>.</ns0:p><ns0:p>Impatiens has been a subject of numerous studies regarding distribution <ns0:ref type='bibr' target='#b118'>(Zhou et al. 2019)</ns0:ref>, ecology <ns0:ref type='bibr' target='#b1'>(Abrahamson, Hershey 1977;</ns0:ref><ns0:ref type='bibr' target='#b17'>Boyer et al. 2016)</ns0:ref>, physiology <ns0:ref type='bibr' target='#b66'>(Nanda, Kumar 1983;</ns0:ref><ns0:ref type='bibr' target='#b97'>Tooke et al. 2005)</ns0:ref>, biochemistry <ns0:ref type='bibr' target='#b89'>(Sreelakshmi et al. 2018)</ns0:ref>, biology <ns0:ref type='bibr' target='#b42'>(Jacquemart et al. 2015)</ns0:ref>, pollination <ns0:ref type='bibr' target='#b0'>(Abrahamczyk et al. 2017)</ns0:ref>, morphology <ns0:ref type='bibr' target='#b7'>(Akiyama, Ohba 2000;</ns0:ref><ns0:ref type='bibr' target='#b45'>Janssens et al. 2018)</ns0:ref>, systematics <ns0:ref type='bibr'>(Chen et al. 2007a, b;</ns0:ref><ns0:ref type='bibr' target='#b36'>Gogoi et al. 2018)</ns0:ref>, phylogeny and evolution <ns0:ref type='bibr' target='#b43'>(Janssens et al. 2007;</ns0:ref><ns0:ref type='bibr' target='#b75'>Ruchisansakun et al. 2015)</ns0:ref>, and other (see <ns0:ref type='bibr'>Adamowski 2016 onwards)</ns0:ref>. Despite the plethora of publications on various attributes of Impatiens, this genus requires further attention and research. Impatiens is taxonomically one of the most difficult groups to classify and remains a major challenge due to the enormous species richness and extraordinary morphological variability, with plants ranging from annuals growing only several centimeters high and bearing a single flower to subshrubs four meters high <ns0:ref type='bibr'>(Hooker 1904</ns0:ref><ns0:ref type='bibr'>(Hooker -1906;;</ns0:ref><ns0:ref type='bibr' target='#b37'>Grey-Wilson 1980;</ns0:ref><ns0:ref type='bibr' target='#b36'>Gogoi et al. 2018;</ns0:ref><ns0:ref type='bibr' target='#b76'>Ruchisansakun et al. 2018)</ns0:ref>.</ns0:p><ns0:p>The majority of balsam species grow in hardly accessible mountain ranges and have delicate flowers with complex morphology <ns0:ref type='bibr' target='#b15'>(Bhaskar 2012;</ns0:ref><ns0:ref type='bibr' target='#b111'>Yu 2012;</ns0:ref><ns0:ref type='bibr' target='#b71'>Rahelivololona et al. 2018)</ns0:ref>.</ns0:p><ns0:p>Herbarium specimens of balsams are difficult to prepare due to the succulent nature of the stems.</ns0:p><ns0:p>Specimens require special preparation such as floral dissection <ns0:ref type='bibr' target='#b84'>(Shui et al. 2011)</ns0:ref>; field notes are otherwise of limited value. Flower colors fade quickly and the position of the individual flower parts is often impossible determine from traditionally prepared specimens. One of the taxonomically important features within the genus Impatiens is related to the morphology of seeds. First information on the diversity of the seed coat of Impatiens was reported by <ns0:ref type='bibr' target='#b41'>Hooker and Thomson (1859)</ns0:ref> and <ns0:ref type='bibr' target='#b106'>Warburg and Reiche (1895)</ns0:ref>. Other works were concerned mostly with the shape and size of seeds rather than details of their surface ornamentation <ns0:ref type='bibr' target='#b81'>(Shimizu 1977)</ns0:ref>.</ns0:p><ns0:p>The development of new imaging methods enables the observation and study of ultrasmall-sized structures. Scanning electron microscopy (SEM) has allowed a detailed analysis of seed coat micromorphology of Impatiens seeds <ns0:ref type='bibr' target='#b88'>(Song et al. 2005;</ns0:ref><ns0:ref type='bibr' target='#b19'>Chen et al. 2007a;</ns0:ref><ns0:ref type='bibr' target='#b117'>Zhang et al. 2016)</ns0:ref>. Earlier works focused on seed dimensions were rarely devoted to the ultrastructure of seeds <ns0:ref type='bibr' target='#b83'>(Shimizu 1979;</ns0:ref><ns0:ref type='bibr' target='#b51'>Lu, Chen 1991)</ns0:ref>. The sculpture on seed coats offers a set of characters which can be used to identify a species, and in combination with other morphological data, can provide crucial evidence towards the taxonomy of a genus <ns0:ref type='bibr' target='#b51'>(Lu, Chen 1991;</ns0:ref><ns0:ref type='bibr' target='#b88'>Song et al. 2005;</ns0:ref><ns0:ref type='bibr' target='#b100'>Utami, Shimizu 2005;</ns0:ref><ns0:ref type='bibr' target='#b18'>Cai et al. 2013;</ns0:ref><ns0:ref type='bibr' target='#b114'>Yu et al. 2015)</ns0:ref>.</ns0:p><ns0:p>Seed morphological features of Impatiens have not only been used for solving various taxonomic problems within the genus but also prove to be useful for determining the impact of various environmental factors on the phenotypic variability of balsam species <ns0:ref type='bibr' target='#b9'>(Argyres, Schmitt 1991;</ns0:ref><ns0:ref type='bibr' target='#b79'>Schmitt 1993;</ns0:ref><ns0:ref type='bibr' target='#b52'>Maciejewska-Rutkowska, Janczak 2016)</ns0:ref>. The understanding of environmentally induced variation in an individual plant phenotype is crucial for predicting population responses to environmental changes. This issue is particularly important regarding invasive species which occupy a wide range of habitats in the invaded range <ns0:ref type='bibr' target='#b74'>(Richards et al. 2006)</ns0:ref>.</ns0:p><ns0:p>Despite an increasing number of publications on the surface of Impatiens seeds by SEM (e.g., <ns0:ref type='bibr' target='#b83'>Shimizu 1979;</ns0:ref><ns0:ref type='bibr' target='#b112'>Yu et al. 2007;</ns0:ref><ns0:ref type='bibr' target='#b84'>Shui et al. 2011;</ns0:ref><ns0:ref type='bibr'>Xia et al. 2019 a.o.)</ns0:ref>, there is still a lack of information on the seed micromorphology of the majority of species. In fact, a detailed understanding of the seed morphology of the entire genus Impatiens is missing, despite major studies using novel imaging methods (e.g., <ns0:ref type='bibr' target='#b115'>Yuan et al. 2004;</ns0:ref><ns0:ref type='bibr' target='#b75'>Ruchisansakun et al. 2015;</ns0:ref><ns0:ref type='bibr' target='#b71'>Rahelivololona et al. 2018)</ns0:ref>. As yet, only about 170 species have been investigated, which is about one fifth of all known balsams (Maciejewska-Rutkowska, Janczak 2016).</ns0:p><ns0:p>One of the species with morphologically undescribed seeds is Impatiens capensis (jewelweed, orange balsam), an annual plant native to eastern North America <ns0:ref type='bibr' target='#b57'>(Meusel et al. 1978)</ns0:ref>, which is currently spreading across Europe. Today I. capensis is considered as naturalized in several European countries <ns0:ref type='bibr' target='#b55'>(Matthews et al. 2015)</ns0:ref>, including Poland, where the species is locally established and invasive due to its rapid spread in the secondary range and high competitiveness in relation to native species, even perennials <ns0:ref type='bibr' target='#b96'>(Tokarska-Guzik et al. 2012)</ns0:ref>. In Poland, it was found for the first time in 1987 <ns0:ref type='bibr' target='#b67'>(Pawlaczyk, Adamowski 1991)</ns0:ref>, and it is currently spreading in the Western Pomerania region <ns0:ref type='bibr' target='#b68'>(Popiela et al. 2015;</ns0:ref><ns0:ref type='bibr'>M Myśliwy, 2017, personal observations)</ns0:ref>. The species occurs in the area of the Szczecin Lagoon and enters alder forests, willow shrubs, rushes and riparian tall herb fringe communities <ns0:ref type='bibr' target='#b67'>(Pawlaczyk, Adamowski 1991;</ns0:ref><ns0:ref type='bibr' target='#b64'>Myśliwy et al. 2009;</ns0:ref><ns0:ref type='bibr'>M Myśliwy, 2014, personal observations)</ns0:ref>. It also appears in moist anthropogenic habitats, e.g., along roadside ditches (M Myśliwy, 2017, personal observations).</ns0:p><ns0:p>Impatiens capensis is an annual plant growing from 0.5-1.5 m or more in height. The flowers are 2.5-3.0 cm long and orange with darker patches in the most common f. capensis. The lower sepal forms a light-orange nectar spur, 5-9 mm long, which is bent at 180° to lie parallel to the sepal-sac <ns0:ref type='bibr' target='#b119'>(Zika 2006)</ns0:ref>. Besides color, it differs from the predominantly Eurasiatic I. nolitangere in that the lower sepal is more rapidly constricted into the spur and the position of the spur <ns0:ref type='bibr' target='#b120'>(Zika 2009)</ns0:ref>. The fruit is a five-valved capsule, 2.0-2.5 cm long and 0.3-0.5 cm wide, with explosive dehiscence ejecting the seeds <ns0:ref type='bibr' target='#b62'>(Moore 1968;</ns0:ref><ns0:ref type='bibr' target='#b36'>Gleason, Cronquist 1991;</ns0:ref><ns0:ref type='bibr' target='#b26'>Day et al. 2012)</ns0:ref>.</ns0:p><ns0:p>The seeds are laterally compressed, prolate spheroid, with four strong ribs of 5-5.6 × 2.7-3.1 mm <ns0:ref type='bibr' target='#b16'>(Bojňanský, Fargašová 2007)</ns0:ref>. The weight ranges from 6.4 to 26.9 mg <ns0:ref type='bibr' target='#b86'>(Simpson et al. 1985)</ns0:ref>. <ns0:ref type='bibr' target='#b78'>Schemske (1978)</ns0:ref> recorded 11.5 mg for cleistogamous seeds and 13.3 mg for chasmogamous ones, and <ns0:ref type='bibr' target='#b104'>Waller (1982)</ns0:ref> 10.6 mg. The seed surface is wrinkled or rough, lusterless, dark-brown, with some roundish and paler spots <ns0:ref type='bibr' target='#b16'>(Bojňanský, Fargašová 2007)</ns0:ref>.</ns0:p><ns0:p>Numerous studies (several hundred; see Adamowski 2016 onward and the literature cited therein) have been devoted to the ecology, biology, and genetics of this species (e.g., <ns0:ref type='bibr' target='#b8'>Antlfinger 1989;</ns0:ref><ns0:ref type='bibr' target='#b80'>Schmitt et al. 1985;</ns0:ref><ns0:ref type='bibr' target='#b28'>Donohue, Schmitt 1999;</ns0:ref><ns0:ref type='bibr' target='#b29'>Donohue et al. 2000;</ns0:ref><ns0:ref type='bibr' target='#b120'>Zika 2009;</ns0:ref><ns0:ref type='bibr' target='#b92'>Tabak, von Wettberg 2008;</ns0:ref><ns0:ref type='bibr' target='#b26'>Day et al. 2012</ns0:ref>). However, a review of the available literature showed a scarcity of data on seed size and a complete lack of information describing the morphological variation of the seed coat of I. capensis <ns0:ref type='bibr' target='#b78'>(Schemske 1978;</ns0:ref><ns0:ref type='bibr' target='#b104'>Waller 1982;</ns0:ref><ns0:ref type='bibr' target='#b86'>Simpson et al. 1985;</ns0:ref><ns0:ref type='bibr' target='#b16'>Bojňanský, Fargašová 2007)</ns0:ref>.</ns0:p><ns0:p>The aim of our work has been to characterize the micromorphological traits and ultrastructure of I. capensis seeds from various habitats and growing conditions and their morphological variability. Anthropogenic changes in habitats were expected as important factors affecting seed micromorphology and ultrastructure.</ns0:p></ns0:div>
<ns0:div><ns0:head>Materials and methods</ns0:head></ns0:div>
<ns0:div><ns0:head>Study sites</ns0:head><ns0:p>Seeds were collected from August to September 2018 (to avoid seasonal variability) from eight populations of I. capensis in Poland. We sampled the entire Polish range of this species from all types of habitats, from natural (alder carrs, hydrophilous tall herb communities along rivers, near water seepages, and along the banks of the Szczecin Lagoon) to anthropogenic (tall herb communities along roadside ditches, transformed forests along artificial canals) (Table <ns0:ref type='table' target='#tab_1'>1</ns0:ref>, Fig. <ns0:ref type='figure'>1</ns0:ref>). The studied populations were also subject to different lighting conditions, which were scored using a 3-point scale: plants which grew in willow forests and the understory of alder carrs were strongly shaded (3), while those from tall herb communities were partly shaded by solitary trees (2) or exposed to full sun (1). As the height of I. capensis specimens, the location of capsules within the plant (main stem vs. branches), and their derivation from flowers of various types (cleistogamous vs. chasmogamous) may affect seeds weight <ns0:ref type='bibr' target='#b104'>(Waller 1982)</ns0:ref>, the seeds for our study were collected always from the main stems of 8-10 plants of similar (average for the population) height and from capsules derived from chasmogamous flowers, to minimize the bias. Species nomenclature was adopted from Euro+Med PlantBase (Euro+Med PlantBase).</ns0:p></ns0:div>
<ns0:div><ns0:head>Biometric and SEM analysis</ns0:head><ns0:p>From 24 to 30 mature seeds were used from each population for biometric analysis. We measured four quantified seed traits: seed length (SL), seed width (SW), seed circumference (SC), and seed area (SA). The seeds were measured as previously described in <ns0:ref type='bibr' target='#b72'>Rewicz et al. (2017)</ns0:ref>. In order to describe the seed mass, we used 15 seeds from each population. The seeds were weighed with an Ohaus PA 21. We determined roundness by the formula: R = 4 × area/π [Major axis]^2 as defined by <ns0:ref type='bibr' target='#b31'>Ferreira and Wayne (2010)</ns0:ref>.</ns0:p><ns0:p>We used eight seeds from each population for SEM. The seeds were air-dried and sputtercoated with a 4-nm-thick layer of gold (Leica EM ACE200). The SEM work was performed on a Phenom Pro X Scanning Electron Microscope at the Department of Invertebrate Zoology and Hydrobiology, University of Lodz, Poland. The 3D models of the seed surface were generated using the dedicated software 3D Roughness Reconstruction for Phenom. SEM micrographs were analyzed as previously described in <ns0:ref type='bibr' target='#b72'>Rewicz et al. (2017)</ns0:ref>. Seed shape terminology and types of seed surfaces were adopted from <ns0:ref type='bibr' target='#b11'>Barthlott (1981)</ns0:ref>.</ns0:p></ns0:div>
<ns0:div><ns0:head>Soil properties</ns0:head><ns0:p>In order to characterize habitat conditions at each locality, five soil core samples (0-20 cm depth) were collected and then mixed together into one composite sample. Soil samples were dried at room temperature, and passed through a sieve to remove fractions larger than 1 mm. The following physicochemical soil parameters were determined <ns0:ref type='bibr' target='#b12'>(Bednarek et al. 2011)</ns0:ref>, as first described by <ns0:ref type='bibr' target='#b63'>Myśliwy (2019)</ns0:ref>: organic matter content defined as the loss on ignition (LOI) -soil samples annealed at 550ºC (%); grain composition (the content of sand, silt, clay) -Bouyoucos's sedimentation method, modified by Casagrande and Prószyński; granulometric categories according to the PSSS (2009) classification; soil reaction (pH) -the potentiometric method, in 1-M solution of KCl; soil calcium carbonate (CaCO 3 ) content (%) -the Scheibler's method; organic carbon (C org ) content (%), and total nitrogen (N tot ) content (%) were determined by an elemental analyzer CHNS/O FlashSmart (Thermo Scientific), and the C/N ratio; the content of available forms of soil nutrients (mg/100 g soil): calcium (Ca) and sodium (Na) were determined spectrophotometrically (Ca -AAS and Na -EAS) on ICE3000; potassium (K) and phosphorus (P) -measured according to the Egner-Riehm method; magnesium (Mg) -measured by Schachtschabel's method; soil moisture content, hand-felt assessed directly in the field using a 4point scale recommended by the Soil Science Society of Poland ( <ns0:ref type='formula'>2017</ns0:ref>): (1) dry (the soil crumbles and turns to dust, it is neither cool nor moist to touch; it darkens visibly after wetting),</ns0:p><ns0:p>(2) fresh (the soil feels cool, but no moisture is felt; darkens after wetting), (3) moist (the soil moistens fingers and tissue paper, but water does not leak when squeezed; clayey, loamy, and some dusty soils are plastic; does not darken after wetting), ( <ns0:ref type='formula'>4</ns0:ref>) wet (water leaks from the soil when squeezed, aggregates, soil smears). <ns0:ref type='table' target='#tab_1'>2020:02:45912:3:0:NEW 9 Sep 2020)</ns0:ref> Manuscript to be reviewed</ns0:p></ns0:div>
<ns0:div><ns0:head>PeerJ reviewing PDF | (</ns0:head></ns0:div>
<ns0:div><ns0:head>Data analysis</ns0:head><ns0:p>The five following basic characteristic traits were calculated: arithmetic average (x), minimum and maximum values (min/max), coefficient of variation (CV), and standard deviation (SD). The distribution of the data was not normal; statistical analysis was based on the Kruskal-Wallis test (for p ≤0.05), which is a nonparametric alternative to ANOVA <ns0:ref type='bibr' target='#b116'>(Zar 1984)</ns0:ref>.</ns0:p><ns0:p>Correlation between pairs of morphological characters was evaluated using Spearman's correlation coefficient and the values were adopted after Meissner ( <ns0:ref type='formula'>2010</ns0:ref>), (correlation: less than 0.20 -very poor; 0.21-0.39 -weak; 0.40-0.69 -moderate; 0.70-0.89 -strong; and above 0.89very strong).</ns0:p><ns0:p>The cluster analysis based on the nearest neighbor method was performed using the matrix on the population's mean values. As the dataset required a linear response model <ns0:ref type='bibr' target='#b48'>(Jongman et al. 1995)</ns0:ref>, the Redundancy Analysis (RDA) was used to relate the variability of morphological traits of seeds to environmental variables. The variables C org and N tot were excluded from the RDA as they were strongly correlated with organic matter content (LOI). The </ns0:p></ns0:div>
<ns0:div><ns0:head>Results</ns0:head></ns0:div>
<ns0:div><ns0:head>Biometric analysis</ns0:head><ns0:p>Seeds from the G (Police) population were the largest, with average values of length (SL) 4.60 mm, width (SW) 2.71 mm, circumference (SC) 11.65 mm, and area (SA) 9.26 mm 2 ; comparatively large seeds were also obtained from the E population (Święta); the B (Lubin) population had the shortest (mean SL 3.88 mm) and narrowest seeds (mean SW 2.03 mm) (Table <ns0:ref type='table' target='#tab_0'>2</ns0:ref>).</ns0:p><ns0:p>The minimum values of analyzed traits were also recorded in the B (Lubin) population (SL 3.16 mm, SW 1.12 mm, SC 7.27 mm, SA 2.43 mm 2 ). The maximum values of length (5.74 mm), circumference (14.59 mm), and area (13.54 mm 2 ) were recorded in the G population (Police). Manuscript to be reviewed A very strong Spearman correlation (r = 0.94) was observed between the seed area and circumference (Table <ns0:ref type='table'>3</ns0:ref>). The most variable features were the seed area (CV = 21.76%) and width (CV = 15.35%). The variation of seed traits ranged insignificantly from 6.19% (H population) to 10.28% (B) for SL; from 7.30% (D) to 20.57% (B) for SW; from 6.22% (H) to 11.42% (B) for SC; and from 13.88% (D) to 26.06% (B) for SA, respectively.</ns0:p><ns0:p>The G (Police: 11.42 mg) and E (Święta: 9.82 mg) populations are characterized by the heaviest seeds. The lightest seeds were observed in the following populations: B (Lubin: 6.52 mg) and H (Trzebieradz: 6.92 mg) (Fig. <ns0:ref type='figure' target='#fig_5'>2</ns0:ref>, Table <ns0:ref type='table' target='#tab_0'>2</ns0:ref>).</ns0:p><ns0:p>The Kruskal-Wallis test showed that the I. capensis populations differed significantly in each of the analyzed traits. The conducted post hoc test (DunnTest) showed that the populations from: Police (G), followed by Czarnocin (D), Święta (E), and Trzebieradz (H) showed the greatest variation in terms of studied traits among all the populations (Table <ns0:ref type='table'>4</ns0:ref>).</ns0:p><ns0:p>The similarity analysis using Euclidean's distances showed two main clusters (Fig. <ns0:ref type='figure' target='#fig_6'>3</ns0:ref>).</ns0:p><ns0:p>The first cluster included six populations of I. capensis (A-D, F, H), all derived from natural habitats, while the other cluster groups two populations (E, G) from anthropogenic habitats, where the examined plants were the highest (Table <ns0:ref type='table' target='#tab_1'>1</ns0:ref>). According to the dendrogram (Fig. <ns0:ref type='figure' target='#fig_6'>3</ns0:ref>), populations C and F are the closest to each other; both were associated with river valleys (Dziwna and Oder rivers, respectively) and close to the river bed, hence under the influence of flooding. The D and A populations were growing in tall herb communities on the banks of the Szczecin Lagoon. The most distinct populations in the first cluster (H and B) were also found on the banks of the Szczecin Lagoon, but they had the lowest average height and differed in habitat conditions from the other populations of this cluster (Table <ns0:ref type='table' target='#tab_1'>1</ns0:ref>).</ns0:p><ns0:p>The investigated morphological parameter of seed shape, roundness, showed statistically significant differences between the populations (p < 0.05). For roundness, the highest value was recorded at D: Czarnocin (0.58) (tall herbs on the bank of the Szczecin Lagoon) and the lowest at H: Trzebieradz (0.47) (alder carr) (Table <ns0:ref type='table'>5</ns0:ref>).</ns0:p></ns0:div>
<ns0:div><ns0:head>Biometric variability of seeds and its relationship with environmental conditions</ns0:head><ns0:p>All environmental variables included in the RDA accounted for 35.6% of the total variation. The results of stepwise forward selection of variables indicated that five variables: anthropogenic disturbances (Anthrop), carbonates (CaCO 3 ), loose sand presence (LoSa), potassium (K), and soil moisture content (Moist) were statistically significant and varied between the studied populations of I. capensis (Fig. <ns0:ref type='figure' target='#fig_7'>4</ns0:ref>). Along the gradient represented by Axis I, the highest correlation between the sample position and environmental variables (the so-called interset correlation) was typical of anthropogenic disturbances and CaCO 3 , followed by the degree of shading and soil Ca, while the C/N ratio was most closely correlated with Axis II, followed by soil content of P, Na, K, and organic soil.</ns0:p><ns0:p>The location of population H (Trzebieradz) in the ordination space (the upper part of the RDA diagram) was associated with the highest C/N ratio, the highest soil moisture and shading, as well as with the lowest soil pH and the lowest soil content of CaCO 3 , Ca, P, and K (Fig. <ns0:ref type='figure' target='#fig_7'>4</ns0:ref>, Table <ns0:ref type='table'>6</ns0:ref>). At the same time, the H population was dominated by short specimens (Table <ns0:ref type='table' target='#tab_1'>1</ns0:ref>), with one of the lightest seeds and average values of biometric traits (Table <ns0:ref type='table' target='#tab_0'>2</ns0:ref>). In contrast, populations characterized by the longest, widest, and heaviest seeds, from E (Święta) and G (Police), located in the right-hand side of the RDA diagram, were also related to a relatively high C/N ratio, but unlike the previous population, they were associated with a low level of soil moisture as well as the highest anthropogenic disturbances (Fig. <ns0:ref type='figure' target='#fig_7'>4</ns0:ref>), and consisted of the tallest specimens (Table <ns0:ref type='table' target='#tab_1'>1</ns0:ref>).</ns0:p><ns0:p>Population D (Czarnocin) occupied the bottom part of the diagram and was distinct in its organic soil, with the highest content of organic matter (LOI), as well as P, K, Mg, and Na content in the soil, while having the lowest C/N ratio (Table <ns0:ref type='table'>6</ns0:ref>). The lowest values of the seed biometric traits were found for population B (Lubin), located in the left-hand part of the RDA diagram (Fig. <ns0:ref type='figure' target='#fig_7'>4</ns0:ref>, Table <ns0:ref type='table' target='#tab_0'>2</ns0:ref>), and associated with high soil pH and the highest content of soil carbonates and calcium, as well as a low level of soil moisture (Table <ns0:ref type='table'>6</ns0:ref>). The other populations (A: Podgrodzie; C: Unin; F: Szczecin-Zdroje) were also on the left side of the diagram, but closer to the center (Fig. <ns0:ref type='figure' target='#fig_7'>4</ns0:ref>). Neither their seed biometric traits nor habitat conditions were distinct (Tables <ns0:ref type='table' target='#tab_0'>2, 6</ns0:ref>).</ns0:p></ns0:div>
<ns0:div><ns0:head>Seed surface ultrastructure</ns0:head><ns0:p>The studied seeds of I. capensis were round in shape, with a lusterless, rough, and darkbrown surface, without roundish and paler spots (Fig. <ns0:ref type='figure'>5</ns0:ref>). The seeds had four strong, clear ribs, the apex and bottom narrowed. Each rib was built of rows of 4-5 cells and had a darker color than the surface between them (Fig. <ns0:ref type='figure'>5 H</ns0:ref>). The seed coat is composed of two types of epidermal cells (Fig. <ns0:ref type='figure'>5 E, H</ns0:ref>) creating a net-like pattern. The cells of the seed surface between the ribs were: elongated with straight anticlinal walls (Fig. <ns0:ref type='figure'>5 E</ns0:ref>), raised cell boundaries between the cells (Fig. <ns0:ref type='figure'>5</ns0:ref> G), slightly concave outer periclinal walls (Fig. <ns0:ref type='figure'>5 F, G</ns0:ref>) and a micropapillate secondary sculpture on the edges of anticyclic walls (Fig. <ns0:ref type='figure'>5 F</ns0:ref>). Near the ribs, there were rows of 4-7 isodiametric cells (Fig. <ns0:ref type='figure'>5 I, K</ns0:ref>) with straight anticlinal walls (Fig. <ns0:ref type='figure'>5</ns0:ref> L), with raised cell boundaries (Fig. <ns0:ref type='figure'>5 M</ns0:ref>) and concave outer periclinal walls. Seeds from all studied populations did not differ in their ultrastructure (Fig. <ns0:ref type='figure'>5 A -D</ns0:ref>).</ns0:p></ns0:div>
<ns0:div><ns0:head>Discussion</ns0:head><ns0:p>SEM gives us the means for studying the morphological characters of seeds and their ultrastructural characteristics which helps or identifying and determining the taxonomic delimitation of various angiosperm groups, as demonstrated for Brasssicaceae <ns0:ref type='bibr' target='#b93'>(Tantaway et al. 2004</ns0:ref>), Caryophyllaceae <ns0:ref type='bibr' target='#b98'>(Ullah et al. 2019a</ns0:ref><ns0:ref type='bibr' target='#b99'>(Ullah et al. , 2019b))</ns0:ref>, Poaceae (Martín-Gómez et al. 2019b), Cyperaceae <ns0:ref type='bibr' target='#b108'>(Więcław et al. 2017)</ns0:ref>, Ranunculaceae <ns0:ref type='bibr' target='#b25'>(Constantinidis et al. 2001;</ns0:ref><ns0:ref type='bibr' target='#b72'>Rewicz et al. 2017;</ns0:ref><ns0:ref type='bibr' target='#b53'>Martín-Gómez et al. 2019a;</ns0:ref><ns0:ref type='bibr' target='#b39'>Hadidchi et al. 2020)</ns0:ref>, Rosaceae (Ballian, Mujagić-Pašić 2013), Onagraceae <ns0:ref type='bibr' target='#b5'>(Akbari, Azizian 2006)</ns0:ref>, and Orchidaceae <ns0:ref type='bibr' target='#b32'>(Gamarra et al. 2007</ns0:ref><ns0:ref type='bibr' target='#b33'>(Gamarra et al. , 2010;;</ns0:ref><ns0:ref type='bibr' target='#b73'>Rewicz et al. 2016)</ns0:ref>. Although seed morphology alone does not provide universally applicable key characters for species identification, it can be as helpful as many other characters used in taxonomy.</ns0:p><ns0:p>Members of Balsaminaceae have a diverse and elaborately sculptured seed coat.</ns0:p><ns0:p>Unfortunately, till now seed morphology has been observed only for a small number of Impatiens species, which has limited the use of the morphological traits of seeds in taxonomy and classification (e.g. <ns0:ref type='bibr' target='#b88'>Song et al. 2005;</ns0:ref><ns0:ref type='bibr' target='#b100'>Utami, Shimizu 2005;</ns0:ref><ns0:ref type='bibr' target='#b19'>Chen et al. 2007a;</ns0:ref><ns0:ref type='bibr' target='#b112'>Yu et al. 2007;</ns0:ref><ns0:ref type='bibr' target='#b47'>Jin et al. 2008;</ns0:ref><ns0:ref type='bibr' target='#b84'>Shui et al. 2011)</ns0:ref>.</ns0:p><ns0:p>We here provide new data about the seed morphology and seed coat sculpture of I. capensis, as well as new information about the area of its distribution in Poland. Our studies have shown maximum seed length (5.74 mm) and width (3.21 mm) beyond the values reported elsewhere. <ns0:ref type='bibr' target='#b16'>Bojňanský and Fargašová (2007)</ns0:ref> found seeds of I. capensis to be 5-5.6 mm long and 2.7-3.1 mm wide. Our ultrastructural studies have shown two types of cells on between the ribs and on the ribs, that have previously not been described (Fig. <ns0:ref type='figure'>5</ns0:ref>). The occurrence of several types of epidermal cells on the seeds of members of Impatiens was previously noted, for instance, three types of epidermal cells have been reported in Impatiens aconitoides <ns0:ref type='bibr' target='#b84'>Shui et al. (2011)</ns0:ref>. We were not able to confirm in any of our studied populations the presence of roundish spots on the Manuscript to be reviewed surface of seeds (Fig. <ns0:ref type='figure' target='#fig_8'>6</ns0:ref>) as reported by <ns0:ref type='bibr' target='#b16'>Bojňanský and Fargašová (2007)</ns0:ref>, which may be due to the different geographical origin of the examined seeds: our seeds of I. capensis are from wildgrowing populations from various habitats, while those studied by <ns0:ref type='bibr' target='#b16'>Bojňanský and Fargašová (2007)</ns0:ref> were obtained from cultivation and of unknown origin.</ns0:p><ns0:p>The analysis of SEM micrographs of I. noli-tangere seeds closely related to I. capensis <ns0:ref type='bibr' target='#b114'>(Yu et al. 2015)</ns0:ref> has shown that seed coats of this species vary significantly depending on the geographical origin of the seeds <ns0:ref type='bibr' target='#b100'>(Utami, Shimizu 2005;</ns0:ref><ns0:ref type='bibr'>Chen et al. 2007 a;</ns0:ref><ns0:ref type='bibr' target='#b47'>Jin et al. 2008</ns0:ref>). On the other hand, the comparison of the seed micromorphology of I. capensis has not shown similarity to seed coat ornamentation of the aforementioned I. noli-tangere (with narrow and ellipsoid seeds, fine reticulate subtype, testa cells with reticulate thickened outer walls; <ns0:ref type='bibr' target='#b88'>Song et al. 2005;</ns0:ref><ns0:ref type='bibr' target='#b100'>Utami, Shimizu 2005;</ns0:ref><ns0:ref type='bibr' target='#b19'>Chen et al. 2007a;</ns0:ref><ns0:ref type='bibr' target='#b47'>Jin et al. 2008)</ns0:ref>. Despite the fact that both species are closely related and may be confused <ns0:ref type='bibr' target='#b120'>(Zika 2009;</ns0:ref><ns0:ref type='bibr' target='#b114'>Yu et al. 2015)</ns0:ref>, their seeds clearly differ morphologically. The new data presented here may be useful in the identification of these species. In turn, there is no information about the seed morphology of I. pallida, which is sympatric and synchronic species to I. capensis <ns0:ref type='bibr' target='#b77'>(Rust 1977)</ns0:ref>, which makes this subject even more difficult. Elucidating the overall variation in seed coat micromorphology and to implement this feature to taxonomy of I. capensis will require more samplings, also within the native range of orange jewelweed as well as other closely related species and this eventually should become the basis for further comparisons and studies. Seed ultrastructure appears to be a constant feature within a taxonomic unit <ns0:ref type='bibr' target='#b90'>(Stace 1992)</ns0:ref> and, as morphological studies show, seed shape and size are highly diverse at the genus and species levels <ns0:ref type='bibr' target='#b112'>(Yu et al. 2007;</ns0:ref><ns0:ref type='bibr' target='#b47'>Jin et al. 2008;</ns0:ref><ns0:ref type='bibr' target='#b84'>Shui et al. 2011;</ns0:ref><ns0:ref type='bibr' target='#b98'>Ullah et al. 2019a</ns0:ref><ns0:ref type='bibr' target='#b99'>Ullah et al. , 2019b;;</ns0:ref><ns0:ref type='bibr' target='#b39'>Hadidchi et al. 2020</ns0:ref>). Both statements have been proven for I. capensis in Poland.</ns0:p><ns0:p>Data concerning the size, shape and structure of seeds not only have been used as an important tool for solving various taxonomic problems within the genus Impatiens but also provide results useful for determining the impact of various environmental factors on the phenotypic variability of these species <ns0:ref type='bibr' target='#b13'>(Bell et al. 1991;</ns0:ref><ns0:ref type='bibr' target='#b9'>Argyres, Schmitt 1991;</ns0:ref><ns0:ref type='bibr' target='#b79'>Schmitt 1993;</ns0:ref><ns0:ref type='bibr' target='#b22'>Chmura et al. 2013;</ns0:ref><ns0:ref type='bibr' target='#b52'>Maciejewska-Rutkowska, Janczak 2016)</ns0:ref>.</ns0:p><ns0:p>Environmental heterogeneity is indicated as a major factor driving morphological changes <ns0:ref type='bibr' target='#b65'>(Nakazato et al. 2008)</ns0:ref>. Seeds are sensitive to changes in biotic and abiotic conditions <ns0:ref type='bibr' target='#b58'>(Moles et al. 2005)</ns0:ref>. According to <ns0:ref type='bibr' target='#b85'>Silvertown (1989)</ns0:ref>, the correlation between seed size and the PeerJ reviewing PDF | (2020:02:45912:3:0:NEW 9 Sep 2020)</ns0:p><ns0:p>Manuscript to be reviewed place where plant is growing is an adaptative feature. Bigger seeds occur in habitats with stable environmental conditions, where seedlings may grow slowly. Small seeds are generally produced by plants with a short life cycle, growing mainly in disturbed habitats.</ns0:p><ns0:p>Orange balsam is known for colonizing a wide range of habitats <ns0:ref type='bibr' target='#b78'>(Schemske 1978;</ns0:ref><ns0:ref type='bibr' target='#b103'>Waller 1980)</ns0:ref>. Moreover, <ns0:ref type='bibr' target='#b86'>Simpson et al. (1985)</ns0:ref> have shown that I. capensis vegetative and reproductive growth parameters reflect habitat differences. Light availability <ns0:ref type='bibr' target='#b86'>(Simpson et al. 1985)</ns0:ref> as well as soil moisture and pH <ns0:ref type='bibr' target='#b103'>(Waller 1980</ns0:ref>) have been reported to affect its growth patterns. Our studies indicate that five environmental variables were statistically significant and were able to serve to discern the studied populations in terms of seed size and weight: anthropogenic disturbances (which may serve as a proxy for habitat fertility), carbonates (CaCO 3 ), loose sand presence, potassium (K), and soil moisture (Fig. <ns0:ref type='figure' target='#fig_7'>4</ns0:ref>). Populations G (Police) and E (Święta), occurring in the most disturbed anthropogenic habitats (artificial canal and roadside), have the heaviest seeds as a result of growth under favorable environmental conditions (neutral or slightly acidic soil with a relatively high C/N ratio). In turn, population B (Lubin) with the smallest and lightest seeds was associated with high soil pH, and the highest content of soil carbonates and calcium.</ns0:p><ns0:p>Interestingly, <ns0:ref type='bibr' target='#b104'>Waller (1982)</ns0:ref> reported that the higher nodes of I. capensis individuals tended to produce heavier seeds. In <ns0:ref type='bibr' target='#b104'>Waller's (1982)</ns0:ref> opinion, the position effect probably leads to a greater mean seed size for higher plants. <ns0:ref type='bibr' target='#b107'>Werner and Platt (1976)</ns0:ref> stated that populations growing at higher plant densities often produce larger seeds. Our results are consistent with both studies, as the largest and heaviest seeds were obtained from populations G (Police) and E (Święta), formed by the highest plants, growing in large numbers and densities.</ns0:p><ns0:p>Another important factor shaping a diverse array of plant traits, including morphological features, is climate <ns0:ref type='bibr' target='#b65'>(Nakazato et al. 2008;</ns0:ref><ns0:ref type='bibr' target='#b24'>Colautti, Barrett 2013;</ns0:ref><ns0:ref type='bibr' target='#b101'>van Boheemen et al. 2019)</ns0:ref>.</ns0:p><ns0:p>Temperature and precipitation gradients are the main climatic factors driving the adaptive diversification of species <ns0:ref type='bibr' target='#b65'>(Nakazato et al. 2008)</ns0:ref>. As it seems, climatic conditions have had a limited effect on the investigated seed parameters till now, due to a small area of secondary distribution of I. capensis in Poland <ns0:ref type='bibr' target='#b3'>(Adamowski et al. 2018;</ns0:ref><ns0:ref type='bibr' /> Manuscript to be reviewed by environmental factors <ns0:ref type='bibr' target='#b27'>(Dlugosh, Parker 2008;</ns0:ref><ns0:ref type='bibr'>Colautti, Barret 2013;</ns0:ref><ns0:ref type='bibr' target='#b60'>Molina-Montenegro et al. 2018;</ns0:ref><ns0:ref type='bibr' target='#b101'>van Boheemen et al. 2019)</ns0:ref>.</ns0:p><ns0:p>Phenotypic plasticity has been considered to be the primary feature enabling aliens to colonize new, environmentally diverse areas <ns0:ref type='bibr' target='#b74'>(Richards et al. 2006;</ns0:ref><ns0:ref type='bibr' target='#b61'>Molina-Montenegro et al. 2010)</ns0:ref>. However, recent research has indicated that alien plants can evolve quickly in newly occupied areas, so both phenotypic plasticity and evolution of reproductive features could be relevant factors for successful invasions <ns0:ref type='bibr' target='#b35'>(Geng et al. 2007;</ns0:ref><ns0:ref type='bibr' target='#b60'>Molina-Montenegro et al. 2018</ns0:ref>). An evolutionary explanation for plant invasiveness implies that seed and fruit traits are crucial for invasive plants since they are related to dispersal strategies and mechanisms to cope with environmental stress. Some research reports have indicated that native and invasive populations employ different strategies for growth and reproduction <ns0:ref type='bibr' target='#b23'>(Chun et al. 2007;</ns0:ref><ns0:ref type='bibr' target='#b61'>Molina-Montenegro et al. 2010</ns0:ref><ns0:ref type='bibr'>, 2018)</ns0:ref>. Results by <ns0:ref type='bibr' target='#b60'>Molina-Montenegro et al. (2018)</ns0:ref> suggest that some seed traits of invasive plant species with rapid adaptive capacity can evolve leading to maximizing their establishment in new environments and such features can be heritable.</ns0:p><ns0:p>Due to the scarcity of data we could not point out the presence of morphological differentiation between native and invasive populations of I. capensis, and we have not been able to determine whether the seed traits are evolving. However, I. capensis, classified as an invasive species in Poland, can be suspected, while adapting and occupying new territories and competing with native species, to develop specific adaptations, contributing to its success and spread in the new environments.</ns0:p></ns0:div>
<ns0:div><ns0:head>Conclusions</ns0:head><ns0:p>New data on seed morphology and seed coat sculpture of I. capensis is provided. The presented results are useful for the identification of this species when occurring together with other closely related species. These details on seed coat ornamentation are here described for the first time.</ns0:p><ns0:p>Further studies on the developmental variation of seed coat sculpture, especially of species closely related to I. capensis, may provide a better understanding of the evolutionary relationships of the different types of sculpture.</ns0:p><ns0:p>We provide new information on the plasticity of seeds of I. capensis. There are only few papers on the phenotypic variability of species of Impatiens. Data on the morphology of seeds can prove useful for determining the impact of various environmental factors on morphological traits and show whether a given feature is stable or susceptible to environmental change.</ns0:p><ns0:p>Our results suggest that certain habitat variables, especially anthropogenic disturbances and individual soil properties, contribute in shaping the morphological variation of seeds of I. capensis. In turn, the seed coat sculpture has turned out to be a stable feature within the secondary range of this species in Poland. Manuscript to be reviewed Biometric comparison of seed traits of Impatiens capensis Meerb.</ns0:p><ns0:p>Seed length (SL), seed width (SW), seed circumference (SC), seed area (SA), variation coefficient CV), standard deviation (SD), minimum/maximum (Min/Max), arithmetic average (x), A-H as in Table <ns0:ref type='table' target='#tab_1'>1</ns0:ref>.</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:02:45912:3:0:NEW 9 Sep 2020)</ns0:p><ns0:p>Manuscript to be reviewed Manuscript to be reviewed Manuscript to be reviewed properties see Tables <ns0:ref type='table' target='#tab_1'>1 and 6</ns0:ref>, respectively. </ns0:p></ns0:div><ns0:figure xml:id='fig_0'><ns0:head>Monte</ns0:head><ns0:label /><ns0:figDesc>Carlo permutation test with the forward selection of environmental variables was applied to determine the importance and statistical significance of variables in explaining the variability in seeds. The software packages Canoco v.4.5 (ter Braak, Šmilauer 2002), MVSP 3.2 (Kovach 2010), and STATISTICA PL. ver. 13.1 (Stat-Soft Inc. 2011) were used for all analyses (van Emden 2008; Lepš, Šmilauer 2003).</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_1'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:02:45912:3:0:NEW 9 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_2'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:02:45912:3:0:NEW 9 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_3'><ns0:head /><ns0:label /><ns0:figDesc>Fig. 1) and short time of residence of little over 30 years. Although this investigated plant has only a few localities inhabiting only a relatively small area in Poland, rapid expansion across environmental gradients has been reported for several plants introduced to a new area and species can evolve quite quickly driven PeerJ reviewing PDF | (2020:02:45912:3:0:NEW 9 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_5'><ns0:head>Figure 2</ns0:head><ns0:label>2</ns0:label><ns0:figDesc>Figure 2</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_6'><ns0:head>Figure 3</ns0:head><ns0:label>3</ns0:label><ns0:figDesc>Figure 3</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_7'><ns0:head>Figure 4</ns0:head><ns0:label>4</ns0:label><ns0:figDesc>Figure 4</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_8'><ns0:head>Figure 6 (</ns0:head><ns0:label>6</ns0:label><ns0:figDesc>Figure 6</ns0:figDesc></ns0:figure>
<ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='37,42.52,270.37,525.00,371.25' type='bitmap' /></ns0:figure>
<ns0:figure type='table' xml:id='tab_0'><ns0:head>Table 2 (on next page)</ns0:head><ns0:label>2</ns0:label><ns0:figDesc /><ns0:table /></ns0:figure>
<ns0:figure type='table' xml:id='tab_1'><ns0:head>Table 1 .</ns0:head><ns0:label>1</ns0:label><ns0:figDesc>Distribution map of Impatiens capensis Meerb. in Europe (C), range in Poland (B), sites of the studied populations (A) (prepared by Adamowski & Myśliwy). Satellite map data ©2019 Google, Modifed using CorelDRAW 18 . Explanation of symbols see Table Satellite map data ©2019 Google, Modifed using CorelDRAW 18 . Explanation of symbols see</ns0:figDesc><ns0:table><ns0:row><ns0:cell>1</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell /><ns0:cell>A</ns0:cell><ns0:cell>B</ns0:cell><ns0:cell>C</ns0:cell><ns0:cell>D</ns0:cell><ns0:cell>E</ns0:cell><ns0:cell>F</ns0:cell><ns0:cell>G</ns0:cell><ns0:cell>H</ns0:cell><ns0:cell>x</ns0:cell></ns0:row><ns0:row><ns0:cell cols='2'>Weight (mg) 7.66</ns0:cell><ns0:cell>6.52</ns0:cell><ns0:cell>8.16</ns0:cell><ns0:cell>7.82</ns0:cell><ns0:cell>9.82</ns0:cell><ns0:cell>8.62</ns0:cell><ns0:cell>11.42</ns0:cell><ns0:cell>6.92</ns0:cell><ns0:cell>8.37</ns0:cell></ns0:row><ns0:row><ns0:cell>SL (mm)</ns0:cell><ns0:cell>4.05</ns0:cell><ns0:cell>3.88</ns0:cell><ns0:cell>4.23</ns0:cell><ns0:cell>4.11</ns0:cell><ns0:cell>4.46</ns0:cell><ns0:cell>4.17</ns0:cell><ns0:cell>4.60</ns0:cell><ns0:cell>4.41</ns0:cell><ns0:cell>4.24</ns0:cell></ns0:row><ns0:row><ns0:cell>Min-max</ns0:cell><ns0:cell>3.50-</ns0:cell><ns0:cell>3.16-</ns0:cell><ns0:cell>3.68-</ns0:cell><ns0:cell>3.59-</ns0:cell><ns0:cell>3.85-</ns0:cell><ns0:cell>3.43-</ns0:cell><ns0:cell>3.88-</ns0:cell><ns0:cell>3.59-</ns0:cell><ns0:cell>3.16-</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>4.64</ns0:cell><ns0:cell>4.48</ns0:cell><ns0:cell>4.70</ns0:cell><ns0:cell>4.75</ns0:cell><ns0:cell>5.26</ns0:cell><ns0:cell>4.73</ns0:cell><ns0:cell>5.74</ns0:cell><ns0:cell>4.82</ns0:cell><ns0:cell>5.74</ns0:cell></ns0:row><ns0:row><ns0:cell>SD</ns0:cell><ns0:cell>0.26</ns0:cell><ns0:cell>0.40</ns0:cell><ns0:cell>0.29</ns0:cell><ns0:cell>0.30</ns0:cell><ns0:cell>0.32</ns0:cell><ns0:cell>0.40</ns0:cell><ns0:cell>0.40</ns0:cell><ns0:cell>0.27</ns0:cell><ns0:cell>0.40</ns0:cell></ns0:row><ns0:row><ns0:cell>CV</ns0:cell><ns0:cell>6.50</ns0:cell><ns0:cell>10.28</ns0:cell><ns0:cell>6.95</ns0:cell><ns0:cell>7.35</ns0:cell><ns0:cell>7.21</ns0:cell><ns0:cell>9.71</ns0:cell><ns0:cell>8.75</ns0:cell><ns0:cell>6.19</ns0:cell><ns0:cell>9.34</ns0:cell></ns0:row><ns0:row><ns0:cell>SW (mm)</ns0:cell><ns0:cell>2.23</ns0:cell><ns0:cell>2.03</ns0:cell><ns0:cell>2.36</ns0:cell><ns0:cell>2.56</ns0:cell><ns0:cell>2.60</ns0:cell><ns0:cell>2.40</ns0:cell><ns0:cell>2.71</ns0:cell><ns0:cell>2.23</ns0:cell><ns0:cell>2.39</ns0:cell></ns0:row><ns0:row><ns0:cell>Min-max</ns0:cell><ns0:cell>1.53-</ns0:cell><ns0:cell>1.12-</ns0:cell><ns0:cell>1.78-</ns0:cell><ns0:cell>2.14-</ns0:cell><ns0:cell>2.19-</ns0:cell><ns0:cell>1.88-</ns0:cell><ns0:cell>2.15-</ns0:cell><ns0:cell>1.71-</ns0:cell><ns0:cell>1.12-</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>2.82</ns0:cell><ns0:cell>2.61</ns0:cell><ns0:cell>3.00</ns0:cell><ns0:cell>2.94</ns0:cell><ns0:cell>3.33</ns0:cell><ns0:cell>2.99</ns0:cell><ns0:cell>3.21</ns0:cell><ns0:cell>2.93</ns0:cell><ns0:cell>3.33</ns0:cell></ns0:row><ns0:row><ns0:cell>SD</ns0:cell><ns0:cell>0.28</ns0:cell><ns0:cell>0.42</ns0:cell><ns0:cell>0.33</ns0:cell><ns0:cell>0.19</ns0:cell><ns0:cell>0.31</ns0:cell><ns0:cell>0.32</ns0:cell><ns0:cell>0.30</ns0:cell><ns0:cell>0.30</ns0:cell><ns0:cell>0.37</ns0:cell></ns0:row><ns0:row><ns0:cell>CV</ns0:cell><ns0:cell>12.50</ns0:cell><ns0:cell>20.57</ns0:cell><ns0:cell>13.98</ns0:cell><ns0:cell>7.30</ns0:cell><ns0:cell>12.02</ns0:cell><ns0:cell>13.19</ns0:cell><ns0:cell>11.04</ns0:cell><ns0:cell>13.51</ns0:cell><ns0:cell>15.35</ns0:cell></ns0:row><ns0:row><ns0:cell>SC (mm)</ns0:cell><ns0:cell>10.00</ns0:cell><ns0:cell>9.55</ns0:cell><ns0:cell>10.51</ns0:cell><ns0:cell>10.61</ns0:cell><ns0:cell>11.21</ns0:cell><ns0:cell>10.49</ns0:cell><ns0:cell>11.65</ns0:cell><ns0:cell>10.75</ns0:cell><ns0:cell>10.61</ns0:cell></ns0:row><ns0:row><ns0:cell>Min-max</ns0:cell><ns0:cell>8.63-</ns0:cell><ns0:cell>7.27-</ns0:cell><ns0:cell>9.13-</ns0:cell><ns0:cell>9.35-</ns0:cell><ns0:cell>9.77-</ns0:cell><ns0:cell>8.61-</ns0:cell><ns0:cell>10.17-</ns0:cell><ns0:cell>8.9-</ns0:cell><ns0:cell>7.27-</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>11.94</ns0:cell><ns0:cell>10.98</ns0:cell><ns0:cell>11.70</ns0:cell><ns0:cell>12.70</ns0:cell><ns0:cell>13.27</ns0:cell><ns0:cell>12.10</ns0:cell><ns0:cell>14.59</ns0:cell><ns0:cell>12.20</ns0:cell><ns0:cell>14.59</ns0:cell></ns0:row><ns0:row><ns0:cell>SD</ns0:cell><ns0:cell>0.77</ns0:cell><ns0:cell>1.09</ns0:cell><ns0:cell>0.67</ns0:cell><ns0:cell>0.69</ns0:cell><ns0:cell>0.89</ns0:cell><ns0:cell>0.89</ns0:cell><ns0:cell>1.04</ns0:cell><ns0:cell>0.67</ns0:cell><ns0:cell>1.02</ns0:cell></ns0:row><ns0:row><ns0:cell>CV</ns0:cell><ns0:cell>7.65</ns0:cell><ns0:cell>11.42</ns0:cell><ns0:cell>6.39</ns0:cell><ns0:cell>6.50</ns0:cell><ns0:cell>7.90</ns0:cell><ns0:cell>8.53</ns0:cell><ns0:cell>8.97</ns0:cell><ns0:cell>6.22</ns0:cell><ns0:cell>9.64</ns0:cell></ns0:row><ns0:row><ns0:cell>SA (mm 2 )</ns0:cell><ns0:cell>6.52</ns0:cell><ns0:cell>5.79</ns0:cell><ns0:cell>7.15</ns0:cell><ns0:cell>7.71</ns0:cell><ns0:cell>8.42</ns0:cell><ns0:cell>6.74</ns0:cell><ns0:cell>9.26</ns0:cell><ns0:cell>7.25</ns0:cell><ns0:cell>7.44</ns0:cell></ns0:row><ns0:row><ns0:cell>Min-max</ns0:cell><ns0:cell>4.72-</ns0:cell><ns0:cell>2.43-</ns0:cell><ns0:cell>5.10-</ns0:cell><ns0:cell>5.83-</ns0:cell><ns0:cell>6.15-</ns0:cell><ns0:cell>4.82-</ns0:cell><ns0:cell>6.74-</ns0:cell><ns0:cell>5.27-</ns0:cell><ns0:cell>2.43-</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>9.53</ns0:cell><ns0:cell>7.69</ns0:cell><ns0:cell>9.10</ns0:cell><ns0:cell>1.23</ns0:cell><ns0:cell>12.16</ns0:cell><ns0:cell>1.21</ns0:cell><ns0:cell>13.54</ns0:cell><ns0:cell>9.61</ns0:cell><ns0:cell>13.54</ns0:cell></ns0:row><ns0:row><ns0:cell>SD</ns0:cell><ns0:cell>1.14</ns0:cell><ns0:cell>1.51</ns0:cell><ns0:cell>1.21</ns0:cell><ns0:cell>1.07</ns0:cell><ns0:cell>1.48</ns0:cell><ns0:cell>1.27</ns0:cell><ns0:cell>1.63</ns0:cell><ns0:cell>1.08</ns0:cell><ns0:cell>1.62</ns0:cell></ns0:row><ns0:row><ns0:cell>CV</ns0:cell><ns0:cell>17.54</ns0:cell><ns0:cell>26.06</ns0:cell><ns0:cell>16.92</ns0:cell><ns0:cell>13.88</ns0:cell><ns0:cell>17.63</ns0:cell><ns0:cell>17.58</ns0:cell><ns0:cell>17.65</ns0:cell><ns0:cell>14.96</ns0:cell><ns0:cell>21.76</ns0:cell></ns0:row></ns0:table><ns0:note>PeerJ reviewing PDF | (2020:02:45912:3:0:NEW 9 Sep 2020)</ns0:note></ns0:figure>
<ns0:note place='foot'>PeerJ reviewing PDF | (2020:02:45912:3:0:NEW 9 Sep 2020)Manuscript to be reviewed</ns0:note>
<ns0:note place='foot' n='2'>PeerJ reviewing PDF | (2020:02:45912:3:0:NEW 9 Sep 2020)Manuscript to be reviewed</ns0:note>
</ns0:body>
" | "Łódź, 07.09.2020
Dear Editor,
I encourage you this last version be reviewed by a professional English editor. I read your letter and it was indicated that a native English reviewed both versions. However I found a number of problems. To mention only an example, in the Abstract it is mentioned 'the presented', instead of 'present'. Please indicate in Acknowledements this information.
Thanks for your comments. Our article has been checked by a native speaker. We added Acknowledgments to the manuscript.
" | Here is a paper. Please give your review comments after reading it. |
9,803 | "<ns0:abstract xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:p>In this study we combine information from aerial LIDAR and hemispherical images taken in the field with ForeStereo -a forest inventory device-to assess the vulnerability and to design conservation strategies for endangered Mediterranean fir forests based on the mapping of fire risk and canopy structure spatial variability. We focused on the largest continuous remnant population of the endangered tree species Abies pinsapo Boiss, spanning 252 ha in Sierra de las Nieves National Park (South Andalusia, Spain). We established 49 sampling plots over the study area. Stand structure variables were derived from ForeStereo device, an approximal sensing technology for tree diameter, height and crown dimensions and stand crown cover and basal area retrieval from stereoscopic hemispherical images photogrammetry. With this information we developed regression models with airborne LIDAR data (spatial resolution of 0.5 points•m -2</ns0:p><ns0:p>). Thereafter, six fuel models were fitted to the plots according to the UCO40 classification criteria, and then the entire area was classified using the Nearest Neighbor algorithm on Sentinel imagery (overall accuracy of 0.56 and a Kappa Coefficient of 0.46). FlamMap software was used for fire simulation scenarios based on fuel models, stand structure, and terrain data. Besides the fire simulation, we analyzed canopy structure to assess the status and vulnerability of this fir population. The assessment shows a secondary growth forest that has an increasing presence of fuel models with potential for high fire spread rate fire and burn probability. Our methodological approach has the potential to be integrated as a support tool for adaptive management and conservation of A. pinsapo across its whole distribution area (< than 4000 ha), as well as for other endangered circunmediterranean fir forests, as.</ns0:p><ns0:p>A. numidica de Lannoy and A. pinsapo marocana Trab. in North Africa.</ns0:p></ns0:div>
</ns0:abstract>
<ns0:body xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:head n='1.'>INTRODUCTION</ns0:head><ns0:p>Endemic conifer species are more abundant in Mediterranean-type climate regions from the Northern Hemisphere than those from the Southern Hemisphere, which have been linked to the selective pressure of cold and/or drought conditions that led to the development of ecophysiological advantages in conifers over angiosperms on oligotrophic soils. Meanwhile, the Mediterranean-type climate regions of South Africa and Southwestern Australia have been more climatically and tectonically stable, which resulted in lower diversity and ancient lineages of conifers. The Mediterranean Basin has 32 endemic conifer species ─this is more than 25% of the 122 total conifer flora <ns0:ref type='bibr' target='#b80'>(Rundel, 2019)</ns0:ref>. In particular, the genus Abies Mill. experienced an outstanding speciation from the late Neogene in the Mediterranean Basin that gave rise to nine species and one natural hybrid <ns0:ref type='bibr' target='#b52'>(Linares 2011)</ns0:ref>. Past climate changes have led to population migrations, and to a shrinkage and fragmentation of ancestral Mediterranean firs, further exacerbated by human impacts. This resulted in Circummediterranean endemic firs of high paleogeographic interest, since they are established in relict restricted-range masses with relevant vulnerability to global warming effects <ns0:ref type='bibr' target='#b38'>(Liepelt, et al., 2010)</ns0:ref>. Adaptive management of these forests to protect them from the increasing fire risk is essential for their survival.</ns0:p><ns0:p>Future climate extreme events, such as severe droughts, mega-fires, and disease infestations threaten these relict Mediterranean firs populations <ns0:ref type='bibr' target='#b84'>(Sánchez-Salguero, et al., 2017)</ns0:ref>. It is well known that fire has influenced the landscape and terrestrial life as far back as the beginning of land plants <ns0:ref type='bibr' target='#b12'>(Bowman, et al., 2009;</ns0:ref><ns0:ref type='bibr' target='#b69'>Pausas & Keeley, 2009;</ns0:ref><ns0:ref type='bibr' target='#b29'>He, et al., 2012)</ns0:ref>. Although many PeerJ reviewing PDF | (2020:02:46252:1:1:NEW 26 May 2020)</ns0:p><ns0:p>Manuscript to be reviewed conifers had developed adaptive traits to live in fire-prone environments, this is not the case of the genus Abies. The firs developed traits proper of the humid areas where they thrive which have rendered them as poorly resistant (thin bark) and resilient (recruitment failure in open spaces) to fires <ns0:ref type='bibr' target='#b24'>(Furyaev, et al., 1983)</ns0:ref> <ns0:ref type='bibr' target='#b89'>(Vega, 1999)</ns0:ref>.</ns0:p><ns0:p>Remote sensing plays a relevant role to study the affection and prevention of the effects of global warming in relict Mediterranean firs forests. Spectral imagery has been employed for the early detection of forest infestations <ns0:ref type='bibr' target='#b33'>(Immitzer & Atzberger, 2014)</ns0:ref>, to estimate evapotranspiration <ns0:ref type='bibr' target='#b20'>(Dzikiti, et al., 2019)</ns0:ref> and to study photosynthetic activity <ns0:ref type='bibr' target='#b19'>(de Sousa, et al., 2017)</ns0:ref>. Meanwhile, 3D point cloud from laser scanner (LIDAR) has been employed in fire management <ns0:ref type='bibr' target='#b16'>(Chuvieco & Kasischke, 2007)</ns0:ref>, to assess forest volume and biomass <ns0:ref type='bibr' target='#b87'>(Van Ardt, et al., 2008)</ns0:ref>, and canopy structure <ns0:ref type='bibr' target='#b1'>(Adamic, et al., 2017)</ns0:ref> <ns0:ref type='bibr' target='#b65'>(Mura, et al., 2015)</ns0:ref>. Also, point cloud can be used for ecological purposes, such as light availability for species distribution modeling <ns0:ref type='bibr' target='#b93'>(Wüest, et al., 2020)</ns0:ref> and forest changes in ecotones <ns0:ref type='bibr' target='#b92'>(Wang, et al., 2020)</ns0:ref>. Airborne LIDAR has shown better suitability to map crown and canopy heights <ns0:ref type='bibr' target='#b90'>(Wang & Glenn, 2008)</ns0:ref>, although in highly dense forests the point cloud finds difficulties to reach the ground, and thus mapping understory vegetation may be inaccurate. However, terrestrial LIDAR has a great potential for estimating shrubs and understory biomass, while there is a lack of points for a precise estimation of crown heights when the canopy cover is high <ns0:ref type='bibr' target='#b31'>(Hilker, et al., 2012)</ns0:ref>.</ns0:p><ns0:p>Mapping fire risk is a relevant tool for landscape planning in the Mediterranean Basin and remote sensing is becoming essential for high precision fuel modeling. Fuel moisture, one of the main factors that affect fuel flammability, is frequently obtained by satellite and meteorological data <ns0:ref type='bibr' target='#b15'>(Chuvieco, et al., 2010)</ns0:ref>. <ns0:ref type='bibr' target='#b94'>Yebra, et al., (2018)</ns0:ref> Manuscript to be reviewed calibrated with field data across Australia to estimate fuel moisture content and flammability using radiative transfer model inversion.</ns0:p><ns0:p>One of the most common algorithms used to obtain burn probability is the Minimum Travel Time (MTT), based on the Huygens' principle <ns0:ref type='bibr' target='#b21'>(Finney, 2002)</ns0:ref>. Several studies have previously applied MTT through FlamMap software to assess fire risk in different ecosystems. In Greece, <ns0:ref type='bibr'>Mitsopoulos, et al. (2015)</ns0:ref> worked on a wildland-urban interface using orthoimages photointerpretation to assess the spatial extent of fuel types and found limitations in automated classification due to poor spectral resolution. <ns0:ref type='bibr'>Maillinis, et al. (2016)</ns0:ref> worked on Holy Mount Athos using Landsat 8 and RapidEye satellite imagery for fuel mapping. In Italy, <ns0:ref type='bibr' target='#b81'>Salis et al. (2015)</ns0:ref> analyzed the whole Sardinia island with MTT to study spatiotemporal patterns of fire risk, considering historical ignition locations. In Spain, <ns0:ref type='bibr' target='#b3'>Alcasena et al. (2019)</ns0:ref> mapped fire risk of the whole Catalonia region using LIDAR derived data to assess vegetation structure, and assuming <ns0:ref type='bibr' target='#b85'>Scott & Burgan (2005)</ns0:ref> fuel model classification. They applied MTT through FlamMap to obtain 150 m resolution fire scenarios and found high priority areas to reduce fire risk, especially Pinus halepensis Mill. dense forests. MTT have been also employed to estimate burn probability in Sierra de Aracena Natural Park (South Spain) in the frame of an economic impact analysis on the area <ns0:ref type='bibr' target='#b63'>(Molina et al., 2017)</ns0:ref>.</ns0:p><ns0:p>FARSITE software is commonly used to predict the fire spread of specific events. <ns0:ref type='bibr' target='#b82'>Salis et al. (2016)</ns0:ref> worked with FARSITE simulations through aerial image and land use maps in several Euro-Mediterranean countries, following an east-west gradient. They improved FARSITE predictions using custom fuel models instead of the standard ones. Thus, accurate fuel models are essential in burn probability and fire risk modeling.</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:02:46252:1:1:NEW 26 May 2020)</ns0:p><ns0:p>Manuscript to be reviewed LIDAR technology provides an unprecedented tool for fuel and forest structure characterization in forest management. In Spain, <ns0:ref type='bibr' target='#b26'>González-Olabarria, et al. (2012)</ns0:ref> analyzed fire risk based on airborne LIDAR (2 points•m -2 ) and <ns0:ref type='bibr' target='#b4'>Anderson's (1982)</ns0:ref> fuel types in a mixed coniferous forest of Pinus nigra Arnold and pinaster Ait. They reported flame length > 20 m, and a limitation in their approach for precise shrub fuel mapping due to the lack of points reaching the understory strata. <ns0:ref type='bibr' target='#b11'>Botequim, et al. (2019)</ns0:ref> also reported this limitation when using LIDAR (0.5 points•m -2 ) to develop accurate canopy models in a mixed forest of Pinus pinea L. with Pinus pinaster and Quercus pyrenaica Willd. In this case they followed the UCO40 fuel model classification and reported low spread rate and flame length in their simulations.</ns0:p><ns0:p>Therefore, LIDAR data needs to be implemented in regression models supported by field sampling to eventually characterize the forest structure. For this purpose, hemispherical images are an alternative to traditional field sampling. This technique has been used in forest ecology for more than 50 years, but its widespread adoption was limited due to constraints related to image processing capacity <ns0:ref type='bibr'>(Chianucci, 2019)</ns0:ref>. However, technical improvements allowed reducing the time for image processing as well as better image quality acquisition, which coupled with the widespread of popular digital cameras, increased the ease of obtaining and storing hemispherical images, becoming an important tool for fieldwork <ns0:ref type='bibr' target='#b28'>(Hall, et al., 2017)</ns0:ref>. In this sense, ForeStereo, a device for forest inventorying developed by the Forest Research Centre of the Spanish National Institute for Agriculture and Food Research and Technology (INIA-CIFOR), obtains, by the analysis of two stereoscopic hemispherical images collected at sampling locations, stand and tree variables in a cost-effective way, reducing time and costs <ns0:ref type='bibr'>(Rodriguez García et al. 2014)</ns0:ref>.</ns0:p><ns0:p>Although most of the works that have applied LIDAR to Abies forests have focused on A. alba Mill., no previous works have used this technology to study the rest of Mediterranean fir forests, except <ns0:ref type='bibr'>Aragón, et al., (2019)</ns0:ref> and <ns0:ref type='bibr' target='#b18'>Cortés-Molino, et al., (2017)</ns0:ref> who worked on A. pinsapo Boiss forests, using the point cloud for tree identification and a basic vegetation landscape analysis, respectively. It must be highlighted that this species, as well as A. numidica de Lannoy, is established in small isolated masses, becoming highly vulnerable to human impacts <ns0:ref type='bibr' target='#b38'>(Liepelt, et al., 2010)</ns0:ref>. The combination of remote sensing technology such as laser scanning and proximal sensing as ForeStereo can contribute to the monitoring of these relict forests, through the acquisition of high precision stand structure data.</ns0:p><ns0:p>Abies pinsapo forests are restricted to three small areas in southern Spain <ns0:ref type='bibr'>(A. pinsapo pinsapo)</ns0:ref> and two in Morocco (Abies pinsapo marocana), totaling less than 8.000 ha <ns0:ref type='bibr' target='#b42'>(Linares, 2008)</ns0:ref>.</ns0:p><ns0:p>Forest fires have reduced the size of populations of this fir. This affection is remarkable in Sierra Bermeja (Málaga, Spain), where Pinus pinaster Ait regrowth better in burned areas, replacing the fir. In the annual time series 1817-1997 in this location, the longest continuous period without fires extended for 34 years <ns0:ref type='bibr' target='#b89'>(Vega, 1999)</ns0:ref>. Rodríguez y Silva (1996) points the low flammability of A. pinsapo, but due to its thin bark the resistance to fire is very low. He also claims that fire spread tends to decrease inside pinsapo forests because of the high canopy closure, the microclimatic conditions where they grow, and the low presence of understory. <ns0:ref type='bibr' target='#b6'>Arista (1995)</ns0:ref> studied the structure and dynamics of this specie in the Natural Park of Sierra de Grazalema (Cádiz, Spain) and found that the relationship between age and size of pinsapo in this population is weak. Acute symptoms of tree growth decline and forest dieback due to stand stagnation and climate change have already been reported in the population of the Natural Park of Sierra de las Nieves (Málaga, Spain) <ns0:ref type='bibr'>(Linares & Carreira, 2009)</ns0:ref>. Pinus halepensis. could take advantage in some areas of pinsapo fir declining growth, turning the pure masses into mixed ones <ns0:ref type='bibr' target='#b54'>(Linares, et al., 2011a)</ns0:ref>. However, reducing intraspecific competition through thinning can decrease the stagnation of A. pinsapo, inducing longer growing seasons <ns0:ref type='bibr' target='#b45'>(Linares, et al., 2009a)</ns0:ref>, and increasing the adaptive capacity to decreasing water availability <ns0:ref type='bibr' target='#b47'>(Linares, et al., 2009b)</ns0:ref>.</ns0:p><ns0:p>This work aims to combine the use of LIDAR and hemispherical images in one of the most relevant A. pinsapo population, located in a protected area in Málaga (Spain), to assess vulnerability through (i) mapping fire risk and (ii) analyzing canopy structure variability and its possible links to reported declining growth symptoms.</ns0:p></ns0:div>
<ns0:div><ns0:head n='2.'>MATERIAL AND METHODS</ns0:head></ns0:div>
<ns0:div><ns0:head n='2.1'>Study area</ns0:head><ns0:p>The study location is a valley in the municipality of Yunquera, in Sierra de las Nieves National Park (Fig 1 <ns0:ref type='figure'>.</ns0:ref>), in a transition between the upper and lower Mesomediterranean bioclimatic band.</ns0:p><ns0:p>The annual rainfall is around 1500 mm and the maximum average temperature of the warmest month (August) is 33.6 ºC (S. <ns0:ref type='bibr'>Rivas-Martínez & Rivas-Saenz, 1996</ns0:ref><ns0:ref type='bibr'>-2020)</ns0:ref>. The forest is placed in a steep valley of about 250 ha. At the southern border there is a crest that was the limit of a harsh wildfire that happened in 1991 which burned 9000 ha <ns0:ref type='bibr' target='#b67'>(Narváez, 1991)</ns0:ref>. The east part is limited by crop fields. This, together with summer weather conditions and important touristic pressure in Sierra de las Nieves National Park, makes the risk of wildfire especially high. In terms of the type of vegetation formation, the studied forest belongs to the association Paeonio broteroi-Abietetum pinsapo <ns0:ref type='bibr' target='#b9'>(Asensi & Rivas-Martínez, 1976</ns0:ref>) composed mainly by pinsapo fir, forming pure forests in the upper and shaded parts of the valley. The incidence of the root-rot fungi Heterobasidion abietinum Niemelä & Korhonen is very high <ns0:ref type='bibr' target='#b49'>(Linares, et al., 2010)</ns0:ref>. In sunny and low-altitude spots, the forest is mixed with Pinus halepensis and scrubs of Juniperus spp and Cistus spp. <ns0:ref type='table' target='#tab_1'>PDF | (2020:02:46252:1:1:NEW 26 May 2020)</ns0:ref> Manuscript to be reviewed</ns0:p></ns0:div>
<ns0:div><ns0:head>PeerJ reviewing</ns0:head></ns0:div>
<ns0:div><ns0:head n='2.2'>Fieldwork: ForeStereo inventory</ns0:head><ns0:p>The access to field sites was approved by the Andalusian Regional Government ('Consejería de Medio Ambiente y Ordenación del Territorio) with the approval code: PNSN/AU/10-2018. The purpose of the fieldwork was to classify local fuel models and collect forest structure data, mainly canopy cover, crown and stand heights. The valley was sampled in Spring 2018 with 49 plots of 8 m radius, using stereoscopic hemispherical images for obtaining tree metrics such as stand height (Ho), Canopy Base Height (CBH), Canopy Cover (CC), Canopy Bulk Density (CBD) and basal area (G). Scrubs cover and height were also assessed by the line-intersect method to support the classification of the fuel models. Each sampling plot was accurately geolocated using a high precision GNSS receiver, supported by an RTK terrestrial station deployed in the upper part of the valley.</ns0:p><ns0:p>Forest inventory was implemented with ForeStereo, a device developed by the Forest Research Centre of the Spanish National Institute for Agriculture and Food Research and Technology (INIA-CIFOR). ForeStereo is equipped with two fish-eye cameras. At sampling locations, three pairs of hemispherical stereoscopic images (Fig. <ns0:ref type='figure'>2</ns0:ref>) with different exposures are taken. The matching process, compiled in a MatLab® software package, consists of four main steps as detailed in Sánchez-González ( <ns0:ref type='formula'>2016</ns0:ref>): (i) a supervised segmentation of tree stems and crowns;</ns0:p><ns0:p>(ii) correspondence of features between the two images and photogrammetric retrieval of tree dimensions; (iii) tree variable modeling and (iv) stand variable estimation, which requires instrumental bias and occlusions correction <ns0:ref type='bibr'>(Montes, 2019)</ns0:ref>. ForeStereo was used to estimate tree height, crown base height, crown volume and diameter at breast height (DBH) for each tree, number of stems per hectare and crown cover (CC) and stand basal area (G). Tree metrics from the hemispherical images were compared with airborne LIDAR output data to develop regression models. Thirty-two random plots were chosen to adjust the models and seventeen to assess their predictive capability.</ns0:p><ns0:p>As the geometry of ForeStereo system and image projections is known, no additional data calibration is needed to carry out photogrammetric retrieval of tree variables. Accuracy of ForeStereo estimated through the Root Mean Squared Error (RMSE) ranges from 0.015 to 0.057 m for DBH <ns0:ref type='bibr' target='#b77'>(Rodríguez-García, et al., 2014;</ns0:ref><ns0:ref type='bibr'>Sánchez-González, et al., 2016)</ns0:ref>, 2.59 to 6.4 m for tree height <ns0:ref type='bibr' target='#b77'>(Rodríguez-García, et al., 2014;</ns0:ref><ns0:ref type='bibr' target='#b60'>Marino et al. 2018</ns0:ref>) and 3.1 and 0.6 m for crown base height and crown diameter respectively <ns0:ref type='bibr' target='#b60'>(Marino et al, 2018)</ns0:ref> and 11.6 m2/ha for G <ns0:ref type='bibr'>(Sánchez-González, 2016)</ns0:ref>.</ns0:p><ns0:p>With the ForeStereo data we were able to estimate stand height at each plot following the Assmann's criteria <ns0:ref type='bibr'>(1970)</ns0:ref>, while the canopy cover was obtained directly from the hemispherical images and the Canopy Base Height (CBH) was averaged for each plot. The calculation of the Canopy Bulk Density (CBD) was based on equations reported by <ns0:ref type='bibr' target='#b78'>Ruiz-Peinado, et al., (2011)</ns0:ref> for A. pinsapo, in which tree height and stem diameter are used to calculate thin branches and needles biomass. Therefore, CBD is estimated by dividing biomass between crown volume (which is obtained from ForeStereo estimations).</ns0:p></ns0:div>
<ns0:div><ns0:head n='2.3'>LIDAR data.</ns0:head><ns0:p>The LIDAR data were obtained in 2015 by the Spanish National Geographic Institute, through the Plan Nacional de Ortofotografía Aérea (PNOA) project. The pixel size is 0.25 m and the point cloud density is 0.5 points•m -2 . FUSION software was employed for point cloud processing and data extraction <ns0:ref type='bibr' target='#b61'>(McGaughey & Carson, 2003)</ns0:ref>. A correlation matrix between ForeStereo tree data and all LIDAR metrics obtained with FUSION was useful to detect which LIDAR metric Manuscript to be reviewed was most suitable to build the regression models in 32 random plots. We chose the models with less root-mean-square (RMSE) and higher adjusted-R in the remaining 17 plots to predict ForeStereo tree metrics with the LIDAR point cloud. Height break for LIDAR metrics was 4m, while it was 0.25 m for the total vegetation height to avoid high shrubs influence and ground points, respectively. These two height breaks were tested to inquire which one was better to fit the models.</ns0:p><ns0:p>General canopy structure traits such as canopy cover and height were analyzed from the regression models obtained to detect declining growth symptoms in this A. pinsapo forest. These variables were chosen due to the airborne LIDAR accuracy to estimate them <ns0:ref type='bibr' target='#b2'>(Ahmed, et al., 2015;</ns0:ref><ns0:ref type='bibr' target='#b7'>Arumãe & Lang, 2018)</ns0:ref>. Also, we tested if canopy bulk density were consistent with the results from this analysis.</ns0:p></ns0:div>
<ns0:div><ns0:head n='2.4'>Fuel models and fire scenarios</ns0:head><ns0:p>To simulate fire risk, we first ascribed field plots to any of the UCO40 fuel models classification, which uses specific criteria and traits tuned to the case of Mediterranean environments and thus performs better than the widely used Prometheus or Rothermel models (Rodríguez y Silva & <ns0:ref type='bibr'>Molina Martínez, 2010;</ns0:ref><ns0:ref type='bibr'>2012)</ns0:ref>. We identified a total of 6 fuel models across the plots (Table <ns0:ref type='table'>1</ns0:ref>).</ns0:p><ns0:p>The UCO4 procedure is based on the fuel models classification of <ns0:ref type='bibr' target='#b85'>Scott & Burgan, (2005)</ns0:ref> Manuscript to be reviewed Ecognition® software was used to segment the place of study with the Nearest Neighbor algorithm in an Object-Based Image Analysis (OBIA) <ns0:ref type='bibr' target='#b25'>(Gao, et al., 2007)</ns0:ref>. To carry out this segmentation, we used the raster layers resulting from the regression models validated previously (CBD, CBH, G, Hv) along with NDVI data from Sentinel-2 images (2015) and terrain models obtained from the LIDAR point cloud (topography, aspect, and slope). Later, an error matrix was calculated to evaluate the accuracy of the fuel model classification.</ns0:p><ns0:p>The following raster layers, generated from the regression models, were incorporated: terrain elevation, aspect, slope, Ho, CBH, CBD, CC, and fuel models.</ns0:p><ns0:p>The initial fuel moisture file (.FMD) used is based on <ns0:ref type='bibr' target='#b85'>Scott & Burgan (2005)</ns0:ref> suggestions. It was assumed low moisture content (two-third cured) in fuel models more common in shaded spots (M8, HR7, and R4) and very low content (fully cured) was chosen for fuel models typically found in sunny spots (M3, M9 and HPM4). Also, we provide a Weather Stream file (.WXS) used for dead fuel moisture conditioning (Table <ns0:ref type='table' target='#tab_1'>2</ns0:ref>). This file modifies initial dead fuel moisture due to weather data such as temperature, relative humidity, cloud cover and hourly precipitation in a period established. Conditioning also implies adjusting initial dead fuel moisture to site factors (elevation, slope, aspect, and canopy cover) thanks to the uploaded previous raster layers in FlamMap <ns0:ref type='bibr' target='#b23'>(Finney, 2006)</ns0:ref>. The .WXS file was built based on in-site meteorological data collected by the University of Jaén in 2014 (temperature, precipitation, and relative humidity) and the Spanish Meteorological Agency (AEMET) data for wind and cloud cover values. We chose the warmest day of 2014 for the conditioning period between the 10:00 to 19:00 hours. The conditioned fuel moistures at the end of this period was the final fuel moisture used for the simulations PeerJ reviewing PDF | (2020:02:46252:1:1:NEW 26 May 2020)</ns0:p></ns0:div>
<ns0:div><ns0:head>Manuscript to be reviewed</ns0:head><ns0:p>As model outputs we obtained three different fire scenarios: Burn probability based on 200 random ignition points using MTT algorithm, flame length, and spread rate, both calculated for each cell. The purpose of simulating fire scenarios was to detect vulnerability areas and to assess how exposed is the pinsapo forest to this risk for conservation planning. For this reason, we did not simulate specific events or spotfires using Farsite software. In this point, FlamMap software is more appropriate, since it calculates spread rate and flame length for each landscape cell without a temporal component and use MTT to simulate 200 random fires to detect the probability of a point to get burned <ns0:ref type='bibr' target='#b23'>(Finney, 2006</ns0:ref><ns0:ref type='bibr' target='#b26'>) (González-Olabarria, et al., 2012)</ns0:ref>. The simulations were set under two prevailing wind conditions: west winds (called 'Ponent' in the area) and east winds (called 'Levant'), both for a frequent speed of 13 km•h -1 .</ns0:p></ns0:div>
<ns0:div><ns0:head n='3.'>RESULTS</ns0:head><ns0:p>We found that Percentage of first returns above 4 m (FR4), Percentage of all returns above 4 m (AR4), Percentage of all returns above mean (ARM) and Percentage of first returns above 0.25m (FR0.25) were the LIDAR metrics that best fit the ForeStereo data (Table <ns0:ref type='table'>3</ns0:ref>). All the significant regression models were obtained with 95% confidence in the seventeen validation plots. Basal Area (G) was the only variable with an acceptable fitting with a height break of 0.25 m. The rest of the variables were better modeled above 4 m. Canopy Cover (CC) had the best fitting model with an RMSE less than 20%, while the highest error was found in modeling the Canopy Base Height (CBH) with an RMSE of 83.3%. This high difference could be due to low point cloud density (minimum 0.5 points/m 2 guaranteed) as well as by the fact that airborne LIDAR PeerJ reviewing PDF | (2020:02:46252:1:1:NEW 26 May 2020) Manuscript to be reviewed produces higher accuracy for variables related to the top of the canopy <ns0:ref type='bibr' target='#b31'>(Hilker, et al., 2012)</ns0:ref>.</ns0:p><ns0:p>Results for dominant height was acceptable (RMSE of 0.55), as finding the top of the crowns with ForeStereo in high-density forest can be harsh.</ns0:p><ns0:p>The error matrix for the fuel model classification with Nearest Neighbor algorithm shows an overall accuracy of 0.56 and a Kappa Coefficient of 0.46 (Table <ns0:ref type='table'>4</ns0:ref>). The most abundant fuel model is M9, with a 30.5% of cover surface, followed by, M3 with 27.4%, M8 with 23%, HPM4 with 7.19%, HR7 with 6.12% and R4 with 5.73%.</ns0:p><ns0:p>Once the corresponding layers were created based on the regression models, the fuel model classification, and the data derived from terrain elevations (Fig. <ns0:ref type='figure'>3</ns0:ref>), fire simulations were obtained from FlamMap (Fig. <ns0:ref type='figure'>4</ns0:ref>). The final landscape file keeps the same pixel resolution of the input data (10 m). We found worse burn probability and spread rate under Levant wind condition but similar flame length scenario in both Levant and Ponient. Burn probability is higher under Levant winds, with a mean value of 0.078, than under Ponent winds, which mean value is 0.061.</ns0:p><ns0:p>Spread rate shows a mean value of 41.82 m• min -1 in Levant and more than 50% of the landscape is ≥50 m•min -1 . In contrast, Ponent wind shows a mean value of 33.34 m• min -1 , and only 36% percent of the landscape is ≥50 m• min -1 . Flame length resulted in similar values in both wind directions.</ns0:p><ns0:p>Regarding the canopy structure analysis to detect symptoms of declining growth, the estimated variables CBH and G were not considered due to RMSEs far above 60%. Instead, we use Ho and CC (Fig. <ns0:ref type='figure'>5</ns0:ref>), as well as CBD, for such assessment. The following results are estimated only in the landscape cells with a vegetation height above 4 m. Canopy heights (Ho) ranges from 4 to 18 m, with a mean value of 9.1 m and a standard deviation (SD) of 3.28. It is remarkable that the mode of canopy heights falls below 5 m in this forest area not affected by fire since the mid-XX century and under long-term, no management policy. On the other hand, canopy height classes between 7 and 12 m showed similar frequencies (high equitability). Lastly, canopy heights higher than 15 m were present but with rather low frequency.</ns0:p><ns0:p>CC has a mean value of 64.5% and an SD of 18. We found low frequency for values between 0-30% (<2% of the area) because most areas with low CC correspond to land covered by shrubs with less than 4 m heights. Almost 25% of the land shows canopy cover values above 80%, which means that areas with near to full cover are relatively abundant. Also, 66% of the area is between a CC of 40-80%.</ns0:p><ns0:p>Lastly, CBD mean value is 0.16 kg• m -3 with a SD of 0.1. Our results showed a high canopy density since >60% of the area has values over 0.1 kg• m -3 and >30% is over 0.3 kg • min -1 .</ns0:p><ns0:p>During the fieldwork, we found several gaps opened due to die-back process (Fig. <ns0:ref type='figure'>6</ns0:ref>).</ns0:p></ns0:div>
<ns0:div><ns0:head n='4.'>DISCUSSION</ns0:head><ns0:p>Endangered Circummediterranean firs are highly vulnerable to climate change effects in the isolated areas where they remain <ns0:ref type='bibr' target='#b84'>(Sánchez-Salguero, et al., 2017)</ns0:ref>. Abies nebrodensis Mattei is currently the rarest conifer in the European flora, with only 34 mature trees able to reproduce sexually in the wild <ns0:ref type='bibr' target='#b68'>(Pasta, et al., 2019)</ns0:ref>. The recovery of this species and the protection of the other ones to avoid a similar decrease is an urgent matter that demands the best techniques available to support the traditional field survey.</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:02:46252:1:1:NEW 26 May 2020)</ns0:p></ns0:div>
<ns0:div><ns0:head>Manuscript to be reviewed</ns0:head><ns0:p>We propose in this work a methodology that combines the use of LIDAR with ForeStereo, UCO40 fuel models and FlamMap simulations to significantly reduce the fieldwork effort and time of inspection, increasing the efficiency of massive data capture in forest management</ns0:p><ns0:p>The application of this methodology in the study area obtained fire simulations which showed that East wind conditions ('Levant') resulted in worse fire scenarios than West ones ('Ponent') as illustrated in Fig. <ns0:ref type='figure'>4</ns0:ref>. Spread rate seems to be more influenced by topography and wind condition <ns0:ref type='bibr' target='#b82'>(Salis, et al., 2016)</ns0:ref> than flame length, which appears to be more conditioned to the fuel characterization. Low spread rate and flame length are found in areas with HR7 and R4 models, because they correspond to high-density A. pinsapo stands, which is a coherent behavior according to Rodríguez y Silva <ns0:ref type='bibr'>(1996)</ns0:ref>.</ns0:p><ns0:p>Similar results can be found in the Euro-Mediterranean work of <ns0:ref type='bibr' target='#b82'>Salis, et al, (2016)</ns0:ref>, in which the maximum spread rate simulated in the Attica region (Greece), Budoni (Italy), Fresnedoso de Ibor and Navalmoral (Spain) is between 50-110 m• min -1 . In their work, the worst flame length scenario is located in Fresnedoso de Ibor (Spain) and Penteli (Greece) with a range between 25-50 m. They also found a higher spread rate and low flame length mainly in areas with herbaceous vegetation, but also in forest and shrublands in steep mountains exposed to wind (as it happens in our work). In these areas flame length is also higher than in lands with herbaceous vegetation. However, both wind conditions generate two remarkable foci of fire risk, well highlighted in the burn probability map (Fig <ns0:ref type='figure'>4</ns0:ref>). One is in the north-west part of the valley, in south-facing slopes (180º N) where highly dense and tall (>1.80 m) patches of the shrub Juniperus spp on steep terrain represent ideal conditions for a high fire spread rate (>50m•min -1 ), while the flame length will depend more on the wind (Ponent >15 m, Levant >30 m). It is not surprising that this PeerJ reviewing PDF | (2020:02:46252:1:1:NEW 26 May 2020)</ns0:p><ns0:p>Manuscript to be reviewed condition corresponds to the M9 fuel model (Table <ns0:ref type='table'>1</ns0:ref>), in which massive shrubs formation dominates the fire behavior. The other focus is placed in the east part of the valley, due to the occurrence of fuel models for which fire behavior is mainly determined by the concomitance of dense shrub cover and very steep slopes (>75º).</ns0:p><ns0:p>González-Olabarria (2012) obtained a lower fire risk landscape in a carefully managed evenaged forest of Pinus nigra. and P. pinaster, while our study corresponds to a non-managed forest of pinsapo firs. This contrast strengthens the idea of an urgent need for adaptive management of these endemic fir forests, leaving behind the traditional paradigm of no-management in biological conservation. The kind of fuel models appears to be determinant for the fire scenarios obtained.</ns0:p><ns0:p>The distribution of canopy structure features depicted in Fig. <ns0:ref type='figure'>5</ns0:ref> highlights (i) that the most frequent stand height barely reaches 5 m, the mean value is just 9.1 m and figures higher than 12-15 m are rare despite that A. pinsapo can reach up to 30-35 m of height <ns0:ref type='bibr'>(López-Quintanilla, et al., 2013)</ns0:ref> ; (ii) that canopy cover has an average value of 64.5%, well below full-cover in an area under a 'set aside' and 'no management' strategy since the late 1960s, and that patches with 90%-100% cover account for less than 10% of the whole area; and (iii) that there is an overall very high variability for both stand height and canopy cover values across the landscape, with a relatively high evenness in both variable distributions, especially regarding tree height. All these results indicate a lack of old-growth stands in the study area, and the predominance of secondary forests, which is consistent with previous studied based on field surveys <ns0:ref type='bibr' target='#b53'>(Linares, et al, 2011b;</ns0:ref><ns0:ref type='bibr'>2013)</ns0:ref>. <ns0:ref type='bibr'>comparison (1957-2007)</ns0:ref> and fractal analysis of digital panchromatic aerial photographs on the same area <ns0:ref type='bibr' target='#b40'>(Linares, et al. 2006;</ns0:ref><ns0:ref type='bibr'>2009)</ns0:ref>, revealed a process of simultaneous stand densification and expansion of A. pinsapo at the landscape level in the last decades, as a consequence of strict protection since the 1960s of an area, mostly covered at the time by bare soils and open scrublands, with a few sparsely distributed and small stands and isolated trees of A. pinsapo.</ns0:p><ns0:p>Increasing competition due to the densification of these regenerating even-aged stands led to stand stagnation in the 1980s, which acted as a predisposing factor for the climate changeinduced forest decline symptoms reported since 1994-95, associated to a series of very intense drought spells which acted as an inciting factor <ns0:ref type='bibr'>(Linares & Carreira, 2009)</ns0:ref>. Finally, tree growth decline and loss of vigor led to the expansion of the root-rot fungus pathogen Heterobasidion abietinum, <ns0:ref type='bibr' target='#b49'>(Linares, et al., 2010)</ns0:ref> which acted as a contributing factor (Manion, 1981) causing widespread mortality and extensive formation of forest-gaps in the last two decades (> 1/3 of the previous basal area lost). This multifactorial forest decline and dieback process increases the production of HR7 and R4 fuel models, as shown in Fig. <ns0:ref type='figure'>6</ns0:ref>. Under the prevalent 'nomanagement' policy, these new open areas are, eventually, being invaded by dense shrubs, as Manuscript to be reviewed supported by our LIDAR and ForeStereo data, increasing their importance in the fire behavior and promoting fuel models with high fire spread rate as M9. The fuel model classification resulted in a remarkable presence of M9 (Fig. <ns0:ref type='figure'>3</ns0:ref>), covering 30.5% of the study area. This suggests that this invasion process could be taking place and is already in an advanced phase. Also, the CBD values point to a high exposure to crown fires <ns0:ref type='bibr' target='#b5'>(Arellano, et al., 2017)</ns0:ref> and could explain the forecasted high flame length in some areas.</ns0:p><ns0:p>As explained in the introduction section, fire intensity in pinsapo forests is known to be low, but the above-mentioned current invasion process of the mortality gaps by the surrounding dense shrubs could invert this tendency.</ns0:p><ns0:p>However, it must be highlighted that the efficacy of employing fire simulations in risk management strongly depends on input data of high accuracy and precision, due to the complex heterogeneity of forest landscapes (Rodríguez y Silva & <ns0:ref type='bibr' target='#b76'>Molina-Martínez, 2012)</ns0:ref>. Although we precisely determined shrubs composition and structure, in a set of training field plots, the low LIDAR point cloud density available hindered reliable mapping of understory vegetation which thus may restrict the accuracy of the obtained fire risk simulations. The combined use of LIDAR, both terrestrial and airborne, could be the best option to map fuel models and canopy data such as Canopy Base Height and Canopy Bulk Density, for an increased accuracy. Nevertheless, ForeStereo showed to be a relevant alternative to terrestrial LIDAR for calculating stand structure. This work attempts to set a precedent and the first approach to fire risk analysis in Abies pinsapo forests using LIDAR. Also, it stands out the significant potential of this method to study the ecological structure of endangered fir species populations and to broaden the understanding of their conservation status. Most of the current work with LIDAR data focuses on Manuscript to be reviewed forests with commercial interest, but few studies have employed this technology to understand the structure of the populations of endangered species forests and their vulnerability to fire risk.</ns0:p></ns0:div>
<ns0:div><ns0:head n='5.'>CONCLUSIONS</ns0:head><ns0:p>Our results show a high fire risk for the largest remaining continuous forest of the relic and endangered A. pinsapo tree species. Such risk, especially under east wind conditions (Levant), was found to be associated with a remarkable presence of shrub-dominated fuel models (M9).</ns0:p><ns0:p>Using aerial LIDAR and ForeStereo data to assess stand structure spatial variability in the area, we found symptoms of stand stagnation and forest decline under the current no-management conservation policy. This process together with climate change trends triggers the formation of mortality-gaps that are eventually invaded by shrubs, increasing the production of the M9 fuel model which in turn worsen fire risk. These findings stress the need for proactive adaptive management of A. pinsapo forests, including: (i) the creation of bare patches through shrub clearing, (ii) a reinforcement of the firewalls in the west part of the valley and (iii) to promote grazing and trampling levels by wild ungulates (or domestic livestock if they were insufficient) to reduce shrub fuel load without compromising A. pinsapo. We also support the adequacy of thinning treatments for canopy structural diversity enhancement as an essential tool to avoid stand stagnation <ns0:ref type='bibr' target='#b45'>(Linares et al., 2009a)</ns0:ref>, <ns0:ref type='bibr' target='#b47'>(Linares, et al., 2009b)</ns0:ref>, <ns0:ref type='bibr' target='#b35'>(Lechuga, et al., 2017;</ns0:ref><ns0:ref type='bibr'>2019)</ns0:ref> and high CBD values, to reduce the probability of crown fires thus increasing resilience to wildfires <ns0:ref type='bibr'>(Koontz, et al, 2020)</ns0:ref> as well asto reduce climate change-induced tree mortality <ns0:ref type='bibr' target='#b45'>(Linares, et al., 2009a)</ns0:ref>.</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:02:46252:1:1:NEW 26 May 2020)</ns0:p></ns0:div>
<ns0:div><ns0:head>Manuscript to be reviewed</ns0:head><ns0:p>The importance of the A.pinsapo populations in Sierra de las Nieves is one of the main reasons that inspired Spanish national policy to upgrade this protected area into a National Park.</ns0:p><ns0:p>Although the models obtained have a low accuracy due to technical limitations, this study shows a preliminary estimate, a first step to deepen into pinsapo forests risk factors using remote and proximal sensing as essential tools to support conservation management. This method can also be extended to the monitoring of other endangered Western Mediterranean relict fir forests such as those of A. numidica and A. pinsapo marocana in North Africa and implementing it in their conservation strategies.</ns0:p></ns0:div>
<ns0:div><ns0:head n='6.'>ACKNOWLEDGMENTS</ns0:head><ns0:p>We thank Víctor Lechuga and Antonio Román (University of Jaén), for assistance during fieldworks, Fernando Montes Pita, (INIA-CIFOR) for facilitating the ForeStereo deviceessential for the fieldwork-; José Luis López Quintanilla (coordinator of the regional government plan for the restoration and conservation of Abies pinsapo habitats) for providing relevant information of the place about the study area and species. Manuscript to be reviewed </ns0:p></ns0:div><ns0:figure xml:id='fig_0'><ns0:head /><ns0:label /><ns0:figDesc>employed satellite analysis from MODIS PeerJ reviewing PDF | (2020:02:46252:1:1:NEW 26 May 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_1'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:02:46252:1:1:NEW 26 May 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_2'><ns0:head /><ns0:label /><ns0:figDesc>, but adapted for Southern Spain climate conditions through providing hybrid model types that represents fuel treatments and their evolution. Shrub and climate characteristics have shown different behavior between American and Mediterranean fuel models, so parameters such as fuel load or fuelbed depth have to be adjusted (Rodríguez y Silva & Molina Martínez, 2012). PeerJ reviewing PDF | (2020:02:46252:1:1:NEW 26 May 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_3'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:02:46252:1:1:NEW 26 May 2020) Manuscript to be reviewed We found a canopy density (CBD) too high that can increase the risk of harsh crown fires (Arellano 2017). These canopy density values in a well below full-cover area, together with low Ho suggests two possible explanations: (i) the high CBD values correspond to full-cover patches with older stands where gaps are still not open and/or (ii) shrubs strata are increasing their height above 4 m, interlocking with the canopy. Both are compatible with different phases of forest decline and the gap opening process. The low stand height values we found even in the patches with older stands and high cover values could indicate symptoms of stand stagnation in such patches. A multi-temporal</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_4'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:02:46252:1:1:NEW 26 May 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_5'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:02:46252:1:1:NEW 26 May 2020)</ns0:figDesc></ns0:figure>
<ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='39,42.52,204.37,525.00,247.50' type='bitmap' /></ns0:figure>
<ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='42,42.52,178.87,525.00,371.25' type='bitmap' /></ns0:figure>
<ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='43,42.52,178.87,525.00,238.50' type='bitmap' /></ns0:figure>
<ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='44,42.52,229.87,525.00,393.75' type='bitmap' /></ns0:figure>
<ns0:figure type='table' xml:id='tab_0'><ns0:head>Fuel loading (tn/ha) Fuel type 1h 10h 100h LiveH LiveW Moisture of extinction (%) Fuel bed depth (cm)</ns0:head><ns0:label /><ns0:figDesc /><ns0:table><ns0:row><ns0:cell cols='5'>Predominance of shrubs in the fire behaviour</ns0:cell><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>M3</ns0:cell><ns0:cell cols='3'>11.47 2.88 3.37</ns0:cell><ns0:cell>0</ns0:cell><ns0:cell>6.1</ns0:cell><ns0:cell>15</ns0:cell><ns0:cell>82.29</ns0:cell></ns0:row><ns0:row><ns0:cell>M8</ns0:cell><ns0:cell cols='2'>11.23 6.1</ns0:cell><ns0:cell>3.47</ns0:cell><ns0:cell>0</ns0:cell><ns0:cell>7.27</ns0:cell><ns0:cell>30</ns0:cell><ns0:cell>121.92</ns0:cell></ns0:row><ns0:row><ns0:cell>M9</ns0:cell><ns0:cell cols='3'>34.71 9.86 4.93</ns0:cell><ns0:cell>0</ns0:cell><ns0:cell>18.89</ns0:cell><ns0:cell>15</ns0:cell><ns0:cell>182.88</ns0:cell></ns0:row><ns0:row><ns0:cell cols='7'>Pine-needle litter with shrubs and/or grassland under forest canopy</ns0:cell></ns0:row><ns0:row><ns0:cell>HPM4</ns0:cell><ns0:cell cols='3'>17.63 13.23 1.17</ns0:cell><ns0:cell>0</ns0:cell><ns0:cell>11.13</ns0:cell><ns0:cell>20</ns0:cell><ns0:cell>76.2</ns0:cell></ns0:row><ns0:row><ns0:cell cols='7'>Predominance of pine-needle with branches and other canopy debris</ns0:cell></ns0:row><ns0:row><ns0:cell>HR7</ns0:cell><ns0:cell>0.73</ns0:cell><ns0:cell cols='2'>3.76 3.47</ns0:cell><ns0:cell>0</ns0:cell><ns0:cell>0</ns0:cell><ns0:cell>25</ns0:cell><ns0:cell>18.28</ns0:cell></ns0:row><ns0:row><ns0:cell cols='5'>Predominance of canopy debris accumulation</ns0:cell><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>R4</ns0:cell><ns0:cell>1.57</ns0:cell><ns0:cell cols='2'>5.16 6.29</ns0:cell><ns0:cell>0</ns0:cell><ns0:cell>0</ns0:cell><ns0:cell>25</ns0:cell><ns0:cell>82</ns0:cell></ns0:row></ns0:table><ns0:note>PeerJ reviewing PDF | (2020:02:46252:1:1:NEW 26 May 2020)</ns0:note></ns0:figure>
<ns0:figure type='table' xml:id='tab_1'><ns0:head>Table 2 (on next page)</ns0:head><ns0:label>2</ns0:label><ns0:figDesc>Dead fuel moisture conditioningWeather stream file (.WXS) provided to Flammap for conditioning the dead fuel moisture.</ns0:figDesc><ns0:table><ns0:row><ns0:cell>1</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>Date</ns0:cell><ns0:cell>T (ºC)</ns0:cell><ns0:cell>RH (%)</ns0:cell><ns0:cell>PP (mm)</ns0:cell><ns0:cell cols='2'>Wind SP (m/s) Wind dir. (º)</ns0:cell><ns0:cell>Cloud (%)</ns0:cell></ns0:row><ns0:row><ns0:cell>08/27/14 10:00</ns0:cell><ns0:cell>28</ns0:cell><ns0:cell>33</ns0:cell><ns0:cell>0</ns0:cell><ns0:cell>2</ns0:cell><ns0:cell>180</ns0:cell></ns0:row><ns0:row><ns0:cell>08/27/14 11:00</ns0:cell><ns0:cell>29</ns0:cell><ns0:cell>32</ns0:cell><ns0:cell>0</ns0:cell><ns0:cell>2</ns0:cell><ns0:cell>180</ns0:cell></ns0:row><ns0:row><ns0:cell>08/27/14 12:00</ns0:cell><ns0:cell>30</ns0:cell><ns0:cell>29</ns0:cell><ns0:cell>0</ns0:cell><ns0:cell>1</ns0:cell><ns0:cell>180</ns0:cell></ns0:row><ns0:row><ns0:cell>08/27/14 13:00</ns0:cell><ns0:cell>31</ns0:cell><ns0:cell>27</ns0:cell><ns0:cell>0</ns0:cell><ns0:cell>2</ns0:cell><ns0:cell>180</ns0:cell></ns0:row><ns0:row><ns0:cell>08/27/14 14:00</ns0:cell><ns0:cell>30</ns0:cell><ns0:cell>27</ns0:cell><ns0:cell>0</ns0:cell><ns0:cell>2</ns0:cell><ns0:cell>180</ns0:cell></ns0:row><ns0:row><ns0:cell>08/27/14 15:00</ns0:cell><ns0:cell>29</ns0:cell><ns0:cell>28</ns0:cell><ns0:cell>0</ns0:cell><ns0:cell>2</ns0:cell><ns0:cell>180</ns0:cell></ns0:row><ns0:row><ns0:cell>08/27/14 16:00</ns0:cell><ns0:cell>29</ns0:cell><ns0:cell>29</ns0:cell><ns0:cell>0</ns0:cell><ns0:cell>2</ns0:cell><ns0:cell>180</ns0:cell></ns0:row><ns0:row><ns0:cell>08/27/14 17:00</ns0:cell><ns0:cell>28</ns0:cell><ns0:cell>30</ns0:cell><ns0:cell>0</ns0:cell><ns0:cell>1</ns0:cell><ns0:cell>180</ns0:cell></ns0:row><ns0:row><ns0:cell>08/27/14 18:00</ns0:cell><ns0:cell>27</ns0:cell><ns0:cell>31</ns0:cell><ns0:cell>0</ns0:cell><ns0:cell>2</ns0:cell><ns0:cell>180</ns0:cell></ns0:row><ns0:row><ns0:cell>08/27/14 19:00</ns0:cell><ns0:cell>26</ns0:cell><ns0:cell>32</ns0:cell><ns0:cell>0</ns0:cell><ns0:cell>2</ns0:cell><ns0:cell>180</ns0:cell></ns0:row><ns0:row><ns0:cell>2</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>3</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>4</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>5</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>6</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>7</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row></ns0:table><ns0:note>PeerJ reviewing PDF | (2020:02:46252:1:1:NEW 26 May 2020)Manuscript to be reviewed</ns0:note></ns0:figure>
</ns0:body>
" | "Att.: Editor of PeerJ
Dear Dr. Editor,
We thank you for your editorial concern and for finding merit in our manuscript submitted for publication in the journal PeerJ entitled “Using ForeStereo and LIDAR data to assess fire and canopy structure-related risks of Abies pinsapo Boiss forests” (Manuscript-46252). We also thank the reviewers for their insightful comments that have improved the quality of our research.
We here submit a fully revised version of the manuscript after addressing all the major and minor points raised by the reviewers. We have we also made additional changes:
• Canopy Bulk Density (CBD) equation was wrong, we corrected it (Table 3).
• We simplified the number of fuel models from 11 to 6 to have enough validation plots.
• Different fuel moisture settings were set and a fuel moisture conditioning was applied to obtain more realistic simulations.
• We used Grammarly tool to correct misspellings and errors in writing.
Therefore, we obtained new fire scenarios from FlamMap outputs.
We also enclose a letter responding to all points or criticisms- A detailed list of responses to the reviewers’ comments is provided below (note that our responses are in blue characters).
We hope that this new and improved version of the manuscript will reach the standards required for publication in PeerJ.
I look forward to hearing from you.
Sincerely,
A. Cortés-Molino (on behalf of the co-authors)
----------------------------------------------
REVIEWER 1: Hsiao-Lung Pan
Basic reporting
Here are my suggestions:
1. Many punctuation errors, phrase errors, and typos in the article need to be fixed. Free grammar and spelling check tools are available on the internet, they help identify wrong spellings and grammar issues, please go over your manuscript and check again.
DONE. The manuscript has been fully reviewed for punctuation and phrase errors using Grammarly tool.
2. Line 82. I suggest using a stronger mood than “might” to introduce RS application in your study.
DONE. “might” has been removed from the sentence and the role of RS as an essential, not only potential, tool has been clearly stated. See Line 91 of the revised manuscript: “Remote sensing plays a relevant role to study…”
3. Line 125 and 126 cite works of Linares, et al. in 2009, but these are two different articles 2009, please change them into 2009a and 2009b to tell them apart. Another works of Linares, et al. in 2011 also need to change.
DONE. The first work cited is renamed as Linares et al., 2009a and the second as Linares et al., 2009b. The same change for Linares, et al., 2011a and 2011b.
4. Line132. 2.1 study place might be better if it changes into study area or study site.
We keep the reference to “population” as an ecological organization level in the phrasing in lines 132-133 “…in one of the most relevant A. pinsapo population, located in…”, since we want here to stress that A. pinsapo forests at the Study area constitute the larger and one of the most important remnant populations of the species. We have changed the title of section 2.1 from “Study place” to “Study area”
5. Line 139. “…Paeonio broteroi-Abietetum pinsapo Asensi & Rivas-Martínez, 1976 composed…” should be “…Paeonio broteroi-Abietetum pinsapo (Asensi & Rivas-Martínez, 1976) composed…”.
DONE. We added the parenthesis to (Asensi & Rivas-Martínez, 1976). Line 190 of new manuscript version.
6. Line 156. megapixels would be clearer than “Mp”.
DONE. Sentence changed to “The ForeStereo is composed of two Megapixels cameras”
7. Line 151 “1.2 Fieldwork: ForeStereo inventory” should be “2.2 Fieldwork…”
DONE. The numeration was changed.
8. Line 177. The full name of the PONA project is needed.
DONE. We specified the full name of Plan Nacional de Ortofotografía Aérea (PNOA) project.
9. Line 186. Citation of Scott & Burgan (2005) is listed in the reference.
DONE. Scott & Burgan (2005) added to the references (see line 663 of the new manuscript version).
10. Please change the symbology of the lower map in Fig 1 and leave the outline of the boundary, so the image can be seen clearly.
DONE. We removed the blue color and leave only the outline boundary. Also, the firewalls are indicated with white arrows.
11. Table 1. The Italic-type description of fuel model categories needs to align altogether.
DONE. Description of fuel model categories are aligned.
12. Table 2. The LFCC4, AR4, LFCCM, and LFCC0.25 are not explained in the article for its meaning and how to measure them.
We changed the abbreviations and added into the abbreviation’s glossary. Now are: “Percentage of first returns above 4 m (FR4), Percentage of all returns above 4 m (AR4), Percentage of all returns above mean (ARM) and Percentage of first returns above 0.25m (FR0.25)” In lines 289-293 of the new manuscript we add this information. Also, in lines 232-238 is explained how we obtain these data.
13. Fig 3. It will nicer for readers to tell which plots are for training and validation, respectively, if different symbols or colors are applied for two groups of plots.
DONE. We think you refer in fact to Figure 1. Different color has been assigned to training (yellow) and validation (red) plots in the Figure. See new version of Fig.1
14. Fig 4. I see the contrast between east and west wind scenarios, but the difference would be more poping-out if both cases use the same levels from low to high. For example, when mapping the rate of spread, use a classification that encompasses slow speed on the left and fast speed on the right and apply colors from blue to red. The title of “Rate of Spread” on both sides are different. Please change them.
DONE. New fig 4 was obtained with different fire scenarios, due to the changes in the CBD calculation and the reduction of fuel models. The new fig 4 has colors from blue to red and the title “Spread rate” has been changed.
15. The second and third paragraph (line156-165) in 2.2 seems more adhere to the introduction of ForeStereo, so I suggest moving them to line 106 before the introduction of LiDAR.
We made a brief introduction to ForeStereo in 145-149 as an example of hemispherical images device. Then, in 2.2 we considered a wider explanation of ForeStereo specifications and how we worked with it in the field.
16. The last paragraph of 2.1 Study place seems more adhere to the topic of fieldwork, I suggest moving it to the top of 2.2.
DONE. The purpose of the fieldwork was explained before the ForeStereo paragraph in 2.2.
17. The error matrix of fuel model classification should be present in the article or supporting material or supplement.
DONE. The error matrix of the 6 fuel models classification is provided in Table 4.
18. Is there any measure besides cleaning shrubs that can be taken based on the simulation?
Yes, we added this paragraph in lines 427-435: “These findings stress the need for proactive adaptive management of A. pinsapo forests, including: (i) the creation of bare patches through shrub clearing, (ii) a reinforcement of the firewalls in the west part of the valley and (iii) to promote grazing and trampling levels by wild ungulates (or domestic livestock if they were insufficient) to reduce shrub fuel load without compromising A. pinsapo. We also support the adequacy of thinning treatments for canopy structural diversity enhancement as an essential tool to avoid stand stagnation (Linares et al., 2009a), (Linares, et al., 2009b), (Lechuga, et al., 2017; 2019) and high CBD values, in order to reduce the probability of crown fires thus increasing resilience to wildfires (Koontz, et al, 2020) as climate change induced tree mortality (Linares, et al., 2009a).”
Experimental design
What were the metrics acquired by ForeStereo and how were they calibrated with field measurement by investigators? And What were the metrics from LiDAR used in regression? These metrics and procedures are not so clearly stated in the Material and Methods section that readers can’t figure out the analysis process until they read the results. These are especially important for future readers who do not know anything about the ForeStereo and the related procedures.
We now explain the metrics acquired by ForeStereo in lines: 197-199. Due to ForeStereo functioning, no further calibration is needed when using ForeStereo in forest inventory. Geometrical relationships formulations allow ForeStereo to calculate the distance to the tree and diameters at different heights, from which different variables are derived. Therefore, the acquisition of a pair of images is enough for the software to run and obtain calibrated photogrammetric measurements.
In 2.3 is explained how the LIDAR data is modelled with ForeStereo data, also in lines 289-299 LIDAR metrics used are detailed.
It is quite straightforward that the east wind causes worse scenarios because the topography seems to dominate in the fire environment. Will this depart from your result a lot when a different fuel moisture condition is applied? If that is the case, this simplicity (line 195-196) only demonstrates the very dry condition in this area, and if this condition is frequent in the study area, the result is convincing, otherwise, the rationale of this simplicity is needed.
Summer is the season with the highest fire risk. The fuel moisture conditioning applied in the .WXD file (Table 2) is based on the microclimatic data collected in the area by the University of Jaén researchers. These data are coherent with the prevalent summer weather conditions in the area of the study.
Validity of the findings
You mentioned in the line 118-119 that the weak relationship between age and size. Is this because of declining growth? And your finding (in Fig 5) concert with their finding?
A poor relationship between age and size for Abies pinsapo is common in these natural forests where competition intensity is very asymmetric between trees (see Arista, M. 1995. The structure and dynamics of an Abies pinsapo forest in Southern Spain. Forest Ecology and Management 74:81-89). Even within a given stand, relationships between age and size will differ a lot among dominant, codominant and suppressed trees. And of course, in stagnating stand the mean slope of the relationship will be lower than that in non-stagnating stands. The fact that we found a high evenness in Fig. 5 for tree-heights between 6-12 m in these mainly secondary forests that re-growth following strict protection in the 1960s (thus near even-aged stands, putting aside the scattered old trees and some remnant small old stands) is consistent with Arista’s findings.
Please, see the comment on the issue in lines 366 to 375 of the new manuscript version
REVIEWER 2: ANONYMOUS
Basic reporting
See Comments for the authors
Experimental design
See Comments for the authors
Validity of the findings
See Comments for the authors
Comments for the author
GENERAL COMMENTS
The manuscript entitled “Using ForeStereo and LIDAR data to assess fire and canopy structure-related risks of Abies pinsapo Boiss forests” has the goal to combine fieldwork data with LIDAR information to characterize Abies pinsapo forests in a study area of southern Spain and to assess potential fire behavior and risk in the area.
The subject of the manuscript falls within the scope of the journal. Overall, the topic investigated is relevant and of great importance for Mediterranean fire-prone areas, particularly for biodiversity protection and forest management point of views.
The study cannot be considered original, as it combines previously developed methods already published in similar studies; nonetheless, it is an interesting work for the relevance of the Abies forests of the area.
Thanks for your positive comments.
I am not a big expert on ForeStereo and LIDAR data methodologies, so my comments will mostly focus on the fire risk part, in which several points need to be carefully revised or clarified.
The Introduction section should be enriched by adding some sentences and references to previous works that analyzed similar topics.
DONE. See, for instance, lines 110 – 123 in the Introduction of the new manuscript version, where results and references to previous works using a similar methodology have been introduced.
In the present version, the fire risk part is absent from the Introduction. Furthermore, there is need to better focus the Introduction on the main goals of the work (“Using ForeStereo and LIDAR data to assess fire and canopy structure-related risks of Abies pinsapo Boiss forests”), that is some redundant parts or sentences could be omitted or shortened.
We agree with the referee’s comment. We have given more weight to the issue of fire-risk in the Inroduction, since that issue is core to our work (the term “fire risk” appears now nine times in this section). In terms of the topics related with fire risk that have been introduced, we can cite the followings: (i) that fire risk is critical specially regarding the conservation of endangered relic tree-species with narrow distribution ranges, and in the context of ongoing climate change (lines 80-83, and 84-90 of new version), (ii) that fire-risk mapping requires high precision fuel modeling (lines 104-109), which link with the emphasis on using LIDAR data and appropriate calibration with field-based observations (e.g., Forestereo), and (iii) that fire-risk mapping have been performed in previous studies in the Mediterranean basin, and that fuel models specifically calibrated for the Mediterranean basin conditions perform better than other fuel models developed for temperate regions or even the California Mediterranean-type region (see paragraphs starting with line 110 in the new version).
The Material and Methods section has some parts that should be better addressed and/or require additional explanations from the authors. Overall, the fire risk methods are not sufficiently complete to allow replication of the work, and the calibration of the FlamMap model is not evident. In other words, it seems that the authors used a model without any preliminary calibration and validation phase.
The description of the methods used in the fire risk assessment has been enlarged. For instance, the rationale of the UCO4 procedure for fuel models classification has been clarified. The procedure for the parametrization of the FlamMap model has been made explicit (e.g., the model parametrization step consisting on the so-called fuel moisture conditioning was carried out based on local meteorological data)
Flammap is a fire behavior simulation model widely used and accepted, that has been already validated –so it´s why it is widely used- by comparing its forecasted values with fire behavior observed for instance in prescribed-fire assays (see Finney (2006) and Jahdi et al., (2015) Evaluating fire modelling systems in recent wildfires of the Golestan National Park, Iran. Forestry: An International Journal of Forest Research, Volume 89, Issue 2, April 2016, Pages 136–149https://doi.org/10.1093/forestry/cpv045). A new validation phase has no sense in the case of our study, which aims at risk assessment using the Flammap tool rather than on developing the tool itself (which was developed and validated by others). A validation step in our case would require an actual fire in the study area which is just what the risk assessment try to help to avoid. In the eventuality (“God forbid!”) that a wildfire take place in the future in these relic forests, our simulations could then be compared with the real case and, if major mismatches of our model simulations with reality are found, then a report on weaks points and needs for further development and adjustment of the FlamMap model will be generated.
The Results and Discussion sections need to be strengthened and revised and should be improved taking into consideration the Specific Comments below.
The length of the text is not extended (less than 350 rows), and could be reinforced to clarify methods and results, as well as to better focus the Introduction section.
The length of the Results sections has been more than doubled, by introducing more information (e.g., which were the LIDAR metrics that best fit the Forestero data) and specifying quantitative aspects of the simulation results presented in Figure 3 (e.g., burn probability under the Levant and Ponient scenarios, % of the landscape showing spread rate simulation values above a certain threshold, and so on).
Regarding the Discussion section, it has been enlarged by 30%. Our results are now compared more thoughtfully to those reported in other studies addressing a similar issue (e.g., obtained fire spread rates and the main factor controlling them, compared with those in other studies in the Mediterranean region), and the scope of results and derivatives of them have been strengthen (e.g., consequences in terms of increasing fire-risk of the shrub encroachment processes that is being triggered in the area due to climate change-induced tree mortality). On the other hand, some information (results) provided in the Discussion section in the original manuscript has been moved to the Result section.
Please, see also our responses above, related to changes in the Introduction and Method sections.
We hope these changes and major revision of the whole text have improve the quality of the manuscript and the fluency of the text.
The use of English could be improved to increase the text fluency and to remove some minor grammatical wobbles.
DONE. We used the tool Grammarly to check the grammar mistakes, also the Material and Methods section is more detailed now.
I recommend a major revision of the manuscript before publication.
SPECIFIC COMMENTS
L35: Please consider using a more generic description of the methods instead of using “ForeStereo”. Alternatively, please describe in few words the mean characteristics/goals of “ForeStereo”
DONE. A sentence has been added in the abstract section in order to clarify the meaning of Forestereo.
L62-135: The fire risk part is absent in the Introduction section, although the authors used a propagation model to characterize fire risk in the study area. There is need for providing at least the state of the art of the application of FlamMap or other models at the Spanish, European or Mediterranean level, as well as the potential limitations/strengths. In addition, the authors should justify the selection of FlamMap rather than other fire models (e.g.: FARSITE, Wildfire Analyst, etc.). At the EU scale, some suggested works are the following ones: Botequim et al. 2019 (https://www.publish.csiro.au/wf/WF19001), Alcasena et al. 2018 (https://www.sciencedirect.com/science/article/pii/S0301479718311551), Mallinis et al. 2016 (https://www.mdpi.com/1999-4907/7/2/46), Salis et al. 2016 (https://www.publish.csiro.au/wf/WF15081).
Likewise, please mention other previous works that (at least at the Spanish level) used remote sensing/LIDAR data to characterize surface and crown fuels and then inform fire spread/risk models.
DONE. The introduction section has been reorganized; we added a fire risk introduction with a state of the art in lines 104-137 of the new manuscript.
L140: “The forest is placed in a steep valley of 252.59 ha”. I would suggest using “of about 250 ha”.
DONE. The sentenced has changed to “about 250 ha”.
L137-157: The description of the “Study place” (or better “Study Area”?) section should be improved and modified. The second paragraph (L149-157) should be merged with another section, because it contains methodological details rather than a description of the study area. The first paragraph (L137-148) could provide some other details, as for instance the main climate (e.g.: annual rain, winds, maximum temperatures in the summer season) and the fire regime of the study area.
DONE. The name of the 2.1 section was changed to Study area. We also included a brief description of ForeStereo in the introduction section (lines 145-149) and explain in 2.2 the fieldwork and the ForeStereo specifications. Some climate information are provided in 2.1 and a fire event is reported in the south part of the valley (Narváez, 1991).
L192-196: Please clarify in a sentence how the fuel models classification of Scott & Burgan was adapted to Southern Spain climate conditions
DONE. Lines 272-275 provide this clarification
L202-204: Please provide the final resolution of the landscape file
DONE. In line 306 we provide the landscape resolution -10 m- the same as the input raster data.
L204-205: This is something that needs a clear explanation. As the authors know, fuel moisture files are mandatory files for FlamMap simulations, and so it is not possible to run FlamMap without these files. Please provide in a Table the moisture values associated with each fuel model for your simulations.
DONE. Fuel moisture file .FMD is eventually used for the new simulations. Lines 262-274 explains its configuration and we provided Table 2 for the fuel conditioning.
L205-207: Please specify if spot fires were simulated or not. If the answer is yes, please consider to provide the spot probability and spotting delay values used for FlamMap runs. Plus, please specify the simulation time (unlimited per each fire ignition, or how many hours of propagation?)
DONE. In lines 275-282 we added the following explanation: “As model outputs we obtained three different fire scenarios: Burn probability based on 200 random ignition points using MTT algorithm, flame length, and spread rate, both calculated for each cell. The purpose of simulating fire scenarios was to detect vulnerability areas and to assess how exposed is the pinsapo forest to this risk for conservation planning. For this reason, we did not simulate specific events or spotfires using Farsite software. In this point, FlamMap software is more appropriate, since it calculates spread rate and flame length for each landscape cell without a temporal component and use MTT to simulate a number of random fires to detect the probability of a point to get burned [ CITATION Fin06 \l 3082 ] (González-Olabarria, et al., 2012).”
L226-228: The fire simulation results are not described. The reference to Figure 4 is not enough. The next lines (L244-255) are probably more appropriate in the Results part rather than in the Discussion.
DONE. The results (lines 289-326) section is improved to describe the information obtained.
L245-248: The authors reported that “Spread rate shows an average value of 54.85 m∙ min-1 in Levant and 22.85 m∙ min-1 in Ponent winds. More than 50% of the area is affected by a flame length above 30 m in Levant winds, whereas in Ponent winds this is only the 16.1% of the area”. The spread rate values are very high, particularly because wind speed was moderate (16 km h-1). For instance, 54.85 m min-1 corresponds to about 3.3 km h-1, that is a lot. How do the authors explain these values so high? Are these spread rates common in the surrounding mountain areas? The same consideration is valid for flame length, which is on average very high. Please provide your comments on this. Furthermore, please compare your fire behavior results with those reported in similar studies carried out in the Iberic Peninsula or in similar Mediterranean-climate areas, and improve the discussion of the study in the light of this comparison.
DONE. New simulations are provided with a better fuel moisture file, fuel moisture conditioning, a simplification of the fuel models and a corrected CBD model. All these changes resulted in lower values for spread rate and flame length (Fig 4) and we compared our results with other works in the discussion section.
Figures 3 and 4. Please consider using easier classification values in the legend (e.g.: 20 rather than 21.4, 40 rather than 42.9, and so on). This could help readers. In Figure 4, please use the same scales when comparing BP, SR and FL values.
DONE. Figs 3 and 4 were modified with the new simulation outputs and the changes suggested.
" | Here is a paper. Please give your review comments after reading it. |
9,804 | "<ns0:abstract xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:p>In this study we combine information from aerial LIDAR and hemispherical images taken in the field with ForeStereo -a forest inventory device-to assess the vulnerability and to design conservation strategies for endangered Mediterranean fir forests based on the mapping of fire risk and canopy structure spatial variability. We focused on the largest continuous remnant population of the endangered tree species Abies pinsapo Boiss, spanning 252 ha in Sierra de las Nieves National Park (South Andalusia, Spain). We established 49 sampling plots over the study area. Stand structure variables were derived from ForeStereo device, an approximal sensing technology for tree diameter, height and crown dimensions and stand crown cover and basal area retrieval from stereoscopic hemispherical images photogrammetry. With this information, we developed regression models with airborne LIDAR data (spatial resolution of 0.5 points•m -2</ns0:p><ns0:p>). Thereafter, six fuel models were fitted to the plots according to the UCO40 classification criteria, and then the entire area was classified using the Nearest Neighbor algorithm on Sentinel imagery (overall accuracy of 0.56 and a KIA-Kappa Coefficient of 0.46). FlamMap software was used for fire simulation scenarios based on fuel models, stand structure, and terrain data.</ns0:p><ns0:p>Besides the fire simulation, we analyzed canopy structure to assess the status and vulnerability of this fir population. The assessment shows a secondary growth forest that has an increasing presence of fuel models with the potential for high fire spread rate fire and burn probability. Our methodological approach has the potential to be integrated as a support tool for the adaptive management and conservation of A. pinsapo across its whole distribution area (< than 4000 ha), as well as for other endangered circum-Mediterranean fir forests, as. A. numidica de Lannoy and A. pinsapo marocana Trab. in North Africa</ns0:p></ns0:div>
</ns0:abstract>
<ns0:body xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:head n='1.'>INTRODUCTION</ns0:head><ns0:p>Endemic conifer species are more numerous in Mediterranean-type climate regions in the Northern Hemisphere than in the Southern Hemisphere, which has been linked to the selective pressure of cold and/or drought conditions that led to the development of ecophysiological advantages for conifers over angiosperms on oligotrophic soils. Meanwhile, the Mediterraneantype climate regions of South Africa and Southwestern Australia have been more climatically and tectonically stable, which resulted in lower diversity and persistence of ancient lineages of conifers. The Mediterranean Basin has 32 endemic conifer species, accounting for more than 25% of the total conifer flora of 122 species <ns0:ref type='bibr' target='#b74'>(Rundel, 2019)</ns0:ref>. In particular, the genus Abies Mill. experienced extensive speciation from the late Neogene that gave rise to nine species and one natural hybrid in the Mediterranean Basin <ns0:ref type='bibr' target='#b48'>(Linares 2011)</ns0:ref>. Past climate changes have led to population migrations, and to shrinkage and fragmentation of ancestral Mediterranean fir populations, further exacerbated by human impacts. This resulted in circum-Mediterranean endemic firs of high paleogeographic interest, since they are established in relict restricted-range populations with relevant vulnerability to global warming effects <ns0:ref type='bibr' target='#b34'>(Liepelt, et al., 2010)</ns0:ref>.</ns0:p><ns0:p>Adaptive management of these forests to protect them from the increasing fire risk is essential for their survival.</ns0:p><ns0:p>Extreme climate events, such as severe droughts, mega-fires, and disease infestations threaten these relict Mediterranean fir populations <ns0:ref type='bibr' target='#b78'>(Sánchez-Salguero, et al., 2017)</ns0:ref>. It is well known that fire has influenced the landscape and terrestrial life as far back as the beginning of land plants <ns0:ref type='bibr' target='#b7'>(Bowman, et al., 2009;</ns0:ref><ns0:ref type='bibr' target='#b65'>Pausas & Keeley, 2009;</ns0:ref><ns0:ref type='bibr' target='#b24'>He, et al., 2012)</ns0:ref>. Although many conifers have developed adaptive traits to live in fire-prone environments, this is not the case for the genus Abies. The firs developed traits appropriate for the humid areas where they thrive, which has rendered them neither resistant (thin bark) nor resilient (recruitment failure in open spaces) to fires <ns0:ref type='bibr' target='#b19'>(Furyaev, et al., 1983)</ns0:ref> <ns0:ref type='bibr' target='#b84'>(Vega, 1999)</ns0:ref>.</ns0:p><ns0:p>Remote sensing is useful for assessment and development of measures for mitigation of the effects of global warming in relict Mediterranean fir forests. Spectral imagery has been employed for the early detection of forest pathogen infestations <ns0:ref type='bibr' target='#b29'>(Immitzer & Atzberger, 2014)</ns0:ref>, to estimate evapotranspiration <ns0:ref type='bibr' target='#b15'>(Dzikiti, et al., 2019)</ns0:ref>, and to study photosynthetic activity <ns0:ref type='bibr' target='#b14'>(de Sousa, et al., 2017)</ns0:ref>. Meanwhile, 3D point cloud data from laser scanning (LIDAR) have been employed in fire management <ns0:ref type='bibr' target='#b11'>(Chuvieco & Kasischke, 2007)</ns0:ref> and to assess forest volume, biomass <ns0:ref type='bibr' target='#b82'>(Van Ardt, et al., 2008)</ns0:ref>, and canopy structure <ns0:ref type='bibr'>(Adamic, et al., 2017)</ns0:ref> <ns0:ref type='bibr' target='#b61'>(Mura, et al., 2015)</ns0:ref>. Also, the point cloud can be used for ecological purposes, such as assessing light availability for species distribution modeling <ns0:ref type='bibr' target='#b87'>(Wüest, et al., 2020)</ns0:ref> and forest changes in ecotones <ns0:ref type='bibr' target='#b86'>(Wang, et al., 2020)</ns0:ref>. Airborne LIDAR has shown better suitability for mapping crown and canopy heights <ns0:ref type='bibr' target='#b85'>(Wang & Glenn, 2008)</ns0:ref>, although in high density forests the point cloud may not reach the ground, and thus mapping understory vegetation may be inaccurate. However, terrestrial LIDAR has a great potential for estimating shrub and understory biomass, although there are insufficient points for a precise estimation of crown heights when the canopy cover is high <ns0:ref type='bibr' target='#b27'>(Hilker, et al., 2012)</ns0:ref>.</ns0:p><ns0:p>Mapping fire risk with the support of remote sensing tools is becoming essential for landscape planning in the Mediterranean Basin. High-precision fuel moisture and flammability spatial modeling is achieved by combining satellite and meteorological data into radiative transfer models <ns0:ref type='bibr' target='#b10'>(Chuvieco, et al., 2006;</ns0:ref><ns0:ref type='bibr' target='#b89'>Yebra, et al., 2018)</ns0:ref>. Burn probability is then assessed through algorithms such as the Minimum Travel Time (MTT) based on the Huygens' principle <ns0:ref type='bibr' target='#b17'>(Finney, 2002)</ns0:ref>. Several studies have previously applied MTT through FlamMap software on fuel spatial models to assess fire risk in Mediterranean-type ecosystems of Greece <ns0:ref type='bibr' target='#b58'>(Mitsopoulos et al. 2015;</ns0:ref><ns0:ref type='bibr' target='#b53'>Mallinis et al. 2016)</ns0:ref>, Italy <ns0:ref type='bibr' target='#b75'>(Salis et al. 2015)</ns0:ref> and Spain <ns0:ref type='bibr' target='#b59'>(Molina et al., 2017;</ns0:ref><ns0:ref type='bibr'>Alcasena et al. 2019)</ns0:ref>. In this last study, fire risk and highly vulnerable areas were mapped for the whole Catalonia region by applying the <ns0:ref type='bibr' target='#b80'>Scott & Burgan (2005)</ns0:ref> fuel model classification on vegetation structure data and running MTT through FlamMap to obtain 150 m resolution fire scenarios.</ns0:p><ns0:p>Alternatively, fire spread from specific ignition events can be forecasted, for example, <ns0:ref type='bibr' target='#b76'>Salis et al. (2016)</ns0:ref> used the FARSITE software to derive fire spread simulations for several Euro-Mediterranean countries along an east-west gradient. All these studies agree that accurate and customized fuel models are key for assessing burn probability and fire risk.</ns0:p><ns0:p>In this respect, airborne LIDAR technology provides an unprecedented tool for fuel and canopy structure characterization in forest ecosystems. However, several studies highlighted limitations of this technology for accurate understory fuel mapping due to the lack of points reaching the ground <ns0:ref type='bibr' target='#b21'>(González-Olabarria, et al. 2012;</ns0:ref><ns0:ref type='bibr' target='#b6'>Botequim, et al. 2019)</ns0:ref>. Therefore, LIDAR data need to be implemented in regression models supported by field sampling to eventually characterize the forest structure. For this purpose, hemispherical images are an alternative to traditional field sampling. This technique has been used in forest ecology for more than 50 years, but its widespread adoption was limited due to constraints related to image processing capacity PeerJ reviewing PDF | (2020:02:46252:2:1:NEW 28 Aug 2020)</ns0:p><ns0:p>Manuscript to be reviewed <ns0:ref type='bibr'>(Chianucci, 2019)</ns0:ref>. However, technical improvements allowed reducing the time for image processing as well as better image quality acquisition. The widespread proliferation of digital cameras has increased the ease of obtaining and storing hemispherical images, which have become an important tool for fieldwork <ns0:ref type='bibr' target='#b23'>(Hall, et al., 2017)</ns0:ref>. ForeStereo, a forest inventorying device patented by the Forest Research Centre of the Spanish National Institute for Agriculture and Food Research and Technology (CIFOR-INIA), allows one to obtain stand and tree variables in a cost-effective way by processing pairs of stereoscopic hemispherical images taken at a sampling location <ns0:ref type='bibr'>(Rodriguez García et al. 2014)</ns0:ref>.</ns0:p><ns0:p>Most studies applying LIDAR to circum-Mediterranean fir forests have focused on the most widely distributed Abies alba Mill., whereas those focusing on other species such as the relicts A. pinsapo Boiss and A. numidica de Lannoy, which are becoming increasingly vulnerable to global change impacts <ns0:ref type='bibr' target='#b34'>(Liepelt, et al., 2010)</ns0:ref>, are very scant. <ns0:ref type='bibr'>Aragón et al. (2019)</ns0:ref> and <ns0:ref type='bibr' target='#b13'>Cortés-Molino et al. (2017)</ns0:ref> studied A. pinsapo Boiss forests using LIDAR, but only for basic tree identification and vegetation landscape analysis, respectively. Now, the combination of remote sensing technology such as laser scanning and proximal sensing (e.g., ForeStereo) can contribute to the monitoring of these relict forests through the acquisition of high-precision stand structure data.</ns0:p><ns0:p>Abies pinsapo is restricted to a few areas in southern Spain (A. pinsapo pinsapo) and northern Morocco (Abies pinsapo marocana), totaling less than 8000 ha <ns0:ref type='bibr' target='#b38'>(Linares, 2008)</ns0:ref>. Forest fires have markedly reduced the size of populations of this fir; in some localities the longest timespan without fires in the period 1817-1997 was just 34 years <ns0:ref type='bibr' target='#b84'>(Vega, 1999)</ns0:ref>. Thus, fire is considered the most important threat for the conservation and survival of this endangered species <ns0:ref type='bibr' target='#b52'>(López-Quintanilla et al. 2013)</ns0:ref>. A. pinsapo shows a very low resistance to fire due to its thin bark, despite its relatively low fuel flammability and low fire spread rates in dense stands, due to PeerJ reviewing PDF | (2020:02:46252:2:1:NEW 28 Aug 2020)</ns0:p><ns0:p>Manuscript to be reviewed sparse understory and relatively humid conditions (Rodríguez y Silva 1996). Additionally, acute symptoms of tree growth decline and forest dieback due to stand stagnation and climate change have already been reported in some populations <ns0:ref type='bibr'>(Linares & Carreira, 2009)</ns0:ref>, where Pinus halepensis is increasing in abundance, turning previously pure A. pinsapo stands into mixed ones <ns0:ref type='bibr' target='#b50'>(Linares, et al., 2011a)</ns0:ref>.</ns0:p><ns0:p>Our work aimed to combine the use of LIDAR and hemispherical images to study one of the most significant A. pinsapo populations, located in a protected area in Málaga (Spain), to assess vulnerability through (i) mapping fire risk and (ii) analyzing canopy structure variability and its possible links to reported declining growth symptoms.</ns0:p></ns0:div>
<ns0:div><ns0:head n='2.'>MATERIAL AND METHODS</ns0:head></ns0:div>
<ns0:div><ns0:head n='2.1'>Study area</ns0:head><ns0:p>The study location is a steep valley of about 250 ha in area in the municipality of Yunquera, in Sierra de las Nieves National Park (Fig 1 <ns0:ref type='figure'>.</ns0:ref>), in a transition between the upper and lower Mesomediterranean bioclimatic band. The annual rainfall is around 1500 mm and the average daily maximum temperature of the warmest month (August) is 33.6 ºC (S. <ns0:ref type='bibr'>Rivas-Martínez & Rivas-Saenz, 1996</ns0:ref><ns0:ref type='bibr'>-2020)</ns0:ref>. At the southern border of the valley there is a crest that was the limit of a severe wildfire in 1991 that burned 9000 ha <ns0:ref type='bibr' target='#b63'>(Narváez, 1991)</ns0:ref>. The eastern part is bordered by crop fields. This, together with summer weather conditions and significant touristic pressure in Sierra de las Nieves National Park, makes the risk of wildfire especially high. This forest belongs to the Paeonio broteroi-Abietetum pinsapo <ns0:ref type='bibr' target='#b4'>(Asensi & Rivas-Martínez, 1976</ns0:ref>) vegetation association composed mainly of pinsapo fir, forming single-species stands in the upper and shaded parts of the valley. The incidence of the root-rot fungus Heterobasidion abietinum PeerJ reviewing PDF | (2020:02:46252:2:1:NEW 28 Aug 2020)</ns0:p></ns0:div>
<ns0:div><ns0:head>Manuscript to be reviewed</ns0:head><ns0:p>Niemelä & Korhonen is very high <ns0:ref type='bibr' target='#b45'>(Linares, et al., 2010)</ns0:ref>. In sunny and low-altitude spots, the forest includes Pinus halepensis and shrubs of Juniperus spp and Cistus spp.</ns0:p></ns0:div>
<ns0:div><ns0:head n='2.2'>Fieldwork: ForeStereo inventory</ns0:head><ns0:p>The purpose of the fieldwork was to classify local fuel models and collect forest structure data, mainly canopy cover, and crown and stand heights. The valley was sampled in Spring 2018 with 49 plots of 8 m radius, using stereoscopic hemispherical images for obtaining tree metrics such as stand height (Ho), Canopy Base Height (CBH), Canopy Cover (CC), Canopy Bulk Density (CBD) and basal area (G). Shrub cover and height were also assessed by the line-intersect method to support the classification of the fuel models. Each sampling plot was accurately geolocated using a high precision GNSS receiver, supported by an RTK terrestrial station deployed in the upper part of the valley.</ns0:p><ns0:p>Access to field sites was approved by the Andalusian Regional Government (Consejería de Medio Ambiente y Ordenación del Territorio) with the approval code: PNSN/AU/10-2018.</ns0:p><ns0:p>Forest inventory was derived using ForeStereo, a device developed by the Forest Research Centre of the Spanish National Institute for Agriculture and Food Research and Technology (INIA-CIFOR). ForeStereo is equipped with two upward-looking fish-eye cameras. At each sampling location three pairs of hemispherical stereoscopic images (Fig. <ns0:ref type='figure'>2</ns0:ref>) with different exposures are taken. The matching process, compiled in a MatLab® software package, consists of four main steps as detailed in Sánchez-González ( <ns0:ref type='formula'>2016</ns0:ref>): (i) a supervised segmentation of tree stems and crowns; (ii) correspondence of features between the two images and photogrammetric retrieval of tree dimensions; (iii) tree variable modeling and (iv) stand variable estimation, which requires correction of instrumental bias and occlusions <ns0:ref type='bibr'>(Montes, 2019)</ns0:ref>. ForeStereo was used to estimate tree height, crown base height, crown volume and diameter at breast height (DBH) for each tree, number of stems per hectare and crown cover (CC) and stand basal area (G). Tree metrics from the hemispherical images were compared with airborne LIDAR output data to develop regression models. Thirty-two of the plots were randomly chosen to adjust the models, and the remaining seventeen plots were used to assess the models' predictive capability.</ns0:p><ns0:p>Because the geometry of the ForeStereo system and image projections is known, no additional data calibration is needed to carry out photogrammetric retrieval of tree variables. Accuracy of ForeStereo estimated through the Root Mean Squared Error (RMSE) ranges from 0.015 to 0.057 m for DBH <ns0:ref type='bibr' target='#b70'>(Rodríguez-García, et al., 2014;</ns0:ref><ns0:ref type='bibr'>Sánchez-González, et al., 2016)</ns0:ref>, 2.59 to 6.4 m for tree height <ns0:ref type='bibr' target='#b70'>(Rodríguez-García, et al., 2014;</ns0:ref><ns0:ref type='bibr' target='#b56'>Marino et al. 2018</ns0:ref>) and 3.1 and 0.6 m for crown base height and crown diameter respectively <ns0:ref type='bibr' target='#b56'>(Marino et al, 2018)</ns0:ref>, and 11.6 m 2 /ha for G <ns0:ref type='bibr'>(Sánchez-González, 2016)</ns0:ref>.</ns0:p><ns0:p>With the ForeStereo data we were able to estimate stand height at each plot following the Assmann's criteria <ns0:ref type='bibr'>(1970)</ns0:ref>, whereas the canopy cover was obtained directly from the hemispherical images, and the Canopy Base Height (CBH) was averaged for each plot. The calculation of the Canopy Bulk Density (CBD) was based on equations reported by <ns0:ref type='bibr' target='#b72'>Ruiz-Peinado, et al., (2011)</ns0:ref> for A. pinsapo, in which tree height and stem diameter are used to calculate thin branch and needle biomass. Therefore, CBD is estimated by dividing biomass by crown volume (which is obtained from ForeStereo estimates).</ns0:p></ns0:div>
<ns0:div><ns0:head n='2.3'>LIDAR data.</ns0:head><ns0:p>The LIDAR data were obtained in 2015 by the Spanish National Geographic Institute, through the 'Plan Nacional de Ortofotografía Aérea (PNOA)' project. The point cloud density is 0.5 points•m -2 . FUSION software was employed for point cloud processing and data extraction <ns0:ref type='bibr' target='#b57'>(McGaughey & Carson, 2003)</ns0:ref>. A correlation matrix between ForeStereo tree data and all LIDAR metrics obtained with FUSION was useful to detect which LIDAR metric was most suitable to build the regression models in 32 random plots. We applied the models with less rootmean-square error (RMSE) and higher adjusted-R in the remaining 17 plots to predict ForeStereo tree metrics from the LIDAR point cloud. Height break for LIDAR metrics was 4 m, whereas it was 0.25 m for the total vegetation height to avoid high shrubs influence and ground points, respectively. These two height breaks were tested to inquire which one was the better to fit the models.</ns0:p><ns0:p>General canopy structure traits such as canopy cover and height were analyzed from the regression models obtained to detect symptoms of declining growth in this A. pinsapo forest.</ns0:p><ns0:p>These variables were chosen because airborne LIDAR can estimate them accurately <ns0:ref type='bibr'>(Ahmed, et al., 2015;</ns0:ref><ns0:ref type='bibr' target='#b2'>Arumãe & Lang, 2018)</ns0:ref>. Also, we tested whether canopy bulk density was consistent with the results from this analysis.</ns0:p></ns0:div>
<ns0:div><ns0:head n='2.4'>Fuel models and fire scenarios</ns0:head><ns0:p>To simulate fire risk, we first classified field plots according to the UCO40 fuel models, which use specific criteria and traits appropriate for Mediterranean environments and thus perform better than the widely used Prometheus or Rothermel models (Rodríguez y Silva <ns0:ref type='bibr' target='#b68'>& Molina Martínez, 2010;</ns0:ref><ns0:ref type='bibr'>2012)</ns0:ref>. We identified a total of 6 fuel models across the plots (Table <ns0:ref type='table' target='#tab_1'>1</ns0:ref>, Fig. <ns0:ref type='figure'>3</ns0:ref>).</ns0:p><ns0:p>The UCO4 procedure is based on the fuel models classification of <ns0:ref type='bibr' target='#b80'>Scott & Burgan, (2005)</ns0:ref> Manuscript to be reviewed behaviors between American and Mediterranean fuel models, so parameters such as fuel load and fuelbed depth must be adjusted (Rodríguez y Silva <ns0:ref type='bibr' target='#b69'>& Molina Martínez, 2012)</ns0:ref>.</ns0:p><ns0:p>Ecognition® software was used to segment the place of study using the Nearest Neighbor algorithm in an Object-Based Image Analysis (OBIA) <ns0:ref type='bibr' target='#b20'>(Gao, et al., 2007)</ns0:ref>. To carry out this segmentation, we used the raster layers resulting from the regression models validated previously (CBD, CBH, G, Hv) along with NDVI data from Sentinel-2 images (2015) and terrain models obtained from the LIDAR point cloud (topography, aspect, and slope). Later, a confusion matrix was calculated to evaluate the accuracy of the fuel model classification.</ns0:p><ns0:p>The following raster layers generated from the regression models were incorporated: terrain elevation, aspect, slope, Ho, CBH, CBD, CC, and fuel models.</ns0:p><ns0:p>The initial fuel moisture file (.FMD) used was based on <ns0:ref type='bibr' target='#b80'>Scott & Burgan (2005)</ns0:ref> suggestions. We assumed a low moisture content (two-third cured) for the fuel more commonly found in northfacing slopes (fuel models M8, HR7 and R4; see Table <ns0:ref type='table' target='#tab_1'>1</ns0:ref>) and very low moisture content (fully cured) for the fuel models more frequently found in south-facing slopes (M3, M9 and HPM4). A Weather Stream file (.WXS) with typical values of circadian change under summer weather conditions in the area was used for dead fuel moisture conditioning (Table <ns0:ref type='table' target='#tab_3'>2</ns0:ref>). This file modifies initial dead fuel moisture based on changes during a given period in weather variables such as temperature, relative humidity, cloud cover and hourly precipitation. Conditioning also implies adjusting initial dead fuel moisture to site factors (elevation, slope, aspect, and canopy cover), based on the corresponding raster layers previously uploaded in FlamMap <ns0:ref type='bibr' target='#b18'>(Finney, 2006)</ns0:ref>. The .WXS file was built from meteorological data that are continuously recorded in situ by the University of Jaén (values of temperature, precipitation and relative humidity) and from records of the Spanish Meteorological Agency-AEMET (values of wind and cloud cover). We chose the warmest day of 2014 for the conditioning period between the 10:00 to 19:00 hours. The conditioned fuel moistures at the end of this period were the final fuel moistures used for the simulations.</ns0:p><ns0:p>We obtained three different datasets as model outputs: (1) Burn probability based on 200 random ignition points using the MTT algorithm, (2) flame length and (3) flame spread rate, calculated for each cell. The purpose of simulating fire scenarios was to detect vulnerable areas and to assess for conservation planning how exposed the pinsapo forest is to this risk. For this reason, we did not simulate specific events or spotfires using Farsite software. Instead, FlamMap software is more appropriate, because it calculates spread rate and flame length for each landscape cell without a temporal component, and uses MTT to simulate 200 random fires to predict the probability of a point to be burned <ns0:ref type='bibr' target='#b18'>(Finney, 2006</ns0:ref><ns0:ref type='bibr' target='#b21'>) (González-Olabarria, et al., 2012)</ns0:ref>.</ns0:p><ns0:p>The simulations were set under two prevailing wind conditions: west winds (locally called 'Ponent') and east winds (called 'Levant'), both for a typical speed of 13 km•h -1 .</ns0:p></ns0:div>
<ns0:div><ns0:head n='3.'>RESULTS</ns0:head><ns0:p>We found that the LIDAR metrics that best fit the ForeStereo data were (Table <ns0:ref type='table' target='#tab_4'>3)</ns0:ref> Manuscript to be reviewed RMSE less than 20%, while the greatest error was found in modeling the Canopy Base Height (CBH) with an RMSE of 83.3%. This high difference could be due to low point cloud density (minimum 0.5 points/m 2 guaranteed) as well as the fact that airborne LIDAR produces better accuracy for variables related to the top of the canopy <ns0:ref type='bibr' target='#b27'>(Hilker, et al., 2012)</ns0:ref>. Results for dominant height were acceptable (RMSE of 0.55), because finding the top of the crowns with ForeStereo in high-density forests can be difficult.</ns0:p><ns0:p>The error matrix for the fuel model classification using the Nearest Neighbor algorithm shows an overall accuracy of 0.56 and a Kappa Coefficient (KIA) of 0.46 (Table <ns0:ref type='table'>4</ns0:ref>). The most frequent fuel model in the study area was M9 (30.5% of land cover), followed by M3 with 27.4% and M8 with 23%. In all of the fuel models shrubs play a predominant role in the fire behavior. The least frequent fuel models were HPM4 (7.19% of land cover; fire behavior mainly controlled by needle litter together with understory shrubs and/or grasses), HR7 (6.12% of land cover; coniferneedles and branches and other canopy debris play a predominant role) and R4 (5.73% of land cover; predominance of canopy debris accumulation in fire behavior).</ns0:p><ns0:p>Once the corresponding layers were created based on the regression models, the fuel model classifications and the terrain elevation data (Fig. <ns0:ref type='figure'>4</ns0:ref>), fire simulations under two wind conditions were obtained from FlamMap (Fig. <ns0:ref type='figure' target='#fig_4'>5</ns0:ref>). The final landscape file keeps the same pixel resolution of the input data (10 m). We found higher burn probabilities and spread rate under Levant wind conditions, but similar flame length scenarios under both Levant and Ponient winds. Burn probability was higher under Levant winds, with a mean value of 0.078, than under Ponent winds, with a mean value of 0.061. Fire spread rate showed a mean value of 41.82 m• min -1 under Levant wind conditions, and more than 50% of the landscape showed spread rates ≥50 m•min -1 . In contrast, Ponent winds resulted in lower fire spread rates (mean value of 33.34 m• PeerJ reviewing PDF | (2020:02:46252:2:1:NEW 28 Aug 2020)</ns0:p><ns0:p>Manuscript to be reviewed min -1 ) and a considerably lower fraction of the landscape (36%) was affected by spread rates ≥50 m• min -1 . Flame length values were similar for the fire simulations under both wind directions.</ns0:p><ns0:p>Regarding the canopy structure analysis to detect symptoms of forest decline and dieback, the estimated variables CBH and G were not considered due to RMSEs far above 60%. Instead, we use Ho and CC (Fig. <ns0:ref type='figure' target='#fig_5'>6</ns0:ref>), as well as CBD, for such assessment. The following results were estimated only in the landscape cells with a vegetation height above 4 m, in order to exclude non-forest patches of shrubs, as well as forest gaps recently opened due to tree mortality (Fig. <ns0:ref type='figure' target='#fig_6'>7</ns0:ref>).</ns0:p><ns0:p>Canopy heights (Ho) ranged from 4 to 18 m, with a mean value of 9.1 m and a standard deviation (SD) of 3.28. It is remarkable that the mode of canopy heights falls below 5 m in this forest area not affected by fire since the mid-20th century and under long-term, no management policy. On the other hand, canopy height classes between 7 and 12 m showed similar frequencies (high equitability). Lastly, canopy heights higher than 15 m were present, but with rather low frequency.</ns0:p><ns0:p>Canopy cover had a mean value of 64.5% and an SD of 18. We found low frequencies for values between 0-30% (<2% of the area) because most areas with low CC corresponded to land covered by shrubs with less than 4 m heights. Almost 25% of the land showed CC values above 80%, which means that areas with near to full cover are relatively common. Also, 66% of the area had a CC of 40-80%. Lastly, CBD mean value was 0.16 kg• m -3 with a SD of 0.1. Our results showed a high canopy density because >60% of the area had values over 0.1 kg• m -3 and >30% was over 0.3 kg • m -3 .</ns0:p></ns0:div>
<ns0:div><ns0:head n='4.'>DISCUSSION</ns0:head><ns0:p>PeerJ reviewing PDF | (2020:02:46252:2:1:NEW 28 Aug 2020)</ns0:p><ns0:p>Manuscript to be reviewed Endangered circum-Mediterranean firs are highly vulnerable to climate change effects in the isolated areas where they remain <ns0:ref type='bibr' target='#b78'>(Sánchez-Salguero, et al., 2017)</ns0:ref>. Abies nebrodensis Mattei is currently the rarest conifer in the European flora, with only 34 mature trees able to reproduce sexually in the wild <ns0:ref type='bibr' target='#b64'>(Pasta, et al., 2019)</ns0:ref>. The recovery of this species and the protection of the other ones to avoid a similar decrease is an urgent matter that demands the best techniques available to support the traditional field survey.</ns0:p><ns0:p>We propose a methodology that combines the use of LIDAR with ForeStereo, UCO40 fuel models, and FlamMap simulations to significantly reduce the effort and time required for fieldwork, increasing the efficiency of the massive data capture required in forest management.</ns0:p><ns0:p>The application of this methodology in the study area obtained fire simulations that showed that east wind conditions ('Levant') resulted in worse fire scenarios than west winds ('Ponent') as illustrated in Fig. <ns0:ref type='figure' target='#fig_4'>5</ns0:ref>. Spread rate appears to be more influenced by topography and wind conditions <ns0:ref type='bibr' target='#b76'>(Salis, et al., 2016)</ns0:ref> than by flame length, which appears to be more influenced by fuel characteristics. Low spread rate and flame length were found in areas with HR7 and R4 models because they correspond to high-density A. pinsapo stands, consistent with the findings of Rodríguez y Silva <ns0:ref type='bibr'>(1996)</ns0:ref>.</ns0:p><ns0:p>Similar results can be found in the Euro-Mediterranean study of <ns0:ref type='bibr' target='#b76'>Salis et al. (2016)</ns0:ref>, in which the maximum spread rates simulated in the Attica region (Greece), Budoni (Italy), and Fresnedoso de <ns0:ref type='bibr'>Ibor and Navalmoral (Spain)</ns0:ref> are between 50-110 m• min -1 . In their study, the worst flame length scenarios are located in Fresnedoso de Ibor (Spain) and Penteli (Greece) with a range between 25-50 m. They also found a higher spread rate and low flame length mainly in areas with herbaceous vegetation, but also in forest and shrublands in steep mountains exposed to wind PeerJ reviewing PDF | (2020:02:46252:2:1:NEW 28 Aug 2020)</ns0:p><ns0:p>Manuscript to be reviewed (as in our study). In these areas flame length is also higher than in lands with herbaceous vegetation.</ns0:p><ns0:p>However, both wind conditions generate two remarkable foci of fire risk, well highlighted in the burn probability map (Fig <ns0:ref type='figure' target='#fig_4'>5</ns0:ref>). One is in the north-west part of the valley, on south-facing slopes (180º N) where very dense and tall (>1.80 m) patches of the shrub Juniperus spp on steep terrain represent ideal conditions for a high fire spread rate (>50 m•min -1 ), whereas the flame length will depend more on the wind (Ponent >15 m, Levant >30 m). It is not surprising that this condition corresponds to the M9 fuel model (Table <ns0:ref type='table' target='#tab_1'>1</ns0:ref>), in which massive shrub formations dominate the fire behavior. The other focus is in the eastern part of the valley, due to the occurrence of fuel models for which fire behavior is mainly determined by the combination of dense shrub cover and very steep slopes (>75º).</ns0:p><ns0:p>González-Olabarria (2012) observed a lower fire risk landscape in a carefully managed evenaged forest of Pinus nigra. and P. pinaster, whereas our study corresponds to a non-managed forest of pinsapo firs. This contrast strengthens the argument for the urgent need for adaptive management of these endemic fir forests, abandoning the traditional paradigm of nonmanagement in biological conservation. The kind of prevailing fuel model appears to be determinant for the fire scenarios obtained, and the current 'don't touch' management strategy, together with the invasion by shrubs into forest mortality gaps, seem to promote high fire risk fuel models in the area.</ns0:p><ns0:p>The distribution of canopy structure features depicted in Fig. <ns0:ref type='figure' target='#fig_5'>6</ns0:ref> highlights: (i) that the most frequent stand height barely reaches 5 m, the mean value is just 9.1 m and figures higher than 12-15 m are rare despite the fact that A. pinsapo can reach up to 30-35 m in height <ns0:ref type='bibr' target='#b52'>(López-Quintanilla, et al., 2013)</ns0:ref>; (ii) that canopy cover has an average value of 64.5%, well below the full-cover criterion under a 'set aside' and 'no management' strategy since the late 1960s, and patches with 90%-100% cover account for less than 10% of the whole area; and (iii) that there is an overall very high variability for both stand height and canopy cover values across the landscape, with a relatively high evenness in both variable distributions, especially regarding tree height. All these results indicate a lack of old-growth stands in the study area, and the predominance of secondary forests, which is consistent with previous studies based on field surveys <ns0:ref type='bibr' target='#b49'>(Linares, et al, 2011b;</ns0:ref><ns0:ref type='bibr'>2013)</ns0:ref>.</ns0:p><ns0:p>We found considerably high values of canopy density (CBD: Fig. <ns0:ref type='figure'>4</ns0:ref>), which can increase the risk of severe crown fires (Arellano 2017). These canopy density values in a well below full-cover area, together with low Ho suggests two possible explanations: (i) the high CBD values correspond to full-cover patches with older stands where gaps are still not open and/or (ii) shrub strata are increasing their height above 4 m, interlocking with the canopy. Both are compatible with different phases of forest decline and the gap opening process.</ns0:p><ns0:p>The low stand height values we found, even in the patches with older stands and high cover values, could indicate symptoms of stand stagnation in such patches. A multi-temporal comparison and fractal analysis of digital panchromatic aerial photographs of the same area <ns0:ref type='bibr' target='#b36'>(Linares, et al. 2006;</ns0:ref><ns0:ref type='bibr'>2009)</ns0:ref>, revealed a process of simultaneous stand densification and expansion of A. pinsapo at the landscape level in the last decades. This is a consequence of strict protection since the 1960s of an area mostly covered at the time by bare soils and open scrublands, with a few sparsely distributed and small stands and isolated trees of A. pinsapo.</ns0:p><ns0:p>Increasing competition due to the densification of these regenerating even-aged stands led to stand stagnation in the 1980s, which acted as a predisposing factor for the climate change-PeerJ reviewing PDF | (2020:02:46252:2:1:NEW 28 Aug 2020)</ns0:p><ns0:p>Manuscript to be reviewed induced forest decline symptoms reported since 1994-95, associated with a series of very intense drought spells that acted as an inciting factor <ns0:ref type='bibr'>(Linares & Carreira, 2009)</ns0:ref>. Finally, tree growth decline and loss of vigor led to the expansion of the root-rot fungus pathogen Heterobasidion abietinum, <ns0:ref type='bibr' target='#b45'>(Linares, et al., 2010)</ns0:ref>, which acted as a contributing factor (Manion, 1981) causing widespread mortality and extensive formation of forest-gaps in the last two decades (> 1/3 of the previous basal area lost). This multifactorial forest decline and dieback process increases the production of HR7 and R4 fuel models, as shown in Fig. <ns0:ref type='figure' target='#fig_6'>7</ns0:ref>. Under the prevalent 'nomanagement' policy, these new open areas are, eventually, being invaded by dense shrubs, as supported by our LIDAR and ForeStereo data. This increases their importance in the fire behavior and promotes fuel models with high fire spread rate such as the UCO M9 fuel type. The fuel model classification revealed a remarkable contribution of M9 (Fig. <ns0:ref type='figure'>4</ns0:ref>), covering 30.5% of the study area. This suggests that shrub invasion is taking place and is already in an advanced phase. Also, the CBD values point to a high exposure to crown fires <ns0:ref type='bibr' target='#b1'>(Arellano, et al., 2017)</ns0:ref> and could explain the forecasted high flame length in some areas.</ns0:p><ns0:p>As explained in the Introduction section, fire intensity in pinsapo forests is known to be low, but the above-mentioned current invasion into the mortality gaps by the surrounding dense shrubs could invert this tendency.</ns0:p><ns0:p>However, it must be highlighted that the efficacy of employing fire simulations in risk management strongly depends on input data of high accuracy and precision, due to the complex Manuscript to be reviewed both terrestrial and airborne, could be the best option to map fuel models and canopy data such as Canopy Base Height and Canopy Bulk Density, for increased accuracy. Nevertheless, ForeStereo was shown to be a useful alternative to terrestrial LIDAR for calculating stand structure. Our study attempts to set a precedent as the first approach to fire risk analysis in Abies pinsapo forests using LIDAR. Also, it demonstrates the significant potential of this method for study of the ecological structure of populations of endangered fir species, and to broaden the understanding of their conservation status. Most of the current work with LIDAR data focuses on forests with commercial interest, and few studies have employed this technology to understand the structure of the populations of endangered species forests and their vulnerability to fire risk.</ns0:p></ns0:div>
<ns0:div><ns0:head n='5.'>CONCLUSIONS</ns0:head><ns0:p>Our results show a high fire risk for the largest remaining continuous forest of the relict and endangered A. pinsapo tree species. Such risk, especially under east wind conditions (Levant), was found to be associated with a remarkable presence of shrub-dominated fuel models (M9).</ns0:p><ns0:p>Using aerial LIDAR and ForeStereo data to assess stand structure spatial variability in the area, we found symptoms of stand stagnation and forest decline under the current no-management conservation policy. This process together with climate change trends triggers the formation of mortality-gaps that are eventually invaded by shrubs, increasing the production of the M9 fuel model, which in turn worsen fire risk. These findings stress the need for proactive adaptive management of A. pinsapo forests, including: (i) the creation of bare patches through shrub PeerJ reviewing PDF | (2020:02:46252:2:1:NEW 28 Aug 2020)</ns0:p><ns0:p>Manuscript to be reviewed clearing, (ii) a reinforcement of the firewalls in the west part of the valley and (iii) promotion of grazing and trampling levels by wild ungulates (or domestic livestock if they were insufficient) to reduce shrub fuel load without compromising A. pinsapo. We also support the efficacy of thinning treatments for canopy structural diversity enhancement as an essential tool to avoid stand stagnation <ns0:ref type='bibr' target='#b41'>(Linares et al., 2009a)</ns0:ref>, <ns0:ref type='bibr' target='#b43'>(Linares, et al., 2009b)</ns0:ref>, <ns0:ref type='bibr' target='#b31'>(Lechuga, et al., 2017;</ns0:ref><ns0:ref type='bibr'>2019)</ns0:ref> and high CBD values, to reduce the probability of crown fires and thus increase resilience to wildfires <ns0:ref type='bibr'>(Koontz, et al, 2020)</ns0:ref>, as well as to reduce climate change-induced tree mortality <ns0:ref type='bibr' target='#b41'>(Linares, et al., 2009a)</ns0:ref>.</ns0:p><ns0:p>The importance of the A.pinsapo populations in Sierra de las Nieves is one of the main reasons that inspired Spanish national policy to upgrade this protected area into a National Park.</ns0:p><ns0:p>Although the models obtained have low accuracy due to technical limitations, our study provides a preliminary estimate, a first step to assess pinsapo forest risk factors using remote and proximal sensing as essential tools to support conservation management. These methods can also be extended to the monitoring of other endangered Western Mediterranean relict fir forests such as those of A. numidica and A. pinsapo marocana in North Africa and can be implemented in their conservation strategies.</ns0:p></ns0:div>
<ns0:div><ns0:head n='6.'>ACKNOWLEDGMENTS</ns0:head><ns0:p>We thank Víctor Lechuga and Antonio Román (University of Jaén), for assistance during fieldwork, Fernando Montes Pita, (INIA-CIFOR) for facilitating the ForeStereo device, which was essential for the fieldwork, and José Luis López Quintanilla (coordinator of the regional PeerJ reviewing PDF | (2020:02:46252:2:1:NEW 28 Aug 2020)</ns0:p><ns0:p>Manuscript to be reviewed government plan for the restoration and conservation of Abies pinsapo habitats) for providing relevant information about the study area and species.</ns0:p><ns0:p>Dr. Eric C. Henry kindly revised the English style and usage. This research did not receive any specific grant from funding agencies in the public, commercial or not-for-profit sectors.</ns0:p></ns0:div>
<ns0:div><ns0:head n='7.'>REFERENCES</ns0:head><ns0:p>Adamic, M., Diaci, J., <ns0:ref type='bibr'>Rozman, A., & Hladnik, D. (2017)</ns0:ref> Manuscript to be reviewed Manuscript to be reviewed Manuscript to be reviewed Best regression models between remotely sensed LIDAR data and field-based </ns0:p></ns0:div><ns0:figure xml:id='fig_0'><ns0:head /><ns0:label /><ns0:figDesc>, but adapted for southern Spain climate conditions through providing hybrid model types that represent fuel traits and their evolution. Shrub and climate characteristics have shown different PeerJ reviewing PDF | (2020:02:46252:2:1:NEW 28 Aug 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_1'><ns0:head /><ns0:label /><ns0:figDesc>: Percentage of first returns above 4 m (x), Percentage of all returns above 4 m (y), Percentage of all returns above mean (z) and Percentage of first returns above 0.25 m (d). All the significant regression models were obtained with 95% confidence in the seventeen validation plots. Basal Area (G) was the only variable with an acceptable fit using a height break of 0.25 m. The rest of the variables were better modeled above 4 m. Canopy Cover (CC) had the best fitting model with an PeerJ reviewing PDF | (2020:02:46252:2:1:NEW 28 Aug 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_2'><ns0:head /><ns0:label /><ns0:figDesc>heterogeneity of forest landscapes (Rodríguez y Silva & Molina-Martínez, 2012). Although we precisely determined shrub composition and structure in a set of training field plots, the low LIDAR point cloud density available hindered reliable mapping of understory vegetation, which thus may restrict the accuracy of the obtained fire risk simulations. The combined use of LIDAR, PeerJ reviewing PDF | (2020:02:46252:2:1:NEW 28 Aug 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_3'><ns0:head>Figure 1</ns0:head><ns0:label>1</ns0:label><ns0:figDesc>Figure 1</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_4'><ns0:head>Figure 5 Fire</ns0:head><ns0:label>5</ns0:label><ns0:figDesc>Figure 5</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_5'><ns0:head>Figure 6 Frequency</ns0:head><ns0:label>6</ns0:label><ns0:figDesc>Figure 6</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_6'><ns0:head>Figure 7</ns0:head><ns0:label>7</ns0:label><ns0:figDesc>Figure 7</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_7'><ns0:head>Fuel</ns0:head><ns0:label /><ns0:figDesc>model classification obtained following the UCO40 criteria (Rodríguez y Silva & Molina-Martínez, 2012), based on Scott & Burgan (2005). Dead fuel models are classified by the time lag: the time required for the moisture content of a fuel to respond to within 2/3 of the new equilibrium moisture content. Larger diameter fuels have longer time lags, so they respond slower to environmental changes (Anderson 1982). LiveH: Live herbaceous fuel, LiveW: Live wood fuel. Moisture of extinction (%): moisture content that prevents flame from propagating. PeerJ reviewing PDF | (2020:02:46252:2:1:NEW 28 Aug 2020) Manuscript to be reviewed Fuel loading (tn/ha)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_8'><ns0:head /><ns0:label /><ns0:figDesc>Forestereo data, used to map spatial distribution of the main forest structure variables. CC: canopy cover; Ho: stand height; CBH: canopy base height; CBD: canopy bulk density; G: basal area. x: Percentage of LIDAR first returns above 4 m, y: Percentage of all returns above 4 m, z: Percentage of all returns above mean, d: Percentage of first returns above 0.25 m.</ns0:figDesc></ns0:figure>
<ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='37,42.52,224.62,525.00,247.50' type='bitmap' /></ns0:figure>
<ns0:figure type='table' xml:id='tab_0'><ns0:head /><ns0:label /><ns0:figDesc>. Long-term use of uneven-aged silviculture in mixed mountain Dinaric forests: A comparison of old-growth and managed stands. Forestry, 90(2), 279-291. https://doi.org/10.1093/forestry/cpw052 Ahmed, O. S., E., F. S., Wulder, M. A., & White, J. C. (2015). Characterizing stand-level forest canopy cover and height using Landsat time series, samples of airborne LIDAR, and the Random Forest algorithm. ISPRS Journal of Photogrammetry and Remote Sensing, 101, United States Department of Agriculture, Forest Service, Intermountain Forest and Range Experiment Station, Ogden, UT. 26 pAragón, J. F.,</ns0:figDesc><ns0:table><ns0:row><ns0:cell>89-101. https://doi.org/10.1016/j.isprsjprs.2014.11.007</ns0:cell></ns0:row><ns0:row><ns0:cell>Alcasena, F. J., Ager, A. A., Bailey, J. D., Pineda, N., Vega-García, C. (2019). Towards a</ns0:cell></ns0:row><ns0:row><ns0:cell>comprehensive wildfire management strategy for Mediterranean areas: Framework</ns0:cell></ns0:row><ns0:row><ns0:cell>development and implementation in Catalonia, Spain. Journal of Environmental</ns0:cell></ns0:row><ns0:row><ns0:cell>Management. 231 303-320 https://doi.org/10.1016/j.jenvman.2018.10.027</ns0:cell></ns0:row><ns0:row><ns0:cell>Anderson, H.E. (1982). Aids to determining fuel models for estimating fire behavior. General</ns0:cell></ns0:row><ns0:row><ns0:cell>Technical Report INT-122,</ns0:cell></ns0:row></ns0:table></ns0:figure>
<ns0:figure type='table' xml:id='tab_1'><ns0:head>Table 1 (on next page)</ns0:head><ns0:label>1</ns0:label><ns0:figDesc /><ns0:table /></ns0:figure>
<ns0:figure type='table' xml:id='tab_2'><ns0:head>Time lag Fuel type 1h 10h 100h LiveH LiveW Moisture of extinction (%) Fuel bed depth (cm)</ns0:head><ns0:label /><ns0:figDesc /><ns0:table><ns0:row><ns0:cell cols='3'>Predominance of shrubs</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>M3</ns0:cell><ns0:cell cols='2'>11.47 2.88 3.37</ns0:cell><ns0:cell>0</ns0:cell><ns0:cell>6.10</ns0:cell><ns0:cell>15</ns0:cell><ns0:cell>82.29</ns0:cell></ns0:row><ns0:row><ns0:cell>M8</ns0:cell><ns0:cell cols='2'>11.23 6.10 3.47</ns0:cell><ns0:cell>0</ns0:cell><ns0:cell>7.27</ns0:cell><ns0:cell>30</ns0:cell><ns0:cell>121.92</ns0:cell></ns0:row><ns0:row><ns0:cell>M9</ns0:cell><ns0:cell cols='2'>34.71 9.86 4.93</ns0:cell><ns0:cell>0</ns0:cell><ns0:cell>18.89</ns0:cell><ns0:cell>15</ns0:cell><ns0:cell>182.88</ns0:cell></ns0:row><ns0:row><ns0:cell cols='6'>Pine-needle litter with shrubs and/or grassland under forest canopy</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>HPM4</ns0:cell><ns0:cell cols='2'>17.63 13.23 1.17</ns0:cell><ns0:cell>0</ns0:cell><ns0:cell>11.13</ns0:cell><ns0:cell>20</ns0:cell><ns0:cell>76.2</ns0:cell></ns0:row><ns0:row><ns0:cell cols='6'>Predominance of pine-needle with branches and other canopy debris</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>HR7</ns0:cell><ns0:cell>0.73</ns0:cell><ns0:cell>3.76 3.47</ns0:cell><ns0:cell>0</ns0:cell><ns0:cell>0</ns0:cell><ns0:cell>25</ns0:cell><ns0:cell>18.28</ns0:cell></ns0:row><ns0:row><ns0:cell cols='4'>Predominance of canopy debris accumulation</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>R4</ns0:cell><ns0:cell>1.57</ns0:cell><ns0:cell>5.16 6.29</ns0:cell><ns0:cell>0</ns0:cell><ns0:cell>0</ns0:cell><ns0:cell>25</ns0:cell><ns0:cell>82</ns0:cell></ns0:row></ns0:table><ns0:note>PeerJ reviewing PDF | (2020:02:46252:2:1:NEW 28 Aug 2020)</ns0:note></ns0:figure>
<ns0:figure type='table' xml:id='tab_3'><ns0:head>Table 2 (on next page)</ns0:head><ns0:label>2</ns0:label><ns0:figDesc>Dead fuel moisture conditioningWeather stream file (.WXS) showing typical summer weather circadian change in the area, used as input to Flammap software for quantifying the moisture of dead fuel.</ns0:figDesc><ns0:table><ns0:row><ns0:cell>1</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>Date</ns0:cell><ns0:cell>T (ºC)</ns0:cell><ns0:cell>RH (%)</ns0:cell><ns0:cell>PP (mm)</ns0:cell><ns0:cell cols='2'>Wind SP (m/s) Wind dir. (º)</ns0:cell><ns0:cell>Cloud (%)</ns0:cell></ns0:row><ns0:row><ns0:cell>08/27/14 10:00</ns0:cell><ns0:cell>28</ns0:cell><ns0:cell>33</ns0:cell><ns0:cell>0</ns0:cell><ns0:cell>2</ns0:cell><ns0:cell>180</ns0:cell></ns0:row><ns0:row><ns0:cell>08/27/14 11:00</ns0:cell><ns0:cell>29</ns0:cell><ns0:cell>32</ns0:cell><ns0:cell>0</ns0:cell><ns0:cell>2</ns0:cell><ns0:cell>180</ns0:cell></ns0:row><ns0:row><ns0:cell>08/27/14 12:00</ns0:cell><ns0:cell>30</ns0:cell><ns0:cell>29</ns0:cell><ns0:cell>0</ns0:cell><ns0:cell>1</ns0:cell><ns0:cell>180</ns0:cell></ns0:row><ns0:row><ns0:cell>08/27/14 13:00</ns0:cell><ns0:cell>31</ns0:cell><ns0:cell>27</ns0:cell><ns0:cell>0</ns0:cell><ns0:cell>2</ns0:cell><ns0:cell>180</ns0:cell></ns0:row><ns0:row><ns0:cell>08/27/14 14:00</ns0:cell><ns0:cell>30</ns0:cell><ns0:cell>27</ns0:cell><ns0:cell>0</ns0:cell><ns0:cell>2</ns0:cell><ns0:cell>180</ns0:cell></ns0:row><ns0:row><ns0:cell>08/27/14 15:00</ns0:cell><ns0:cell>29</ns0:cell><ns0:cell>28</ns0:cell><ns0:cell>0</ns0:cell><ns0:cell>2</ns0:cell><ns0:cell>180</ns0:cell></ns0:row><ns0:row><ns0:cell>08/27/14 16:00</ns0:cell><ns0:cell>29</ns0:cell><ns0:cell>29</ns0:cell><ns0:cell>0</ns0:cell><ns0:cell>2</ns0:cell><ns0:cell>180</ns0:cell></ns0:row><ns0:row><ns0:cell>08/27/14 17:00</ns0:cell><ns0:cell>28</ns0:cell><ns0:cell>30</ns0:cell><ns0:cell>0</ns0:cell><ns0:cell>1</ns0:cell><ns0:cell>180</ns0:cell></ns0:row><ns0:row><ns0:cell>08/27/14 18:00</ns0:cell><ns0:cell>27</ns0:cell><ns0:cell>31</ns0:cell><ns0:cell>0</ns0:cell><ns0:cell>2</ns0:cell><ns0:cell>180</ns0:cell></ns0:row><ns0:row><ns0:cell>08/27/14 19:00</ns0:cell><ns0:cell>26</ns0:cell><ns0:cell>32</ns0:cell><ns0:cell>0</ns0:cell><ns0:cell>2</ns0:cell><ns0:cell>180</ns0:cell></ns0:row><ns0:row><ns0:cell>2</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>3</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>4</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>5</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>6</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>7</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row></ns0:table><ns0:note>PeerJ reviewing PDF | (2020:02:46252:2:1:NEW 28 Aug 2020) Manuscript to be reviewed PeerJ reviewing PDF | (2020:02:46252:2:1:NEW 28 Aug 2020)</ns0:note></ns0:figure>
<ns0:figure type='table' xml:id='tab_4'><ns0:head>Table 3 (on next page)</ns0:head><ns0:label>3</ns0:label><ns0:figDesc /><ns0:table /></ns0:figure>
</ns0:body>
" | "Att.: Editor of PeerJ
Dear Dr. Editor,
We thank you for your editorial concern in our manuscript to PeerJ entitled “Using ForeStereo and LIDAR data to assess fire and canopy structure-related risks of Abies pinsapo Boiss forests” (Manuscript-46252). We also thank the reviewers for their new comments that have improved the quality of this new version of the manuscript.
We here submit the revised version of the manuscript after addressing the major and minor points raised by the reviewers. Some additional changes are included:
• A new figure to depict every fuel model identified in the plots.
• The picture of the new Fig 7 (former Fig 6) is replaced with a new image which better illustrates the declining process and the shrubs invasion.
We also enclose a letter responding to all points or criticisms. A detailed list of responses to the reviewers’ comments is provided below (note that our responses are in blue characters).
We hope that this new and improved version of the manuscript will reach the standards required for publication in PeerJ.
Sincerely,
A. Cortés-Molino (on behalf of the co-authors)
----------------------------------------------
Editor comments (Cho-ying Huang)
I received two sets of comments from the first round of review. Both of them were thorough, but comments from the anonymous one were more in-depth, and those from the other one were not too much. I returned your revision to both of the original reviewers but failed to obtain one that I was hoping to have. Therefore, I have to send your ms to a third reviewer before making my final decision. The third reviewer raises some critical issues, which I concur with them. On top of those, I feel that the Introduction section is way too lengthy (starting from L108) and should be downsized substantially. Therefore, a 'major revisions' tag is given for this revision.
Thanks for your suggestion. We agree and have substantially reduce the length of the introduction from line 104 and onward.
--------------------------------------------
REVIEWER 1: Hsiao-Lung Pan
For a better manuscript, please check again the punctuation, typo, and redundant words with Grammarly before publishing. Some instances are listed, including but not limited to these.
1. Line 42. 'With this information',' we developed regression models...'
2. Line 49. '...with 'the' potential for high fire spread rate fire and burn probability.'
3. Line 52. circunmediterranean or circum-Mediterranean?
4. Line 77.'...experienced an outstanding speciation...'. Please delete 'an'.
5. Line 79. 'Past climate changes have led to population migrations, and to shrinkage and fragmentation...'
6. Line 93. '...detection of forest infestations (Immitzer & Atzberger, 2014) to estimate evapotranspiration...'
7. Line 97. '... 'the' point cloud...'
All the minor changes have been addressed by using Grammarly and then with deep reading. Moreover, the style and English usage of the full text of the manuscript has been revised by an English-native speaker (who is botanist).
--------------------------------------------
REVIEWER 3: Anonymous
Basic reporting
The article must be use clear, unambiguous, technically correct text.
Done. We checked the full text for a more precise language.
Experimental design
Methods should be described with sufficient information to be reproducible by another investigator.
Done. We checked the Materials & Methods section for this purpose.
Validity of the findings
Speculation is welcome but should be identified by proving.
Done. We make every effort to improve the discussion section, the reasoning on which speculations are based, and have highlighted in each case the figuress and tables in which results are shown that prove or suggest indications supporting speculation.
Comments for the Author
General Comments
This study may contribute to existing literature on 'combine ForeStereo and LIDAR data to assess fire in endangered fir forests'. Research objectives are provided a new procedure to map fire risk and canopy structure spatial variability. The overall level of the paper is good, but there have some questions need to be clearly demonstrated the methods and purpose of the research. The study used ForeStereo and LiDAR data for characterizing the forest structure. The author should be noted that two methods might have measurement error, due to they focused different scale and had unequal variance. There is lack of the results to prove the consistency of two methods in the manuscript. I believe that the MS needs some clarifications and improvements before being considered for publication.
We understand that future research is needed to overcome our technical limitations. We believe that we provided enough information in the text about the accuracy and the limitations of both methods employed:
• Lines 192-197: The accuracy of the ForeStereo based on several previous studies is provided: “As the geometry of the ForeStereo system and image projections is known, no additional data calibration is needed to carry out photogrammetric retrieval of tree variables. Accuracy of ForeStereo estimated through the Root Mean Squared Error (RMSE) ranges from 0.015 to 0.057 m for DBH (Rodríguez-García, et al., 2014; Sánchez-González, et al., 2016), 2.59 to 6.4 m for tree height (Rodríguez-García, et al., 2014; Marino et al. 2018) and 3.1 and 0.6 m for crown base height and crown diameter respectively (Marino et al, 2018) and 11.6 m2/ha for G (Sánchez-González, 2016).”
• Lines 268-269: We mentioned “All the significant regression models were obtained with 95% confidence in the seventeen validation plots”.
• Table 3 has all the percentage of variance explained (R2) and the RMSE of the regression models. We review in Lines 266-276 this table and mention in Lines 271-276: “Canopy Cover (CC) had the best fitting model with an RMSE less than 20%, while the highest error was found in modeling the Canopy Base Height (CBH) with an RMSE of 83.3%. This high difference could be due to low point cloud density (minimum 0.5 points/m2 guaranteed) as well as by the fact that airborne LIDAR produces better accuracy for variables related to the top of the canopy (Hilker, et al., 2012). Results for dominant height was acceptable (RMSE of 0.55), as finding the top of the crowns with ForeStereo in high-density forest can be harsh.”
• We provided an entire confusion matrix in table 4 (cited in Lines 277-278). The overall accuracy and the KIA values are reasonable due to the high heterogeneity of the forest landscape (in contrast with commercial managed forests).
• We discarded CBH and G for the canopy structure analysis due to the high RMSE obtained, as we mentioned in Lines 296-297.
• We also mentioned in lines 392-397 the technical limitations that implies fire simulation: “However, it must be highlighted that the efficacy of employing fire simulations in risk management strongly depends on input data of high accuracy and precision, due to the complex heterogeneity of forest landscapes (Rodríguez y Silva & Molina-Martínez, 2012). Although we precisely determined shrubs composition and structure, in a set of training field plots, the low LIDAR point cloud density available hindered reliable mapping of understory vegetation which thus may restrict the accuracy of the obtained fire risk simulations. The combined use of LIDAR, both terrestrial and airborne, could be the best option to map fuel models and canopy data such as Canopy Base Height and Canopy Bulk Density, for an increased accuracy”
• The purpose of this work is to set a precedent and a first approach to fire risk analysis in Abies pinsapo as said in Lines 400-406: “This work attempts to set a precedent and the first approach to fire risk analysis in Abies pinsapo forests using LIDAR. Also, it stands out the significant potential of this method to study the ecological structure of endangered fir species populations and to broaden the understanding of their conservation status. Most of the current work with LIDAR data focuses on forests with commercial interest, but few studies have employed this technology to understand the structure of the populations of endangered species forests and their vulnerability to fire risk”.
• In Lines 427-430 we highlight again the limitations of our work: “Although the models obtained have a low accuracy due to technical limitations, this study shows a preliminary estimate, a first step to deepen into pinsapo forests risk factors using remote and proximal sensing as essential tools to support conservation management.”
A few additional comments are given below:
1. The author did a lot of analysis works and output results, but did not discussed tables and figures completely, e.g. Fig 3, Fig 4, Fig 5, table 2 etc. I suggest that author don't just report data. The author should be good to have more discussion about the generalization of the research findings.
In the first revision we modified substantially the discussion section, increasing the mentions to tables and figures. We worked more on it in this revision and hope you find it more satisfactory.
2.The figure captions is too simple. e.g. Fig 5, What is A and B?? I will recommend the author should be described figure captions more details and more information.
Done. We follow your suggestion, now the captions have more detailed information.
3.L206-209. The author has sampled 49 plots for obtaining tree metrics and for classifying the fuel models. This number of plots is enough for statistics analysis. The author didn't mention why fixed the train model plots (yellow points) and validation plots (red ones). I will recommend splitting a dataset randomly into training and test datasets divided it into smaller sets for building up and validating a model.
We appreciate your recommendation. The plots are currently splitted randomly into train and validation (mentioned in Line 190-191). During the previous revision process, we made the following changes:
• Canopy Bulk Density (CBD) equation was wrong, we corrected it (Table 3).
• We simplified the number of fuel models from 11 to 6 to have enough validation plots and because the differences between some fuel models are low and LIDAR data cannot recognize these little variations.
• Different fuel moisture settings were set, and a fuel moisture conditioning was applied to obtain more realistic simulations (as required by previous suggestions).
Due to all these changes, we decided to remake all the calculations and fire simulations to ensure there was no more mistakes like we found in CBD and to fit the simplification in the fuel models number. We fixed the number of fuel models but not the splitting of the plots, which was done with ‘random sample’ function of MS Excel.
Please, also note that forestereo plots name are different from the final plot names (we could not use forestereo in no-canopy plots).
4. L277-281. The author used Scott & Burgan (2005) suggestions to distinguish initial fuel moisture. The author didn't mention what meaning of shaded spots (M8, HR7, and R4), very low content (fully cured) and sunny spots (M3, M9 and HPM4).
Done. We rewritten these lines in 239-242 “We assumed a low moisture content (two-third cured) for the fuel more commonly found in north-facing slopes (fuel models M8, HR7 and R4; see Table 1) and very low moisture content (fully cured) for the fuel models more frequently found in south-facing slopes (M3, M9 and HPM4).”
5.L244, The author says the LiDAR pixel size is 0.25 m and the point cloud density is 0.5 points∙m-2 in ms. It means that 1 m2 could not rasterize to 1 pixel. Please check it.
Done. We removed the mention to the pixel size since LIDAR data quality is given by point cloud density.
6. L309-L315, The author showed that results of LIDAR metrics and ForeStereo data. The model fitting is not so good. The Canopy Cover (CC) had the best fitting model with an RMSE less than 20%, but others showed great variance. The results might lead to failure analysis and inference. The limitations and the applicability of this work to other regions should also be discussed.
Done. Please, see the respond to general comments where we detailed all the text citations about our technical limitations.
7. The ms is not easy to read. I will recommend revising the ms structure.
Done. We revised the structure and hope is easier to read now.
Additionally, I've found some minor concerns:
1.Table 1 , Table 1 and Table 3. please unify the digit after the decimal point in the manuscript.
Done. Table 1 data has the digits we found in the referenced source and we found useful to write the R2 with two digits after the decimal point. However, we unified the digit after the decimal for every column so each column values has the same digit.
2. Table 2, Please correct 'tn' to 'ton'
Done; tn is now ton in Table 1.
3. Please unify Kappa Coefficient (L46) and KIA
Done. We checked it and write: “(overall accuracy of 0.56 and a KIA-Kappa Coefficient of 0.46).”
4. The predictive model formula in Table 4 should be use Mathematical equation format.
Done. The previous abbreviations used in the formula (FR4, AR4,…) are replaced by math format in Table 3.
" | Here is a paper. Please give your review comments after reading it. |
9,805 | "<ns0:abstract xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:p>In this study we combine information from aerial LIDAR and hemispherical images taken in the field with ForeStereo -a forest inventory device-to assess the vulnerability and to design conservation strategies for endangered Mediterranean fir forests based on the mapping of fire risk and canopy structure spatial variability. We focused on the largest continuous remnant population of the endangered tree species Abies pinsapo Boiss, spanning 252 ha in Sierra de las Nieves National Park (South Andalusia, Spain). We established 49 sampling plots over the study area. Stand structure variables were derived from ForeStereo device, an approximal sensing technology for tree diameter, height and crown dimensions and stand crown cover and basal area retrieval from stereoscopic hemispherical images photogrammetry. With this information, we developed regression models with airborne LIDAR data (spatial resolution of 0.5 points•m -2</ns0:p><ns0:p>). Thereafter, six fuel models were fitted to the plots according to the UCO40 classification criteria, and then the entire area was classified using the Nearest Neighbor algorithm on Sentinel imagery (overall accuracy of 0.56 and a KIA-Kappa Coefficient of 0.46). FlamMap software was used for fire simulation scenarios based on fuel models, stand structure, and terrain data.</ns0:p><ns0:p>Besides the fire simulation, we analyzed canopy structure to assess the status and vulnerability of this fir population. The assessment shows a secondary growth forest that has an increasing presence of fuel models with the potential for high fire spread rate fire and burn probability. Our methodological approach has the potential to be integrated as a support tool for the adaptive management and conservation of A. pinsapo across its whole distribution area (< than 4000 ha), as well as for other endangered circum-Mediterranean fir forests, as. A. numidica de Lannoy and A. pinsapo marocana Trab. in North Africa</ns0:p></ns0:div>
</ns0:abstract>
<ns0:body xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:head n='1.'>INTRODUCTION</ns0:head><ns0:p>Endemic conifer species are more numerous in Mediterranean-type climate regions in the Northern Hemisphere than in the Southern Hemisphere, which has been linked to the selective pressure of cold and/or drought conditions that led to the development of ecophysiological advantages for conifers over angiosperms on oligotrophic soils. Meanwhile, the Mediterraneantype climate regions of South Africa and Southwestern Australia have been more climatically and tectonically stable, which resulted in lower diversity and persistence of ancient lineages of conifers. The Mediterranean Basin has 32 endemic conifer species, accounting for more than 25% of the total conifer flora of 122 species <ns0:ref type='bibr' target='#b76'>(Rundel, 2019)</ns0:ref>. In particular, the genus Abies Mill. experienced extensive speciation from the late Neogene that gave rise to nine species and one natural hybrid in the Mediterranean Basin <ns0:ref type='bibr' target='#b51'>(Linares 2011)</ns0:ref>. Past climate changes have led to population migrations, and to shrinkage and fragmentation of ancestral Mediterranean fir populations, further exacerbated by human impacts. This resulted in circum-Mediterranean endemic firs of high paleogeographic interest, since they are established in relict restricted-range populations with relevant vulnerability to global warming effects <ns0:ref type='bibr' target='#b38'>(Liepelt, et al., 2010)</ns0:ref>.</ns0:p><ns0:p>Adaptive management of these forests to protect them from the increasing fire risk is essential for their survival.</ns0:p><ns0:p>Extreme climate events, such as severe droughts, mega-fires, and disease infestations threaten these relict Mediterranean fir populations <ns0:ref type='bibr'>(Sánchez-Salguero, et al., 2017)</ns0:ref>. It is well known that fire has influenced the landscape and terrestrial life as far back as the beginning of land plants <ns0:ref type='bibr' target='#b12'>(Bowman, et al., 2009;</ns0:ref><ns0:ref type='bibr' target='#b67'>Pausas & Keeley, 2009;</ns0:ref><ns0:ref type='bibr' target='#b29'>He, et al., 2012)</ns0:ref>. Although many conifers have developed adaptive traits to live in fire-prone environments, this is not the case for the genus Abies. The firs developed traits appropriate for the humid areas where they thrive, which has rendered them neither resistant (thin bark) nor resilient (recruitment failure in open spaces) to fires <ns0:ref type='bibr' target='#b23'>(Furyaev, et al., 1983)</ns0:ref> <ns0:ref type='bibr' target='#b86'>(Vega, 1999)</ns0:ref>.</ns0:p><ns0:p>Remote sensing is useful for assessment and development of measures for mitigation of the effects of global warming in relict Mediterranean fir forests. Spectral imagery has been employed for the early detection of forest pathogen infestations <ns0:ref type='bibr' target='#b33'>(Immitzer & Atzberger, 2014)</ns0:ref>, to estimate evapotranspiration <ns0:ref type='bibr' target='#b19'>(Dzikiti, et al., 2019)</ns0:ref>, and to study photosynthetic activity <ns0:ref type='bibr' target='#b18'>(de Sousa, et al., 2017)</ns0:ref>. Meanwhile, 3D point cloud data from laser scanning (LIDAR) have been employed in fire management <ns0:ref type='bibr' target='#b15'>(Chuvieco & Kasischke, 2007)</ns0:ref> and to assess forest volume, biomass <ns0:ref type='bibr' target='#b84'>(Van Ardt, et al., 2008)</ns0:ref>, and canopy structure <ns0:ref type='bibr' target='#b0'>(Adamic, et al., 2017)</ns0:ref> <ns0:ref type='bibr' target='#b64'>(Mura, et al., 2015)</ns0:ref>. Also, the point cloud can be used for ecological purposes, such as assessing light availability for species distribution modeling <ns0:ref type='bibr' target='#b90'>(Wüest, et al., 2020)</ns0:ref> and forest changes in ecotones <ns0:ref type='bibr' target='#b88'>(Wang, et al., 2020)</ns0:ref>. Airborne LIDAR has shown better suitability for mapping crown and canopy heights <ns0:ref type='bibr' target='#b87'>(Wang & Glenn, 2008)</ns0:ref>, although in high density forests the point cloud may not reach the ground, and thus mapping understory vegetation may be inaccurate. However, terrestrial LIDAR has a great potential for estimating shrub and understory biomass, although there are insufficient points for a precise estimation of crown heights when the canopy cover is high <ns0:ref type='bibr' target='#b31'>(Hilker, et al., 2012)</ns0:ref>.</ns0:p><ns0:p>Mapping fire risk with the support of remote sensing tools is becoming essential for landscape planning in the Mediterranean Basin. High-precision fuel moisture and flammability spatial modeling is achieved by combining satellite and meteorological data into radiative transfer models <ns0:ref type='bibr' target='#b14'>(Chuvieco, et al., 2006;</ns0:ref><ns0:ref type='bibr' target='#b91'>Yebra, et al., 2018)</ns0:ref>. Burn probability is then assessed through algorithms such as the Minimum Travel Time (MTT) based on the Huygens' principle <ns0:ref type='bibr' target='#b21'>(Finney, 2002)</ns0:ref>. Several studies have previously applied MTT through FlamMap software on fuel spatial models to assess fire risk in Mediterranean-type ecosystems of Greece <ns0:ref type='bibr' target='#b61'>(Mitsopoulos et al. 2015;</ns0:ref><ns0:ref type='bibr' target='#b57'>Mallinis et al. 2016)</ns0:ref>, Italy <ns0:ref type='bibr' target='#b77'>(Salis et al. 2015)</ns0:ref> and Spain <ns0:ref type='bibr' target='#b62'>(Molina et al., 2017;</ns0:ref><ns0:ref type='bibr' target='#b2'>Alcasena et al. 2019)</ns0:ref>. In this last study, fire risk and highly vulnerable areas were mapped for the whole Catalonia region by applying the <ns0:ref type='bibr' target='#b82'>Scott & Burgan (2005)</ns0:ref> fuel model classification on vegetation structure data and running MTT through FlamMap to obtain 150 m resolution fire scenarios.</ns0:p><ns0:p>Alternatively, fire spread from specific ignition events can be forecasted, for example, <ns0:ref type='bibr' target='#b78'>Salis et al. (2016)</ns0:ref> used the FARSITE software to derive fire spread simulations for several Euro-Mediterranean countries along an east-west gradient. All these studies agree that accurate and customized fuel models are key for assessing burn probability and fire risk.</ns0:p><ns0:p>In this respect, airborne LIDAR technology provides an unprecedented tool for fuel and canopy structure characterization in forest ecosystems. However, several studies highlighted limitations of this technology for accurate understory fuel mapping due to the lack of points reaching the ground <ns0:ref type='bibr' target='#b25'>(González-Olabarria, et al. 2012;</ns0:ref><ns0:ref type='bibr' target='#b10'>Botequim, et al. 2019)</ns0:ref>. Therefore, LIDAR data need to be implemented in regression models supported by field sampling to eventually characterize the forest structure. For this purpose, hemispherical images are an alternative to traditional field sampling. This technique has been used in forest ecology for more than 50 years, but its widespread adoption was limited due to constraints related to image processing capacity Manuscript to be reviewed <ns0:ref type='bibr'>(Chianucci, 2019)</ns0:ref>. However, technical improvements allowed reducing the time for image processing as well as better image quality acquisition. The widespread proliferation of digital cameras has increased the ease of obtaining and storing hemispherical images, which have become an important tool for fieldwork <ns0:ref type='bibr' target='#b27'>(Hall, et al., 2017)</ns0:ref>. ForeStereo, a forest inventorying device patented by the Forest Research Centre of the Spanish National Institute for Agriculture and Food Research and Technology (CIFOR-INIA), allows one to obtain stand and tree variables in a cost-effective way by processing pairs of stereoscopic hemispherical images taken at a sampling location <ns0:ref type='bibr'>(Rodriguez García et al. 2014)</ns0:ref>.</ns0:p><ns0:p>Most studies applying LIDAR to circum-Mediterranean fir forests have focused on the most widely distributed Abies alba Mill., whereas those focusing on other species such as the relicts A. pinsapo Boiss and A. numidica de Lannoy, which are becoming increasingly vulnerable to global change impacts <ns0:ref type='bibr' target='#b38'>(Liepelt, et al., 2010)</ns0:ref>, are very scant. <ns0:ref type='bibr'>Aragón et al. (2019)</ns0:ref> and <ns0:ref type='bibr' target='#b17'>Cortés-Molino et al. (2017)</ns0:ref> studied A. pinsapo Boiss forests using LIDAR, but only for basic tree identification and vegetation landscape analysis, respectively. Now, the combination of remote sensing technology such as laser scanning and proximal sensing (e.g., ForeStereo) can contribute to the monitoring of these relict forests through the acquisition of high-precision stand structure data.</ns0:p><ns0:p>Abies pinsapo is restricted to a few areas in southern Spain (A. pinsapo pinsapo) and northern Morocco (Abies pinsapo marocana), totaling less than 8000 ha <ns0:ref type='bibr' target='#b42'>(Linares, 2008)</ns0:ref>. Forest fires have markedly reduced the size of populations of this fir; in some localities the longest timespan without fires in the period 1817-1997 was just 34 years <ns0:ref type='bibr' target='#b86'>(Vega, 1999)</ns0:ref>. Thus, fire is considered the most important threat for the conservation and survival of this endangered species <ns0:ref type='bibr' target='#b55'>(López-Quintanilla et al. 2013)</ns0:ref>. A. pinsapo shows a very low resistance to fire due to its thin bark, despite its relatively low fuel flammability and low fire spread rates in dense stands, due to Manuscript to be reviewed sparse understory and relatively humid conditions (Rodríguez y Silva 1996). Additionally, acute symptoms of tree growth decline and forest dieback due to stand stagnation and climate change have already been reported in some populations <ns0:ref type='bibr'>(Linares & Carreira, 2009)</ns0:ref>, where Pinus halepensis is increasing in abundance, turning previously pure A. pinsapo stands into mixed ones <ns0:ref type='bibr' target='#b53'>(Linares, et al., 2011a)</ns0:ref>.</ns0:p><ns0:p>Our work aimed to combine the use of LIDAR and hemispherical images to study one of the most significant A. pinsapo populations, located in a protected area in Málaga (Spain), to assess vulnerability through (i) mapping fire risk and (ii) analyzing canopy structure variability and its possible links to reported declining growth symptoms.</ns0:p></ns0:div>
<ns0:div><ns0:head n='2.'>MATERIAL AND METHODS</ns0:head></ns0:div>
<ns0:div><ns0:head n='2.1'>Study area</ns0:head><ns0:p>The study location is a steep valley of about 250 ha in area in the municipality of Yunquera, in Sierra de las Nieves National Park (Fig 1 <ns0:ref type='figure'>.</ns0:ref>), in a transition between the upper and lower Mesomediterranean bioclimatic band. The annual rainfall is around 1500 mm and the average daily maximum temperature of the warmest month (August) is 33.6 ºC (S. <ns0:ref type='bibr'>Rivas-Martínez & Rivas-Saenz, 1996</ns0:ref><ns0:ref type='bibr'>-2020)</ns0:ref>. At the southern border of the valley there is a crest that was the limit of a severe wildfire in 1991 that burned 9000 ha <ns0:ref type='bibr' target='#b65'>(Narváez, 1991)</ns0:ref>. The eastern part is bordered by crop fields. This, together with summer weather conditions and significant touristic pressure in Sierra de las Nieves National Park, makes the risk of wildfire especially high. This forest belongs to the Paeonio broteroi-Abietetum pinsapo <ns0:ref type='bibr' target='#b8'>(Asensi & Rivas-Martínez, 1976)</ns0:ref> </ns0:p></ns0:div>
<ns0:div><ns0:head>Manuscript to be reviewed</ns0:head><ns0:p>Niemelä & Korhonen is very high <ns0:ref type='bibr' target='#b49'>(Linares, et al., 2010)</ns0:ref>. In sunny and low-altitude spots, the forest includes Pinus halepensis and shrubs of Juniperus spp and Cistus spp.</ns0:p></ns0:div>
<ns0:div><ns0:head n='2.2'>Fieldwork: ForeStereo inventory</ns0:head><ns0:p>The purpose of the fieldwork was to classify local fuel models and collect forest structure data, mainly canopy cover, and crown and stand heights. The valley was sampled in Spring 2018 with 49 plots of 8 m radius (50.27 m 2 ), using stereoscopic hemispherical images for obtaining tree metrics such as stand height (Ho), Canopy Base Height (CBH), Canopy Cover (CC), Canopy Bulk Density (CBD) and basal area (G). Shrub cover and height were also assessed by the lineintersect method to support the classification of the fuel models. Each sampling plot was accurately geolocated using a high precision GNSS receiver, supported by an RTK terrestrial station deployed in the upper part of the valley. We assumed a maximum error of 1m in each plot, due to the difficulty of getting GNSS coverage in high dense canopy.</ns0:p><ns0:p>Access to field sites was approved by the Andalusian Regional Government (Consejería de Medio Ambiente y Ordenación del Territorio) with the approval code: PNSN/AU/10-2018.</ns0:p><ns0:p>Forest inventory was derived using ForeStereo, a device developed by the Forest Research Centre of the Spanish National Institute for Agriculture and Food Research and Technology (INIA-CIFOR). ForeStereo is equipped with two upward-looking fish-eye cameras. At each sampling location three pairs of hemispherical stereoscopic images (Fig. <ns0:ref type='figure'>2</ns0:ref>) with different exposures are taken. The matching process, compiled in a MatLab® software package, consists of four main steps as detailed in Sánchez-González ( <ns0:ref type='formula'>2016</ns0:ref>): (i) a supervised segmentation of tree stems and crowns; (ii) correspondence of features between the two images and photogrammetric retrieval of tree dimensions; (iii) tree variable modeling and (iv) stand variable estimation, which requires correction of instrumental bias and occlusions <ns0:ref type='bibr'>(Montes, 2019)</ns0:ref>. ForeStereo was used to PeerJ reviewing <ns0:ref type='table' target='#tab_3'>PDF | (2020:02:46252:3:0:NEW 15 Sep 2020)</ns0:ref> Manuscript to be reviewed estimate tree height, crown base height, crown volume and diameter at breast height (DBH) for each tree, number of stems per hectare and crown cover (CC) and stand basal area (G). Tree metrics from the hemispherical images were compared with airborne LIDAR output data to develop regression models. Thirty-two of the plots were randomly chosen to adjust the models, and the remaining seventeen plots were used to assess the models' predictive capability.</ns0:p><ns0:p>Because the geometry of the ForeStereo system and image projections is known, no additional data calibration is needed to carry out photogrammetric retrieval of tree variables. Accuracy of ForeStereo estimated through the Root Mean Squared Error (RMSE) ranges from 0.015 to 0.057 m for DBH <ns0:ref type='bibr' target='#b73'>(Rodríguez-García, et al., 2014;</ns0:ref><ns0:ref type='bibr'>Sánchez-González, et al., 2016)</ns0:ref>, 2.59 to 6.4 m for tree height <ns0:ref type='bibr' target='#b73'>(Rodríguez-García, et al., 2014;</ns0:ref><ns0:ref type='bibr' target='#b59'>Marino et al. 2018</ns0:ref>) and 3.1 and 0.6 m for crown base height and crown diameter respectively <ns0:ref type='bibr' target='#b59'>(Marino et al, 2018)</ns0:ref>, and 11.6 m 2 /ha for G <ns0:ref type='bibr'>(Sánchez-González, 2016)</ns0:ref>.</ns0:p><ns0:p>With the ForeStereo data we were able to estimate stand height at each plot following the Assmann's criteria <ns0:ref type='bibr'>(1970)</ns0:ref>, whereas the canopy cover was obtained directly from the hemispherical images, and the Canopy Base Height (CBH) was averaged for each plot. The calculation of the Canopy Bulk Density (CBD) was based on equations reported by <ns0:ref type='bibr' target='#b74'>Ruiz-Peinado, et al., (2011)</ns0:ref> for A. pinsapo, in which tree height and stem diameter are used to calculate thin branch and needle biomass. Therefore, CBD is estimated by dividing biomass by crown volume (which is obtained from ForeStereo estimates).</ns0:p></ns0:div>
<ns0:div><ns0:head n='2.3'>LIDAR data.</ns0:head><ns0:p>The LIDAR data were obtained in 2015 by the Spanish National Geographic Institute, through the 'Plan Nacional de Ortofotografía Aérea (PNOA)' project. The point cloud density is 0.5 points•m -2 . FUSION software was employed for point cloud processing and data extraction <ns0:ref type='bibr' target='#b60'>(McGaughey & Carson, 2003)</ns0:ref>. A correlation matrix between ForeStereo tree data and all LIDAR metrics obtained with FUSION was useful to detect which LIDAR metric was most suitable to build the regression models in 32 random plots. We selected the best correlation results (R<0.5, p-value < 0.05) to test linear, power and exponential regression models. The models with less root-mean-square error (RMSE) and higher adjusted-R in the remaining 17 plots were chosen to predict ForeStereo tree metrics from the LIDAR point cloud. Height break for LIDAR metrics was 4 m, whereas it was 0.25 m for the total vegetation height to avoid high shrubs influence and ground points, respectively. These two height breaks were tested to inquire which one was the better to fit the models.</ns0:p><ns0:p>General canopy structure traits such as canopy cover and height were analyzed from the regression models obtained to detect symptoms of declining growth in this A. pinsapo forest.</ns0:p><ns0:p>These variables were chosen because airborne LIDAR can estimate them accurately <ns0:ref type='bibr' target='#b1'>(Ahmed, et al., 2015;</ns0:ref><ns0:ref type='bibr' target='#b6'>Arumãe & Lang, 2018)</ns0:ref>. Also, we tested whether canopy bulk density was consistent with the results from this analysis.</ns0:p></ns0:div>
<ns0:div><ns0:head n='2.4'>Fuel models and fire scenarios</ns0:head><ns0:p>To simulate fire risk, we first classified field plots according to the UCO40 fuel models, which use specific criteria and traits appropriate for Mediterranean environments and thus perform better than the widely used Prometheus or Rothermel models (Rodríguez y Silva <ns0:ref type='bibr' target='#b70'>& Molina Martínez, 2010;</ns0:ref><ns0:ref type='bibr'>2012)</ns0:ref>. We identified a total of 6 fuel models across the plots (Table <ns0:ref type='table' target='#tab_0'>1</ns0:ref>, Fig. <ns0:ref type='figure'>3</ns0:ref>).</ns0:p><ns0:p>The UCO4 procedure is based on the fuel models classification of <ns0:ref type='bibr' target='#b82'>Scott & Burgan, (2005)</ns0:ref>, but Ecognition® software was used to segment the place of study using the Nearest Neighbor algorithm in an Object-Based Image Analysis (OBIA) <ns0:ref type='bibr' target='#b24'>(Gao, et al., 2007)</ns0:ref>. To carry out this segmentation, we used the raster layers resulting from the regression models validated previously (CBD, CBH, G, Hv) along with NDVI data from Sentinel-2 images (2015) and terrain models obtained from the LIDAR point cloud (topography, aspect, and slope). Later, a confusion matrix was calculated to evaluate the accuracy of the fuel model classification.</ns0:p><ns0:p>The following raster layers generated from the regression models were incorporated: terrain elevation, aspect, slope, Ho, CBH, CBD, CC, and fuel models.</ns0:p><ns0:p>The initial fuel moisture file (.FMD) used was based on <ns0:ref type='bibr' target='#b82'>Scott & Burgan (2005)</ns0:ref> suggestions. We assumed a low moisture content (two-third cured) for the fuel more commonly found in northfacing slopes (fuel models M8, HR7 and R4; see Table <ns0:ref type='table' target='#tab_0'>1</ns0:ref>) and very low moisture content (fully cured) for the fuel models more frequently found in south-facing slopes (M3, M9 and HPM4). A Weather Stream file (.WXS) with typical values of circadian change under summer weather conditions in the area was used for dead fuel moisture conditioning (Table <ns0:ref type='table' target='#tab_2'>2</ns0:ref>). This file modifies initial dead fuel moisture based on changes during a given period in weather variables such as temperature, relative humidity, cloud cover and hourly precipitation. Conditioning also implies adjusting initial dead fuel moisture to site factors (elevation, slope, aspect, and canopy cover), based on the corresponding raster layers previously uploaded in FlamMap <ns0:ref type='bibr' target='#b22'>(Finney, 2006)</ns0:ref>. The Manuscript to be reviewed .WXS file was built from meteorological data that are continuously recorded in situ by the University of Jaén (values of temperature, precipitation and relative humidity) and from records of the Spanish Meteorological Agency-AEMET (values of wind and cloud cover). We chose the warmest day of 2014 (available recorded data) for the conditioning period between the 10:00 to 19:00 hours without precipitations or any cloud cover. The conditioned fuel moistures at the end of this period were the final fuel moistures used for the simulations.</ns0:p><ns0:p>We obtained three different datasets as model outputs: (1) Burn probability based on 200 random ignition points using the MTT algorithm, (2) flame length and (3) flame spread rate, calculated for each cell. The purpose of simulating fire scenarios was to detect vulnerable areas and to assess for conservation planning how exposed the pinsapo forest is to this risk. For this reason, we did not simulate specific events or spotfires using Farsite software. Instead, FlamMap software is more appropriate, because it calculates spread rate and flame length for each landscape cell without a temporal component, and uses MTT to simulate 200 random fires to predict the probability of a point to be burned <ns0:ref type='bibr' target='#b22'>(Finney, 2006</ns0:ref><ns0:ref type='bibr' target='#b25'>) (González-Olabarria, et al., 2012)</ns0:ref>.</ns0:p><ns0:p>The simulations were set under two prevailing wind conditions: west winds (locally called 'Ponent') and east winds (called 'Levant'), both for a typical speed of 13 km•h -1 .</ns0:p></ns0:div>
<ns0:div><ns0:head n='3.'>RESULTS</ns0:head><ns0:p>We found that the LIDAR metrics that best fit the ForeStereo data were (Table <ns0:ref type='table' target='#tab_3'>3</ns0:ref>): Percentage of first returns above 4 m (x), Percentage of all returns above 4 m (y), Percentage of all returns above mean (z) and Percentage of first returns above 0.25 m (d). All the significant regression models were obtained with 95% confidence in the seventeen validation plots. Basal Area (G) was the only variable with an acceptable fit using a height break of 0.25 m. The rest of the variables were better modeled above 4 m. Canopy Cover (CC) had the best fitting model with an RMSE less than 20%, while the greatest error was found in modeling the Canopy Base Height (CBH) with an RMSE of 83.3%. This high difference could be due to low point cloud density (minimum 0.5 points•m -2 guaranteed) as well as the fact that airborne LIDAR produces better accuracy for variables related to the top of the canopy <ns0:ref type='bibr' target='#b31'>(Hilker, et al., 2012)</ns0:ref>, such as canopy cover or canopy height, than variables under the canopy. Results for dominant height were acceptable (RMSE of 0.55), because finding the top of the crowns with ForeStereo in high-density forests can be difficult.</ns0:p><ns0:p>The error matrix for the fuel model classification using the Nearest Neighbor algorithm shows an overall accuracy of 0.56 and a Kappa Coefficient (KIA) of 0.46 (Table <ns0:ref type='table'>4</ns0:ref>). The most frequent fuel model in the study area was M9 (30.5% of land cover), followed by M3 with 27.4% and M8 with 23%. In all the fuel models, shrubs play a predominant role in the fire behavior. The least frequent fuel models were HPM4 (7.19% of land cover; fire behavior mainly controlled by needle litter together with understory shrubs and/or grasses), HR7 (6.12% of land cover; coniferneedles and branches and other canopy debris play a predominant role) and R4 (5.73% of land cover; predominance of canopy debris accumulation in fire behavior).</ns0:p><ns0:p>Once the corresponding layers were created based on the regression models, the fuel model classifications and the terrain elevation data (Fig. <ns0:ref type='figure'>4</ns0:ref>), fire simulations under two wind conditions were obtained from FlamMap (Fig. <ns0:ref type='figure' target='#fig_11'>5</ns0:ref>). The final landscape file keeps the same pixel resolution of the input data (10 m). We found higher burn probabilities and spread rate under Levant wind conditions, but similar flame length scenarios under both Levant and Ponient winds. Burn probability was higher under Levant winds, with a mean value of 0.078, than under Ponent Manuscript to be reviewed winds, with a mean value of 0.061. Fire spread rate showed a mean value of 41.82 m• min -1 under Levant wind conditions, and more than 50% of the landscape showed spread rates ≥50 m•min -1 . In contrast, Ponent winds resulted in lower fire spread rates (mean value of 33.34 m• min -1 ) and a considerably lower fraction of the landscape (36%) was affected by spread rates ≥50 m• min -1 . Flame length values were similar for the fire simulations under both wind directions.</ns0:p><ns0:p>Regarding the canopy structure analysis to detect symptoms of forest decline and dieback, the estimated variables CBH and G were not considered due to RMSEs far above 60%. Instead, we use Ho and CC (Fig. <ns0:ref type='figure' target='#fig_12'>6</ns0:ref>), as well as CBD, for such assessment. The following results were estimated only in the landscape cells with a vegetation height above 4 m, in order to exclude non-forest patches of shrubs, as well as forest gaps recently opened due to tree mortality (Fig. <ns0:ref type='figure' target='#fig_13'>7</ns0:ref>).</ns0:p><ns0:p>Canopy heights (Ho) ranged from 4 to 18 m, with a mean value of 9.1 m and a standard deviation (SD) of 3.28. It is remarkable that the mode of canopy heights falls below 5 m in this forest area not affected by fire since the mid-20th century and under long-term, no management policy. On the other hand, canopy height classes between 7 and 12 m showed similar frequencies (high equitability). Lastly, canopy heights higher than 15 m were present, but with rather low frequency.</ns0:p><ns0:p>Canopy cover had a mean value of 64.5% and an SD of 18. We found low frequencies for values between 0-30% (<2% of the area) because most areas with low CC corresponded to land covered by shrubs with less than 4 m heights. Almost 25% of the land showed CC values above 80%, which means that areas with near to full cover are relatively common. Also, 66% of the area had a CC of 40-80%. Lastly, CBD mean value was 0.16 kg• m -3 with a SD of 0.1. Our results showed a high canopy density because >60% of the area had values over 0.1 kg• m -3 and >30% was over 0.3 kg • m -3 .</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:02:46252:3:0:NEW 15 Sep 2020)</ns0:p><ns0:p>Manuscript to be reviewed</ns0:p></ns0:div>
<ns0:div><ns0:head n='4.'>DISCUSSION</ns0:head><ns0:p>Endangered circum-Mediterranean firs are highly vulnerable to climate change effects in the isolated areas where they remain <ns0:ref type='bibr'>(Sánchez-Salguero, et al., 2017)</ns0:ref>. Abies nebrodensis Mattei is currently the rarest conifer in the European flora, with only 34 mature trees able to reproduce sexually in the wild <ns0:ref type='bibr' target='#b66'>(Pasta, et al., 2019)</ns0:ref>. The recovery of this species and the protection of the other ones to avoid a similar decrease is an urgent matter that demands the best techniques available to support the traditional field survey.</ns0:p><ns0:p>We propose a methodology that combines the use of LIDAR with ForeStereo, UCO40 fuel models, and FlamMap simulations to significantly reduce the effort and time required for fieldwork, increasing the efficiency of the massive data capture required in forest management.</ns0:p><ns0:p>The application of this methodology in the study area obtained fire simulations that showed that east wind conditions ('Levant') resulted in worse fire scenarios than west winds ('Ponent') as illustrated in Fig. <ns0:ref type='figure' target='#fig_11'>5</ns0:ref>. Spread rate appears to be more influenced by topography and wind conditions <ns0:ref type='bibr' target='#b78'>(Salis, et al., 2016)</ns0:ref> than by flame length, which appears to be more influenced by fuel characteristics. Low spread rate and flame length were found in areas with HR7 and R4 models because they correspond to high-density A. pinsapo stands, consistent with the findings of Rodríguez y Silva <ns0:ref type='bibr'>(1996)</ns0:ref>.</ns0:p><ns0:p>Similar results can be found in the Euro-Mediterranean study of <ns0:ref type='bibr' target='#b78'>Salis et al. (2016)</ns0:ref>, in which the maximum spread rates simulated in the Attica region (Greece), Budoni (Italy), and Fresnedoso de <ns0:ref type='bibr'>Ibor and Navalmoral (Spain)</ns0:ref> are between 50-110 m• min -1 . In their study, the worst flame length scenarios are located in Fresnedoso de Ibor (Spain) and Penteli (Greece) with a range Manuscript to be reviewed between 25-50 m. They also found a higher spread rate and low flame length mainly in areas with herbaceous vegetation, but also in forest and shrublands in steep mountains exposed to wind (as in our study). In these areas flame length is also higher than in lands with herbaceous vegetation.</ns0:p><ns0:p>However, both wind conditions generate two remarkable foci of fire risk, well highlighted in the burn probability map (Fig <ns0:ref type='figure' target='#fig_11'>5</ns0:ref>). One is in the north-west part of the valley, on south-facing slopes (180º N) where very dense and tall (>1.80 m) patches of the shrub Juniperus spp on steep terrain represent ideal conditions for a high fire spread rate (>50 m•min -1 ), whereas the flame length will depend more on the wind (Ponent >15 m, Levant >30 m). It is not surprising that this condition corresponds to the M9 fuel model (Table <ns0:ref type='table' target='#tab_0'>1</ns0:ref>), in which massive shrub formations dominate the fire behavior. The other focus is in the eastern part of the valley, due to the occurrence of fuel models for which fire behavior is mainly determined by the combination of dense shrub cover and very steep slopes (>75º).</ns0:p><ns0:p>González-Olabarria (2012) observed a lower fire risk landscape in a carefully managed evenaged forest of Pinus nigra. and P. pinaster, whereas our study corresponds to a non-managed forest of pinsapo firs. This contrast strengthens the argument for the urgent need for adaptive management of these endemic fir forests, abandoning the traditional paradigm of nonmanagement in biological conservation. The kind of prevailing fuel model appears to be determinant for the fire scenarios obtained, and the current 'don't touch' management strategy, together with the invasion by shrubs into forest mortality gaps, seem to promote high fire risk fuel models in the area.</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:02:46252:3:0:NEW 15 Sep 2020)</ns0:p></ns0:div>
<ns0:div><ns0:head>Manuscript to be reviewed</ns0:head><ns0:p>The distribution of canopy structure features depicted in Fig. <ns0:ref type='figure' target='#fig_12'>6</ns0:ref> highlights: (i) that the most frequent stand height barely reaches 5 m, the mean value is just 9.1 m and figures higher than 12-15 m are rare despite the fact that A. pinsapo can reach up to 30-35 m in height <ns0:ref type='bibr' target='#b55'>(López-Quintanilla, et al., 2013)</ns0:ref>; (ii) that canopy cover has an average value of 64.5%, well below the full-cover criterion under a 'set aside' and 'no management' strategy since the late 1960s, and patches with 90%-100% cover account for less than 10% of the whole area; and (iii) that there is an overall very high variability for both stand height and canopy cover values across the landscape, with a relatively high evenness in both variable distributions, especially regarding tree height. All these results indicate a lack of old-growth stands in the study area, and the predominance of secondary forests, which is consistent with previous studies based on field surveys <ns0:ref type='bibr' target='#b52'>(Linares, et al, 2011b;</ns0:ref><ns0:ref type='bibr'>2013)</ns0:ref>.</ns0:p><ns0:p>We found considerably high values of canopy density (CBD: Fig. <ns0:ref type='figure'>4</ns0:ref>), which can increase the risk of severe crown fires (Arellano 2017). These canopy density values in a well below full-cover area, together with low Ho suggests two possible explanations: (i) the high CBD values correspond to full-cover patches with older stands where gaps are still not open and/or (ii) shrub strata are increasing their height above 4 m, interlocking with the canopy. Both are compatible with different phases of forest decline and the gap opening process.</ns0:p><ns0:p>The low stand height values we found, even in the patches with older stands and high cover values, could indicate symptoms of stand stagnation in such patches. A multi-temporal comparison and fractal analysis of digital panchromatic aerial photographs of the same area <ns0:ref type='bibr' target='#b41'>(Linares, et al. 2006;</ns0:ref><ns0:ref type='bibr'>2009)</ns0:ref>, revealed a process of simultaneous stand densification and expansion of A. pinsapo at the landscape level in the last decades. This is a consequence of Manuscript to be reviewed scrublands, with a few sparsely distributed and small stands and isolated trees of A. pinsapo.</ns0:p><ns0:p>Increasing competition due to the densification of these regenerating even-aged stands led to stand stagnation in the 1980s, which acted as a predisposing factor for the climate changeinduced forest decline symptoms reported since 1994-95, associated with a series of very intense drought spells that acted as an inciting factor <ns0:ref type='bibr'>(Linares & Carreira, 2009)</ns0:ref>. Finally, tree growth decline and loss of vigor led to the expansion of the root-rot fungus pathogen Heterobasidion abietinum, <ns0:ref type='bibr' target='#b49'>(Linares, et al., 2010)</ns0:ref>, which acted as a contributing factor <ns0:ref type='bibr'>(Manion, 1981)</ns0:ref> causing widespread mortality and extensive formation of forest-gaps in the last two decades (> 1/3 of the previous basal area lost). This multifactorial forest decline and dieback process increases the production of HR7 and R4 fuel models, as shown in Fig. <ns0:ref type='figure' target='#fig_13'>7</ns0:ref>. Under the prevalent 'nomanagement' policy, these new open areas are, eventually, being invaded by dense shrubs, as supported by our LIDAR and ForeStereo data. This increases their importance in the fire behavior and promotes fuel models with high fire spread rate such as the UCO M9 fuel type. The fuel model classification revealed a remarkable contribution of M9 (Fig. <ns0:ref type='figure'>4</ns0:ref>), covering 30.5% of the study area. This suggests that shrub invasion is taking place and is already in an advanced phase. Also, the CBD values point to a high exposure to crown fires <ns0:ref type='bibr' target='#b5'>(Arellano, et al., 2017)</ns0:ref> and could explain the forecasted high flame length in some areas.</ns0:p><ns0:p>As explained in the Introduction section, fire intensity in pinsapo forests is known to be low, but the above-mentioned current invasion into the mortality gaps by the surrounding dense shrubs could invert this tendency.</ns0:p><ns0:p>However, it must be highlighted that the efficacy of employing fire simulations in risk management strongly depends on input data of high accuracy and precision, due to the complex heterogeneity of forest landscapes (Rodríguez y Silva & <ns0:ref type='bibr' target='#b71'>Molina-Martínez, 2012)</ns0:ref>. Although we Manuscript to be reviewed precisely determined shrub composition and structure in a set of training field plots, the low LIDAR point cloud density available hindered reliable mapping of understory vegetation, which thus may restrict the accuracy of the obtained fire risk simulations. The combined use of LIDAR, both terrestrial and airborne, could be the best option to map fuel models and canopy data such as Canopy Base Height and Canopy Bulk Density, for increased accuracy. Nevertheless, ForeStereo was shown to be a useful alternative to terrestrial LIDAR for calculating stand structure. Our study attempts to set a precedent as the first approach to fire risk analysis in Abies pinsapo forests using LIDAR. Also, it demonstrates the significant potential of this method for study of the ecological structure of populations of endangered fir species, and to broaden the understanding of their conservation status. Most of the current work with LIDAR data focuses on forests with commercial interest, and few studies have employed this technology to understand the structure of the populations of endangered species forests and their vulnerability to fire risk.</ns0:p></ns0:div>
<ns0:div><ns0:head n='5.'>CONCLUSIONS</ns0:head><ns0:p>Our results show a high fire risk for the largest remaining continuous forest of the relict and endangered A. pinsapo tree species. Such risk, especially under east wind conditions (Levant), was found to be associated with a remarkable presence of shrub-dominated fuel models (M9).</ns0:p><ns0:p>Using aerial LIDAR and ForeStereo data to assess stand structure spatial variability in the area, we found symptoms of stand stagnation and forest decline under the current no-management conservation policy. This process together with climate change trends triggers the formation of Manuscript to be reviewed mortality-gaps that are eventually invaded by shrubs, increasing the production of the M9 fuel model, which in turn worsen fire risk. These findings stress the need for proactive adaptive management of A. pinsapo forests, including: (i) the creation of bare patches through shrub clearing, (ii) a reinforcement of the firewalls in the west part of the valley and (iii) promotion of grazing and trampling levels by wild ungulates (or domestic livestock if they were insufficient) to reduce shrub fuel load without compromising A. pinsapo. We also support the efficacy of thinning treatments for canopy structural diversity enhancement as an essential tool to avoid stand stagnation <ns0:ref type='bibr' target='#b45'>(Linares et al., 2009a)</ns0:ref>, <ns0:ref type='bibr' target='#b47'>(Linares, et al., 2009b)</ns0:ref>, <ns0:ref type='bibr' target='#b35'>(Lechuga, et al., 2017;</ns0:ref><ns0:ref type='bibr'>2019)</ns0:ref> and high CBD values, to reduce the probability of crown fires and thus increase resilience to wildfires <ns0:ref type='bibr'>(Koontz, et al, 2020)</ns0:ref>, as well as to reduce climate change-induced tree mortality <ns0:ref type='bibr' target='#b45'>(Linares, et al., 2009a)</ns0:ref>.</ns0:p><ns0:p>The importance of the A.pinsapo populations in Sierra de las Nieves is one of the main reasons that inspired Spanish national policy to upgrade this protected area into a National Park.</ns0:p><ns0:p>Although the models obtained have low accuracy due to technical limitations, our study provides a preliminary estimate, a first step to assess pinsapo forest risk factors using remote and proximal sensing as essential tools to support conservation management. These methods can also be extended to the monitoring of other endangered Western Mediterranean relict fir forests such as those of A. numidica and A. pinsapo marocana in North Africa and can be implemented in their conservation strategies.</ns0:p></ns0:div>
<ns0:div><ns0:head n='6.'>ACKNOWLEDGMENTS</ns0:head><ns0:p>PeerJ reviewing PDF | (2020:02:46252:3:0:NEW 15 Sep 2020)</ns0:p><ns0:p>Manuscript to be reviewed Manuscript to be reviewed Dead fuel models are classified by the time lag: the time required for the moisture content of a fuel to respond to within 2/3 of the new equilibrium moisture content. Larger diameter fuels have longer time lags, so they respond slower to environmental changes <ns0:ref type='bibr' target='#b4'>(Anderson 1982)</ns0:ref>. Manuscript to be reviewed Manuscript to be reviewed </ns0:p></ns0:div><ns0:figure xml:id='fig_0'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:02:46252:3:0:NEW 15 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_1'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:02:46252:3:0:NEW 15 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_2'><ns0:head /><ns0:label /><ns0:figDesc>vegetation association composed mainly of pinsapo fir, forming single-species stands in the upper and shaded parts of the valley. The incidence of the root-rot fungus Heterobasidion abietinum PeerJ reviewing PDF | (2020:02:46252:3:0:NEW 15 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_3'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:02:46252:3:0:NEW 15 Sep 2020) Manuscript to be reviewed adapted for southern Spain climate conditions through providing hybrid model types that represent fuel traits and their evolution. Shrub and climate characteristics have shown different behaviors between American and Mediterranean fuel models, so parameters such as fuel load and fuelbed depth must be adjusted (Rodríguez y Silva & Molina Martínez, 2012).</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_4'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:02:46252:3:0:NEW 15 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_5'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:02:46252:3:0:NEW 15 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_6'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:02:46252:3:0:NEW 15 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_7'><ns0:head /><ns0:label /><ns0:figDesc>strict protection since the 1960s of an area mostly covered at the time by bare soils and open PeerJ reviewing PDF | (2020:02:46252:3:0:NEW 15 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_8'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:02:46252:3:0:NEW 15 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_9'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:02:46252:3:0:NEW 15 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_10'><ns0:head>Figure 1</ns0:head><ns0:label>1</ns0:label><ns0:figDesc>Figure 1</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_11'><ns0:head>Figure 5 Fire</ns0:head><ns0:label>5</ns0:label><ns0:figDesc>Figure 5</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_12'><ns0:head>Figure 6 Frequency</ns0:head><ns0:label>6</ns0:label><ns0:figDesc>Figure 6</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_13'><ns0:head>Figure 7</ns0:head><ns0:label>7</ns0:label><ns0:figDesc>Figure 7</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_14'><ns0:head>Fuel</ns0:head><ns0:label /><ns0:figDesc>model classification obtained following the UCO40 criteria (Rodríguez y Silva & Molina-Martínez, 2012), based on Scott & Burgan (2005).</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_15'><ns0:head /><ns0:label /><ns0:figDesc>LiveH: Live herbaceous fuel, LiveW: Live wood fuel. Moisture of extinction (%): moisture content that prevents flame from propagating. PeerJ reviewing PDF | (2020:02:46252:3:0:NEW 15 Sep 2020) Manuscript to be reviewed Fuel loading (tn/ha)</ns0:figDesc></ns0:figure>
<ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='37,42.52,224.62,525.00,247.50' type='bitmap' /></ns0:figure>
<ns0:figure type='table' xml:id='tab_0'><ns0:head>Table 1 (on next page)</ns0:head><ns0:label>1</ns0:label><ns0:figDesc /><ns0:table /></ns0:figure>
<ns0:figure type='table' xml:id='tab_1'><ns0:head>Time lag Fuel type 1h 10h 100h LiveH LiveW Moisture of extinction (%) Fuel bed depth (cm)</ns0:head><ns0:label /><ns0:figDesc /><ns0:table><ns0:row><ns0:cell cols='3'>Predominance of shrubs</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>M3</ns0:cell><ns0:cell cols='2'>11.47 2.88 3.37</ns0:cell><ns0:cell>0</ns0:cell><ns0:cell>6.10</ns0:cell><ns0:cell>15</ns0:cell><ns0:cell>82.29</ns0:cell></ns0:row><ns0:row><ns0:cell>M8</ns0:cell><ns0:cell cols='2'>11.23 6.10 3.47</ns0:cell><ns0:cell>0</ns0:cell><ns0:cell>7.27</ns0:cell><ns0:cell>30</ns0:cell><ns0:cell>121.92</ns0:cell></ns0:row><ns0:row><ns0:cell>M9</ns0:cell><ns0:cell cols='2'>34.71 9.86 4.93</ns0:cell><ns0:cell>0</ns0:cell><ns0:cell>18.89</ns0:cell><ns0:cell>15</ns0:cell><ns0:cell>182.88</ns0:cell></ns0:row><ns0:row><ns0:cell cols='6'>Pine-needle litter with shrubs and/or grassland under forest canopy</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>HPM4</ns0:cell><ns0:cell cols='2'>17.63 13.23 1.17</ns0:cell><ns0:cell>0</ns0:cell><ns0:cell>11.13</ns0:cell><ns0:cell>20</ns0:cell><ns0:cell>76.2</ns0:cell></ns0:row><ns0:row><ns0:cell cols='6'>Predominance of pine-needle with branches and other canopy debris</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>HR7</ns0:cell><ns0:cell>0.73</ns0:cell><ns0:cell>3.76 3.47</ns0:cell><ns0:cell>0</ns0:cell><ns0:cell>0</ns0:cell><ns0:cell>25</ns0:cell><ns0:cell>18.28</ns0:cell></ns0:row><ns0:row><ns0:cell cols='4'>Predominance of canopy debris accumulation</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>R4</ns0:cell><ns0:cell>1.57</ns0:cell><ns0:cell>5.16 6.29</ns0:cell><ns0:cell>0</ns0:cell><ns0:cell>0</ns0:cell><ns0:cell>25</ns0:cell><ns0:cell>82</ns0:cell></ns0:row></ns0:table><ns0:note>PeerJ reviewing PDF | (2020:02:46252:3:0:NEW 15 Sep 2020)</ns0:note></ns0:figure>
<ns0:figure type='table' xml:id='tab_2'><ns0:head>Table 2 (on next page)</ns0:head><ns0:label>2</ns0:label><ns0:figDesc>Dead fuel moisture conditioningWeather stream file (.WXS) showing typical summer weather circadian change in the area, used as input to Flammap software for quantifying the moisture of dead fuel.</ns0:figDesc><ns0:table><ns0:row><ns0:cell>1</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>Date</ns0:cell><ns0:cell>T (ºC)</ns0:cell><ns0:cell>RH (%)</ns0:cell><ns0:cell>PP (mm)</ns0:cell><ns0:cell cols='2'>Wind SP (m/s) Wind dir. (º)</ns0:cell><ns0:cell>Cloud (%)</ns0:cell></ns0:row><ns0:row><ns0:cell>08/27/14 10:00</ns0:cell><ns0:cell>28</ns0:cell><ns0:cell>33</ns0:cell><ns0:cell>0</ns0:cell><ns0:cell>2</ns0:cell><ns0:cell>180</ns0:cell></ns0:row><ns0:row><ns0:cell>08/27/14 11:00</ns0:cell><ns0:cell>29</ns0:cell><ns0:cell>32</ns0:cell><ns0:cell>0</ns0:cell><ns0:cell>2</ns0:cell><ns0:cell>180</ns0:cell></ns0:row><ns0:row><ns0:cell>08/27/14 12:00</ns0:cell><ns0:cell>30</ns0:cell><ns0:cell>29</ns0:cell><ns0:cell>0</ns0:cell><ns0:cell>1</ns0:cell><ns0:cell>180</ns0:cell></ns0:row><ns0:row><ns0:cell>08/27/14 13:00</ns0:cell><ns0:cell>31</ns0:cell><ns0:cell>27</ns0:cell><ns0:cell>0</ns0:cell><ns0:cell>2</ns0:cell><ns0:cell>180</ns0:cell></ns0:row><ns0:row><ns0:cell>08/27/14 14:00</ns0:cell><ns0:cell>30</ns0:cell><ns0:cell>27</ns0:cell><ns0:cell>0</ns0:cell><ns0:cell>2</ns0:cell><ns0:cell>180</ns0:cell></ns0:row><ns0:row><ns0:cell>08/27/14 15:00</ns0:cell><ns0:cell>29</ns0:cell><ns0:cell>28</ns0:cell><ns0:cell>0</ns0:cell><ns0:cell>2</ns0:cell><ns0:cell>180</ns0:cell></ns0:row><ns0:row><ns0:cell>08/27/14 16:00</ns0:cell><ns0:cell>29</ns0:cell><ns0:cell>29</ns0:cell><ns0:cell>0</ns0:cell><ns0:cell>2</ns0:cell><ns0:cell>180</ns0:cell></ns0:row><ns0:row><ns0:cell>08/27/14 17:00</ns0:cell><ns0:cell>28</ns0:cell><ns0:cell>30</ns0:cell><ns0:cell>0</ns0:cell><ns0:cell>1</ns0:cell><ns0:cell>180</ns0:cell></ns0:row><ns0:row><ns0:cell>08/27/14 18:00</ns0:cell><ns0:cell>27</ns0:cell><ns0:cell>31</ns0:cell><ns0:cell>0</ns0:cell><ns0:cell>2</ns0:cell><ns0:cell>180</ns0:cell></ns0:row><ns0:row><ns0:cell>08/27/14 19:00</ns0:cell><ns0:cell>26</ns0:cell><ns0:cell>32</ns0:cell><ns0:cell>0</ns0:cell><ns0:cell>2</ns0:cell><ns0:cell>180</ns0:cell></ns0:row><ns0:row><ns0:cell>2</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>3</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>4</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>5</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>6</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>7</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row></ns0:table><ns0:note>PeerJ reviewing PDF | (2020:02:46252:3:0:NEW 15 Sep 2020) Manuscript to be reviewed PeerJ reviewing PDF | (2020:02:46252:3:0:NEW 15 Sep 2020)</ns0:note></ns0:figure>
<ns0:figure type='table' xml:id='tab_3'><ns0:head>Table 3 (on next page)</ns0:head><ns0:label>3</ns0:label><ns0:figDesc>Best regression models between LIDAR data (independent variable) and field-based Forestereo data (dependent variable), used to map spatial distribution of the main forest structure variables To find which LIDAR data best suit to each field data, a correlation matrix was done. The best results (R> 0.5 and p-value < 0.05) were tested by using linear, power and exponential regression models. The models with highest R</ns0:figDesc><ns0:table /><ns0:note>2 and lower RMSE were selected. CC: canopy cover; Ho: stand height; CBH: canopy base height; CBD: canopy bulk density; G: basal area. x: Percentage of LIDAR first returns above 4 m, y: Percentage of all returns above 4 m, z: Percentage of all returns above mean, d: Percentage of first returns above 0.25 m.</ns0:note></ns0:figure>
</ns0:body>
" | "
Att.: Editor of PeerJ
Dear, Dr. Editor,
We thank you for the minor revisions’ decision and the positive comments of this work. The responses of the new comments are provided below.
I hope that this revised version of the manuscript, according to the criticisms of the reviewer, accomplish all the requirements to be published in PeerJ.
Yours sincerely,
Álvaro Cortés-Molino (on behalf of the co-authors)
-------------------------------------
Comments for the Author
General Comments
This is a potentially very interesting article. The treated topic is actual and very important for assessing fire and canopy structure-related risks. It discussed ' a forest inventory device, to assess fire risk and canopy structure spatial variability by using aerial LIDAR and hemispherical images'. The Response letter line numbers are didn't matching with revision MS. The author should notice next time. But, the revision appropriately addressed many of the comments raised by the reviewers. However, I believe that the MS needs some clarifications and improvements before being considered for publication. I haven't big issue to raise, only few moderate-major points. A few additional comments are given below:
1. L204-205. The author had splitted randomly into train and validation, but Figure 1 still shown Yellow points are fitting plots, and red ones are validation plots. Please check it.
Done. Thanks for pointing this misunderstood. Fitting plots are now called training plots.
2.L184, The 49 plots are used 8 m radius, and author says the LiDAR point cloud density is 0.5 points∙m-2 in ms. Please specify accuracy of GPS device. The author mention that some high difference could be due to low point cloud density (minimum 0.5 points/m2 guaranteed) as well as the fact that airborne LIDAR produces better accuracy for variables related to the top of the canopy (Hilker, et al., 2012). Why canopy cover has high R2. Please specify it.
Done. We assumed a maximum error of 1m for every plot: Lines () “Each sampling plot was accurately geolocated using a high precision GNSS receiver, supported by an RTK terrestrial station deployed in the upper part of the valley. We assumed a maximum error of 1m in each plot, due to the difficulty of getting GNSS signal in high dense canopy.”
Canopy cover is one of the variables related to the top of the canopy which aerial LIDAR produces better accuracy. It is explained in lines (): “The rest of the variables were better modeled above 4 m. Canopy Cover (CC) had the best fitting model with an RMSE less than 20%, while the greatest error was found in modeling the Canopy Base Height (CBH) with an RMSE of 83.3%. This high difference could be due to low point cloud density (minimum 0.5 points·m-2 guaranteed) as well as the fact that airborne LIDAR produces better accuracy for variables related to the top of the canopy (Hilker, et al., 2012), such as canopy cover or canopy height, than variables under the canopy”
3. My second suggestion is about the regression model (Table 3). There should be mention that method of model fit and scatterplots to show relationship derived predictive model on y-axis and field collected data on the X-axis. The model method is incomplete, insufficient and incorrect. A complete description should provide how to model fitted. I would like to know how LiDAR attributes are associated with field drive observations. Please add the statistical analysis.
Done. Thanks for the suggestion. We improved the Table 3 description to better clarify how we did this step. We hope is now clearer:
“To find which LIDAR data best suit to each field data, a correlation matrix was done. The best results (R> 0.5 and p-value < 0.05) were tested by using linear, power and exponential regression models. The models with highest R2 and lower RMSE were selected.
CC: canopy cover; Ho: stand height; CBH: canopy base height; CBD: canopy bulk density; G: basal area. x: Percentage of LIDAR first returns above 4 m, y: Percentage of all returns above 4 m, z: Percentage of all returns above mean, d: Percentage of first returns above 0.25 m.”
We also addressed this in Material & Methods section in lines (): “A correlation matrix between ForeStereo tree data and all LIDAR metrics obtained with FUSION was useful to detect which LIDAR metric was most suitable to build the regression models in 32 random plots. We selected the best correlation results (R<0.5, p-value < 0.05) to test linear, power and exponential regression models. The models with less root-mean-square error (RMSE) and higher adjusted-R in the remaining 17 plots were chosen to predict ForeStereo tree metrics from the LIDAR point cloud. Height break for LIDAR metrics was 4 m, whereas it was 0.25 m for the total vegetation height to avoid high shrubs influence and ground points, respectively. These two height breaks were tested to inquire which one was the better to fit the models.”
Additionally, I've found some minor concerns:
1. The aspect 0° and 360° should be the same. Please check it.
2. Table 2, the PP(mm) and cloud(%) are all zero, please addressing in 2.2 Fieldwork subsection.
Done. Aspect 0º and 360º are now the same in the aspect map. The climatic conditions do not correspond to the fieldwork days, but for the warmest day recorded in the valley. We addressed it in 2.4 section and clarified it: “This file modifies initial dead fuel moisture based on changes during a given period in weather variables such as temperature, relative humidity, cloud cover and hourly precipitation. Conditioning also implies adjusting initial dead fuel moisture to site factors (elevation, slope, aspect, and canopy cover), based on the corresponding raster layers previously uploaded in FlamMap (Finney, 2006). The .WXS file was built from meteorological data that are continuously recorded in situ by the University of Jaén (values of temperature, precipitation and relative humidity) and from records of the Spanish Meteorological Agency-AEMET (values of wind and cloud cover). We chose the warmest day of 2014 (available recorded data) for the conditioning period between the 10:00 to 19:00 hours without precipitations or any cloud cover. The conditioned fuel moistures at the end of this period were the final fuel moistures used for the simulations.”
" | Here is a paper. Please give your review comments after reading it. |
9,806 | "<ns0:abstract xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:p>Background. Globally, and in the Cape Floristic Region of South Africa, extreme fires have become more common in recent years. Such fires pose societal and ecological threats and have inter alia been attributed to climate change and modification of fuels due to alien plant invasions. Understanding the flammability of different types of indigenous and invasive alien vegetation is essential to develop fire risk prevention and mitigation strategies. We assessed the flammability of 30 species of indigenous and invasive alien plants commonly occurring in coastal fynbos and thicket shrublands in relation to varying fire weather conditions. Methods. Fresh plant shoots were sampled and burnt experimentally across diverse fire weather conditions to measure flammability in relation to fire weather conditions, live fuel moisture, fuel load and vegetation grouping (fynbos, thicket, and invasive alien plants). Flammability measures considered were: burn intensity, completeness of burn, time-to-ignition, and the likelihood of spontaneous ignition. We also investigated whether the drying of plant shoots (simulating drought conditions) differentially affected the flammability of vegetation groups.</ns0:p><ns0:p>Results. Fire weather conditions enhanced all measures of flammability, whereas live fuel moisture reduced burn intensity and completeness of burn. Live fuel moisture was not significantly correlated with fire weather, suggesting that the mechanism through which fire weather enhances flammability is not fuel moisture. It furthermore implies that the importance of live fuel moisture for flammability of evergreen shrublands rests on inter-specific and inter-vegetation type differences in fuel moisture, rather than short-term intra-specific fluctuation in live fuel moisture in response to weather conditions. Fuel load significantly increased burn intensity, while reducing ignitability. Although fire weather, live fuel moisture, and fuel load had significant effects on flammability measures, vegetation and species differences accounted for most of the variation. Flammability was generally highest in invasive alien plants, intermediate in fynbos, and lowest in thicket. Fynbos ignited rapidly and burnt completely, whereas thicket was slow to ignite and burnt incompletely. Invasive alien plants were slow to ignite, but burnt with the highest intensity, potentially due to volatile organic composition. The drying of samples resulted in increases in all measures of flammability that were comparable among vegetation groups. Flammability, and by implication fire risk, should thus not increase disproportionately in one vegetation group compared to another under drought conditions -unless the production of dead fuels is disproportionate among vegetation groups. Thus, we suggest that the dead:live fuel ratio is a potentially useful indicator of flammability of evergreen shrublands and that proxies for this ratio need to be</ns0:p></ns0:div>
</ns0:abstract>
<ns0:body xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:head>Introduction</ns0:head><ns0:p>Flammability is the ability of vegetation (fuel) to burn <ns0:ref type='bibr' target='#b39'>(Fernandes and Cruz, 2012;</ns0:ref><ns0:ref type='bibr' target='#b44'>Gill and Zylstra, 2005)</ns0:ref> and is a measure of fire behavior (fire intensity/severity) used in vegetation fire risk mitigation studies <ns0:ref type='bibr' target='#b50'>(Keeley, 2009)</ns0:ref>. Vegetation flammability may result from climatic and weather effects <ns0:ref type='bibr' target='#b15'>(Bond and Midgley, 1995;</ns0:ref><ns0:ref type='bibr' target='#b66'>Mutch, 1970;</ns0:ref><ns0:ref type='bibr' target='#b95'>Snyder, 1984)</ns0:ref>. For example, in arid areas, dryness limits fuel accumulation and fires follow episodic rain, whereas in temperate areas, fuel loads are not limiting but fires follow the drying of those fuels <ns0:ref type='bibr' target='#b21'>(Bradstock, 2010;</ns0:ref><ns0:ref type='bibr' target='#b71'>Pausas and Bradstock, 2007)</ns0:ref>, however dry conditions may also result in an increase in fire risk caused by the availability of dried fuels <ns0:ref type='bibr' target='#b77'>(Piñol et al., 1998)</ns0:ref>. Fire-prone vegetation groups may furthermore have evolved traits that enhance their flammability and improve vegetation fitness in fire-dependent communities <ns0:ref type='bibr' target='#b15'>(Bond and Midgley, 1995)</ns0:ref>. Correspondingly, species with high flammability traits may burn intensely, such that itself and the neighbour die, thereby facilitating recruitment -the 'kill thy neighbour' hypothesis <ns0:ref type='bibr' target='#b15'>(Bond and Midgley, 1995)</ns0:ref>. Flammability traits may thus provide resilience associated with fire tolerance <ns0:ref type='bibr' target='#b15'>(Bond and Midgley, 1995;</ns0:ref><ns0:ref type='bibr' target='#b23'>Calitz et al., 2015)</ns0:ref>. Fire is accordingly one of the main determining factors of the ecology and distribution of Manuscript to be reviewed ecosystems of the world, and is important for maintaining plant diversity <ns0:ref type='bibr' target='#b10'>(Bond, 1997;</ns0:ref><ns0:ref type='bibr' target='#b13'>Bond et al., 2003;</ns0:ref><ns0:ref type='bibr' target='#b11'>Bond and Keeley, 2005)</ns0:ref>.</ns0:p><ns0:p>Flammability is also affected by weather conditions <ns0:ref type='bibr' target='#b10'>(Bond, 1997;</ns0:ref><ns0:ref type='bibr' target='#b53'>Keeley and Syphard, 2017)</ns0:ref>. Fire danger indices -based on ambient temperature, relative humidity, wind speed, and rainfall -are commonly used to rate the fire-proneness of weather conditions <ns0:ref type='bibr' target='#b35'>(Dowdy et al., 2009;</ns0:ref><ns0:ref type='bibr' target='#b67'>Noble et al., 1980;</ns0:ref><ns0:ref type='bibr' target='#b93'>Sirca et al., 2018)</ns0:ref>. Flammability is also influenced by fuel properties such as the amount of flammable plant material (fuel load), packing ratio and chemical composition <ns0:ref type='bibr' target='#b17'>(Brooks et al., 2004;</ns0:ref><ns0:ref type='bibr' target='#b19'>Burger and Bond, 2015;</ns0:ref><ns0:ref type='bibr' target='#b30'>Curran et al., 2017)</ns0:ref>. For instance, greater fuel loads or volatile substances can increase fire intensity <ns0:ref type='bibr' target='#b3'>(Baeza et al., 2002;</ns0:ref><ns0:ref type='bibr' target='#b90'>Saura-Mas et al., 2010)</ns0:ref>.</ns0:p><ns0:p>Globally, extreme fires have become more common in recent years. Examples include the shrublands of California, Australia, Europe <ns0:ref type='bibr' target='#b62'>(Montenegro et al., 2004;</ns0:ref><ns0:ref type='bibr' target='#b88'>San-Miguel-Ayanz et al., 2013)</ns0:ref>, and more recently, South Africa <ns0:ref type='bibr' target='#b54'>(Kraaij et al., 2018)</ns0:ref>. These fires have been accredited to the combinations of climate change (in the form of weather conditions more conducive to fire and extended droughts), increased ignitions, expanded wildland-urban interface areas linked to increasing human populations, and changes in fuels that are often human-induced <ns0:ref type='bibr' target='#b0'>(Archibald et al., 2008;</ns0:ref><ns0:ref type='bibr' target='#b62'>Montenegro et al., 2004;</ns0:ref><ns0:ref type='bibr' target='#b98'>Syphard et al., 2017;</ns0:ref><ns0:ref type='bibr' target='#b101'>Turco et al., 2017;</ns0:ref><ns0:ref type='bibr' target='#b103'>van Wilgen, 1984)</ns0:ref>. Fuels accumulate far beyond normal levels when humans suppress fires to safeguard assets, and due to invasion by invasive alien plants (hereafter IAPs) <ns0:ref type='bibr' target='#b54'>(Kraaij et al., 2018;</ns0:ref><ns0:ref type='bibr' target='#b83'>Radeloff et al., 2005;</ns0:ref><ns0:ref type='bibr' target='#b91'>Scott et al., 1998)</ns0:ref>. The IAPs may affect flammability by altering the fuel structure, fuel distribution (horizontal or vertical fuel continuity), live fuel moisture, chemical contents and fuel load <ns0:ref type='bibr' target='#b17'>(Brooks et al., 2004;</ns0:ref><ns0:ref type='bibr' target='#b31'>Davies and Nafus, 2013;</ns0:ref><ns0:ref type='bibr' target='#b85'>Richardson and van Wilgen, 2004)</ns0:ref>. Extreme fires are also known to occur in shrublands after severe droughts due to the increase of dead (~dry) to live fuel ratios <ns0:ref type='bibr' target='#b52'>(Keeley et al., 2012;</ns0:ref><ns0:ref type='bibr' target='#b53'>Keeley and Syphard, 2017;</ns0:ref><ns0:ref type='bibr' target='#b54'>Kraaij et al., 2018)</ns0:ref>.</ns0:p><ns0:p>Along the southern Cape coast of South Africa, fynbos and thicket shrublands occur interspersed (Supplemental Figure <ns0:ref type='figure' target='#fig_10'>S1</ns0:ref>) despite displaying different fire dynamics and fuel structural traits <ns0:ref type='bibr' target='#b25'>(Campbell et al., 1981;</ns0:ref><ns0:ref type='bibr' target='#b61'>Moll et al., 1984)</ns0:ref>. Fynbos ecosystems commonly support fires that consume surface and canopy fuels and comprise species that readily burn to open recruitment opportunities (gaps) post-fire <ns0:ref type='bibr' target='#b18'>(Buhk et al., 2007;</ns0:ref><ns0:ref type='bibr' target='#b33'>Deacon et al., 1992)</ns0:ref>. However, thicket mostly does not exhibit high flammability traits <ns0:ref type='bibr' target='#b23'>(Calitz et al., 2015)</ns0:ref>, and recruitment from seed largely occurs in inter-fire periods <ns0:ref type='bibr' target='#b28'>(Cowling et al., 1997;</ns0:ref><ns0:ref type='bibr' target='#b75'>Pierce and Cowling, 1984)</ns0:ref>. In Manuscript to be reviewed and residential properties <ns0:ref type='bibr' target='#b38'>(Fares et al., 2017;</ns0:ref><ns0:ref type='bibr' target='#b54'>Kraaij et al., 2018)</ns0:ref>. The extreme nature of these fires has been attributed to extensive IAP fuels, an expansive wildland-urban interface area, an unprecedented regional drought preceding the fires which likely greatly increased litter fuels in thicket, fynbos and stands of IAPs, and very high fire danger weather conditions at the time of the fires <ns0:ref type='bibr' target='#b54'>(Kraaij et al., 2018;</ns0:ref><ns0:ref type='bibr' target='#b79'>Preston, 2017)</ns0:ref>. The 2017 Knysna fires called for improved understanding of potential differences in flammability among vegetation groups, including IAPs occurring in this region. An analysis of satellite image derived proxies for burn severity showed it to be higher, but completeness of burn lower, in IAPs than in indigenous fynbos and thicket vegetation <ns0:ref type='bibr' target='#b54'>(Kraaij et al. 2018</ns0:ref>). However, the findings have not been verified with field observations <ns0:ref type='bibr' target='#b54'>(Kraaij et al., 2018)</ns0:ref>. Other studies have experimentally compared the flammability of species from several biomes (both fire-prone and fire-resistant) <ns0:ref type='bibr' target='#b19'>(Burger and Bond, 2015;</ns0:ref><ns0:ref type='bibr' target='#b23'>Calitz et al., 2015)</ns0:ref>, however, no study has compared the flammability of indigenous vegetation with that of IAPs, nor under varying fire weather conditions.</ns0:p><ns0:p>The primary aim of our study was to compare the flammability of live plant material amongst three vegetation groups -IAPs, fynbos, and thicket -and under varying fire weather conditions. Flammability measures considered were burn intensity, completeness of burn, and ignitability (time-to-ignition and likelihood of spontaneous ignition), while fuel traits considered were live fuel moisture and fuel load. A secondary aim was to assess the flammability of partially dried plant material as a crude proxy for drought effects, to ascertain whether drying of fuels (~drought) would differentially affect the flammability of the vegetation groups of interest.</ns0:p><ns0:p>Study results will inform fire risk management in the southern Cape landscapes and elsewhere with similar fuel traits and characteristics.</ns0:p></ns0:div>
<ns0:div><ns0:head>Materials & Methods</ns0:head></ns0:div>
<ns0:div><ns0:head>Study area</ns0:head><ns0:p>This study was conducted along the southern Cape coast of South Africa within the Cape Floristic Region close to the city of George (33.964°S, 22.534°E). The climate is moderated by the maritime influence with average minimum and maximum temperatures ranging from 7-19⁰C in June and 15-26⁰C in January an annual average rainfall of approximately 800 mm throughout the year <ns0:ref type='bibr' target='#b9'>(Bond, 1981)</ns0:ref>. The area experiences weather conditions suitable for fires at any time of the year and fires are often associated with hot, dry katabatic ('berg') winds <ns0:ref type='bibr' target='#b57'>(Kraaij et al., 2013;</ns0:ref><ns0:ref type='bibr' target='#b103'>van Wilgen, 1984)</ns0:ref>.</ns0:p><ns0:p>The vegetation of the study area is classified as Southern Cape Dune Fynbos <ns0:ref type='bibr' target='#b64'>(Mucina and Rutherford, 2006;</ns0:ref><ns0:ref type='bibr' target='#b75'>Pierce and Cowling, 1984)</ns0:ref>, which consists of medium-dense Manuscript to be reviewed sclerophyllous fynbos (~fine-leaved) shrublands up to 2 m in height, interspersed with dense clumps of subtropical mesophyllous thicket shrubs or trees up to 4 m in height (Supplemental Figure <ns0:ref type='figure' target='#fig_10'>S1</ns0:ref>) <ns0:ref type='bibr' target='#b25'>(Campbell et al., 1981;</ns0:ref><ns0:ref type='bibr' target='#b55'>Kraaij et al., 2011;</ns0:ref><ns0:ref type='bibr' target='#b75'>Pierce and Cowling, 1984)</ns0:ref>. Both fynbos and thicket are evergreen. Fynbos shrublands are fire-prone and flammable while smaller areas of thicket vegetation seldom burn <ns0:ref type='bibr' target='#b42'>(Geldenhuys, 1994)</ns0:ref>. The persistence of fynbos-thicket mosaics requires fire at appropriate intervals (15-25 years) since thicket becomes dominant in the prolonged absence of fire <ns0:ref type='bibr' target='#b58'>(Kraaij and van Wilgen, 2014;</ns0:ref><ns0:ref type='bibr'>Strydom et al., submitted)</ns0:ref>. The area contains extensive invasions of alien shrubs and trees, commonly of the genera Acacia, Eucalyptus, and Pinus, that co-occur with, and potentially replace, the native vegetation (Supplemental Figure <ns0:ref type='figure' target='#fig_10'>S1</ns0:ref>) <ns0:ref type='bibr' target='#b2'>(Baard and Kraaij, 2014;</ns0:ref><ns0:ref type='bibr' target='#b104'>van Wilgen et al., 2016)</ns0:ref>.</ns0:p></ns0:div>
<ns0:div><ns0:head>Data collection</ns0:head></ns0:div>
<ns0:div><ns0:head>Live plant samples</ns0:head><ns0:p>We experimentally measured the flammability of plant shoots (i.e. plant stems) of species from three vegetation groups, namely IAPs, fynbos, and thicket. Fine fuels such as plant shoots (hereafter samples) are the primary carriers or vectors of fire spread <ns0:ref type='bibr' target='#b65'>(Murray et al., 2013)</ns0:ref>. Our experiments were thus focused on plant shoots, with one stem constituting one sample.</ns0:p><ns0:p>Sampling was done over 21 occasions (February -November 2018) that were specifically selected to represent varying fire weather conditions. On each occasion, we collected two live plant samples of 30 species across three vegetation groups (10 species per vegetation group; details in Supplemental Table <ns0:ref type='table'>S1</ns0:ref>) common in the study area. One sample was used for flammability experiments, while the other was used for live fuel moisture measurements. For each species, samples of approximately 70 cm in length that were representative of the fuel structure characteristic of the species were sourced. On each sampling occasion, samples from all 30 species were collected and burnt to ensure that flammability was measured under comparable conditions. Sample collection either started at 9h00 and subsequent burning at 12h00 or at 11h00 and 14h00 (respectively) to incorporate additional variation in fire weather conditions. Samples were kept in closed plastic containers after collection prior to burning, and burning was completed within four hours of sample collection to minimise moisture loss. The order in which samples of different species were burnt was also randomised among the different burning occasions to not consistently expose particular species to longer periods of moisture loss prior to burning. For each occasion, the Canadian fire weather index was computed based on the temperature, relative humidity, rainfall (over the past 24 hours), and wind speed (Van Wagner 1987) at the time that burning commenced. This index integrates drought and other PeerJ reviewing <ns0:ref type='table'>PDF | (2020:04:48231:1:1:NEW 10 Sep 2020)</ns0:ref> atmospheric effects that are relevant to fire behavior and fuel moisture, and it was shown to be the best performing fire danger index in Mediterranean ecosystems <ns0:ref type='bibr' target='#b93'>(Sirca et al., 2018)</ns0:ref>. The input weather measures were obtained from a weather station located on the George Campus of Nelson Mandela University ('Saasveld NMMU CW373' on the Vital Weather online platform: www.vitalweather.co.za) where the experimental burning was conducted.</ns0:p><ns0:p>Samples used for flammability were burnt outdoors using an approach similar to that of <ns0:ref type='bibr' target='#b23'>Calitz et al. (2015)</ns0:ref> and <ns0:ref type='bibr' target='#b30'>Curran et al. (2017)</ns0:ref>. Plant flammability was measured using the method and equipment described by <ns0:ref type='bibr' target='#b49'>Jaureguiberry et al. (2011)</ns0:ref>, the apparatus comprises a metal barrel (85 cm x 60 cm) that is horizontally orientated with the top removable half that is used for wind protection <ns0:ref type='bibr' target='#b3'>(Baeza et al., 2002)</ns0:ref>. The metal barrel is connected to a grill thermometer, removable gas cylinder and a blowtorch <ns0:ref type='bibr' target='#b30'>(Curran et al., 2017;</ns0:ref><ns0:ref type='bibr' target='#b49'>Jaureguiberry et al., 2011)</ns0:ref>. Each sample was placed on the barrel cavity grill to pre-heat at 230°C for two minutes to imitate the heating and drying effect of an approaching fire. If the samples had not spontaneously ignited within two minutes, it was ignited at the top of the shoot by exposing it to the blow torch for a period of five seconds <ns0:ref type='bibr' target='#b23'>(Calitz et al., 2015)</ns0:ref>. Advantages of using this apparatus are that it preserves the architectural arrangement of plant material <ns0:ref type='bibr' target='#b49'>(Jaureguiberry et al., 2011)</ns0:ref>. It further enables a more realistic comparison of relative canopy flammability among species than methods that use only smaller plant components (i.e. twigs or leaves) <ns0:ref type='bibr' target='#b19'>(Burger and Bond, 2015;</ns0:ref><ns0:ref type='bibr' target='#b49'>Jaureguiberry et al., 2011)</ns0:ref>.</ns0:p><ns0:p>Four aspects associated with species-level flammability were measured and recorded (largely after <ns0:ref type='bibr' target='#b23'>Calitz et al., 2015 and</ns0:ref><ns0:ref type='bibr' target='#b49'>Jaureguiberry et al., 2011)</ns0:ref>. Firstly, burn intensity taken as the maximum temperature (cf. <ns0:ref type='bibr' target='#b50'>Keeley, 2009)</ns0:ref> reached by a sample while burning, measured using an infrared thermometer (Major Tech 695; maximum recordable temperature: 800°C) after <ns0:ref type='bibr' target='#b49'>Jaureguiberry et al. (2011)</ns0:ref>. Secondly, the completeness of burn, calculated as the proportion of the pre-burn wet mass of the samples that was consumed by the fire (mass was measured using an electronic scale). Thirdly, time-to-ignition, measured as the time elapsed between placement of the samples on the grill and spontaneous ignition (appearance of the first flame); samples that required to be ignited with the blow torch were therefore excluded from this measures' dataset. For every sample, we recorded whether it spontaneously ignited within the two minutes (pre-heating duration was consistent as there were many samples) of pre-heating or not, this binomial response comprising the fourth measure termed 'spontaneous ignition'.</ns0:p><ns0:p>Live fuel moisture was calculated on a sample shoot similar in dimensions to that of flammability measurements. The fresh material was stored in sealed containers (of known mass) until these were weighed (within less than 3 hours of collection) to obtain wet fuel mass. Manuscript to be reviewed Samples were then oven-dried at 80⁰C for 48 hours and weighed again to obtain dry fuel mass <ns0:ref type='bibr' target='#b87'>(Ruffault et al., 2018;</ns0:ref><ns0:ref type='bibr' target='#b100'>Teie, 2009;</ns0:ref><ns0:ref type='bibr' target='#b110'>Yebra et al., 2019)</ns0:ref>. The live fuel moisture was calculated as the percentage of wet mass comprised of water. Although sample size (shoot length) was standardized, samples nevertheless presented different fuel loads which are directly related to burn intensity <ns0:ref type='bibr' target='#b22'>(Byram 1959)</ns0:ref>. Thus, dry plant mass was used to represent the variable fuel load.</ns0:p><ns0:p>We estimated the dry plant mass of each sample from its pre-burn wet mass and the percentage water content that was calculated for its analogous dried sample.</ns0:p></ns0:div>
<ns0:div><ns0:head>Dried plant samples</ns0:head><ns0:p>To investigate whether simulated drought conditions differentially affected the flammability of the vegetation groups, additional samples (similar to that collected for the flammability experiment's live samples described above) were collected and left to dry under ambient conditions, out of direct sunlight, for a minimum of two weeks but not until leaf loss occurred. Sampling was conducted over five occasions (during February -March 2019) of high fire weather conditions.</ns0:p><ns0:p>The drying duration was standardized for all species to avoid the loss of leaves since certain plants would drop leaves due to drought stress <ns0:ref type='bibr' target='#b27'>(Clarke and McCaig, 1982)</ns0:ref>. Flammability experiments and pre-burn estimations of live fuel moisture were undertaken on these dried samples as described above for live (undried) samples.</ns0:p></ns0:div>
<ns0:div><ns0:head>Data analysis</ns0:head></ns0:div>
<ns0:div><ns0:head>Live plant samples</ns0:head><ns0:p>We assessed flammability (of live samples) in terms of four response variables (burn intensity, completeness of burn, time-to-ignition, and spontaneous ignition) respectively, in relation to the predictor variables (i) fire weather (continuous), (ii) live fuel moisture (continuous), (iii) fuel load (dry plant mass; continuous), (iv) vegetation groups (IAPs, fynbos, thicket; categorical) and (v) species (30 species; categorical) using generalized linear mixed-effects models <ns0:ref type='bibr' target='#b5'>(Bates, 2010;</ns0:ref><ns0:ref type='bibr' target='#b69'>O'Hara, 2009)</ns0:ref> using the lme4 package <ns0:ref type='bibr' target='#b5'>(Bates, 2010)</ns0:ref> in the open-source R software version 3.6.1 (R Development Core Team 2019). Detailed species-level comparisons were not the primary focus of the study and species was therefore included as a random factor, whereas the other predictor variables were included as fixed factors. To test for potential collinearity between fire weather and live fuel moisture, we ran the Spearman-rank correlation test for each respective species. It showed that these variables were not significantly correlated (see Results) and could both be retained in subsequent analyses. Burn intensity was log-transformed (to correct right-skewed distribution), completeness of burn arcsine-transformed (as it was Manuscript to be reviewed expressed as proportions), time-to-ignition square root-transformed (to correct left-skewed distribution), and spontaneous ignition assessed using logistic regression (binomial family, logit link function) (formulae provided in Supplemental Table <ns0:ref type='table'>S2</ns0:ref>). Subsequently, Type II Wald chisquare test <ns0:ref type='bibr' target='#b48'>(Hastie and Pregibon, 1992)</ns0:ref> was computed to determine the significance of fixed factors on the specific models. We incorporated the scale function to the generalized linear mixed-effects models and logistic regression model (using transformed data) to standardize variables of different scales and obtain the relative influence of fixed factors <ns0:ref type='bibr'>(Becker et al., 1988)</ns0:ref>.</ns0:p></ns0:div>
<ns0:div><ns0:head>Dried plant samples</ns0:head><ns0:p>We compared the flammability (in terms of burn intensity, completeness of burn, and time-toignition, respectively) of the dried samples with that of live samples of the same species that was measured on five occasions under comparable fire weather conditions. We calculated the change in flammability between live and dried samples by subtracting the flammability measure of each live sample from that of its dried counterpart. We then used this derived variable as response variable and employed Kruskal-Wallis to test whether the difference in flammability between live and dried samples varied among vegetation groups.</ns0:p></ns0:div>
<ns0:div><ns0:head>Results</ns0:head></ns0:div>
<ns0:div><ns0:head>Live plant samples</ns0:head><ns0:p>Fire weather and live fuel moisture were not significantly correlated within any of the study species (Supplemental Table <ns0:ref type='table'>S1</ns0:ref>). Increasing severity of fire weather significantly increased flammability through increasing burn intensity, increasing completeness of burn, increasing the likelihood of spontaneous ignition, and reducing time-to-ignition (Table <ns0:ref type='table'>1</ns0:ref>, Fig. <ns0:ref type='figure' target='#fig_10'>1</ns0:ref>). Increasing live fuel moisture significantly decreased burn intensity, completeness of burn, and the likelihood of spontaneous ignition. Fuel load significantly increased burn intensity and time-to-ignition.</ns0:p><ns0:p>In considering vegetation groups, flammability was generally highest in IAPs, intermediate in fynbos, and lowest in thicket (Table <ns0:ref type='table'>1</ns0:ref>, Fig. <ns0:ref type='figure' target='#fig_10'>1</ns0:ref>). IAPs burnt at significantly higher intensity than fynbos and thicket. IAPs and fynbos showed significantly higher ignitability (shorter time-to-ignition and a greater likelihood of spontaneous ignition) than thicket.</ns0:p><ns0:p>Amongst the different fixed factors, vegetation groups consistently had the largest influence (i.e. the largest scaled estimates; Table <ns0:ref type='table'>1</ns0:ref>) on all flammability measures. Fire weather had the second largest influence on ignitability, while live fuel moisture had the second largest influence on burn intensity and completeness of burn. The total variance in the flammability measures explained by the models was generally low (24 -40%; conditional R 2 values, Table <ns0:ref type='table'>1</ns0:ref>). The fixed factors combined explained less variation (8 -22%; marginal R 2 values, Table <ns0:ref type='table'>1</ns0:ref>) than species as random factor by itself (12 -20%), except in terms of spontaneous ignition where vegetation groups and fire weather were most influential.</ns0:p></ns0:div>
<ns0:div><ns0:head>Dried plant samples</ns0:head><ns0:p>Drying out of samples under ambient conditions for two weeks resulted in an average reduction in fuel moisture contents of approximately 30% (Fig. <ns0:ref type='figure'>2 A</ns0:ref>), and the extent of this reduction did not differ significantly among vegetation groups (H 2 =1.4, p=0.505). Dried samples exhibited increased flammability compared to their live counterparts, i.e. an average increase in burn intensity of 115°C; an 11% increase in completeness of burn; and a 46 seconds reduction in time-to-ignition (Fig. <ns0:ref type='figure'>2 B -D</ns0:ref>). However, this differential response in flammability between dried and live samples was comparable among the vegetation groups in terms of burn intensity (H 2 =0.8, p=0.666), completeness of burn (H 2 =1.8, p=0.410), and time-to-ignition (H 2 =0.6, p=0.741).</ns0:p></ns0:div>
<ns0:div><ns0:head>Discussion</ns0:head></ns0:div>
<ns0:div><ns0:head>Effects of fuel moisture, fire weather, and fuel load on flammability</ns0:head><ns0:p>Fuel moisture content is widely regarded to be a major determinant of flammability in grassland, shrubland and forested ecosystems with sufficient evidence of its dampening effects on fire behaviour and flammability <ns0:ref type='bibr' target='#b7'>(Bianchi and Defossé, 2015;</ns0:ref><ns0:ref type='bibr' target='#b38'>Fares et al., 2017;</ns0:ref><ns0:ref type='bibr' target='#b73'>Pausas and Paula, 2012)</ns0:ref>. That is why several fire danger indices attempt to account for the moisture contents of dead and live fuels to improve fire danger forecasting <ns0:ref type='bibr' target='#b26'>(Chuvieco et al., 2004;</ns0:ref><ns0:ref type='bibr' target='#b59'>Madula, 2013;</ns0:ref><ns0:ref type='bibr' target='#b86'>Rothermel, 1983;</ns0:ref><ns0:ref type='bibr' target='#b87'>Ruffault et al., 2018;</ns0:ref><ns0:ref type='bibr' target='#b93'>Sirca et al., 2018)</ns0:ref>. Although dead fuel moisture responds closely and rapidly to fire weather, the relation between live fuel moisture and fire weather is more complicated as it depends on plant physiology and medium-to long-term meteorological trends <ns0:ref type='bibr' target='#b7'>(Bianchi and Defossé, 2015;</ns0:ref><ns0:ref type='bibr' target='#b16'>Bowman et al., 2014;</ns0:ref><ns0:ref type='bibr' target='#b26'>Chuvieco et al., 2004;</ns0:ref><ns0:ref type='bibr' target='#b68'>Nolan et al., 2016)</ns0:ref>. Accordingly, live fuel moisture in this study was not significantly correlated with fire weather in any of the study species. Live fuel moisture did significantly (negatively) correlate to burn intensity and completeness of burn, but the magnitude of its influence on flammability relative to the other factors investigated was generally low.</ns0:p><ns0:p>Fire weather significantly enhanced all measures of flammability, however the lack of response of live plant moisture contents to fire weather suggests that the mechanism through which fire weather enhances flammability may not be live fuel moisture. Other studies that have Manuscript to be reviewed investigated fuel moisture-flammability relations (e.g., <ns0:ref type='bibr' target='#b8'>Bianchi et al., 2018)</ns0:ref> have not evidently assessed the effects of fire weather or have manipulated fuel moisture through drying out of fuels beyond natural levels of fluctuation in live fuels <ns0:ref type='bibr' target='#b34'>(Dimitrakopoulos and Papaioannou, 2001)</ns0:ref>.</ns0:p><ns0:p>We argue that the importance of live fuel moisture for flammability of evergreen shrublands rests on inter-specific and inter-vegetation type differences in fuel moisture contents (cf. <ns0:ref type='bibr' target='#b26'>Chuvieco et al., 2004)</ns0:ref>, rather than medium-term intra-specific fluctuation in live fuel moisture in response to weather conditions. The incorporation of satellite-derived proxies for live fuel moisture into fire danger indices is therefore unlikely to be useful in these systems. Although fire weather increased all measures of flammability (and particularly ignitability), it was less influential than vegetation groups (see scaled estimates in Table <ns0:ref type='table'>1</ns0:ref>). The contribution of shortterm weather conditions to the severity of the 2017 Knysna fires was regarded to have been secondary to that of the long-term drought preceding these fires that would have caused a buildup of dead fuels <ns0:ref type='bibr' target='#b54'>(Kraaij et al., 2018)</ns0:ref>. Fire weather is expected to increase in importance in its effects on flammability if cognizance is taken of dry or dead fuels (see below) and when considering stand level fire behaviour. Although flammability experiments at the plant shoot scale are an improvement over those on excised leaves, and although results at the plant shoot and whole-plant scale are often in agreement (Pausa and Moreira, 2012), the scale of experimentation relative to stand or landscape level fire remains inadequate. For instance, particular aspects of fire weather, such as wind speed, greatly influence wildfire spread and spotting behavior <ns0:ref type='bibr' target='#b41'>(Forsyth et al., 2019)</ns0:ref>. Such dynamics cannot be considered using the shootlevel flammability methods used in the current study; this may lead to an underestimation of the importance of fire weather on flammability and, by implication, fire behavior.</ns0:p><ns0:p>Fuel load had varying effects on flammability, depending on the measure considered; it increased burn intensity, but reduced ignitability. These findings support other evidence for positive correlations between the amount of biomass (~fuel load) that vegetation presents and fire intensity or severity <ns0:ref type='bibr' target='#b3'>(Baeza et al., 2002;</ns0:ref><ns0:ref type='bibr' target='#b50'>Keeley, 2009;</ns0:ref><ns0:ref type='bibr' target='#b90'>Saura-Mas et al., 2010)</ns0:ref>, but negative correlations between fuel load and ignitability <ns0:ref type='bibr' target='#b47'>(Guijarro et al., 2002)</ns0:ref>, the rate of spread <ns0:ref type='bibr' target='#b45'>(Grootemaat et al., 2017)</ns0:ref> and completeness of burn <ns0:ref type='bibr' target='#b54'>(Kraaij et al., 2018;</ns0:ref><ns0:ref type='bibr' target='#b106'>van Wilgen et al., 1990)</ns0:ref>. Such contrasting effects on the different aspects of flammability relate to variation in fuel structural traits and emphasize the need to consider flammability in terms of its constituent measures rather than treating it as a composite measure <ns0:ref type='bibr' target='#b36'>(Engber and Varner, 2012;</ns0:ref><ns0:ref type='bibr'>Pausas et al., 2012;</ns0:ref><ns0:ref type='bibr' target='#b89'>Santana and Marrs, 2014)</ns0:ref>.</ns0:p><ns0:p>Although live fuel moisture content, fire weather conditions, and fuel load had significant effects on some of the flammability measures, these factors did not explain a large portion of Manuscript to be reviewed variability in the flammability response. Species, which was assessed as a random factor, often accounted for more variation in flammability than the fixed factors combined. This suggests important species effects on flammability, which warrant more detailed investigation. Our method of placing plant shoots horizontally on the barrel cavity grill with different amounts and sizes of plant parts oriented towards the grill could have introduced additional variation in the flammability response.</ns0:p></ns0:div>
<ns0:div><ns0:head>Vegetation group effects in relation to fire risk</ns0:head><ns0:p>Vegetation group comparisons showed that the flammability of IAPs exceeded that of thicket in terms of all flammability measures and exceeded that of fynbos in terms of burn intensity. These findings support claims <ns0:ref type='bibr' target='#b41'>(Forsyth et al., 2019;</ns0:ref><ns0:ref type='bibr' target='#b96'>Stander, 2019)</ns0:ref> and other evidence <ns0:ref type='bibr' target='#b17'>(Brooks et al., 2004;</ns0:ref><ns0:ref type='bibr' target='#b54'>Kraaij et al., 2018;</ns0:ref><ns0:ref type='bibr' target='#b84'>Richardson and Rejmánek, 2011)</ns0:ref> that invasions by alien plants can add to the severity, intensity, and difficulty of control of wildfires. Fynbos and IAPs were more ignitable than thicket, and thus present higher risks under moderate and high fire weather conditions, whereas thicket presents lower risks under low and moderate fire weather conditions. Accordingly, observations from the 2017 Knysna fires indicated that thicket only becomes ignitable under very high or extreme fire weather conditions but may then burn at intensities exceeding that in fynbos but not that of IAPs <ns0:ref type='bibr' target='#b54'>(Kraaij et al., 2018)</ns0:ref> presumably on account of disparate fuel loads <ns0:ref type='bibr' target='#b50'>(Keeley, 2009;</ns0:ref><ns0:ref type='bibr' target='#b60'>Mandle et al., 2011)</ns0:ref>. In our study, there were no significant differences between the flammability of fynbos and IAPs but completeness of burn appeared to be the highest in fynbos. <ns0:ref type='bibr' target='#b54'>Kraaij et al. (2018)</ns0:ref> also observed that fynbos burnt more completely than thicket and IAPs in the 2017 Knysna fires which suggests that the risk of recurring fire will be lowest in fynbos for some period post-fire, whereas incomplete burning of IAPs will not afford the same level of risk reduction shortly post-fire.</ns0:p></ns0:div>
<ns0:div><ns0:head>Simulated drought conditions</ns0:head><ns0:p>Extremely large and severe fires, including the 2017 Knysna fires, are often associated with preceding droughts <ns0:ref type='bibr' target='#b54'>(Kraaij et al., 2018;</ns0:ref><ns0:ref type='bibr' target='#b81'>Quinn, 1994;</ns0:ref><ns0:ref type='bibr' target='#b88'>San-Miguel-Ayanz et al., 2013;</ns0:ref><ns0:ref type='bibr' target='#b107'>Williams, 2013)</ns0:ref> and the resultant increase in dead fuels <ns0:ref type='bibr' target='#b50'>(Keeley, 2009)</ns0:ref>. The extent and severity to which thicket, normally regarded as a fire-resistant (~poorly ignitable) vegetation <ns0:ref type='bibr' target='#b23'>(Calitz et al., 2015;</ns0:ref><ns0:ref type='bibr' target='#b29'>Cowling and Potts, 2015)</ns0:ref>, burnt in the 2017 Knysna fires was attributed to extreme fire weather conditions and to the preceding severe drought <ns0:ref type='bibr' target='#b54'>(Kraaij et al., 2018)</ns0:ref>. In this study, we confirmed that the drying of fuels as a crude proxy for severe drought effects considerably increased flammability. However, the magnitude of the increase in flammability in response to drying of Manuscript to be reviewed fuels was consistent across vegetation groups. Flammability, and by implication fire risk, is thus unlikely to increase disproportionately in one vegetation group compared to another under extended drought unless the production of dead fuels due to drought would be disproportionate among the vegetation groups. We concede that the proxy for drought conditions could not realistically simulate all potential effects of drought on fuel modification and flammability, and in particular on the dying off of fuels and resultant increase in litter component. Detailed consideration of this aspect was beyond the scope of this study and warrants further investigation. Given the low moisture contents of dead fuels, the ratio of dead to live fuels are likely to be a useful indicator of fire risk in evergreen shrublands <ns0:ref type='bibr' target='#b50'>(Keeley, 2009)</ns0:ref>. Proxies for this ratio should, therefore, be sought for incorporation into fire danger indices.</ns0:p></ns0:div>
<ns0:div><ns0:head>Conclusions</ns0:head><ns0:p>Our experimental burning of shoots of 30 shrub species confirmed that fire weather, live fuel moisture, and fuel load have significant effects on flammability measures. However, vegetation group and species differences accounted for most of the variation in flammability. Flammability was generally highest in invasive alien plants, intermediate in fynbos, and lowest in thicket. The drying of plant shoots resulted in increases in flammability that were comparable among vegetation groups, implying that under drought conditions, fire risk should not increase disproportionately in one vegetation group compared to another, unless the production of dead fuels is disproportionate among vegetation groups.</ns0:p></ns0:div>
<ns0:div><ns0:head>Table 1(on next page)</ns0:head><ns0:p>Output of generalized linear mixed-effects models and logistic regression model that assessed flammability in terms of burn intensity, completeness of burn, time-to-ignition and spontaneous ignition.</ns0:p><ns0:p>Fixed factors included in the generalized linear mixed-effects models (gaussian family, identity function; details in Supplementary 2) and logistic regression model (binomial family, logit link function) were fire weather, fuel moisture, fuel load, and vegetation groups (IAPs, invasive alien plants; Fyn, fynbos; and Thi, thicket), while species was included as a random factor.</ns0:p><ns0:p>Significance codes: *p <0.05, **p <0.01, ***p <0.001 a Chisq statistics and significance levels were obtained from deviance tables (Type II Wald chi-square tests; details in Supplemental Table <ns0:ref type='table'>S3</ns0:ref>). Manuscript to be reviewed Manuscript to be reviewed</ns0:p><ns0:note type='other'>Figure 2</ns0:note><ns0:p>The change (Δ) between live and dried samples in (A) fuel moisture, (B) burn intensity, (C) completeness of burn, and (D) time-to-ignition, compared among vegetation groups.</ns0:p><ns0:p>Live and dried samples were of the same species under comparable fire weather conditions.</ns0:p><ns0:p>Vegetation groups were IAPs, invasive alien plants; Fyn, fynbos; and Thi, thicket. Medians (lines), 25-75 quantile ranges (boxes), 1.5 * interquartile ranges (whiskers), and outliers</ns0:p></ns0:div><ns0:figure xml:id='fig_0'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:04:48231:1:1:NEW 10 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_1'><ns0:head /><ns0:label /><ns0:figDesc>2017, extreme fires occurred in this region around the town of Knysna which burnt indigenous fynbos and thicket vegetation and further caused extensive damage to commercial plantations PeerJ reviewing PDF | (2020:04:48231:1:1:NEW 10 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_2'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:04:48231:1:1:NEW 10 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_3'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:04:48231:1:1:NEW 10 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_4'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:04:48231:1:1:NEW 10 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_5'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:04:48231:1:1:NEW 10 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_6'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:04:48231:1:1:NEW 10 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_7'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:04:48231:1:1:NEW 10 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_8'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:04:48231:1:1:NEW 10 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_9'><ns0:head>b</ns0:head><ns0:label /><ns0:figDesc>Scaled estimates were derived from incorporating the scale function in the generalized linear mixed-effects models and logistic regression model. using the r.squared GLMM function, where conditional R2 indicates the proportion of variance explained by fixed and random factors combined, marginal R 2 indicates the proportion of variance explained by fixed factors alone and R 2 (1|Species) indicates variance explained by the random factor alone. PeerJ reviewing PDF | (2020:04:48231:1:1:NEW 10 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_10'><ns0:head>Figure 1</ns0:head><ns0:label>1</ns0:label><ns0:figDesc>Figure 1</ns0:figDesc></ns0:figure>
</ns0:body>
" | "Dear Editor
We have responded to the reviewer comments below and revised the ms using tracked changes. Our indication of line numbers below refers to those in the revised ms showing tracked changes. We hope that you will find the changes appropriate and thank you for the opportunity to submit a revised ms to PeerJ.
Reviewer 1
Basic reporting
No comment, the manuscript complies with the standards.
Experimental design
Methods require a few clarifications. There are some relevant issues:
- Moisture content variation after sampling and its relation with weather;
- The FWI as a suitable indicator of weather influence on flammability;
- Fire intensity as a temperature;
- Fuel samples for moisture content determination exposed to a temperature lower than desirable;
- Fuel moisture content expressed wrongly.
Details are provided in the general comments section. We have responded to these issues where the reviewer elaborated below.
Validity of the findings
The findings suffer from the problems identified above. More generally, flammability experiments are known to be poor surrogates of 'real' fire conditions. Any finding from an experiment of this type is always subject to strong doubts about its validity. We discussed the limitations of flammability experiments in more detail now (lines 337-346 and 361-364).
Comments for the author
Flammability experiments are a difficult topic, because they tend to be subjective and the researchers involved tend to have poor understanding of fire processes. I did identify a number of issues in the manuscript:
L59-63. Some confusion here between weather and climate. Individual fires are never climate-driven, the fire regime is. Reviewer 2 also queried this statement and we replaced the statement with what was recommended by Reviewer 2 (lines 60-66).
L75-76. This sentence does not make sense. The only reason why RH and temperature are used in fire danger rating is because they determine fuel moisture content. This statement and the following one were omitted (lines 79-81).
L118. Live fuel moisture? Because fire weather determines dead fuel moisture. Yes, we inserted ‘live’ before ‘fuel moisture’ (line 125) and throughout the rest of the ms where appropriate.
L153-154. 'subsequently spreading fire to other plant structures' is not really accurate nor relevant. Better to just say that fine fuels such as plant shoots are the primary carriers/vectors of fire spread. Reworded as suggested (lines 159-162).
L159. What defined a sample? One stem? Yes, this was clarified in lines 158 and 162-163.
L164. The experiments deal with live fuels. Live fuels lose moisture rapidly after collection. So, either you ensured the samples maintained their moisture status or you let the moisture vary after collection. In one case or another the moisture content is not expected to correlate with the current fire weather at the time of the burning. We added the following information on how we attempted to minimize this source of variability (lines 173-177): “Samples were kept in closed plastic containers after collection prior to burning, while burning was completed within four hours of sample collection in order to minimise moisture loss.” Additionally, it has been stated how moisture loss was minimized in the case of the samples used to measure live fuel moisture contents (lines 213-214). Were the experiments carried out outdoors? We added that experiments were undertaken outdoors (line 186).
L166. A Canadian reference should be used, e.g. Van Wagner (1987). We replaced the citations we had with Van Wagner & Forest 1987 (lines 179) and added this source to the reference list (lines 647-648). As the study deals with live fuels the authors should have used other indices of the Canadian system which should be better correlated with live fuel moisture content, namely the Drought Code (DC) or the Buildup Index (BUI), as the FWI integrates this plus the atmospheric influences. Better yet, the authors could have tried to separate the effect of the atmosphere (by using the vapour pressure deficit, currently used as a predictor of live fuel moisture content) and the effect of drought. We chose the Canadian fire weather index as we were interested in the relation between fire weather conditions, different measures of flammability (~fire behavior) and live fuel moisture. We were not solely interested in the relation between fuel moisture and some weather/drought index. As pointed out by the reviewer, the Canadian fire weather index is a more composite index integrating atmospheric influences (short- and medium-term weather) and drought effects that relate to fire behavior and fuel moisture. Furthermore, the Canadian FWI has been shown by Sirca et al. 2018 to have, amongst fire danger indices, the best overall performance in Mediterranean systems. Accordingly, we now added a statement (and Sirca et al. 2018 citation) to justify our use of this index (lines 180-182).
L187. Although unfortunately found quite often in the fire ecology literature, this is extremely wrong: maximum temperature does not reflect fire intensity or the heat flux produced by combustion, as all flames attain the same temperature. Temperatures measured in fires are not the temperature of the fire, which is meaningless by definition, but the temperature of the measuring device. Using the measurements of the study, fire intensity would be expressed by the product of time-to-ignition (assuming it is an acceptable proxy for fire spread rate) and the proportion of burnt fuel, i.e. following the fire intensity concept of Byram (1959). As conceded by the reviewer, several studies (e.g. Burger & Bond 2015 and Calitz et al. 2015; but also Jaureguiberry et al. 2011 whom originally published the method of measuring flammability with the type of device we used) have regarded maximum temperature as a proxy for burn intensity, and these studies have been published in peer-reviewed literature suggesting that the method/rationale is acceptable among fire ecologists. The reviewer furthermore suggests that time-to-ignition and the proportion of burnt fuel be considered when assessing flammability and that these variables should be used to compute fire intensity. We have measured these variables in addition to maximum temperature and are therefore comfortable that the combination of variables we assessed are adequate to provide a comprehensive picture of flammability. Keeping these different variables apart, as opposed to deriving indices that are products of several variables, is the most transparent approach which also allows for comparisons with other studies that may have considered only some of these variables. We have added the citation to Jaureguiberry et al. 2011 to justify our method (lines 203-204).
L200. Current knowledge indicates that drying temperatures lower than 100ºC (Matthews in the Int. J. Wildland Fire) underestimate fuel moisture content. We applied an established drying method and temperature (backed up by citations, see lines 215-216) that are in line with studies globally that assessed plant moisture contents – see Yebra et al. 2019, a global database on live fuel moisture content. We have now added the citation to this global database (line 216, and to the reference list in lines 662-667) to further back up our methods. Furthermore, Matthews (2010) states “Differences between oven-dry masses of fuels dried at 60 and 105°C of up to 3.5% were measured.” In our study, the variance within and amongst species was in excess of 3.5% (see figure below) and thus we don’t believe this has influenced our results.
L201-202. Do you mean that it was calculated on a wet weight basis? The standard is to refer fuel moisture content always on a dry basis ((wet-dry)*dry*100). Both approaches are used in the literature, e.g. Yebra et al. 2019 expressed fuel moisture on a dry basis, but several studies, e.g. Burger & Bond 2015 and Calitz et al. 2015 expressed fuel moisture on a wet basis. We chose the latter in order for our results to be comparable with other local studies in similar vegetation. Expressing fuel moisture on a dry basis would change the absolute values (see below), however, the ordinal structure of the data would remain unchanged. Thus, this would not affect the analyses nor results.
L203. Make the sentence more accurate: fire intensity is directly proportional to fuel load. Keeley (2009) is not the right ref. here and you should cite Byram (1959). Rephrased the statement to say that fuel load is directly related to burn intensity and we replaced Keeley (2009) with Byram (1959) (lines 218-219) and added the latter to the reference list (lines 469-470).
L205. So you had % fuel consumed and fuel load, which enables calculation of fuel consumption (in actual weight) from which fire intensity is calculated. Kindly see our earlier justification for the use of maximum temperature as a proxy for burn intensity.
L203. Reference to running GLMM appears here a 2nd time. Thanks for pointing this out. We have realigned the text in lines 242-250 to delete this repetition.
L293. Also in response to drought (water availability). In response to the other reviewer’s comments we now state that live fuel moisture depends on plant phusiology and medium- to long-term meteorological trends (lines 307-310).
L301-302. Very unlikely! There are other motives but note that the fuel samples were pre-heated and this would totally override any effect that temperature could have had. We omitted this statement (lines 322-323).
L311-312. Flammability experiments are microscale experiments that do not warrant this type of inference. Not sure we understand what inference the reviewer is referring to. The indicated line numbers refer to the statement “Although fire weather increased all measures of flammability (and particularly ignitability), it was less influential than vegetation groups.” This statement does not constitute inference, it simply states our results that are based on the stats on the importance of factors (based on scaled estimates in Table 1). We now added in line 333 “(see scaled estimates in Table 1)” to clarify that this statement is based on our results rather than being an inference. We have acknowledged the limitations of flammability experiments at plant shoot scale elsewhere in the manuscript (lines 337-346 and 361-364)
L317-323. Basically you are undermining the results here by acknowledging that they are representative of real world conditions. We acknowledge the opposite here, i.e. we concede that the scale of experimentation when burning shoots is still inadequate and not representative of real world stand- or landscape-scale fire. This is in line with the reviewer’s concern about the validity of extrapolating flammability experiments to real-world fire. However, we also added reference to a citation showing that flammability experiments at the scale of plant shoot and whole plant are often in agreement (Pausas and Moreira 2012) (lines 339-340).
L350-354. This is totally unwarranted. Burn completeness as an outcome of microscale experiments cannot be compared with that resulting from a wildfire. Under wildfire conditions the amount of residual fuel is mostly a function of fuel size (particle diameter or thickness) and condition (moisture). We rephrased this argument (lines 377-382) to now provide additional evidence of differences in completeness of burn among vegetation types based on a landscape-level fire (Kraaij et al. 2018). The suggestion about differences among vegetation types in risk of recurring fire is now based on this evidence in addition to the similar findings of our flammability experiments.
L373. A more relevant motive is that dead fuel moisture content is much lower under conditions that allow fire spread We have replaced (in lines 401-402) the stated suggestion (that dead fuels respond more rapidly to weather conditions than live fuels) with the motive suggested by the reviewer (the low moisture contents of dead fuels).
Reviewer: Owen Price
Basic reporting
No comment
Experimental design
All good. There should be mention in the discussion of how accurate the method was. We have added a statement to this effect (lines 361-364) (and see later response to more specific comment of the reviewer).
Validity of the findings
There is some confusion about the aims and the conclusions. See later response to more specific suggestions of the reviewer about the hypotheses and study aims (which we have addressed).
Comments for the author
This study examines the flammability of three different vegetation types in the Cape region of South Africa and how flammability responds to fire weather. It is an important contribution and in general the methods are appropriate, it is well written with appropriate references. However, I think some of the logic was loose, which left me confused about what the actual hypothesis and conclusions were. See later response to more specific suggestions of the reviewer about the hypotheses and study aims (which we have addressed). In particular, there seems to be no distinction made between live fuel moisture and dead fuel moisture. We have now clarified throughout the ms whether we referred to live or dead fuel moisture.
Specific points.
Line 60. I think ‘fires’ should be ‘fire regimes’. An individual fire is weather-driven, not climate driven. We replaced this statement with the reviewer’s suggestion below (lines 60-63).
Line 60-63. I think this could be clearer. For example, Bradstock 2010 explains in arid areas, dryness limits fuel and fires follow episodic rain, whereas in temperate forests, fuels are not limiting, and fire follows drying of those fuels. Bradstock, R (2010) A biogeographic model of fire regimes in Australia: current and future implications. Global Ecology and Biogeography 19, 145-158. We replaced this statement with what the reviewer suggested and added the Bradstock (2010) citation (lines 61-65) as well as in the reference list (lines 467-468).
Line 66. Replace ‘would’ with ‘may’. This is only a theory. Done as suggested (line 69).
Line 87. Add ‘and’ before ‘changes’. Done as suggested (line 91).
Line 97. I think here you need to add something about litter fuels. Fires are often an interaction between litter and live fuels. For example, in eucalypt forests, it is generally thought that you cannot have a canopy fire without a corresponding surface fire. If this is the case in your system, then you are not measuring everything you need to about the flammability of these vegetation types. If litter fuels are not important, then say so. We added (lines 102-103) that fynbos typically experiences fires that consume surface and canopy fuels (whereas thicket does not exhibit high flammability). And later in the paragraph where we referred to the fires that did burn both thicket and fynbos, we added that the regional drought likely greatly increased litter fuels in thicket and fynbos and stands of IAP (lines 111-112).
Line 111. Add ‘it’ after ‘showed’. Thanks, we have added ‘it’ (line 115).
Line 117. These aims are not very clear. Surely comparing vegetation types is one of the main aims. Why pose those hypotheses, when broadly speaking, they have been established in previous studies? We rephrased the main aim of our study for it to emphasize the comparison among vegetation types; and we removed the hypotheses that expressed relationships established in other studies (lines 122-129). We accordingly removed reference in the Discussion to these hypotheses (lines 318, 320 and 349). This helped focus the manuscript — thanks for the suggestion.
L146. What form are these invasive species? Presumably small trees, but still taller than the natives? Do they co-occur with the natives, or replace them? We added that these invasive species are shrubs and trees and that they co-occur with, and potentially replace, the native vegetation (lines 153-155). We have furthermore provided (and cited) Supplemental Figure S1 showing pictures of the structure and co-occurrence of these vegetation types in the landscape.
L160. It is important to describe how soon the burning occurred after sampling. E.g. if they were all sampled at once and then burned one by one during the day, then some would have dried out more than others. We added information about this aspect in response to a similar comment of Reviewer 1. We furthermore added that the order in which species were burnt was randomized such that particular species were not consistently exposed to longer period of moisture loss prior to burning (lines 174-178).
Line 204. How was this dry mass calculated? Using the ratios found in the subsequent study of dried plant samples? We improved our explanation of how dry mass was calculated in lines 217-223. It used the % water content that was calculated for analogous dried samples (but not those samples used in the subsequent study of dried samples that simulated drought conditions).
Line 287. This is not true. Dead fuel moisture (e.g. in leaf litter) is responsive to weather. Live fuel moisture responds to drought (as your 2-week drying shows). Those fire danger indices account for dead fuel moisture. See for example Nolan et al 2016. Here is a quote from there: “The first major novel outcome of this study was the formal demonstration that dynamic transformations in fuel moisture associated with major wildfires can occur rapidly for dead fuels and within several weeks to months for live fuels”. Nolan, RH, Boer, MM, de Dios, VR, Caccamo, G, Bradstock, RA (2016) Large-scale, dynamic transformations in fuel moisture drive wildfire activity across southeastern Australia. Geophysical Research Letters 43, 4229-4238. We changed the statement (and related text) to make it clear that dead fuel moisture is more responsive in the short term to fire weather than live fuel moisture and we cited Nolan et al. 2016 (lines 305-313). We however maintain that fire danger indices attempt to account for both dead and live fuel moisture and this statement is backed up by several citations.
Line 293. Similarly, you would not expect live fuel moisture to be related to fire weather. We changed this statement in accordance with the changes made in relation to the previous comment (line 307-311).
Line 308. I don’t think anyone argues the latter. We replaced “short-term” with “medium-term” in this statement for the argument to be in line with changes made in relation to the reviewer’s previous two comments (lines 307-311).
Line 310. I disagree. They are very useful for tracking fire risk through drought, and as you explain later, the 2017 fires were partially caused by drought. They are not a component of hourly or daily fire weather, but they are a component of weekly or monthly flammability fluctuations. We disagree that satellite-derived proxies for LIVE fuel moisture would improve fire danger forecasting in these evergreen sclerophyllous systems, because live fuel moisture is incredibly constant even through very significant droughts. We have actually measured live fuel moisture of fynbos prior to and during the worst drought in recorded history that preceded the 2017 fires and there was virtually no decline in live fuel moisture during this extreme drought (unpublished data which we are planning to publish soon). This is in accordance with the lack of a correlation found in the current study between live fuel moisture and fire weather (expressed as the Canadian FWI which incorporates medium-term drought indices). Hence our argument that severe droughts in these systems likely result in substantial loss of live material which then become flammable DEAD fuels, rather than significant declines in LIVE fuel moisture.
Line 324. The reason that fuel load is negatively related to ignitability is probably to do with leaf traits like thickness and surface area to volume ratio. This has been studied in some detail. This is not my area of research so I don’t know the studies, but I suggest you look one or two up. We now considered fuel load effects on other aspects of flammability (ignitability and rate of spread, with citations) (lines 352-353) and mentioned that it relates to fuel structural traits (lines 354-355).
Line 333. It strikes me that your method can be a source of variation and you should acknowledge that. For example, each time you put a branch on the apparatus, a different amount is touching the plate. Can you show this variation, e.g. in graphs of the actual data? We added a statement to acknowledge this additional potential source of variation (lines 362-365). We could add graphs showing species-level variation in flammability as a supplemental figure, but we felt that species-level investigations and related discussions were beyond the scope of this study and we intend to publish that as a separate paper.
Line 372. I like that you are considering this, but it is complicated. Most (but not all) dead leaves will drop to become litter, and litter has different flammability properties than leaves on the tree (probably a lot less flammable due to packing). This is a comment of the reviewer and does not require any change to the ms.
Figures. It would be nice to have a study area figure, either a map, or photos of the vegetation types or both We did not think that a map was productive use of space, given that the study conducted experimental burning of plant shoots, rather than being confined to a particular geographic space with distinct boundaries. A description of the location of the study region and coordinates were provided in lines 137-138. We have now provided a collage of photos of the different vegetation types showing their fuel structure and co-occurrence in the landscape as Supplemental Figure S1 and referred to it in the text in lines 101, 147 & 154-155.
" | Here is a paper. Please give your review comments after reading it. |
9,807 | "<ns0:abstract xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:p>Background: Pain assessment is a key measure to accompany treatments in a wide range of clinical settings. Low-cost valid and reliable pressure algometer would allow objective pressure pain assessment to assist a variety of health professionals. However, the pressure algometer is often expensive, which limits its daily use in both clinical settings and research. Objectives: The study aimed to assess the instrumental validity and the intra-and inter-rater reliability of an inexpensive digital adapted pressure algometer. Methods: A single rater applied 60 random compressions on a force platform.</ns0:p><ns0:p>The pressure pain thresholds of 20 volunteers were collected twice (three days apart) by two raters. The main outcome measures were the maximal peak force (in KPa) and the pressure pain threshold (adapted pressure algometer vs. force platform). Chronbach's α test was used to assess internal consistency. The standard error of measurement provided estimates of measurement error, and The Bland-Altman method estimated the measurement bias, with lower and upper limits of agreement. Results: No differences were observed when comparing the compression results. The internal consistency was moderate to excellent, with low standard error of measurement values. Moderate to excellent correlations were found, with a low risk of bias for all measures. The results showed both the validity and intra-rater reliability of the adapted pressure algometer.</ns0:p><ns0:p>Inter-rater reliability was moderate. Conclusion: The adapted pressure algometer provides valid intra-rater reliable measures of compressive force and pressure pain threshold, respectively. The results could potentially help to make the use of pressure pain threshold method more widespread among clinicians.</ns0:p></ns0:div>
</ns0:abstract>
<ns0:body xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:head>Introduction</ns0:head><ns0:p>Pain is mostly assessed by patient self-report, using the visual analogue scale of pain. <ns0:ref type='bibr' target='#b0'>1</ns0:ref> Self-reported pain intensity is important and reflects physiological and psychological features.</ns0:p><ns0:p>However, it can be difficult to interpret due to subjectivity and overestimation of pain level <ns0:ref type='bibr' target='#b1'>2</ns0:ref> .</ns0:p><ns0:p>Objective pain assessment is essential to establish a prospective evaluation, comparing baseline results to other timeline assessments or even as a prognostic measure that can predict future outcomes. <ns0:ref type='bibr' target='#b0'>1,</ns0:ref><ns0:ref type='bibr' target='#b3'>3</ns0:ref> The pressure pain threshold has been used to assist the diagnosis of pain by providing a quantified force value of tissue tenderness. <ns0:ref type='bibr' target='#b4'>4</ns0:ref> The pressure pain threshold occurs at the minimum transition point when the applied pressure is sensed as pain. <ns0:ref type='bibr' target='#b5'>5</ns0:ref> The pressure algometer is an equipment used to assess pressure pain threshold on both regional and widespread musculoskeletal pain. <ns0:ref type='bibr' target='#b6'>6</ns0:ref> The equipment includes a system to convert the force applied through a 1cm <ns0:ref type='bibr' target='#b1'>2</ns0:ref> pressure application surface and a readings display in Newtons (N/cm 2 ) or kilograms of force (Kgf/cm 2 ). The pressure algometer enables the rater to semi-objectively quantify the mechanical sensitivity to pain level and the recovery of underlying problems or soreness levels. <ns0:ref type='bibr' target='#b4'>4,</ns0:ref><ns0:ref type='bibr' target='#b7'>7</ns0:ref> Unfortunately, the commercially available pressure algometers are expensive and may require a software to see the results, which may demand more time and training to assess the pressure pain threshold. The validation of an easy-to-read low-cost adapted pressure algometer would enable widespread quantitative measures of pressure pain thresholds in clinical practice routine, <ns0:ref type='bibr' target='#b4'>4,</ns0:ref><ns0:ref type='bibr' target='#b8'>8</ns0:ref> Manuscript to be reviewed benefiting early assessment of pain conditions in low income and developing countries, mainly in primary care. A portable pressure algometer adapted from a hanging scale may be a costeffective alternative to ensure accurate algometry assessments.</ns0:p><ns0:p>The hanging scale is a battery-operated equipment used to weigh objects in a suspended manner.</ns0:p><ns0:p>The equipment uses a load cell, which is a metallic sturdy element, yet elastic enough for a load to deform it. The load cell is attached to a strain gauge, which reads the electrical resistance change when a pressure or traction load is placed in the load cell. The change in electrical resistance is converted to a digital signal by the strain gauge, and the result is readable on a display. <ns0:ref type='bibr' target='#b9'>9</ns0:ref> To determine the instrumental validity of an equipment, the correlation level between the score of a certain instrument and some external criterion has to be confirmed. This criterion has to be a widely accepted measure and considered gold-standard, with the same measurement characteristics of the assessment tool. <ns0:ref type='bibr' target='#b11'>10,</ns0:ref><ns0:ref type='bibr' target='#b12'>11</ns0:ref> The purpose of this study was to examine the instrumental validity, the intra-and inter-rater reliability of a low-cost pressure algometer, adapted from a hanging scale. The validity was assessed by comparing differences in the measurements of a series of random peak forces applied on a force platform, the gold standard for measuring vertical (compression) and horizontal forces. <ns0:ref type='bibr' target='#b8'>8,</ns0:ref><ns0:ref type='bibr' target='#b13'>12</ns0:ref> The current hypothesis is that a low-cost adapted pressure algometer has validity and reliability to be considered as a standard equipment to assess pressure pain threshold.</ns0:p></ns0:div>
<ns0:div><ns0:head>Materials & Methods</ns0:head></ns0:div>
<ns0:div><ns0:head>Equipment</ns0:head><ns0:p>All data was collected at the facilities of the Clinic-School of Physical Therapy -Federal University of Juiz de Fora -in May 2019. The adapted pressure algometer (MED.DOR Ltd., Brazil; maximum compression = 50 Kgf, precision = 0.1 Kgf, 3 digits display) had a 5-cm screw attached to the distal extremity. A 1-cm 2 round rubber application surface was attached to follow the standardization for pressure algometry (Figure <ns0:ref type='figure' target='#fig_4'>1</ns0:ref>).</ns0:p><ns0:p>A two-axis force platform (37 cm × 37 cm; PASCO, Pasport PS-2142, Roseville, USA), collected data using five force beams (sample rate = 1,000 Hz). Four beams in the corner were to measure the vertical force (range: −1,100 N to +4,400 N) and a 5 th beam measured the force in a parallel axis (range: −1,100 N to +1,100 N). The recorded trials were converted to KPa, as 1 Kg/cm 2 = 98.066 KPa.</ns0:p></ns0:div>
<ns0:div><ns0:head>INSERT FIGURE 1 HERE</ns0:head></ns0:div>
<ns0:div><ns0:head>Procedures</ns0:head></ns0:div>
<ns0:div><ns0:head>Instrumental validity</ns0:head><ns0:p>An independent trained rater performed 60 random 3-s pressure trials using the adapted pressure algometer with 3-s apart on the force platform. Data were collected and stored using the PASCO Capstone Software (Version 1.13.4, PASCO Scientific, 2019) and the adapted pressure algometer display readings were recorded through an off-board USB synchronized camera.</ns0:p></ns0:div>
<ns0:div><ns0:head>Intra-and inter-rater reliability</ns0:head><ns0:p>The middle Deltoid muscle's pressure pain threshold of 20 participants (10 women; 22±2 years; 63±13 Kg; 160±10 cm; 23±4 Kg/cm 2 ) were collected twice (3 days apart -day 1 and 2) by 2 trained independent raters. The exclusion criteria included: BMI>28 kg/cm 2 ; any self-reported health issues; 5-day alcohol consumption before the assessments; shoulder pain; previous shoulder surgery or any diagnosed shoulder or cervical impairment. The objectives of the study were explained to the subjects, and they were notified of the benefits and potential risks involved before signing an informed consent form prior to participation. The Federal University of Juiz de Fora ethics committee for human investigation approved the procedures employed in the study (Reference number: 02599418.9.0000.5147). The 4-day training consisted of applying constantprogressive pressure with the adapted pressure algometer on the laboratory-grade load cell Manuscript to be reviewed Biomedical Equipment, Porto Alegre, RS, Brazil). Then, a third rater monitored the pressure for other 2 consecutive days using the same software, but the raters in training did not received any visual feedback. The training aimed to ensure the adequate velocity to apply the pressure using the adapted algometer (1 kg/s).</ns0:p><ns0:p>To evaluate the intra-and inter-rater reliability, the follow positioning was adopted: 1) the participant remained seated with the feet on the floor; 2) the hands rested on the thighs; and 3) the trunk was erect positioned. These sites received progressive 1 kg/s pressure controlled by a metronome until the participant experienced pain <ns0:ref type='bibr' target='#b4'>4</ns0:ref> . An effort was made to standardize the anatomic locations at each session. The same examiner was responsible for palpating and marking the pressure pain threshold site on each subject before any measurements, both on day 1 and day 2. The middle Deltoid's site was topographically determined in the middle of a horizontal line drawn between the acromioclavicular joint and the Deltoid muscle insertion. <ns0:ref type='bibr' target='#b14'>13</ns0:ref> Three measurements were performed for each site, with 10 to 15 seconds apart. The first measurement was discarded. <ns0:ref type='bibr' target='#b15'>14,</ns0:ref><ns0:ref type='bibr' target='#b16'>15</ns0:ref> The participant lifted the opposite hand when the pressure pain threshold was achieved, i.e. when the applied pressure evoked pain. The examiner pressured the 'tare' button to lock the reading, immediately retracting the adapted pressure algometer. Then, the pressure pain threshold reading was registered. <ns0:ref type='bibr' target='#b17'>16</ns0:ref> </ns0:p></ns0:div>
<ns0:div><ns0:head>Statistical analysis</ns0:head><ns0:p>The recorded peaks were extracted. All trials were used for analysis, consisting a total of: 1) Sixty measures (validity analysis -force platform vs. adapted pressure algometer); 2) Eighty measures (reliability analysis). Data were presented as mean and standard deviation. The independent Student's t-test was used to compare differences between measures in the validity process. The intra-and inter-rater differences were compared using the mixed between (rater 1 vs. rater 2) and within-subjects analysis (moment and moment*rater) of variance with repeated measures. All data was reworked using the Holm's post hoc test to avoid multiple testing.</ns0:p><ns0:p>Significance was set at p<0.05. Intraclass correlation coefficients [ICC (2,1) ] were calculated to compare the results between both types of equipment and raters. Poor reliability was indicated by values less than 0.5, moderate reliability ranged from 0.5 to 0.75, values between 0.75 and 0.9 indicated good reliability, and an excellent reliability occurred when the values were greater than 0.90. <ns0:ref type='bibr' target='#b18'>17</ns0:ref> Chronbach's α test was used to assess the expected correlation of both types of equipment measuring the same construct. The standard error of measurement (SEM) was also calculated to provide an estimate of measurement error. A linear regression estimated the coefficient of correlation (r), the adjusted coefficient of determination (r 2 ). The Bland-Altman method estimated the measurement bias, with lower and upper limits of agreement between results. The statistics were performed using the JAMOVI software (JAMOVI project, version 0.9, 2018).</ns0:p></ns0:div>
<ns0:div><ns0:head>Results</ns0:head><ns0:p>Validity: Force platform vs. PA No significant differences were observed in pressure trials (Table <ns0:ref type='table' target='#tab_0'>1</ns0:ref>) between the adapted pressure algometer (405.63±235.34 KPa) and the force platform (434.15±239.18 KPa; p=0.25).</ns0:p><ns0:p>The ICC (2,1) and the Chronbach's α returned values of 0.98 and 0.99, respectively. The SEM returned a value of 0.005 Kgf, and the linear regression showed statically significant results (r=0.99; adjusted r 2 =0.99; p=0.001). The Bland-Altman results showed high levels of agreement (Figure <ns0:ref type='figure' target='#fig_5'>2</ns0:ref>).</ns0:p></ns0:div>
<ns0:div><ns0:head>INSERT FIGURE 2 HERE</ns0:head></ns0:div>
<ns0:div><ns0:head>Intra-and Inter-rater reliability</ns0:head><ns0:p>The pressure pain threshold from both raters showed very low variation over time (Rater 1: Day 1=203±74 KPa, Day 2=206±71.6 KPa; Rater 2: Day 1=214±73.7 KPa, Day 2=215±69.6 KPa).</ns0:p><ns0:p>The intra-rater comparison showed no significant differences (Moment: F=0.05; p=0.83 and Moment*Rater: F=0.01; p=0.93) (Table <ns0:ref type='table' target='#tab_0'>1</ns0:ref>). The ICC (2,1) and Chronbach's α analysis returned relevant values (Rater 1: ICC (2,1) =0.76, and Chronbach's α=0.85; Rater 2: ICC (2,1) =0.73, and Chronbach's α=0.84). The SEM values were low (Rater 1=0.02, and Rater 2=0.01), and moderate values were also obtained in the linear regression analysis (Rater 1: r=0.74; adjusted r 2 =0.52; Rater 2: r=0.73; adjusted r 2 =0.50). The Bland-Altman results showed high levels of agreement (Figure <ns0:ref type='figure' target='#fig_6'>3</ns0:ref>).</ns0:p></ns0:div>
<ns0:div><ns0:head>INSERT TABLE 1 HERE</ns0:head><ns0:p>PeerJ reviewing PDF | (2020:02:46274:1:2:NEW 19 Jun 2020)</ns0:p></ns0:div>
<ns0:div><ns0:head>Manuscript to be reviewed</ns0:head><ns0:p>The Inter-rater reliability showed no differences among measurements (F=0.22; p=0.64) (Table <ns0:ref type='table' target='#tab_0'>1</ns0:ref>), with moderate results for reliability analysis (Day 1: ICC (2,1) =0.56, and Chronbach's α=0.75; Day 2: ICC (2,1) =0.54, and Chronbach's α=0.71). The SEM result showed very low values (Day 1: 0.04, and Day 2: 0.02 Kgf), and moderate values on the linear regression analysis (Day 1: r=0.60; adjusted r 2 =0.33, and Day 2: r=0.55; adjusted r 2 =0.26). The Bland-Altman analysis showed acceptable levels of agreement (Figure <ns0:ref type='figure' target='#fig_6'>3</ns0:ref>).</ns0:p></ns0:div>
<ns0:div><ns0:head>INSERT FIGURE 3 HERE</ns0:head></ns0:div>
<ns0:div><ns0:head>Discussion</ns0:head><ns0:p>This study was designed to examine the instrumental validity and the intra-and inter-rater reliability of a low-cost pressure algometer, adapted from a hanging scale. The results showed no significant differences in peak compressive force recorded from the adapted pressure algometer and the force platform, with statistically significant results for the expected correlation of both equipment in measuring the same construct. These findings support the primary hypothesis, which contended that a low-cost pressure algometer has validity and reliability to be considered as a standard equipment to assess pressure pain threshold. Therefore, the tested device seems to be an alternative to expensive equipment.</ns0:p><ns0:p>Pain has been described as a multidimensional event, involving psychological and physical domains with different patterns depending on the pressure pain threshold site and emotional state. <ns0:ref type='bibr' target='#b19'>18</ns0:ref> These characteristics may impair conclusions and lead to biased clinical reasoning regarding group pain patterns due to intra-group and longitudinal variability in subjects' comorbidities and momentaneous emotional state. Nevertheless, the physical assessment is essential to provide prospective data comparing the effects of intervention for pain management. <ns0:ref type='bibr' target='#b20'>[19]</ns0:ref><ns0:ref type='bibr' target='#b23'>[20]</ns0:ref><ns0:ref type='bibr' target='#b24'>[21]</ns0:ref> Thus, the pressure algometry is important to diagnose some musculoskeletal problems. The diagnosis of fibromyalgia includes the pressure pain threshold as a key assessment to distinguish healthy individuals from those with fibromyalgia. <ns0:ref type='bibr' target='#b25'>22,</ns0:ref><ns0:ref type='bibr' target='#b26'>23</ns0:ref> Neck pain, cranio-cervical headache, and temporomandibular disorders also include the pressure pain threshold as an important component for clinical reasoning about the level of severity, influencing the treatment direction. <ns0:ref type='bibr' target='#b0'>1,</ns0:ref><ns0:ref type='bibr' target='#b5'>5,</ns0:ref><ns0:ref type='bibr' target='#b27'>24</ns0:ref> The validation procedure enables the adapted pressure algometer for clinical assessments in a practice routine, which may directly impact in primary Manuscript to be reviewed and ambulatory care of low-income and developing countries, by adding an objective and inexpensive tool to assess pressure pain threshold.</ns0:p><ns0:p>Previous studies showed acceptable levels of validity and reliability of other digital algometry systems. <ns0:ref type='bibr' target='#b4'>4,</ns0:ref><ns0:ref type='bibr' target='#b6'>6</ns0:ref> Balaguier et al. <ns0:ref type='bibr' target='#b28'>25</ns0:ref> found high reliability between all three pressure pain threshold measures at sites in the lower back. Walton et al. <ns0:ref type='bibr' target='#b0'>1</ns0:ref> found high reliability between measures in the upper fibers of the trapezius. Waller et al. <ns0:ref type='bibr' target='#b29'>26</ns0:ref> found high intra and inter rater-reliability (ICC=0.81-0.99; ICC=0.92-0.95, respectively) using 5 research assistants, each testing 20 painfree subjects at the wrist, leg, cervical and lumbar spine. However, those previous studies used commercial pressure algometers. For clinical and ambulatory settings, the high cost and the user's interface would be an issue to obtain fast objective pain measurements, requiring both training and experience for assessments. Brazilian physiotherapists have an average monthly salary of USD 500, according to the Occupational Brazilian Classification (https://www.salario.com.br/profissao/fisioterapeuta-geral-cbo-223605/). The adapted pressure algometer used in this study had a production cost of USD 10.00, while standard digital equipment cost range from USD 600.00 to USD 1,000.00.</ns0:p><ns0:p>Other protocols to assess pain, such as temporal summation and conditioned pain modulation, also include the pressure algometry. <ns0:ref type='bibr' target='#b30'>27</ns0:ref> The temporal summation is usually used to recognize central sensitization, an augmentation of responsiveness of central neurons to input from unimodal and polymodal receptors. <ns0:ref type='bibr' target='#b30'>27,</ns0:ref><ns0:ref type='bibr' target='#b31'>28</ns0:ref> Temporal summation can be elicited using a digital algometer, by 10 consecutive pressure pulses at pressure pain threshold intensity on the same place. For each pulse, the pressure has to be increased at a rate of 2 kg/s until the previously determined pressure pain threshold, where it was maintained for one second before being released. The assessed subject receives instructions to rate the pain intensity of the 1 st , 5 th , and 10 th pressure pulse according to a visual numerical rating scale. The final score is calculated by subtracting the 1 st score from the last one. The higher the final score, more efficient the nociceptive signaling to the brain. <ns0:ref type='bibr' target='#b32'>29</ns0:ref> The outcome of the processes involved in central sensitization, often a characteristic of chronic conditions, is an increased responsiveness to peripheral stimuli including the mechanical pressure. <ns0:ref type='bibr' target='#b30'>27</ns0:ref> The assessment of temporal summation includes the algometry combined with other measurements, such as subjective scales and validated questionnaires, and provides a multidimensional overview of pain in several situations of both acute and chronic musculoskeletal conditions. The conditioned pain modulation is often Manuscript to be reviewed used to evaluate the efficacy of endogenous pain inhibition, and can be induced by inflating an occlusion cuff on the arm, opposite of the test stimulus, to a painful intensity. The temporal summation procedure is repeated while wearing the cuff. The conditioned pain modulation starts when the cuff inflation is adjusted equal to a level 3 of 10 on the visual numerical rating scale.</ns0:p><ns0:p>Then, the arm is rested on a table while the temporal summation assessment is repeated at the opposite side. The efficacy of conditioned pain modulation is examined by subtracting the rating scale score at the 1 st pressure pulse prior to and during cuff inflation. The efficacy of conditioned pain modulation on temporal summation is assessed by subtracting the rating scale at the 10 th pressure pulse prior to and during cuff inflation. <ns0:ref type='bibr' target='#b31'>28</ns0:ref> As one of the mechanical stimulation for the above mentioned protocols is the pressure pain threshold stimuli, the validity and reliability of the adapted pressure algometer might ensure its usefulness for those protocols.</ns0:p><ns0:p>It is also important to note that no significant differences were found for intra/inter-rater reliability. The pressure pain threshold in body sites other than Deltoid muscle must be assessed to ensure the adapted pressure algometer validity on those sites. However, we hypothesize that they should not give any different results to direct assessment using the adapted pressure algometer, since the standard deviation remained at very low values and the current results gave very good measures compared to the force platform and additional good reliability. In fact, the instrumental validity of an equipment's measures also ensures unbiased assessments. <ns0:ref type='bibr' target='#b33'>30</ns0:ref> Other studies have identified different factors to consider when evaluating the pressure pain threshold, such as gender and obesity. <ns0:ref type='bibr' target='#b34'>31,</ns0:ref><ns0:ref type='bibr' target='#b35'>32</ns0:ref> A review of studies involving induced pain found a consistent pattern of women exhibiting greater pain sensitivity and a reduction in pain inhibition compared to men. <ns0:ref type='bibr' target='#b36'>33</ns0:ref> In addition, the characteristic of pain imposed is an important factor for these differences, since the type of pressure pain has one of the highest effect sizes in the pain report. <ns0:ref type='bibr' target='#b16'>15,</ns0:ref><ns0:ref type='bibr' target='#b37'>34</ns0:ref> It is suggested that interactions between biological and psychosocial factors are responsible for these gender differences, but all studies indicate the need for additional research to elucidate the mechanisms that drive gender differences in pain responses. <ns0:ref type='bibr' target='#b34'>31,</ns0:ref><ns0:ref type='bibr' target='#b36'>33,</ns0:ref><ns0:ref type='bibr' target='#b37'>34</ns0:ref> Some studies suggest that in areas with additional subcutaneous fat, pain thresholds for electrical or pressure stimuli increase and pain sensitivity decreases in obese individuals. <ns0:ref type='bibr' target='#b35'>32,</ns0:ref><ns0:ref type='bibr' target='#b38'>35</ns0:ref> A study have also shown biochemical changes in trigger points with higher levels of inflammatory mediators, catecholamines and cytokines in obese individuals. <ns0:ref type='bibr' target='#b39'>36</ns0:ref> Mechanical stretching of the skin in response to excess fat can lead to a decrease in the density of nociceptive fibers, and obesity is associated with the chemical inhibition of pain with an increase in β endorphin and endogenous opioid peptide. <ns0:ref type='bibr' target='#b35'>32</ns0:ref> The present study balanced the sample concerning gender factor, and all participants were classified as normal according to their body mass index. However, the current sample was chosen only for reliability analysis purposes. Pressure pain threshold as a clinical result is well established, but more studies should take into account sex and body mass index differences to avoid bias in experimental protocols. <ns0:ref type='bibr' target='#b34'>31</ns0:ref> Pressure pain threshold was also positively but poorly correlated with high-density lipoprotein cholesterol. <ns0:ref type='bibr' target='#b41'>37</ns0:ref> A high pressure pain threshold was also found among subjects with hyperglycemia and excessive alcohol consumption. <ns0:ref type='bibr' target='#b41'>37</ns0:ref> In the present study, no blood assessment was performed to exclude those factors. However, the sample was constituted by young adults, decreasing the chance of any important health issues.</ns0:p><ns0:p>Additionally, exclusion criteria included previous excessive alcohol consumption.</ns0:p><ns0:p>Considering its portability, easy assemblage, and the lower cost, the currently tested device seems to be a valid standard equipment for pressure pain threshold assessment. Therefore, the adapted pressure algometer is a valid method to assess compression compared to a force platform. The portability, cost-effectiveness, and friendly user system provide an effective way to measure pressure pain threshold.</ns0:p></ns0:div>
<ns0:div><ns0:head>Conclusions</ns0:head><ns0:p>The current hypothesis is that a low-cost pressure algometer has validity and reliability enough to be considered as a standard equipment to assess pressure pain threshold. The results showed that the low-cost adapted pressure algometer is a valid and reliable tool to assess compression measurements, including the pressure pain threshold. A low-cost pressure pain threshold assessment is possible using the adapted pressure algometer. Future directions are including the assessment in clinical routine to spread the systematic evaluation of pressure pain. Further studies should consider other assessments such as temporal summation and conditioned modulated pain using the pressure algometer. Manuscript to be reviewed Manuscript to be reviewed </ns0:p><ns0:note type='other'>Figure 1</ns0:note><ns0:note type='other'>Figure 2</ns0:note><ns0:note type='other'>Figure 3</ns0:note></ns0:div><ns0:figure xml:id='fig_0'><ns0:head></ns0:head><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:02:46274:1:2:NEW 19 Jun 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_1'><ns0:head>(</ns0:head><ns0:label /><ns0:figDesc>Miotec™ Biomedical Equipment, Porto Alegre, RS, Brazil; maximum tension-compression = 200 kgf, precision = 0.1 kgf, maximum error = measurement = 0.33%) with Miotec™ software for visual feedback (MioTrainer™, Biomedical Equipment, Porto Alegre, RS, Brazil) for 2 nonconsecutive days (3 non-consecutive hours per day). The conversion from analog to digital signals was performed by an A/D board (Miotec™, Biomedical Equipment) with 16-bit resolution input range, a sampling frequency of 2 kHz, common rejection module greater than 100 dB, signal-noise ratio less than 03 μV, Root Mean Square and impedance of 109 Ω. All pieces of information were recorded and processed using the software Miotec Suite™ (Miotec™ PeerJ reviewing PDF | (2020:02:46274:1:2:NEW 19 Jun 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_2'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:02:46274:1:2:NEW 19 Jun 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_3'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:02:46274:1:2:NEW 19 Jun 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_4'><ns0:head>Figure 1 .</ns0:head><ns0:label>1</ns0:label><ns0:figDesc>Figure 1. Adapted pressure algometer -PA.</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_5'><ns0:head>Figure 2 .</ns0:head><ns0:label>2</ns0:label><ns0:figDesc>Figure 2. Bland-Altman plot: instrumental validity.</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_6'><ns0:head>Figure 3 .</ns0:head><ns0:label>3</ns0:label><ns0:figDesc>Figure 3. Bland-Altman plot: Intra-rater reliability.</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_7'><ns0:head>Figure 3 .</ns0:head><ns0:label>3</ns0:label><ns0:figDesc>Figure 3. Bland-Altman plot: Intra-rater reliability. A) Rater 1: Bias = -2.55 (95% confidence interval [CI] = -24.3 to 22.2); lower limit of agreement (LLA) = -106.38 (95% CI = -149 to -63.3); upper limit of agreement (ULA) = 101.28 (95% CI = 58.2 to 144.4). 2). B) Rater 2: Bias = -1.03 (95% CI = -25.9 to 23.9); LLA = -105.28 (95% CI = -148.6 to -62); ULA = 103.22 (95% CI = 59.9 to 146. Inter-rater reliability. C) Day 1: Bias = -10.8 (95% CI: -41.6 to 20); LLA = -139.8 (95% CI: -193.4 to -86.3); ULA = 118.2 (95% CI: 64.7 to 171.8); D) Day 2: Bias = -9.27 (95% CI = -40.7 to 22.1); LLA = -140.77 (95% CI = -195.4 to -86.2); ULA = 122.23 (95% CI = 67.6 to 176.8).</ns0:figDesc></ns0:figure>
<ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='20,42.52,70.87,525.00,393.75' type='bitmap' /></ns0:figure>
<ns0:figure type='table' xml:id='tab_0'><ns0:head>Table 1 :</ns0:head><ns0:label>1</ns0:label><ns0:figDesc>Validity, Intra-and Inter-rater reliability pairwise comparisons.</ns0:figDesc><ns0:table><ns0:row><ns0:cell>95% Confidence Interval</ns0:cell></ns0:row></ns0:table><ns0:note>PeerJ reviewing PDF | (2020:02:46274:1:2:NEW 19 Jun 2020)</ns0:note></ns0:figure>
</ns0:body>
" | "Technical Changes
1: Re-used Text
We noted that the amount of re-used text has increased since you initially submitted your revised manuscript:
https://www.tandfonline.com/doi/abs/10.1080/14763141.2016.1213875?journalCode=rspb20 (e.g., lines 89-92)
http://repositorio.ufjf.br:8080/jspui/bitstream/ufjf/11330/1/arianykleintahara.pdf (e.g., lines 114-116, 118-123, 127-129) . https://pubmed.ncbi.nlm.nih.gov/27330520/ (e.g., lines 150-153)
We are unable to consider your submission unless you address this issue and submit a new version of the manuscript.
We appreciate that the same methods are often reused and there are only so many ways to describe a procedure.
We are, of course, mostly concerned with unacknowledged text overlap, so this is probably best addressed by citing the paper where each method was first used in the Methods section.
A suitable wording might be something like 'Data were collected as previously described in AUTHOR_NAME (DATE). Specifically...'.
For convenience, you might consider uploading your protocols to a service such as protocols.io for ease of citation.
We are sending you a copy of the report that we use to identify areas where text has been reused. Please refer to this document when editing your manuscript.
ANSWER: Thank you for your concern about our manuscript. However, the sentences you highlighted in our text are descriptions of the devices, to ensure reproducibility in further studies. Please, see:
Lines 114-116:
progressive pressure with the adapted pressure algometer on the laboratory-grade load cell (Miotec™ Biomedical Equipment, Porto Alegre, RS, Brazil; maximum tension‐compression = 200 kgf, precision = 0.1 kgf, maximum error = measurement = 0.33%)
Miotec = the company in charge to produce the load cell, city, state, country. This is reported according to ICMJE. This statement includes the specifications of the load cell, again, to ensure reproducibility. This is nos plagiarism, but the correct way to report the specs.
The same occurred in Lines 118-123:
The conversion from analog to digital signals was performed by an A/D board (Miotec™, Biomedical Equipment) with 16‐bit resolution input range, a sampling frequency of 2 kHz, common rejection module greater than 100 dB, signal‐noise ratio less than 03 μV, Root Mean Square and impedance of 109 Ω. All pieces of information were recorded and processed using the software Miotec Suite™ (Miotec™ Biomedical Equipment, Porto Alegre, RS, Brazil).
Again, not plagiarism, but an adequate form to report the equipment's specifications to ensure reproducibility. I already used the same text in several other manuscripts, always reporting the same specs, without characterizing any form of plagiarism.
The lines 89-92, 127-129, and 150-153 were restructured to avoid any additional problem.
2: References
In the reference section, please provide the full author name lists for any references with 'et al.' including, but not limited to, these references:
Zhang Y, Zhang S, Gao Y, et al.
Zhang Y, Zhang S, Gao Y, et al.
Shah JP, Danoff J V., Desai MJ, et al.
ANSWER: All references were doubled-checked.
2: Affiliations
We notice that the author affiliations (Carolina F dos Anjos, Maria C Macedo, Ilha G Fernandes, Esteban A Muñoz, Alexandre Barbosa) you have provided in the system are slightly different to those in the document.
System version:
Department of Physical Therapy, Universidade Federal de Juiz de Fora, Governador Valadares, Minas Gerais, Brazil
Department of Physical Education, Universidade Federal de Juiz de Fora, Governador Valadares, Minas Gerais, Brazil
Manuscript version:
Department of Physical Therapy, Federal University of Juiz de Fora, Governador Valadares, Minas Gerais, Brazil
Department of Physical Education, Federal University of Juiz de Fora, Governador Valadares, Minas Gerais, Brazil
As our system will use the author names/affiliations entered in the system for publication, and the text document for a reference, please ensure that both versions are complete and the same. Please do not include any address information such as street addresses or postal codes.
Please edit the author affiliations using the 'Edit' button to the right of the names here, or edit your manuscript source file and upload it here.
ANSWER: Thank you for you correction. All affiliations were corrected.
3: References
In the reference section, please provide the full author name lists for any references with 'et al.' including, but not limited to, these references:
Walton D, Macdermid J, Nielson W, et al.
González-Fernández M, Ghosh N, Ellison T, et al.
Walton DM, Macdermid JC, Giorgianni AA, et al.
If you have used EndNote, you can change the references using the steps provided on our author instructions.
ANSWER: Thank you for your attention. We opt to use the AMA 10th edition style for now, so all authors’ names were reported.
4: Figures
Figure 3 has multiple parts that should be relabeled. Each figure with multiple parts should have alphabetical (e.g. A, B, C) labels on each part and all parts of each single figure should be submitted together in one file.
In this case: the 4 parts of Figure 3 should be labeled A-D, and the numerical labels (1, 2, 3, and 4) should be removed.
Please provide a replacement figure measuring minimum 900 pixels and maximum 3000 pixels on all sides, saved as PNG, EPS, or PDF (vector images only) file format without excess white space around the images. The file name should be formatted as 'Figure3.png'.
ANSWER: Thank you for your attention. The files were all corrected and uploaded.
Editor comments (Denis Marcellin-Little)
MAJOR REVISIONS
Thank you for your submission. The reviewers have recommended a major revision of the manuscript and have made specific recommendations. Please address each of these points in your revised manuscript and include an itemized reply to each reviewer.
ANSWER: THANK YOU FOR YOUR ATTENTION. WE REWORKED THE STATISTICS AND PERFORMED ALL CORRECTIONS ADDRESSED BY BOTH REVIEWERS.
Reviewer 1 (Pascal Madeleine)
Comments for the Author
Manuscript Number: PeerJ 46274
Title: Instrumental validity and intra/inter-rater reliability of a novel low-cost digital pressure algometer
Main comments
This study aimed to assess the instrumental validity, the intra- and inter-rater reliability of an inexpensive digital adapted pressure algometer. The aim is of interest for clinical research related to musculoskeletal disorders.
ANSWER: THANK YOU SO MUCH FOR YOUR TIME CORRECTING OUR MANUSCRIPT. WE TRIED TO IMPROVE IT ACCORDING TO YOUR GUIDANCE. PLEASE, SEE OUR ANSWERS BELOW EACH OF YOUR COMMENTS.
The paper is generally well written and of interest. There are few flaws that need to be addressed in a revision and this is definitely feasible. It is erroneous to report pressure in Kg or N – the authors should comply with the SI meaning that pressure is reported in Pa (or kPa). Then all results related force expressed in Kg or N needs to be converted into Pa (see e.g. https://www.convertunits.com/from/kg/cm2/to/kpa ). Make also changes elsewhere (abstract, introduction and discussion).
ANSWER: THANK YOU FOR YOUR ATTENTION. ALL DATA WAS CONVERTED, REASSESSED AND THE UNIT CHANGED TO KPA.
More the choice of force platform and load cell should be justified better.
ANSWER: THANK YOU FOR YOUR COMMENT. THE AUTHORS DECIDED TO STAND WITH THE FORCE PLATFORM RESULTS ALONE, AS IT IS THE GOLD STANDARD. HOWEVER, THE LABORATORY-GRADE LOAD CELL DESCRIPTION IS STILL IN THE MANUSCRIPT DUE TO THE RATER’S TRAINING TO APPLY THE PRESSURE FOR THE RELIABILITY ANALYSIS. ANY OTHER ADDITIONAL MEASUREMENTS USING THE LABORATORY-GRADE LOAD CELL WERE DELETED FROM THE MANUSCRIPT.
Finally, it would be good how the re-calibration should be done so potential drift does not influence PPT collected longitudinally.
ANSWER: WE APOLOGIZE, BUT COULD YOU CLARIFY WHAT “RE-CALIBRATION” YOU REFER?
See below for other suggestions for improving the manuscript. Reduce the number of abbreviations.
ANSWER: AGREED. WE SIGNIFICANTLY REDUCED THE NUMBER OF ABBREVIATIONS. PLEASE, CHECK THE CORRECTED MANUSCRIPT.
Specific comments
In general, make clear if PA relates to the expensive or cheap algometer.
ANSWER: THANK YOU FOR YOUR ATTENTION. WE CORRECTED ALL PA MENTIONS THROUGHOUT THE TEXT. PLEASE, REFER TO OUR CORRECTED MANUSCRIPT.
Line 52-53 “quantify the pain level and the recovery of underlying problems or soreness levels” PA does not quantify pain levels – it is more appropriate to write “mechanical sensitivity to pain” see a recent PPT review https://www.ncbi.nlm.nih.gov/pubmed/29403305. It probably better to refer PPT as a semi-objective method (the patient still reports when pressure turns into pain)
ANSWER: AGREED. WE CHANGED THE SENTENCE. PLEASE, CHECK OUR CORRECTED DRAFT.
Line 54 define expensive/inexpensive
ANSWER: TO AVOID ANY MISINTERPRETATION, THE WORD WAS CHANGED TO “LOW-COST”.
Paragraph equipment
It is a bit unclear why both force platform and load cell are used considering the current goal? Express 1N=1/g where g=9.82 ms/s^2
ANSWER: THANK YOU FOR YOUR COMMENT. AS PREVIOUSLY MENTIONED, THE LOAD CELL’S MEASUREMENTS WERE DISCARDED TO FOCUS ON THE FORCE PLATFORM RESULTS (GOLD-STANDARD EQUIPMENT). ALSO, AS YOU SUGGESTED, WE OPT TO SHOW THE RESULTS IN KPA. PLEASE, SEE THE CORRECTED VERSION.
Line Did the rater/participant were familiarized with the experimental procedure?
ANSWER: THANK YOU FOR YOUR ATTENTION. YES, WE DESCRIBED THE RATER’S TRAINING PROTOCOL. THE PARTICIPANTS RECEIVED EXPLANATION ABOUT THE PROCEDURES, AND THE 1ST OF 3 MEASURES WAS DISCARDED.
Making a PPT at another body location
ANSWER: PLEASE, CLARIFY YOUR COMMENT.
Line 109-110 Add demographics of the men
ANSWER: THANK YOU FOR YOUR COMMENT. HOWEVER, THE DEMOGRAPHICS ARE FOR BOTH, MEN AND WOMEN. IN PARENTHESIS WE SPECIFIED THAT FROM 20 PARTICIPANTS, 10 WERE WOMEN AND, CONSEQUENTLY, 10 WERE MEN.
Line 111 IMC?
ANSWER: THANK YOU FOR YOUR ATTENTION. SORRY, OUR MISTAKE. WE CHANGED IMC TO BMI.
Line 130 temporal summation issue
ANSWER: THANK YOU FOR YOUR COMMENT. WE EXCLUDED THOSE PARTICIPANTS WITH ANY POSSIBLE PROBLEM LEADING TO ANY TEMPORAL SUMMATION POSSIBILITY. IT WOULD REALLY BE A PROBLEM IF THE SHOULDER PAIN (OR REFERRED PAIN FROM CERVICAL, FOR EXAMPLE) WAS PRESENT. ALSO, TEMPORAL SUMMATION IS A SIGN OF CENTRAL SENSITIZATION, WHICH IS OFTEN PRESENT IN CHRONIC PATIENTS (https://www.ncbi.nlm.nih.gov/pubmed/20036180). ALSO, THE PROTOCOL TO IDENTIFY THIS PROBLEM WOULD INCLUDE AT LEAST 10 CONSECUTIVE PRESSURES TO EVOKE SUCH SUMMATION EFFECT, WHILE IN HEALTHY SUBJECTS, THE SUMMATION WOULD NOT BE OBSERVED AS A RELEVANT EFFECT. WE ALSO STATE THAT THE PROCEDURE WAS THE SAME FOR ALL PARTICIPANTS, CONSISTING, IF PRESENT, IN A SYSTEMATIC ERROR, MEASURED BY THE SEM, AND THE BLAND-ALTMAN METHOD, AS WE INCLUDED THE LAST 2 MEASUREMENTS TO ASSESS SUCH PROPERTY.
Line 144 why choosing an ICC(1,1)? Method?
ANSWER: THANK YOU FOR YOUR COMMENT. OUR MISTAKE IN REPORTING THE ICC TYPE. THE METHOD WAS ICC (2,1), WHERE EACH SUBJECT IS ASSESSED BY A DIFFERENT SET OF RANDOMLY SELECTED RATERS, AND RELIABILITY IS CALCULATED BY TAKING AN AVERAGE OF THE K RATERS’ MEASUREMENTS. WE CORRECTED IN OUR CURRENT VERSION.
Did you correct for multiple comparison (t-test)
ANSWER: THANK YOU FOR YOUR COMMENT. REVIEWING THE STATISTICS AFTER YOUR COMMENT, WE NOTICED THAT WE HAVE USED THE WRONG VALUES FOR ALL COMPARISONS. WE APOLOGYSE FOR THAT MISTAKE AND THANK YOU FOR YOUR ATTENTION. SO, WE DOUBLED-CHECK OUR ORIGINAL FILES FROM DATA COLLECTION, AND PERFORMED THE STATISTICS ALL OVER AGAIN WITH ALL VALUES PREVIOUSLY CHANGED TO KPA. WE ALSO CHANGED THE RELIABILITY ANALYSIS TO AVOID MULTIPLE TESTING TO ANOVA REPEATED MEASURES. AS WE MEASURED DISTINCT PROPERTIES, THE VALIDITY DIFFERENCES WERE PERFORMED WITH THE INDEPENDENT STUDENT T-TEST, AND THE INTRA- AND INTER-RATER RELIABILITY, WITH THE MIXED ANOVA WITH REPEATED MEASURES.
Results
Report results with the same number of decimals
ANSWER: AGREED AND CORRECTED.
How are the ICC values interpreted? Add a reference (e.g. Landis and Koch 1977)
ANSWER: THANK YOU FOR YOUR COMMENT. WE ADDED A MORE RECENT REFERENCE AND HOW THE ICC VALUES WERE INTERPRETED.
Discussion
Add also studies mentioning the use of PPT in rehabilitation studies (e.g. rehabilitation, training or ergonomics interventions)
Line 203 add threshold at the end of the sentence
ANSWER: THANK YOU FOR YOUR ATTENTION. CORRECTED.
Line 205 add also information concerning spatial aspects, i.e. information the changes in PPT as a function of location – see also line 228 about temporal summation. Here, the authors could consider adding aspect related to spatial summation too.
ANSWER: THANK YOU FOR YOUR COMMENT. HOWEVER, WE HUMBLY UNDERSTAND THAT THE CONCEPT COULD NOT IMPROVE THE RATIONALE FOR THE CURRENT STUDY. THE SPATIAL SUMMATION OR AREA-BASED SUMMATION REFERS TO THE INCREASE IN PAIN EVOKED BY INCREASING THE AREA OF STIMULATION (https://www.sciencedirect.com/science/article/abs/pii/S1526590014009729), WHICH IS NOT RELATED TO THE ADAPTED ALGOMETER VALIDATED IN OUR STUDY AS THE DEVICE HAS A FIXED AREA OF 1-CM2. THE MAIN GOAL IS TO VALIDATE THE DEVICE. SO, OUR OPINION IS THAT OTHER PROTOCOLS HAVE TO BE APPRECIATED IN FURTHER STUDIES, IF NECESSARY. IF YOU HAVE A DIFFERENT VIEWPOINT, PLEASE CLARIFY.
Check references for normal/capital letters in the reference list
ANSWER: THANK YOU FOR YOUR ATTENTION. WE USE THE MENDELEY SOFTWARE TO MANAGE OUR REFERENCES. WE ALSO DOUBLED-CHECKED THE REFERENCES AT THE CURRENT VERSION.
Figures change to SI units (kPa)
ANSWER: AGREED AND CORRECTED.
Add also a table with the measured values PPTs, CI and differences. See Walton et al (2011)
ANSWER: THANK YOU FOR YOUR COMMENT. THE SUGGESTED TABLE IS MERGED IN OUR NEW VERSION. PLEASE, CHECK THE CORRECTED MANUSCRIPT.
Reviewer 2 (Andrea Tomas)
Basic reporting
Grammatically, the paper is poorly written and it is often challenging to extract the meaning from various sections, particularly the introduction and the discussion. I would suggest having someone review the paper for English fluency prior to resubmission.
ANSWER: THANK YOU FOR YOUR COMMENT. WE SUBMITTED TO A NATIVE SPEAKER TO IMPROVE CLARITY. PLEASE, CHECK OUR NEW VERSION.
The flow of the paper is further complicated by a lack of structured organization within each section, making it challenging to follow.
ANSWER: THANK YOU FOR YOUR COMMENT. WE TRIED TO CHANGE THE STRUCTURE TO IMPROVE CLARITY. PLEASE, CHECK THE CORRECTED MANUSCRIPT.
The introduction should be the clearest part of a scientific article and unfortunately this paper's introduction is very confusing. The reader should be presented with thorough background information as well as a clear description of what is being studied in the present paper.
ANSWER: THANK YOU FOR YOUR COMMENT. WE REORGANIZED THE OVERALL INTRODUCTION AND DISCUSSION STRUCTURES. PLEASE, CHECK OUR NEW VERSION.
The fact that a pressure algometer was used on both a load cell and force platform is hinted at but is not made clear in the introduction, and in fact it does not become clear until reading the materials and methods.
ANSWER: THANK YOU FOR YOUR COMMENT. WE DISCARDED THE LOAD CELL ANALYSIS, AS THE RESULTS REMAINED THE SAME FOR BOTH DEVICES AND THE FORCE PLATFORM IS THE GOLD-STANDARD FOR COMPRESSIVE FORCES.
In addition, there is no clear description of how the adapted portable hanging scale is different from currently used pressure algometers. Some of the content of the introduction should be condensed into one sentence to state the potential use of it in low income countries instead of mentioning this multiple times.
ANSWER: THANK YOU FOR YOUR ATTENTION. THE MAIN DIFFERENCES ARE THE COST AND THE EASY-TO-READ DISPLAY, INSTEAD A DEDICATED SOFTWARE TO ASSESS THE RESULTS. WE INCLUDED THIS EXPLANATION IN OUR NEW VERSION. WE ALSO CONDENSED IN ONE PARAGRAPH THE SENTENCES INVOLVING THE POTENTIAL USE OF THE ADAPTED PRESSURE ALGOMETER IN LOW INCOME CONTRIES.
Pressure pain threshold (PPT) is explained well in the beginning of the introduction.
ANSWER: THANK YOU FOR YOUR KIND COMMENT.
There is inconsistency in the manner in which authors of other studies are cited throughout the paper, especially in the discussion.
ANSWER: THANK YOU FOR YOUR ATTENTION. WE USE THE MENDELEY SOFTWARE TO MANAGE OUR REFERENCES. WE ALSO DOUBLED-CHECKED THE REFERENCES AT THE CURRENT VERSION.
Experimental design
The study is very interesting, and indeed, has the potential for a real clinical impact. Unfortunately, the language overcomplicates the description of the study design. I am not sure what is the relevance of reporting the how people were trained on non-consecutive days/hours?
ANSWER: THANK YOU FOR YOUR COMMENT. IT IS IMPORTANT TO ENSURE THAT THE RATERS WERE WELL TRAINED TO APPLY THE PRESSURE WITH THE IDEAL VELOCITY (1 KG/S). SO, WE TRAINED BOTH RATERS FIRSTLY WITH VISUAL FEEDBACK (USING THE LOAD CELL AND A SOFTWARE FOR VISUAL FEEDBACK), AND THEN WE CERTIFIED THAT THEY WERE ABLE TO REPRODUCE THE SAME VELOCITY WITHOUT VISUAL FEEDBACK (MONITORING THE APPLIED PRESSURE WITH THE SAME SOFTWARE), AS EXPLAINED IN DETAIL IN THE METHODS SECTION.
Specific comments in attachment.
Validity of the findings
The hypothesis stated in the discussion is not the same as the hypothesis used in the introduction. Please clarify.
ANSWER: THANK YOU FOR YOUR COMMENT. WE CORRECTED THE HYPOTHESIS IN OUR DISCUSSION SECTION TO ENSURE CLARITY. PLEASE, CHECK OUR CORRECTED VERSION.
The authors should explain clearly the differences between commercially used PA and the equipment they used in this study (adapted PA?).
ANSWER: THANK YOU FOR YOUR COMMENT. THE MAIN DIFFERENCES IS THE PRICE. WE TRIED TO EXPLAIN THE IMPORTANCE OF A LOW-COST EQUIPMENT IN OUR CURRENT VERSION BY AN EXAMPLE OF PHYSIOTHERAPIST’S SALARY IN BRAZIL.
In addition, it is important that they are consistent in referring to their equipment as an adapted PA as in the caption of Figure 1. There are multiple times in this paper where the adapted PA is referred to simply as the PA.
ANSWER: THANK YOU FOR YOUR COMMENT. THE ISSUE WAS CORRECTED THROUGHOUT THE TEXT. PLEASE, SEE OUR CORRECTED VERSION.
Comments for the Author
This is a well designed study with the potential for clinical impact. The problems are not with the methodology, but with the writing and presentation of the findings. Although this paper requires significant editing to improve the language and overall flow, I believe that will make it much more accessible and easier to follow when reading it.
ANSWER: THANK YOU FOR YOUR COMMENT. WE RESTRUCTURED THE SECTIONS ABOVE MENTIONED. WE HOPE ALL CHANGES WOULD BE ENOUGH TO IMPROVE CLARITY.
Review Peer J
Instrumental validity and intra/inter-rater reliability of a novel low-cost digital pressure algometer
Abstract
Line 22- Missing verb…would allow pressure pain assessment to _____a variety….
ANSWER: Verb included. Thank you.
Line 24- Exchange ‘,’ for the word AND
ANSWER: the ‘,’ was changed. Thank you for your attention.
Line 26- It should say 60 random compressions (number before the word)
ANSWER: Corrected.
Line 28- Not sure if use of numerals is correct if value is less than 10
ANSWER: Corrected.
Line 36- Sentence starting with The results….
The description of the conclusion, second sentence in line 36 is unclear, as the findings do not ensure anything. If the conclusion is that the results could potentially help make the use of this method more widespread among clinicians, that is what should be emphasized.
ANSWER: Thank you for your comment. We corrected the sentence as you suggested.
Conclusion should be organized that the findings are grouped together first and that the potential usefulness of this information is stated last.
ANSWER: Thank you for your comment. We restructured the abstract results and conclusion sections. Please, see our corrected version.
Line 38- Remove first ‘,’
ANSWER: Corrected.
Introduction
Line 42- It should say Visual ‘analogue’ scale (not analogical)
ANSWER: Thank you for your attention. Corrected.
Line 45- It is not clearly written. Please rewrite for clarity
ANSWER: Thank you for your comment. The sentence was restructured to provide clarity.
Line 47- Add the word ‘by’ between the word diagnosis of pain and providing
ANSWER: Corrected.
Line 57- The word weighting does not appear to be used correctly here
ANSWER: Thank you for your comment. We changed the sentence to provide clarity. Please, see our new version.
Line 61- Describing the force platform is out of place in this paragraph. As it is written, it is unclear if the force platform is used as a separate comparison or if it is a component of the design with adapted hanging potable scale.
The focus of introduction should be to present relevant background information and to introduce specific method being tested. If a force platform is used for comparison purpose only, this should be discussed in the later section and not discussed here.
ANSWER: Thank you for your comment. The sentence was replaced to improve clarity. Please, check our corrected manuscript.
Line 63- Incorrect use of the word ‘Property’.
ANSWER: As We restructured the sentences, the word was deleted.
Line 64-66- This is redundant as this was already mentioned briefly in Line 54. Consider removing the sentence in Line 54
ANSWER: Thank you for your attention. All sentences were restructured to avoid redundancy. Please, see our corrected manuscript.
Line 67-69- This is also a bit redundant and should be combined with previous statements regarding potential use in low-income and developing countries.
ANSWER: Thank you for your attention. All sentences were restructured to avoid redundancy. Please, see our corrected manuscript.
Line 74- Word ‘Enough’ before the words validity and reliability
ANSWER: The word was removed. Thank you for your attention.
M&M
Line 102- It should be stated that the pressure was applied using pressure algometer.
ANSWER: Thank you. The statement was included.
Line 117- 122- Unclear and confusedly written. Are all the details regarding non-consecutive days/hours relevant to the paper?
ANSWER: Thank you for your comment. The sentences were restructured. We do believe the details are important to provide reproducibility, if necessary. We included the sentence: “The training aimed to ensure the adequate velocity to apply the pressure using the adapted algometer (1 kg/s).”
Line 130-132- Explain the relevance of this statement and also the statement is poorly written.
ANSWER: Thank you. The 1st measurement is often discarded due to inaccuracy (especially if the subject is not used to 1-cm2 stimuli). We followed the same protocol of previous studies. The sentences were restructured to improve clarity. Please, check our new version.
ANSWER:
Results
Line 155 and 162 should say ‘No significant differences’
ANSWER: Corrected. Thank you for your attention.
Line 158, 164, 171, 173, 177, 179- ‘Very good’ is a subjective assessment and should not be used in the Results section, Use ‘statistically significant’ instead if appropriate
ANSWER: Thank you for your comment. We replaced the words ‘very good’ for other (statistically significant, moderate, relevant). Please, check our new version.
Discussion
Line 186 and 187- No significant difference
ANSWER: Corrected. Thank you for your attention.
Line 188- This is not hypothesis as stated
Isometric peak compression force was never stated in the hypothesis in Introduction part.
ANSWER: Thank you for your comment. The sentence was restructured according to the hypothesis.
Line 189- Authors need to be more specific about which equipment they are referring to. In this line PA is referred to as an alternative to an expensive equipment. In line 54 it is stated that PA is often expensive.
ANSWER: Thank you for your attention. All mentions were carefully reviewed and restructured to improve clarity.
Line 190- This should be a new paragraph
ANSWER: Thank you. Corrected.
Line 195-Line 196- ‘Diagnostic tool’ should be used instead of ‘diagnostics’.
ANSWER: Thank you for your comment. The whole sentence was restructured to improve clarity.
Line 215-Reference needed
ANSWER: Thank you for your comment. We removed the sentence to be more concise. Please, refer to our corrected manuscript.
Line 217- Authors mention several studies, but there is only one reference
ANSWER: Our mistake. Thank you for the comment. The sentence was corrected.
Line 215-Authors should specify whether PPT was high or low in the specific correlation
ANSWER: Thank you for your comment. We removed the sentence to be more concise. Please, refer to our corrected manuscript.
Line 223- Same as line 215
ANSWER: Thank you for your comment. We included a reference and changed the sentence to improve clarity.
Line 227- States that the excessive alcohol consumption people were excluded, however in M&M it simply states the length of time that alcohol was not consumed during with no mention of amount
ANSWER: Thank you for your comment. We did not assess the amount of alcohol consumption. However, as mentioned, We excluded those with less than 5-days of alcohol consumption, a considerable amount of time to allow the subject to metabolize the substance and to be included in the analysis.
Line 228- 235- It is unclear what this paragraph adds to the discussion. Consider rewriting to emphasize the contribution that the modified PA adds.
ANSWER: Thank you for your comment. The purpose was to inform that the adapted pressure algometer could be used in other protocols, as the basis of such protocols would be the pressure pain threshold. We were asked to include the conditioned pain modulation as well. As the text was restructured, the final sentence highlights the importance of the present equipment. Please, refer to our corrected manuscript.
Line 238-Line 241- I don’t think it is possible to draw a conclusion that various muscle bellies will be equally valid based upon testing using the force platform and LLC.
ANSWER: Thank you for your comment. Agreed. However, it was not concluded. We do believe that a valid tool would provide accurate measures, if used in an adequate manner. Please, note that the previous sentence was: “The pressure pain threshold in body sites other than Deltoid muscle must be assessed to ensure the adapted pressure algometer validity on those sites”. Please, let us know if more clarification is necessary.
Line 244- Uses authors name when citing. In other places it is used in the format ‘Studies written….’
The format used in line 244 is written in a more professional manner compared to the rest of the paper when other authors were cited. Consider revising.
ANSWER: Thank you for your comment. We rather to explore both ways or citing throughout the text. Please, note that the in text references will be further formatted if the manuscript is accepted, highlighting all authors’ names.
Line 244 -Which 3 measures? Be specific
ANSWER: Thank you. We added ‘pressure pain threshold’. Please, see our corrected manuscript.
Line 254- PA is mentioned here in comparison to force platform but not load cell. Does that matter?
ANSWER: Thank you for your attention. As We restructured the text without the LLC measures, the sentence was not altered.
Line 255- The issues of portability and user system have not previously been mentioned regarding the standard pressure algometers. Based on my experience, standard pressure algometers are portable and easy to use.
ANSWER: Thank you for your comment. Yes, some are portable and some are also easy-to-use. Others need a smartphone application or a software to use (only as an example: http://www.to-nod.com/index.php/pt/, and https://www.instagram.com/medeor.medtech/), requiring other devices and training. Our team has developed courses to educate the Brazilian Physiotherapists to use affordable technologies in physical assessment, and We could notice difficulties to handle inter-connected devices. We also state that the sentence did not make any assumptions about any specific device, but instead, highlighted the characteristics of the adapted pressure algometer, as We corrected changing ‘PA’ for ‘adapted pressure algometer’. We apologize for our mistake.
Conclusion
Line 261- Here the tool is described as adapted PA, however it has been referred to just as PA throughout the paper and the specific adaptations made and differences between this equipment and standard algometer MUST be described in detail.
ANSWER: Thank you for your comment. The entire text was checked to ensure the distinction between the standard and the adapted pressure algometer. Please, see our reviewed manuscript.
Line 262- Incorrect use of ‘Is warranted’- Are you trying to comment on the validation?
ANSWER: Thank you. We restructured the previous sentence to improve clarity.
" | Here is a paper. Please give your review comments after reading it. |
9,808 | "<ns0:abstract xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:p>Background: Pain assessment is a key measure that accompanies treatments in a wide range of clinical settings. A low-cost valid and reliable pressure algometer would allow objective assessment of pressure pain to assist a variety of health professionals. However, the pressure algometer is often expensive, which limits its daily use in both clinical and research settings. Objectives: This study aimed to assess the instrumental validity, and the intra-and inter-rater reliability of an inexpensive digitally adapted pressure algometer.</ns0:p><ns0:p>Methods: A single rater applied 60 random compressions on a force platform. The pressure pain thresholds of 20 volunteers were collected twice (three days apart) by two raters. The main outcome measurements were as follows: the maximal peak force (in kPa) and the pressure pain threshold (adapted pressure algometer vs. force platform).</ns0:p><ns0:p>Cronbach's α test was used to assess internal consistency. The standard error of measurement provided estimates of measurement error, and the measurement bias was estimated with the Bland-Altman method, with lower and upper limits of agreement.</ns0:p><ns0:p>Results: No differences were observed when comparing the compression results (P = 0.51). The validity and internal intra-rater consistencies ranged from 0.84 to 0.99, and the standard error of measurement from 0.005 to 0.04. A very strong (r = 0.73-0.74) to nearperfect (r = 0.99) correlations were found, with a low risk of bias for all measurements.</ns0:p><ns0:p>The results demonstrated both the validity and intra-rater reliability of the digitally adapted pressure algometer. Inter-rater reliability results were moderate (r = 0.55-0.60; Cronbach' s α = 0.71-0.75). Conclusion: The adapted pressure algometer provide valid and reliable measurements of pressure pain threshold. The results support more widespread use of the pressure pain threshold method among clinicians.</ns0:p></ns0:div>
</ns0:abstract>
<ns0:body xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:head>Introduction</ns0:head><ns0:p>Pain has been described as a multidimensional construct involving psychological and physical domains with different patterns depending on the emotional state. <ns0:ref type='bibr' target='#b0'>1</ns0:ref> These characteristics may impair conclusions and lead to biased clinical reasoning regarding the patterns of group pain due to intra-group and longitudinal variability in subjects' co-morbidities and momentaneous emotional state. Nevertheless, physical assessment is essential to provide objective data to compare the prospective effects of interventions for pain management. <ns0:ref type='bibr' target='#b1'>[2]</ns0:ref><ns0:ref type='bibr' target='#b2'>[3]</ns0:ref><ns0:ref type='bibr' target='#b4'>[4]</ns0:ref> Pain is mostly assessed by patient self-reports using the visual analog scale. <ns0:ref type='bibr' target='#b5'>5</ns0:ref> Self-reported pain intensity is important and reflects physiological and psychological features. However, it can be difficult to interpret because of its subjectivity and overestimation of pain level. <ns0:ref type='bibr' target='#b6'>6</ns0:ref> Objective pain assessment is essential to establish a prospective evaluation, compare baseline results to other timeline assessments, or even as a prognostic measure to predict future outcomes. <ns0:ref type='bibr' target='#b5'>5,</ns0:ref><ns0:ref type='bibr' target='#b8'>7</ns0:ref> Thus, pressure algometry is a diagnostic aid used to assess some musculoskeletal problems. The pressure pain threshold has been used to aid the diagnosis of pain by providing a quantified force value of tissue tenderness <ns0:ref type='bibr' target='#b9'>8</ns0:ref> and occurs at the minimum transition point when the applied pressure is sensed as pain. <ns0:ref type='bibr' target='#b10'>9</ns0:ref> The pressure algometer is an instrument used to assess the pressure pain threshold for both regional and widespread musculoskeletal pain. <ns0:ref type='bibr' target='#b11'>10</ns0:ref> This equipment includes a system to convert the force applied through a 1 cm 2 pressure application surface to Newtons Manuscript to be reviewed (N/cm 2 ) or kilograms of force (kgf/cm 2 ), and a readings display. The units can be easily converted to kilopascal (kPa), the international metric for pressure (1 kg/cm 2 = 98.066 kPa). The pressure algometer enables the rater to semi-objectively quantify the mechanical sensitivity to pain level and the recovery of underlying problems or soreness levels. <ns0:ref type='bibr' target='#b9'>8,</ns0:ref><ns0:ref type='bibr' target='#b12'>11</ns0:ref> The instrumental validity of commercial pressure algometers has already been assessed in previous studies. Kinser et al. <ns0:ref type='bibr' target='#b9'>8</ns0:ref> and Vaughan et al. <ns0:ref type='bibr' target='#b13'>12</ns0:ref> manually applied pressure on a force platform to test the reliability and construct validity of pressure algometers. Both studies found high levels of correlation between the force platform and pressure algometers. Other studies have assessed the responsiveness of a pressure algometer to diagnose dysfunctional conditions. Ko et al. <ns0:ref type='bibr' target='#b14'>13</ns0:ref> assessed the correlation between a modified pressure algometer and a commercial algometer to assess the pressure pain threshold of the epigastric region. Unfortunately, commercially available pressure algometers are expensive and may require specific software for reporting and viewing the results, resulting in more time and training required to assess the pressure pain threshold. The validation of an easy-to-read, low-cost, digitally adaptable pressure algometer would enable widespread quantitative measurements of pressure pain thresholds in clinical practice routine, <ns0:ref type='bibr' target='#b9'>8,</ns0:ref><ns0:ref type='bibr' target='#b16'>14</ns0:ref> benefiting early assessment of pain conditions in low-income and developing countries, mainly in primary care. A portable pressure algometer adapted from a hanging scale may be a cost-effective alternative to ensure accurate algometry assessments.</ns0:p><ns0:p>The hanging scale is a battery-operated instrument used to weigh objects suspended from an attachment. The equipment uses a load cell, which is a metallic sturdy element, yet elastic enough for a load to deform it. The load cell is attached to a strain gauge, which reads the change in electrical resistance when a pressure or traction load is placed in the load cell. The change in electrical resistance is converted to a digital signal by the strain gauge, and the result appears on a display. <ns0:ref type='bibr' target='#b17'>15</ns0:ref> Among other measures, the correlation level between the result of a certain instrument and some external criterion must be confirmed to determine the instrumental validity of an equipment. The criterion has to be a widely accepted measure and considered as the goldstandard method, with the same measurement characteristics of the assessment tool. <ns0:ref type='bibr' target='#b18'>16,</ns0:ref><ns0:ref type='bibr' target='#b19'>17</ns0:ref> The purpose of this study was to examine the instrumental validity as well as the intra-and interrater reliability of a low-cost pressure algometer adapted from a hanging scale. Validity was assessed by comparing differences in the measurements of a series of random peak forces applied on a laboratory-grade force platform. Force platforms measure vertical ground reaction forces in response to compressions applied on the surface. They are considered as the goldstandard devices for ground reaction forces owing to their high measurement precision. <ns0:ref type='bibr' target='#b16'>14,</ns0:ref><ns0:ref type='bibr' target='#b20'>18</ns0:ref> The hypothesis is that a low-cost adapted pressure algometer is valid and reliable to be considered an acceptable method to assess the pressure pain threshold.</ns0:p></ns0:div>
<ns0:div><ns0:head>Materials & Methods</ns0:head></ns0:div>
<ns0:div><ns0:head>Equipment</ns0:head><ns0:p>All data were collected at the facilities of the Clinic-School of Physical Therapy, Federal University of Juiz de Fora, in May 2019. The low-cost adapted pressure algometer (MED.DOR Ltd., Brazil; maximum compression = 50 kgf, precision = 0.1 kgf, 3 digits display) had a 5-cm screw attached to the distal extremity. A 1-cm 2 round rubber application surface was attached to follow the standardization for pressure algometry (Figure <ns0:ref type='figure' target='#fig_3'>1</ns0:ref>). The low-cost adapted pressure algometer calibration is checked by placing a known weight (1 kg) on the application surface.</ns0:p><ns0:p>The maximal tolerated difference between the weight and the value on the display is 0.1 kgf. The adapted algometer used in the present study was brand new, and the calibration was checked twice before any measurement.</ns0:p><ns0:p>A two-axis force platform (37 cm × 37 cm; PASCO, Pasport PS-2142, Roseville, USA) collected data using five force beams (sample rate = 1,000 Hz). Four beams in the corner were used to measure the vertical force (range: −1,100 N to +4,400 N) and a 5 th beam measured the force in a parallel axis (range: −1,100 N to +1,100 N). The recorded trials were converted to kPa (1 kg/cm 2 = 98.066 kPa).</ns0:p></ns0:div>
<ns0:div><ns0:head>INSERT FIGURE 1 HERE</ns0:head></ns0:div>
<ns0:div><ns0:head>Procedures</ns0:head></ns0:div>
<ns0:div><ns0:head>Instrumental validity</ns0:head><ns0:p>An independent rater performed 60 random 3-s pressure trials using the adapted pressure algometer on the force platform with an interval of 3 s. Data were collected and stored using the PASCO Capstone Software (Version 1.13.4, PASCO Scientific, 2019), and the low-cost adapted pressure algometer display readings were recorded using an off-board USB synchronized camera.</ns0:p></ns0:div>
<ns0:div><ns0:head>Intra-and inter-rater reliability</ns0:head><ns0:p>Before the assessment of any participant, two independent raters performed a 4-day training protocol. The protocol consisted of applying constant-progressive pressure with a low-cost adapted pressure algometer on a laboratory-grade load cell (Miotec™ Biomedical Equipment, Porto Alegre, RS, Brazil; maximum tension-compression = 200 kgf, precision = 0.1 kgf, maximum error = measurement = 0.33%) with Miotec™ software for visual feedback (MioTrainer™, Biomedical Equipment, Porto Alegre, RS, Brazil) for two nonconsecutive days (3 nonconsecutive hours per day). The conversion from analog to digital signals was performed by an A/D board (Miotec™, Biomedical Equipment) with a 16-bit resolution input range, a sampling frequency of 2 kHz, a common rejection module greater than 100 dB, a signal-noise ratio less than 03 μV, root mean square, and impedance of 109 Ω. All data were recorded and processed using the Miotec Suite™ software (Miotec™ Biomedical Equipment, Porto Alegre, RS, Brazil). An assessor monitored the pressure applied by the raters for two consecutive days using the same software, but the raters did not receive any visual feedback. The training was aimed at ensuring the velocity to apply pressure using a low-cost adapted algometer (1 kg/s).</ns0:p><ns0:p>The independently trained raters collected the middle deltoid muscle's pressure pain threshold of 20 participants (10 women; 22±2 years; 63±13 kg; 160±10 cm; 23±4 kg/cm 2 ). The exclusion criteria for participants included: body mass index >28 kg/cm 2 , any self-reported health issues, alcohol consumption within five days prior to the assessments, shoulder pain, previous shoulder surgery, or any diagnosed shoulder or cervical impairment. The objectives of the study were explained to the subjects, who were notified of the benefits and potential risks involved before signing an informed consent form prior to participation. The Federal University of Juiz de Fora ethics committee for human investigation approved the procedures employed in the study (reference number: 02599418.9.0000.5147).</ns0:p><ns0:p>The pressure pain threshold was collected twice (3 days apart: days 1 and 2). To evaluate the intra-and inter-rater reliability, the following positioning was adopted: 1) the participant remained seated with the feet on the floor, 2) the hands rested on the thighs, and 3) the trunk was erect. The middle deltoid's site received progressive 1 kg/s pressure controlled by a metronome until the participant experienced pain. <ns0:ref type='bibr' target='#b9'>8</ns0:ref> An effort was made to standardize the anatomic locations of each session. The same rater was responsible for palpating and marking the pressure pain Manuscript to be reviewed threshold site on each subject before any measurements, both on days 1 and 2. The middle deltoid's site was topographically determined in the middle of a horizontal line drawn between the acromioclavicular joint and the deltoid muscle insertion. <ns0:ref type='bibr' target='#b21'>19</ns0:ref> Three measurements were performed for each site, with 10 to 15 seconds apart. The first measurement was discarded. <ns0:ref type='bibr' target='#b22'>20,</ns0:ref><ns0:ref type='bibr' target='#b23'>21</ns0:ref> The participant lifted the opposite hand when the pressure pain threshold was achieved, that is, when the applied pressure evoked pain. The examiner pressured the 'tare' button to lock the reading, immediately retracting the adapted pressure algometer. Then, the pressure pain threshold reading was registered. <ns0:ref type='bibr' target='#b24'>22</ns0:ref> </ns0:p></ns0:div>
<ns0:div><ns0:head>Statistical analysis</ns0:head><ns0:p>The recorded peaks were then extracted. All trials were used for analysis, consisting of the following: a total of 1) 60 measurements (validity analysis, force platform vs. adapted pressure algometer) and 2) 80 measurements (reliability analysis). Data are presented as mean values and standard deviations. The independent Student's t-test was used to compare differences between measurements in the validation process. The intra-and inter-rater differences were compared using the mixed between-(rater 1 vs. rater 2) and within-subject analysis (moment and moment*rater) of variance with repeated measures. All data were reworked using Holm's post hoc test to avoid multiple testing. Significance was set at p<0.05. Intraclass correlation coefficients [ICC (2,1) ] were calculated to compare the results between both types of equipment and raters. Poor reliability was indicated by values less than 0.5, moderate reliability between 0.5 to 0.75, good reliability between 0.75 and 0.9, and excellent reliability greater than 0.90. <ns0:ref type='bibr' target='#b25'>23</ns0:ref> Chronbach's α test was used to assess the expected correlation of both types of equipment measuring the same construct. The standard error of measurement (SEM) was also calculated to provide an estimate of measurement error. A linear regression was used to estimate the coefficient of correlation (r) and the adjusted coefficient of determination (r 2 ). The magnitude of the correlation was qualitatively interpreted using the following thresholds: <0.1, trivial; 0.1-0.3, small; 0.3-0.5, moderate; 0.5-0.7, large; 0.7-0.9, very large; and >0.9, nearly perfect. <ns0:ref type='bibr' target='#b27'>24</ns0:ref> The Bland-Altman method estimated the measurement bias, with lower and upper limits of agreement between results. Statistical analyses were performed using JAMOVI software (JAMOVI project, version 0.9, 2018). Manuscript to be reviewed</ns0:p></ns0:div>
<ns0:div><ns0:head>Results</ns0:head><ns0:p>Validity: Force platform vs. PA No significant differences were observed in pressure trials (Table <ns0:ref type='table' target='#tab_0'>1</ns0:ref>) between the adapted pressure algometer (405.63±235.34 kPa) and the force platform (434.15±239.18 kPa; p=0.25).</ns0:p><ns0:p>The ICC (2,1) and Cronbach's α returned values of 0.98 and 0.99, respectively. The SEM returned a value of 0.005 kgf, and the linear regression showed statistically significant results (r=0.99; adjusted r 2 =0.99; p=0.001). The Bland-Altman results showed high levels of agreement (Figure <ns0:ref type='figure' target='#fig_4'>2</ns0:ref>).</ns0:p></ns0:div>
<ns0:div><ns0:head>INSERT FIGURE 2 HERE</ns0:head></ns0:div>
<ns0:div><ns0:head>Intra-and inter-rater reliability</ns0:head><ns0:p>The pressure pain threshold from both raters showed very low variation over time (Rater 1: Day 1=203±74 kPa, Day 2=206±71.6 kPa; Rater 2: Day 1=214±73.7 kPa, Day 2=215±69.6 kPa). The intra-rater comparison showed no significant differences (Moment: F=0.05; p=0.83 and Moment*Rater: F=0.01; p=0.93) (Table <ns0:ref type='table' target='#tab_0'>1</ns0:ref>). The ICC (2,1) and Cronbach's α analysis returned relevant values (Rater 1: ICC (2,1) =0.76, Cronbach's α=0.85; Rater 2: ICC (2,1) =0.73, Cronbach's α=0.84). The SEM values were low (Rater 1=0.02, Rater 2=0.01), and moderate values were also obtained in the linear regression analysis (Rater 1: r=0.74, adjusted r 2 =0.52; Rater 2: r=0.73, adjusted r 2 =0.50). The Bland-Altman results showed high levels of agreement (Figure <ns0:ref type='figure' target='#fig_5'>3</ns0:ref>).</ns0:p></ns0:div>
<ns0:div><ns0:head>INSERT TABLE 1 HERE</ns0:head><ns0:p>The inter-rater reliability showed no differences among measurements (F=0.22; p=0.64) (Table <ns0:ref type='table' target='#tab_0'>1</ns0:ref>), with moderate results for reliability analysis (Day 1: ICC (2,1) =0.56, Cronbach's α=0.75; Day 2: ICC (2,1) =0.54, Cronbach's α=0.71). The SEM results showed very low values (Day 1: 0.04 kgf, Day 2: 0.02 kgf), and moderate values in the linear regression analysis (Day 1: r=0.60, adjusted r 2 =0.33; Day 2: r=0.55, adjusted r 2 =0.26). The Bland-Altman analysis showed acceptable levels of agreement (Figure <ns0:ref type='figure' target='#fig_5'>3</ns0:ref>).</ns0:p></ns0:div>
<ns0:div><ns0:head>INSERT FIGURE 3 HERE</ns0:head><ns0:p>PeerJ reviewing PDF | (2020:02:46274:2:0:NEW 3 Aug 2020)</ns0:p></ns0:div>
<ns0:div><ns0:head>Manuscript to be reviewed</ns0:head></ns0:div>
<ns0:div><ns0:head>Discussion</ns0:head><ns0:p>This study was designed to examine the instrumental validity and the intra-and inter-rater reliability of an adapted low-cost pressure algometer. The results showed no significant differences in the peak compressive force recorded from the adapted pressure algometer and the force platform. A significant correlation was observed between the low-cost adapted algometer and the force platform while measuring the same construct. These findings support the primary hypothesis, which contends that a low-cost pressure algometer is both valid and reliable to be considered as a standard equipment to assess the pressure pain threshold. Therefore, the tested device seems to be an acceptable alternative to expensive instruments.</ns0:p><ns0:p>Previous studies showed acceptable levels of validity and reliability of other digital algometry systems. 8,10 Kinser et al. <ns0:ref type='bibr' target='#b9'>8</ns0:ref> tested the construct validity of a digital pressure algometer using the same protocol as our study by manually applying pressure on a force platform. The authors used ten sets of five applications to 80 N and one additional set of five applications to subsequent levels of progressive 10 N (20, 30, 40, 50, 60, 70, 80, 90, 100, and 110 N). The results showed high correlation between the tested algometer and the force platform (r = 0.99) for both 80 N and incremental trials. Vaughan et al. <ns0:ref type='bibr' target='#b13'>12</ns0:ref> also used the force platform as a reference instrument to test the validity of a digital pressure algometer. The authors applied 300 vertical pressures on the force platform with progressive pressure rates (10, 20, 30, 40, and 50 kPa/s). The result showed an excellent ICC range (0.90 to 0.99) for all comparisons. In general, all previous studies and the present work suggest the validity of the results obtained from a digital pressure algometer. These excellent results could be attributed to the strain gauge-based system used to acquire the signals.</ns0:p><ns0:p>The conversion from analogic (load cell deformation) to electrical-digital signal (strain gauge) is very effective, even in very affordable systems. As the resistance varies in a sturdy element with the applied force, the strain gauge converts the force (in this case, pressure) into a change in electrical resistance that can be measured.</ns0:p><ns0:p>Several studies have evaluated the reliability of distinct pressure algometers as a tool to distinguish healthy individuals from those with musculoskeletal disorders. Balaguier et al. <ns0:ref type='bibr' target='#b28'>25</ns0:ref> found high reliability between all three pressure pain threshold measures at sites in the lower back. Walton et al. <ns0:ref type='bibr' target='#b5'>5</ns0:ref> assessed the intra-rater and inter-rater reliability of an accessible digital algometer in 60 healthy volunteers and 40 individuals with neck pain. The authors tested the upper fibers of the trapezius and tibialis anterior muscles. The intra-rater ICC results in both groups ranged from 0.94 to 0.97 for the trapezius and tibialis anterior muscles. The inter-rater ICC range (0.79-0.90) was lower than that of the intra-rater due to variations between observers. However, both results were considered adequate. Waller et al. <ns0:ref type='bibr' target='#b29'>26</ns0:ref> found high intra-and inter-rater reliability (ICC=0.81-0.99; ICC=0.92-0.95, respectively) using five research assistants. Each assistant tested 20 pain-free subjects at the wrist, leg, cervical, and lumbar spine. The intra-rater SEM ranged between 79 and 100 kPa. However, Van Wilgen et al. <ns0:ref type='bibr' target='#b30'>27</ns0:ref> found lower values for intra-rater reliability compared to inter-rater reliability of pressure algometry in healthy volleyball athletes and those with patellar tendinopathy. The authors found high inter-rater reliability (ICC = 0.93), but only moderate intra-rater reliability (ICC = 0.60) for pain pressure threshold measurements. The authors argued that the lower intra-rater ICC values were probably due to variance within the observer and also within the athletes, as the pain in patellar tendinopathy varies over time. The diagnosis of fibromyalgia utilizes the pressure pain threshold as a key assessment to distinguish healthy individuals from those with fibromyalgia. <ns0:ref type='bibr' target='#b31'>28,</ns0:ref><ns0:ref type='bibr' target='#b32'>29</ns0:ref> Neck pain, cranio-cervical headache, and temporomandibular disorders also include the pressure pain threshold as an important component for clinical reasoning about the level of severity, influencing the treatment direction. <ns0:ref type='bibr' target='#b5'>5,</ns0:ref><ns0:ref type='bibr' target='#b10'>9,</ns0:ref><ns0:ref type='bibr' target='#b33'>30</ns0:ref> However, those previous studies used commercial pressure algometers. For clinical and ambulatory settings, the high cost and the user's interface would be an issue to obtain fast objective pain measurements, requiring both training and experience for assessments. Brazilian physiotherapists have an average monthly salary of USD 500, according to the Occupational Brazilian Classification (https://www.salario.com.br/profissao/fisioterapeuta-geral-cbo-223605/).</ns0:p><ns0:p>The adapted pressure algometer used in this study had a production cost of USD 10.00, while the standard digital equipment cost ranged from USD 600.00 to USD 1,000.00. The validation procedure enables use of the low-cost adapted pressure algometer for clinical assessments in a practice routine, which may directly impact primary and ambulatory care in low-income and developing countries, by adding an objective and inexpensive tool to assess the pressure pain threshold.</ns0:p><ns0:p>Some limitations of the present study must be addressed. The pressure pain threshold in body sites other than the deltoid muscle must be assessed to ensure the validity of the adapted pressure algometer on different sites. However, we hypothesize that they should not give any different results to direct assessment using the adapted pressure algometer, since the standard deviation remained at very low values and the current results gave very good measures compared to the force platform and additional good reliability. The instrumental validity of an equipment's measurements also ensures unbiased assessments. <ns0:ref type='bibr' target='#b38'>34</ns0:ref> Other studies have identified different factors to consider when evaluating the pressure pain threshold, such as gender and obesity. <ns0:ref type='bibr' target='#b39'>35,</ns0:ref><ns0:ref type='bibr' target='#b40'>36</ns0:ref> A review of studies involving induced pain found a consistent pattern of women exhibiting greater pain sensitivity and a reduction in pain inhibition compared to men. <ns0:ref type='bibr' target='#b41'>37</ns0:ref> In addition, the characteristic of pain imposed is an important factor for these differences, since the type of pressure pain has one of the highest effect sizes in the pain report. <ns0:ref type='bibr' target='#b23'>21,</ns0:ref><ns0:ref type='bibr' target='#b42'>38</ns0:ref> It is suggested that interactions between biological and psychosocial factors are responsible for these gender differences, but all studies indicate the need for additional research to elucidate the mechanisms that drive gender differences in pain responses. <ns0:ref type='bibr' target='#b39'>35,</ns0:ref><ns0:ref type='bibr' target='#b41'>37,</ns0:ref><ns0:ref type='bibr' target='#b42'>38</ns0:ref> Some studies suggest that in areas with additional subcutaneous fat, pain thresholds for electrical or pressure stimuli increase and pain sensitivity decreases in obese individuals. <ns0:ref type='bibr' target='#b40'>36,</ns0:ref><ns0:ref type='bibr' target='#b43'>39</ns0:ref> A study has also shown biochemical changes in trigger points with higher levels of inflammatory mediators, catecholamines, and cytokines in obese individuals. <ns0:ref type='bibr' target='#b44'>40</ns0:ref> Mechanical stretching of the skin in response to excess fat can lead to a decrease in the density of nociceptive fibers, and obesity is associated with the chemical inhibition of pain with an increase in β endorphin and endogenous opioid peptide. <ns0:ref type='bibr' target='#b40'>36</ns0:ref> The present study had a balanced cohort with regard to participant sex, and all participants were classified as normal according to their body mass index. However, the current sample was chosen only for reliability analysis. Pressure pain threshold as a clinical result is well established, but more studies should take into account sex and body mass index differences to avoid bias in experimental protocols. <ns0:ref type='bibr' target='#b39'>35</ns0:ref> Pressure pain threshold was also positively but poorly correlated with high-density lipoprotein cholesterol. <ns0:ref type='bibr' target='#b45'>41</ns0:ref> A high pressure pain threshold was also found among subjects with hyperglycemia and excessive alcohol consumption. <ns0:ref type='bibr' target='#b45'>41</ns0:ref> In the present study, no blood assessment was performed to exclude those factors. However, the sample consisted of young adults, decreasing the chance of any important health issues. Additionally, exclusion criteria included previous excessive alcohol consumption.</ns0:p><ns0:p>Considering its portability, easy assemblage, and lower cost, the currently tested device seems to be a valid standard equipment for pressure pain threshold assessment. Therefore, the adapted pressure algometer is a valid device providing similar measurements compared to a force platform. The portability, cost-effectiveness, and friendly user system provide an effective way to measure the pressure pain threshold.</ns0:p></ns0:div>
<ns0:div><ns0:head>Conclusions</ns0:head><ns0:p>The current hypothesis is that a low-cost pressure algometer is valid and reliable enough to be considered as a standard equipment to assess the pressure pain threshold. The results showed that the low-cost adapted pressure algometer is a valid tool compared to a force platform. The lowcost adapted pressure algometer is also reliable for assessing the pressure pain threshold. Future directions include evaluating the low-cost adapted pressure algometer in routine clinical assessments for the systematic evaluation of pressure pain. Further studies should consider other assessments, such as temporal summation and conditioned modulated pain, using a low-cost adapted pressure algometer. Manuscript to be reviewed Manuscript to be reviewed </ns0:p><ns0:note type='other'>Figure 1</ns0:note><ns0:note type='other'>Figure 2</ns0:note><ns0:note type='other'>Figure 3</ns0:note></ns0:div><ns0:figure xml:id='fig_0'><ns0:head></ns0:head><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:02:46274:2:0:NEW 3 Aug 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_1'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:02:46274:2:0:NEW 3 Aug 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_2'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:02:46274:2:0:NEW 3 Aug 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_3'><ns0:head>Figure 1 .</ns0:head><ns0:label>1</ns0:label><ns0:figDesc>Figure 1. Adapted pressure algometer -PA.</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_4'><ns0:head>Figure 2 .</ns0:head><ns0:label>2</ns0:label><ns0:figDesc>Figure 2. Bland-Altman plot: instrumental validity.</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_5'><ns0:head>Figure 3 .</ns0:head><ns0:label>3</ns0:label><ns0:figDesc>Figure 3. Bland-Altman plot: Intra-rater reliability.</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_6'><ns0:head>Figure 3 .</ns0:head><ns0:label>3</ns0:label><ns0:figDesc>Figure 3. Bland-Altman plot: Intra-rater reliability. A) Rater 1: Bias = -2.55 (95% confidence interval [CI] = -24.3 to 22.2); lower limit of agreement (LLA) = -106.38 (95% CI = -149 to -63.3); upper limit of agreement (ULA) = 101.28 (95% CI = 58.2 to 144.4). 2). B) Rater 2: Bias = -1.03 (95% CI = -25.9 to 23.9); LLA = -105.28 (95% CI = -148.6 to -62); ULA = 103.22 (95% CI = 59.9 to 146. Inter-rater reliability. C) Day 1: Bias = -10.8 (95% CI: -41.6 to 20); LLA = -139.8 (95% CI: -193.4 to -86.3); ULA = 118.2 (95% CI: 64.7 to 171.8); D) Day 2: Bias = -9.27 (95% CI = -40.7 to 22.1); LLA = -140.77 (95% CI = -195.4 to -86.2); ULA = 122.23 (95% CI = 67.6 to 176.8).</ns0:figDesc></ns0:figure>
<ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='21,42.52,70.87,525.00,393.75' type='bitmap' /></ns0:figure>
<ns0:figure type='table' xml:id='tab_0'><ns0:head>Table 1 :</ns0:head><ns0:label>1</ns0:label><ns0:figDesc>Validity, Intra-and Inter-rater reliability pairwise comparisons.</ns0:figDesc><ns0:table><ns0:row><ns0:cell>95% Confidence Interval</ns0:cell></ns0:row></ns0:table><ns0:note>PeerJ reviewing PDF | (2020:02:46274:2:0:NEW 3 Aug 2020)</ns0:note></ns0:figure>
</ns0:body>
" | "General comments to Reviewers
Dear Dr. Madeleine and Dr. Tomas
Thank you for your efforts to improve our manuscript’s clarity. We performed several changes to restructure the text according to your guidance. The manuscript was professionally edited to make sure that paragraphs, sentences, grammar, and spelling are clear. The authors acknowledge your work and thanks again for your support.
Cordially,
Alexandre Barbosa, PhD
Correspondent author
Reviewer 1 (Pascal Madeleine)
The MS has improved but it is difficult to see what was changed... moreover, the MS lacks still of scientific rigor... here are points that need further attention:
- the authors need to correct erroneous use og unit's abbreviations: Kg/kg - for PPT, it should be kPa not KPa. check the entire paper
ANSWER: Thank you for your comment. The abbreviations were changed: KPa to kPa; and Kg to kg.
- mention how to transform Newtons (N/cm2) or kilograms of force (Kgf/cm2) per squared cm to kPa in the introduction.
ANSWER: Thank you for your attention. The sentence was added in the Introduction section.
- cite other reliability papers for PPT assessment - make an additional search
ANSWER: Thank you for your comment. We added more references in our introduction section. Please, see the modified manuscript.
- force platform. gold standard for what? not clear a load cell can also measure force!
ANSWER: Thank you for your comment. The sentence was: “The validity was assessed by comparing differences in the measurements of a series of random peak forces applied on a force platform, the gold standard for measuring vertical (compression) and horizontal forces.” We restructured the sentence to ensure clarity: “The validity was assessed by comparing differences in the measurements of a series of random peak forces applied on a laboratory-grade force platform. Force platforms measure vertical ground reaction forces in response to compressions applied on the surface. They are considered gold standard devices to ground reaction forces due to their precision.”
- concerning re-calibration: how often is this necessary? Somedic provide a weight resulting in a 100kPa reading. this is recommended to check every time one use the pressure algometer
ANSWER: Thank you for your concern. The device used in our trials was new, so calibration was not a concern. However, We included the following recalibration statement: “The adapted algometer’s calibration is checked by placing a known weight (1 Kg) on the application surface. The maximal tolerated difference between the weight and the value on the display is 0.1 kgf. The adapted algometer used in the present study was brand new, and the calibration was doubled checked before any measurement.”
- what is meant by compression measurements?
ANSWER: Thank you for your comment. We restructured the sentences to ensure clarity as follows: “The results showed that the low-cost adapted pressure algometer is a valid tool compared to force platform. The low-cost adapted pressure algometer is also reliable to assess the pressure pain threshold.”
Reviewer 2 (Andrea Tomas)
Basic reporting
Overall, the efforts undertaken by the authors to improve this paper appear substantial and it has improved considerably compared to the initial version. However, my primary concern remains that there are frequent and consequential errors in grammar, diction, and syntax that ultimately fail to construct a narrative that is easy to follow. While the structure of the introduction has been improved quite nicely, the discussion remains disjointed and confusing, forcing the reader too often to draw their own conclusions as to why the authors included the examples or references. The “procedures” section is also disorganized. For instance, the section regarding intra- and inter-rater reliability begins by describing which muscle was tested by two raters. It then transitions to describing the training protocol. It then transitions back to describing a third rater, however it is unclear if at this point the authors are still describing the training or have returned to discussing the initial experiment.
ANSWER: Thank you for your comment. The methods section was reorganized to improve clarity. We also restructured the discussion section to present more studies which assessed the validity and the reliability process of other pressure algometers. In this section, the paragraph about the temporal summation and conditioned pain modulation protocols was excluded to improve the flow focused on the validity and reliability process. We kept the limitations at the end of the discussion section and the explanations were all referenced. Our statements about our sample selection to avoid any risk of bias concerning those limitations were also maintained.
Experimental design
My original comments remain true. This is an interesting study with the potential for real clinical impact. The study is well designed. However, it is the description of the study that requires improvement.
ANSWER: Thank you for your comment. We tried to improve the manuscript following your guidance.
Validity of the findings
No comments
Comments for the Author
Specific comments
Line 36-36 – The P values should be reported (at least in parentheses) in addition to the qualifiers “moderate” and “excellent” or else these appear to be subjective descriptions.
ANSWER: Thank you for your comment. We restructured the results in the abstract inserting the values and ranges.
Line 39 – “valid intra-rater reliable measures” does not make sense. It is possible this should read “valid and reliable…” however given the later descriptions in this manuscript, I am not sure that this author’s intent
ANSWER: Thank you for your comment. We restructured the abstract’s final statement to improve clarity.
Line 52 – The word “instrument” is more appropriate than “equipment,” and replace “on” with “for”
ANSWER: Thank you for your attention. The changes were made according to your guidance.
Line 65 – Replace “equipment” with instrument.” The mechanism is more appropriately described as “suspended from an attachment” than “in a suspended manner”
ANSWER: Thank you for your attention. Both changes were made according to your guidance.
Line 70-71 – Confusing description of the validation process. A “score” implies either a standardized or specific rating system, which likely differs depending on what is being tested and in what manner it is being tested.
ANSWER: Thank you for your attention. The word ‘score’ was replaced by ‘result’ to avoid misinterpretation. We also restructured the sentence to improve clarity.
Line 74 – Incorrect use of comma after “validity” – consider “as well as” or similar
ANSWER: Thank you for your attention. The comma was removed and the sentence was restructured using your suggestion.
Line 78-79 – Remove “current” as this is implied by the fact that a hypothesis is being discussed in the introduction. This sentence should state “…algometer is valid and reliable to be considered an acceptable method…” or similar. The purpose of this paper is to describe the use of this tool and demonstrate its use – not to set forth a new standard of care/measurement.
ANSWER: Thank you for your attention. The sentence was restructured as you suggested.
Line 106 – Numerals less than 10 should be written (ex. “two” not “2”)
ANSWER: Corrected.
Line 108 – Rephrase as “alcohol consumption within five days prior to…”
ANSWER: Thank you for your attention. Corrected as you suggested.
Line 113-118 – This is a confusing transition from describing the test to the training protocol (which by definition must have taken place prior to the test being performed). To maintain a structured flow, it is strongly advised that all training descriptions be completed before discussing how the tests were carried out.
ANSWER: Thank you for your guidance. We agree that our description was confuse. This section was restructured to improve clarity following your advice. The training description was relocated to maintain the flow of the ideas. The ‘third rater’ was described as ‘assessor’ to avoid any misinterpretation. Thank you again.
Line 122 – “Pieces of information” should be described as “data”
ANSWER: Corrected.
Line 123-124 – Introduction of a third rater at this time is another example of how the flow of this section is harmed by the writing. This should have been described immediately following the description of the first two raters without transitioning first to describe training. The transition in 124 to raters in training further confuses the issue as it becomes even less clear if the author is describing the training process or the experiment at this time
ANSWER: Thank you. As previously described, We restructured the whole section to improve clarity.
Line 129 – Delete “positioned.” Which sites are being referenced? It is assumed to be the middle deltoid, but this should be stated again
ANSWER: The word ‘positioned’ was deleted. The site description was inserted.
Line 144-146 – Multiple instances of “measures” being used instead of “measurements”
ANSWER: Corrected.
Line 146 – “Validation” should be used in place of “validity”
ANSWER: Thank you for your attention. Corrected as you suggested.
Line 151-153 – This should be stated in the abstract instead of the subjective terms used
ANSWER: Thank you for your comment. We included the ranges for correlation values. The values were included in the abstract to avoid subjective terms. However, some terms are the interpretation of the ranges and widely used in the literature. We kept some terms followed by the numbers in the abstract.
Line 173 – The use of “both” implies two. However, earlier three raters were described. It should be clarified as to which portion of the test this is referring to. In general, the description of the rate training vs. the actual experiment makes interpretation of the results unnecessarily complicated.
ANSWER: Thank you for your attention. As We changed the description of the ‘third rater’ for ‘assessor’, the sentence was not altered.
Line 198-199 – “…statistically significant results for the expected correlation of both equipment in measuring the same construct” should be rephrased using more straightforward language to emphasize the findings
ANSWER: Thank you for your comment. The sentence was restructured as follows: ‘Significant correlation was also found between the low-cost adapted algometer and the force platform while measuring the same construct.’
Line 200 – Replace “has validity and reliability” with “is both valid and reliable enough…”
ANSWER: Corrected.
Line 202 – This is better described as an “acceptable alternative” rather than simply an alternative
ANSWER: Thank you. We inserted the word.
Line 208 – Should “prospective” be “objective” – unclear what author’s intent is
ANSWER: Thank you for your comment. The intent was ‘prospective’. As the treatment evolves, the prospective assessments will provide measurements to evaluate the treatment’s success or not. However, to ensure clarity the sentence was restructured as: ‘Nevertheless, the physical assessment is essential to provide objective data to compare the prospective effects of interventions for pain management’
Line 209 – Pressure algometry is better described as a diagnostic aid, rather that implied as a primary method of obtaining a diagnosis
ANSWER: Thank you for your comment. The sentence was restructured to: ‘Thus, the pressure algometry is a diagnostic aid to assess some musculoskeletal problems’
Lines 212, 232 – Replace “include” with “utilize”
ANSWER: Corrected.
" | Here is a paper. Please give your review comments after reading it. |
9,809 | "<ns0:abstract xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:p>Background: Pain assessment is a key measure that accompanies treatments in a wide range of clinical settings. A low-cost valid and reliable pressure algometer would allow objective assessment of pressure pain to assist a variety of health professionals. However, the pressure algometer is often expensive, which limits its daily use in both clinical and research settings. Objectives: This study aimed to assess the instrumental validity, and the intra-and inter-rater reliability of an inexpensive digitally adapted pressure algometer.</ns0:p><ns0:p>Methods: A single rater applied 60 random compressions on a force platform. The pressure pain thresholds of 20 volunteers were collected twice (three days apart) by two raters. The main outcome measurements were as follows: the maximal peak force (in kPa) and the pressure pain threshold (adapted pressure algometer vs. force platform).</ns0:p><ns0:p>Cronbach's α test was used to assess internal consistency. The standard error of measurement provided estimates of measurement error, and the measurement bias was estimated with the Bland-Altman method, with lower and upper limits of agreement.</ns0:p><ns0:p>Results: No differences were observed when comparing the compression results (P = 0.51). The validity and internal intra-rater consistencies ranged from 0.84 to 0.99, and the standard error of measurement from 0.005 to 0.04 kPa. Very strong (r = 0.73-0.74) to near-perfect (r = 0.99) correlations were found, with a low risk of bias for all measurements. The results demonstrated the validity and intra-rater reliability of the digitally adapted pressure algometer. Inter-rater reliability results were moderate (r = 0.55-0.60; Cronbach' s α = 0.71-0.75). Conclusion: The adapted pressure algometer provide valid and reliable measurements of pressure pain threshold. The results support more widespread use of the pressure pain threshold method among clinicians.</ns0:p></ns0:div>
</ns0:abstract>
<ns0:body xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:head>Introduction</ns0:head><ns0:p>Pain has been described as a multidimensional construct involving psychological and physical domains with different patterns depending on the emotional state. <ns0:ref type='bibr' target='#b0'>1</ns0:ref> These characteristics may impair conclusions and lead to biased clinical reasoning regarding the patterns of group pain due to intra-group and longitudinal variability in subjects' co-morbidities and momentaneous emotional state. Nevertheless, physical assessment is essential to provide objective data to compare the prospective effects of interventions for pain management. <ns0:ref type='bibr' target='#b1'>[2]</ns0:ref><ns0:ref type='bibr' target='#b2'>[3]</ns0:ref><ns0:ref type='bibr' target='#b4'>[4]</ns0:ref> Pain is mostly assessed by patient self-reports using the visual analog scale. <ns0:ref type='bibr' target='#b5'>5</ns0:ref> Self-reported pain intensity is important and reflects physiological and psychological features. However, it can be difficult to interpret because of its subjectivity and overestimation of pain level. <ns0:ref type='bibr' target='#b6'>6</ns0:ref> Objective pain assessment is essential to establish a prospective evaluation, compare baseline results to other timeline assessments, or even as a prognostic measure to predict future outcomes. <ns0:ref type='bibr' target='#b5'>5,</ns0:ref><ns0:ref type='bibr' target='#b8'>7</ns0:ref> Thus, pressure algometry is a diagnostic aid used to assess some musculoskeletal problems. The pressure pain threshold has been used to aid the diagnosis of pain by providing a quantified force value of tissue tenderness and occurs at the minimum transition point when the applied pressure is sensed as pain. <ns0:ref type='bibr' target='#b9'>8,</ns0:ref><ns0:ref type='bibr' target='#b10'>9</ns0:ref> The pressure algometer is an instrument used to assess the pressure pain threshold for both regional and widespread musculoskeletal pain. <ns0:ref type='bibr' target='#b11'>10</ns0:ref> This equipment includes a system to convert the force applied through a 1 cm 2 pressure application surface to Newtons Manuscript to be reviewed (N/cm 2 ) or kilograms of force (kgf/cm 2 ), and a display. The units can be easily converted to kilopascal (kPa), the international metric for pressure (1 kg/cm 2 = 98.066 kPa). The pressure algometer enables the rater to semi-objectively quantify the mechanical sensitivity to pain level and the recovery of underlying problems or soreness levels. <ns0:ref type='bibr' target='#b9'>8,</ns0:ref><ns0:ref type='bibr' target='#b12'>11</ns0:ref> The instrumental validity of commercial pressure algometers has already been assessed in previous studies. Kinser et al. <ns0:ref type='bibr' target='#b9'>8</ns0:ref> and Vaughan et al. <ns0:ref type='bibr' target='#b13'>12</ns0:ref> manually applied pressure on a force platform to test the reliability and construct validity of pressure algometers. Both studies found high levels of correlation between the force platform and pressure algometers. Other studies have assessed the responsiveness of a pressure algometer to diagnose dysfunctional conditions. Ko et al. <ns0:ref type='bibr' target='#b14'>13</ns0:ref> assessed the correlation between a modified pressure algometer and a commercial algometer to assess the pressure pain threshold of the epigastric region. Unfortunately, commercially available pressure algometers are expensive and may require specific software for reporting and viewing the results, resulting in more time and training required to assess the pressure pain threshold. The validation of an easy-to-read, low-cost, digitally adaptable pressure algometer would enable widespread quantitative measurements of pressure pain thresholds in clinical practice routine, benefiting early assessment of pain conditions in low-income and developing countries, mainly in primary care. <ns0:ref type='bibr' target='#b9'>8,</ns0:ref><ns0:ref type='bibr' target='#b16'>14</ns0:ref> A portable pressure algometer adapted from a hanging scale may be a cost-effective alternative to ensure accurate algometry assessments.</ns0:p><ns0:p>The hanging scale is a battery-operated instrument used to weigh objects suspended from an attachment. The equipment uses a load cell, which is a metallic sturdy element, yet elastic enough for a load to deform it. The load cell is attached to a strain gauge, which reads the change in electrical resistance when a pressure or traction load is placed in the load cell. The change in electrical resistance is converted to a digital signal by the strain gauge, and the result appears on a display. <ns0:ref type='bibr' target='#b17'>15</ns0:ref> Among other measures, the correlation level between the result of a certain instrument and some external criterion must be confirmed to determine the instrumental validity of an equipment. The criterion has to be a widely accepted measure and considered as the goldstandard method, with the same measurement characteristics of the assessment tool. <ns0:ref type='bibr' target='#b18'>16,</ns0:ref><ns0:ref type='bibr' target='#b19'>17</ns0:ref> The purpose of this study was to examine the instrumental validity as well as the intra-and interrater reliability of a low-cost pressure algometer adapted from a hanging scale. Validity was assessed by comparing differences in the measurements of a series of random peak forces applied on a laboratory-grade force platform. Force platforms measure vertical ground reaction forces in response to compressions applied on the surface. They are considered as the goldstandard devices for ground reaction forces owing to their high measurement precision. <ns0:ref type='bibr' target='#b16'>14,</ns0:ref><ns0:ref type='bibr' target='#b20'>18</ns0:ref> The hypothesis is that a low-cost adapted pressure algometer is valid and reliable to be considered an acceptable method to assess the pressure pain threshold.</ns0:p></ns0:div>
<ns0:div><ns0:head>Materials & Methods</ns0:head></ns0:div>
<ns0:div><ns0:head>Equipment</ns0:head><ns0:p>All data were collected at the facilities of the Clinic-School of Physical Therapy, Federal University of Juiz de Fora, in May 2019. The low-cost adapted pressure algometer (MED.DOR Ltd., Brazil; maximum compression = 50 kgf, precision = 0.1 kgf, 3-digit display) had a 5-cm screw attached to the distal extremity. A 1-cm 2 round rubber application surface was attached to follow the standardization for pressure algometry (Figure <ns0:ref type='figure' target='#fig_3'>1</ns0:ref>). The low-cost adapted pressure algometer calibration is checked by placing a known weight (1 kg) on the application surface.</ns0:p><ns0:p>The maximal tolerated difference between the weight and the value on the display is 0.1 kgf. The adapted algometer used in the present study was brand new, and the calibration was checked twice before any measurement.</ns0:p><ns0:p>A two-axis force platform (37 cm × 37 cm; PASCO, Pasport PS-2142, Roseville, USA) collected data using five force beams (sample rate = 1,000 Hz). Four beams in the corner were used to measure the vertical force (range: −1,100 N to +4,400 N) and a 5 th beam measured the force in a parallel axis (range: −1,100 N to +1,100 N). The recorded trials were converted to kPa (1 kg/cm 2 = 98.066 kPa).</ns0:p></ns0:div>
<ns0:div><ns0:head>INSERT FIGURE 1 HERE</ns0:head></ns0:div>
<ns0:div><ns0:head>Procedures</ns0:head></ns0:div>
<ns0:div><ns0:head>Instrumental validity</ns0:head><ns0:p>An independent rater performed 60 random 3-s pressure trials using the adapted pressure algometer on the force platform with an interval of 3 s. Data were collected and stored using the PASCO Capstone Software (Version 1.13.4, PASCO Scientific, 2019), and the low-cost adapted pressure algometer display readings were recorded using an off-board USB synchronized camera.</ns0:p></ns0:div>
<ns0:div><ns0:head>Intra-and inter-rater reliability</ns0:head><ns0:p>Before the assessment of any participant, two independent raters performed a 4-day training protocol. The protocol consisted of applying constant-progressive pressure with a low-cost adapted pressure algometer on a laboratory-grade load cell (Miotec™ Biomedical Equipment, Porto Alegre, RS, Brazil; maximum tension-compression = 200 kgf, precision = 0.1 kgf, maximum error = measurement = 0.33%) with Miotec™ software for visual feedback (MioTrainer™, Biomedical Equipment, Porto Alegre, RS, Brazil) for two nonconsecutive days (3 nonconsecutive hours per day). The conversion from analog to digital signals was performed by an A/D board (Miotec™, Biomedical Equipment) with a 16-bit resolution input range, a sampling frequency of 2 kHz, a common rejection module greater than 100 dB, a signal-noise ratio less than 03 μV, root mean square, and impedance of 109 Ω. All data were recorded and processed using the Miotec Suite™ software (Miotec™ Biomedical Equipment, Porto Alegre, RS, Brazil). An assessor monitored the pressure applied by the raters for two consecutive days using the same software, but the raters did not receive any visual feedback. The training was aimed at ensuring the velocity to apply pressure using a low-cost adapted algometer (1 kg/s).</ns0:p><ns0:p>The independently trained raters collected the middle deltoid muscle's pressure pain threshold of 20 participants (10 women; 22±2 years; 63±13 kg; 160±10 cm; 23±4 kg/cm 2 ). The exclusion criteria for participants included: body mass index >28 kg/cm 2 , any self-reported health issues, alcohol consumption within five days prior to the assessments, shoulder pain, previous shoulder surgery, or any diagnosed shoulder or cervical impairment. The objectives of the study were explained to the subjects, who were notified of the benefits and potential risks involved before signing an informed consent form prior to participation. The Federal University of Juiz de Fora ethics committee for human investigation approved the procedures employed in the study (reference number: 02599418.9.0000.5147).</ns0:p><ns0:p>The pressure pain threshold was collected twice (3 days apart: days 1 and 2). To evaluate the intra-and inter-rater reliability, the following positioning was adopted: 1) the participant remained seated with the feet on the floor, 2) the hands rested on the thighs, and 3) the trunk was erect. The middle deltoid's site received progressive 1 kg/s pressure controlled by a metronome until the participant experienced pain. <ns0:ref type='bibr' target='#b9'>8</ns0:ref> An effort was made to standardize the anatomic locations of each session. The same rater was responsible for palpating and marking the pressure pain Manuscript to be reviewed threshold site on each subject before any measurements, both on days 1 and 2. The middle deltoid's site was topographically determined in the middle of a horizontal line drawn between the acromioclavicular joint and the deltoid muscle insertion. <ns0:ref type='bibr' target='#b21'>19</ns0:ref> Three measurements were performed for each site, with 10 to 15 seconds apart. The first measurement was discarded. <ns0:ref type='bibr' target='#b22'>20,</ns0:ref><ns0:ref type='bibr' target='#b23'>21</ns0:ref> The participant lifted the opposite hand when the pressure pain threshold was achieved, that is, when the applied pressed evoked pain. The examiner pressured the 'tare' button to lock the reading, immediately retracting the adapted pressure algometer. Then, the pressure pain threshold reading was registered. <ns0:ref type='bibr' target='#b24'>22</ns0:ref> </ns0:p></ns0:div>
<ns0:div><ns0:head>Statistical analysis</ns0:head><ns0:p>The recorded peaks were then extracted. All trials were used for analysis, consisting of the following: a total of 1) 60 measurements (validity analysis, force platform vs. adapted pressure algometer) and 2) 80 measurements (reliability analysis). Data are presented as mean values and standard deviations. The independent Student's t-test was used to compare differences between measurements in the validation process. The intra-and inter-rater differences were compared using the mixed between-(rater 1 vs. rater 2) and within-subject analysis (moment and moment*rater) of variance with repeated measures. All data were reworked using Holm's post hoc test to avoid multiple testing. Significance was set at p<0.05. Intraclass correlation coefficients [ICC (2,1) ] were calculated to compare the results between both types of equipment and raters. Poor reliability was indicated by values less than 0.5, moderate reliability between 0.5 to 0.75, good reliability between 0.75 and 0.9, and excellent reliability greater than 0.90. <ns0:ref type='bibr' target='#b25'>23</ns0:ref> Chronbach's α test was used to assess the expected correlation of both types of equipment measuring the same construct. The standard error of measurement (SEM) was also calculated to provide an estimate of measurement error. A linear regression was used to estimate the coefficient of correlation (r) and the adjusted coefficient of determination (r 2 ). The magnitude of the correlation was qualitatively interpreted using the following thresholds: <0.1, trivial; 0.1-0.3, small; 0.3-0.5, moderate; 0.5-0.7, large; 0.7-0.9, very large; and >0.9, nearly perfect. <ns0:ref type='bibr' target='#b27'>24</ns0:ref> The Bland-Altman method estimated the measurement bias, with lower and upper limits of agreement between results. Statistical analyses were performed using JAMOVI software (JAMOVI project, version 0.9, 2018). Manuscript to be reviewed</ns0:p></ns0:div>
<ns0:div><ns0:head>Results</ns0:head><ns0:p>Validity: Force platform vs. PA No significant differences were observed in pressure trials (Table <ns0:ref type='table' target='#tab_0'>1</ns0:ref>) between the adapted pressure algometer (405.63±235.34 kPa) and the force platform (434.15±239.18 kPa; p=0.25).</ns0:p><ns0:p>The ICC (2,1) and Cronbach's α returned values of 0.98 and 0.99, respectively. The SEM returned a value of 0.005 kgf, and the linear regression showed statistically significant results (r=0.99; adjusted r 2 =0.99; p=0.001). The Bland-Altman results showed high levels of agreement (Figure <ns0:ref type='figure' target='#fig_4'>2</ns0:ref>).</ns0:p></ns0:div>
<ns0:div><ns0:head>INSERT FIGURE 2 HERE</ns0:head></ns0:div>
<ns0:div><ns0:head>Intra-and inter-rater reliability</ns0:head><ns0:p>The pressure pain threshold from both raters showed very low variation over time (Rater 1: Day 1=203±74 kPa, Day 2=206±71.6 kPa; Rater 2: Day 1=214±73.7 kPa, Day 2=215±69.6 kPa). The intra-rater comparison showed no significant differences (Moment: F=0.05; p=0.83 and Moment*Rater: F=0.01; p=0.93) (Table <ns0:ref type='table' target='#tab_0'>1</ns0:ref>). The ICC (2,1) and Cronbach's α analysis returned relevant values (Rater 1: ICC (2,1) =0.76, Cronbach's α=0.85; Rater 2: ICC (2,1) =0.73, Cronbach's α=0.84). The SEM values were low (Rater 1=0.02, Rater 2=0.01), and moderate values were also obtained in the linear regression analysis (Rater 1: r=0.74, adjusted r 2 =0.52; Rater 2: r=0.73, adjusted r 2 =0.50). The Bland-Altman results showed high levels of agreement (Figure <ns0:ref type='figure' target='#fig_5'>3</ns0:ref>).</ns0:p></ns0:div>
<ns0:div><ns0:head>INSERT TABLE 1 HERE</ns0:head><ns0:p>The inter-rater reliability showed no differences among measurements (F=0.22; p=0.64) (Table <ns0:ref type='table' target='#tab_0'>1</ns0:ref>), with moderate results for reliability analysis (Day 1: ICC (2,1) =0.56, Cronbach's α=0.75; Day 2: ICC (2,1) =0.54, Cronbach's α=0.71). The SEM results showed very low values (Day 1: 0.04 kgf, Day 2: 0.02 kgf), and moderate values in the linear regression analysis (Day 1: r=0.60, adjusted r 2 =0.33; Day 2: r=0.55, adjusted r 2 =0.26). The Bland-Altman analysis showed acceptable levels of agreement (Figure <ns0:ref type='figure' target='#fig_5'>3</ns0:ref>).</ns0:p></ns0:div>
<ns0:div><ns0:head>INSERT FIGURE 3 HERE</ns0:head><ns0:p>PeerJ reviewing PDF | (2020:02:46274:3:0:NEW 16 Sep 2020)</ns0:p></ns0:div>
<ns0:div><ns0:head>Manuscript to be reviewed</ns0:head></ns0:div>
<ns0:div><ns0:head>Discussion</ns0:head><ns0:p>This study was designed to examine the instrumental validity and the intra-and inter-rater reliability of an adapted low-cost pressure algometer. The results showed no significant differences in the peak compressive force recorded from the adapted pressure algometer and the force platform. A significant correlation was observed between the low-cost adapted algometer and the force platform while measuring the same construct. These findings support the primary hypothesis, which contends that a low-cost pressure algometer is both valid and reliable to be considered as a standard equipment to assess the pressure pain threshold. Therefore, the tested device seems to be an acceptable alternative to expensive instruments.</ns0:p><ns0:p>Previous studies showed acceptable levels of validity and reliability of other digital algometry systems. 8,10 Kinser et al. <ns0:ref type='bibr' target='#b9'>8</ns0:ref> tested the construct validity of a digital pressure algometer using the same protocol as our study by manually applying pressure on a force platform. The authors used ten sets of five applications to 80 N and one additional set of five applications to subsequent levels of progressive 10 N (20, 30, 40, 50, 60, 70, 80, 90, 100, and 110 N). The results showed high correlation between the tested algometer and the force platform (r = 0.99) for both 80 N and incremental trials. Vaughan et al. <ns0:ref type='bibr' target='#b13'>12</ns0:ref> also used the force platform as a reference instrument to test the validity of a digital pressure algometer. The authors applied 300 vertical pressures on the force platform with progressive pressure rates (10, 20, 30, 40, and 50 kPa/s). The result showed an excellent ICC range (0.90 to 0.99) for all comparisons. In general, all previous studies and the present work suggest the validity of the results obtained from a digital pressure algometer. These excellent results could be attributed to the strain gauge-based system used to acquire the signals.</ns0:p><ns0:p>The conversion from analogic (load cell deformation) to electrical-digital signal (strain gauge) is very effective, even in very affordable systems. As the resistance varies in a sturdy element with the applied force, the strain gauge converts the force (in this case, pressure) into a change in electrical resistance that can be measured.</ns0:p><ns0:p>Several studies have evaluated the reliability of distinct pressure algometers as a tool to distinguish healthy individuals from those with musculoskeletal disorders. Balaguier et al. <ns0:ref type='bibr' target='#b28'>25</ns0:ref> found high reliability between all three pressure pain threshold measures at sites in the lower back. Walton et al. <ns0:ref type='bibr' target='#b5'>5</ns0:ref> assessed the intra-rater and inter-rater reliability of an accessible digital algometer in 60 healthy volunteers and 40 individuals with neck pain. The authors tested the upper fibers of the trapezius and tibialis anterior muscles. The intra-rater ICC results in both groups ranged from 0.94 to 0.97 for the trapezius and tibialis anterior muscles. The inter-rater ICC range (0.79-0.90) was lower than that of the intra-rater due to variations between observers. However, both results were considered adequate. Waller et al. <ns0:ref type='bibr' target='#b29'>26</ns0:ref> found high intra-and inter-rater reliability (ICC=0.81-0.99; ICC=0.92-0.95, respectively) using five research assistants. Each assistant tested 20 pain-free subjects at the wrist, leg, cervical, and lumbar spine. The intra-rater SEM ranged between 79 and 100 kPa. However, Van Wilgen et al. <ns0:ref type='bibr' target='#b30'>27</ns0:ref> found lower values for intra-rater reliability compared to inter-rater reliability of pressure algometry in healthy volleyball athletes and those with patellar tendinopathy. The authors found high inter-rater reliability (ICC = 0.93), but only moderate intra-rater reliability (ICC = 0.60) for pain pressure threshold measurements. The authors argued that the lower intra-rater ICC values were probably due to variance within the observer and also within the athletes, as the pain in patellar tendinopathy varies over time. The diagnosis of fibromyalgia utilizes the pressure pain threshold as a key assessment to distinguish healthy individuals from those with fibromyalgia. <ns0:ref type='bibr' target='#b31'>28,</ns0:ref><ns0:ref type='bibr' target='#b32'>29</ns0:ref> Neck pain, cranio-cervical headache, and temporomandibular disorders also include the pressure pain threshold as an important component for clinical reasoning about the level of severity, influencing the treatment direction. <ns0:ref type='bibr' target='#b5'>5,</ns0:ref><ns0:ref type='bibr' target='#b10'>9,</ns0:ref><ns0:ref type='bibr' target='#b33'>30</ns0:ref> However, those previous studies used commercial pressure algometers. For clinical and ambulatory settings, the high cost and the user's interface would be an issue to obtain fast objective pain measurements, requiring both training and experience for assessments. Brazilian physiotherapists have an average monthly salary of USD 500, according to the Occupational Brazilian Classification (https://www.salario.com.br/profissao/fisioterapeuta-geral-cbo-223605/).</ns0:p><ns0:p>The adapted pressure algometer used in this study had a production cost of USD 10.00, while the standard digital equipment cost ranged from USD 600.00 to USD 1,000.00. The validation procedure enables use of the low-cost adapted pressure algometer for clinical assessments in a practice routine, which may directly impact primary and ambulatory care in low-income and developing countries, by adding an objective and inexpensive tool to assess the pressure pain threshold.</ns0:p><ns0:p>Some limitations of the present study must be addressed. The pressure pain threshold in body sites other than the deltoid muscle must be assessed to ensure the validity of the adapted pressure algometer on different sites. However, we hypothesize that they should not give any different results to direct assessment using the adapted pressure algometer, since the standard deviation remained at very low values and the current results gave very good measures compared to the force platform and additional good reliability. The instrumental validity of an equipment's measurements also ensures unbiased assessments. <ns0:ref type='bibr' target='#b34'>31</ns0:ref> Other studies have identified different factors to consider when evaluating the pressure pain threshold, such as gender and obesity. <ns0:ref type='bibr' target='#b35'>32,</ns0:ref><ns0:ref type='bibr' target='#b36'>33</ns0:ref> A review of studies involving induced pain found a consistent pattern of women exhibiting greater pain sensitivity and a reduction in pain inhibition compared to men. <ns0:ref type='bibr' target='#b37'>34</ns0:ref> In addition, the characteristic of pain imposed is an important factor for these differences, since the type of pressure pain has one of the highest effect sizes in the pain report. <ns0:ref type='bibr' target='#b23'>21,</ns0:ref><ns0:ref type='bibr' target='#b39'>35</ns0:ref> It is suggested that interactions between biological and psychosocial factors are responsible for these gender differences, but all studies indicate the need for additional research to elucidate the mechanisms that drive gender differences in pain responses. <ns0:ref type='bibr' target='#b35'>32,</ns0:ref><ns0:ref type='bibr' target='#b37'>34,</ns0:ref><ns0:ref type='bibr' target='#b39'>35</ns0:ref> Some studies suggest that in areas with additional subcutaneous fat, pain thresholds for electrical or pressure stimuli increase and pain sensitivity decreases in obese individuals. <ns0:ref type='bibr' target='#b36'>33,</ns0:ref><ns0:ref type='bibr' target='#b40'>36</ns0:ref> A study has also shown biochemical changes in trigger points with higher levels of inflammatory mediators, catecholamines, and cytokines in obese individuals. <ns0:ref type='bibr' target='#b41'>37</ns0:ref> Mechanical stretching of the skin in response to excess fat can lead to a decrease in the density of nociceptive fibers, and obesity is associated with the chemical inhibition of pain with an increase in β endorphin and endogenous opioid peptide. <ns0:ref type='bibr' target='#b36'>33</ns0:ref> The present study had a balanced cohort with regard to participant sex, and all participants were classified as normal according to their body mass index. However, the current sample was chosen only for reliability analysis. Pressure pain threshold as a clinical result is well established, but more studies should take into account sex and body mass index differences to avoid bias in experimental protocols. <ns0:ref type='bibr' target='#b35'>32</ns0:ref> Pressure pain threshold was also positively but poorly correlated with high-density lipoprotein cholesterol. <ns0:ref type='bibr' target='#b42'>38</ns0:ref> A high pressure pain threshold was also found among subjects with hyperglycemia and excessive alcohol consumption. <ns0:ref type='bibr' target='#b42'>38</ns0:ref> In the present study, no blood assessment was performed to exclude those factors. However, the sample consisted of young adults, decreasing the chance of any important health issues. Additionally, exclusion criteria included previous excessive alcohol consumption.</ns0:p><ns0:p>Considering its portability, easy assembly, and lower cost, the currently tested device seems to be a valid standard equipment for pressure pain threshold assessment. Therefore, the adapted pressure algometer is a valid device providing similar measurements compared to a force platform. The portability, cost-effectiveness, and friendly user system provide an effective way to measure the pressure pain threshold.</ns0:p></ns0:div>
<ns0:div><ns0:head>Conclusions</ns0:head><ns0:p>The current hypothesis is that a low-cost pressure algometer is valid and reliable enough to be considered as a standard equipment to assess the pressure pain threshold. The results showed that the low-cost adapted pressure algometer is a valid tool compared to a force platform. The lowcost adapted pressure algometer is also reliable for assessing the pressure pain threshold. Future directions include evaluating the low-cost adapted pressure algometer in routine clinical assessments for the systematic evaluation of pressure pain. Further studies should consider other assessments, such as temporal summation and conditioned modulated pain, using a low-cost adapted pressure algometer. Manuscript to be reviewed Manuscript to be reviewed </ns0:p><ns0:note type='other'>Figure 1</ns0:note><ns0:note type='other'>Figure 2</ns0:note><ns0:note type='other'>Figure 3</ns0:note></ns0:div><ns0:figure xml:id='fig_0'><ns0:head></ns0:head><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:02:46274:3:0:NEW 16 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_1'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:02:46274:3:0:NEW 16 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_2'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:02:46274:3:0:NEW 16 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_3'><ns0:head>Figure 1 .</ns0:head><ns0:label>1</ns0:label><ns0:figDesc>Figure 1. Adapted pressure algometer -PA.</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_4'><ns0:head>Figure 2 .</ns0:head><ns0:label>2</ns0:label><ns0:figDesc>Figure 2. Bland-Altman plot: instrumental validity.</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_5'><ns0:head>Figure 3 .</ns0:head><ns0:label>3</ns0:label><ns0:figDesc>Figure 3. Bland-Altman plot: Intra-rater reliability.</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_6'><ns0:head>Figure 3 .</ns0:head><ns0:label>3</ns0:label><ns0:figDesc>Figure 3. Bland-Altman plot: Intra-rater reliability. A) Rater 1: Bias = -2.55 (95% confidence interval [CI] = -24.3 to 22.2); lower limit of agreement (LLA) = -106.38 (95% CI = -149 to -63.3); upper limit of agreement (ULA) = 101.28 (95% CI = 58.2 to 144.4). 2). B) Rater 2: Bias = -1.03 (95% CI = -25.9 to 23.9); LLA = -105.28 (95% CI = -148.6 to -62); ULA = 103.22 (95% CI = 59.9 to 146. Inter-rater reliability. C) Day 1: Bias = -10.8 (95% CI: -41.6 to 20); LLA = -139.8 (95% CI: -193.4 to -86.3); ULA = 118.2 (95% CI: 64.7 to 171.8); D) Day 2: Bias = -9.27 (95% CI = -40.7 to 22.1); LLA = -140.77 (95% CI = -195.4 to -86.2); ULA = 122.23 (95% CI = 67.6 to 176.8).</ns0:figDesc></ns0:figure>
<ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='21,42.52,70.87,525.00,393.75' type='bitmap' /></ns0:figure>
<ns0:figure type='table' xml:id='tab_0'><ns0:head>Table 1 :</ns0:head><ns0:label>1</ns0:label><ns0:figDesc>Validity, Intra-and Inter-rater reliability pairwise comparisons.</ns0:figDesc><ns0:table><ns0:row><ns0:cell>95% Confidence Interval</ns0:cell></ns0:row></ns0:table><ns0:note>PeerJ reviewing PDF | (2020:02:46274:3:0:NEW 16 Sep 2020)</ns0:note></ns0:figure>
</ns0:body>
" | "Editor comments (Denis Marcellin-Little)
Your manuscript is improved, please update the document according to the input from Reviewer 2 and please make the following changes before resubmission.
ANSWER: Dear Prof. Marcellin-Little, thank you for your effort to improve our manuscript. Please, see our corrected version in accordance to your comments.
- Abstract – add the unit after the reported SEM (kPa).
ANSWER: Corrected.
- Abstract – change “A very strong” to “Very strong”
ANSWER: Corrected.
- Abstract – In the phrase “… demonstrated both the validity and …”, delete the word “both”.
ANSWER: Corrected.
- Line 62. Delete the word “readings”
ANSWER: Corrected.
- Line 102. Change “3 digits” to “3-digit”
ANSWER: Corrected.
- Line 307. Change “assemblage” to “assembly”.
ANSWER: Corrected.
Reviewer 2 (Andrea Tomas)
Basic reporting
This version of the manuscript shows a significant improvement in structure and overall writing. It is clear and much easier to follow than previous versions. There are a few instances in which the literature is referenced in an inconsistent manner (see the attached document for specific lines). These should be addressed prior to publication.
ANSWER: Dear Prof. Tomas, thank you for your kind comment. The specific comments are addressed below.
Line 58 – The superscript reference (8) is used in the middle of the sentence, which is different from how the references appear for example in line 50 (2-4 at the end of the sentence). Make sure to be consistent throughout and to be following the required citation style of the journal.
ANSWER: Corrected.
Line 77 – Reference in middle of sentence again
ANSWER: Corrected.
Line 160 – Recommend writing “pressed” instead of “pressured”
ANSWER: Corrected.
Comments for the Author
The efforts undertaken to improve the English and to provide clarity and structure to this manuscript are clear and I appreciate the effort put forth. The manuscript has improved dramatically as a result.
ANSWER: Thank you for you kind comment. The authors appreciate the reviewer’s efforts to improve our manuscript.
" | Here is a paper. Please give your review comments after reading it. |
9,810 | "<ns0:abstract xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:p>Background. Obesity is at a record high in Gulf Cooperation Council (GCC) countries and is expected to continue increasing. Diet is a major contributor to this disease, but there is inadequate nationally representative dietary research from these countries. The aim was to quantify the number dietary studies using food frequency questionnaires (FFQs) that have been conducted in individual GCC countries and to assess the quality of eligible studies.</ns0:p><ns0:p>Methodology. Four databases (PubMed, Web of Science, MEDLINE, and DOAJ) were searched for keywords; records were screened for eligible studies and data were abstracted on study characteristics (publication year, geographical locations, sample size, units of measurement, number of foods examined, number of Arab foods and key findings). Quality was assessed using an adapted Newcastle-Ottawa Quality Assessment Scale for cross-sectional studies.</ns0:p></ns0:div>
<ns0:div><ns0:head>Results.</ns0:head><ns0:p>Only seven studies were eligible from four of six GCC countries (Saudi Arabia, Bahrain, Kuwait and Qatar). All eligible studies used FFQs, but only 29% used a validated questionnaire, one being in Arabic, and none of the studies used any additional tools to measure diet. Fifty-seven percent of studies made an effort to include local foods. The majority of studies (71%) either measured frequency or quantity of food consumed, but only 29% attempted to account for both frequency and quantity.</ns0:p></ns0:div>
<ns0:div><ns0:head>Conclusions.</ns0:head><ns0:p>The quality of studies varied and major weaknesses of FFQ validity and adaptability have been highlighted. More dietary investigations are needed using validated FFQs that have been adapted to the local GCC diets. Using reference tools will allow for better dietary estimations.</ns0:p></ns0:div>
</ns0:abstract>
<ns0:body xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:head>Introduction</ns0:head><ns0:p>Obesity is an epidemic in the countries of the Gulf Cooperation Council (GCC) (that is, Saudi Arabia, Bahrain, Kuwait, Oman, Qatar, and United Arab Emirates). Approximately one out of every three adults is obese (Body Mass Index ≥30), and the obesity prevalence has been rising in every member country. For example, between 2011 and 2016, the obesity prevalence rose in Saudi Arabia (KSA) from 32.1 to 35.4%, in Bahrain from 27.1% to 29.8%, in Kuwait from 35.1% to 37.9%, in Oman from 23.7% to 27%, in Qatar from 31.8% to 35.1%, and in the United Arab Emirates (UAE) from 28.3% to 31.7% (1). Apart from obesity, the GCC countries are also leading countries in the world in diabetes and cardiovascular disease prevalence <ns0:ref type='bibr' target='#b1'>(2)</ns0:ref><ns0:ref type='bibr' target='#b2'>(3)</ns0:ref><ns0:ref type='bibr' target='#b3'>(4)</ns0:ref>.</ns0:p><ns0:p>There is mounting evidence of a potential causal link between specific dietary factors (such as, fruit, vegetable, processed meat, and trans-fat intake) and the above mentioned chronic conditions <ns0:ref type='bibr' target='#b4'>(5)</ns0:ref><ns0:ref type='bibr' target='#b6'>(6)</ns0:ref><ns0:ref type='bibr' target='#b7'>(7)</ns0:ref>. A recent systematic review of dietary data from 195 countries found that 22% of all adult deaths worldwide are due to unhealthy diet; more than half of diet-related deaths are attributable to a high sodium intake, low intake of whole grains, and low fruit intake <ns0:ref type='bibr' target='#b8'>(8)</ns0:ref>.</ns0:p><ns0:p>Several factors likely contribute to obesity in GCC countries. With increased wealth from oil reserves, these countries have seen rapid economic growth. The urbanization of the landscape has seen a rise in international fast food chains, making it easier and quicker to consume processed foods <ns0:ref type='bibr' target='#b9'>(9,</ns0:ref><ns0:ref type='bibr' target='#b10'>10)</ns0:ref>. This has resulted in a change of diet from traditional, locally produced goods such as wheat, vegetables and dates to fast foods high in fat, sugar and salt content <ns0:ref type='bibr' target='#b11'>(11)</ns0:ref>. Whist all GCC countries have attempted to develop a national plan that addresses nutrition and physical activity, most have not followed up, which makes it difficult to evaluate the impact of such programs <ns0:ref type='bibr' target='#b12'>(12)</ns0:ref>. Changes in lifestyle such as increased use of cars, electrical home appliances, television and gaming devices have resulted in a more sedentary lifestyle <ns0:ref type='bibr' target='#b9'>(9,</ns0:ref><ns0:ref type='bibr' target='#b13'>13)</ns0:ref>. The extremely hot climate found in these countries is also likely to deter outdoor activities with people opting to use cars, even for short journeys <ns0:ref type='bibr' target='#b9'>(9,</ns0:ref><ns0:ref type='bibr' target='#b14'>14)</ns0:ref>. A combination of all these is likely to play a role in the current epidemic.</ns0:p><ns0:p>Given the high prevalence of chronic conditions in the GCC, one would expect that these countries engage extensively in diet and nutrition research. However, dietary studies have been limited; only approximately 1% of global dietary research has come from Arab countries <ns0:ref type='bibr' target='#b15'>(15)</ns0:ref>. Their h-indices [measurement of performance by combining productivity (number of papers) and impact (number of citations)] are much lower than neighbouring non-Arab countries <ns0:ref type='bibr' target='#b16'>(16)</ns0:ref>.</ns0:p><ns0:p>One would similarly expect that assessment tools used in dietary studies from GCC countries would differ from those in European or North American studies as Middle Eastern diets vary a great deal from their western counterparts. For example, date palm fruit is highly consumed in Gulf regions, with daily consumption ranging from 68 -164 g daily <ns0:ref type='bibr' target='#b17'>(17)</ns0:ref><ns0:ref type='bibr' target='#b18'>(18)</ns0:ref><ns0:ref type='bibr' target='#b19'>(19)</ns0:ref>, whereas only 140 g of this fruit is consumed annually in Europe <ns0:ref type='bibr' target='#b20'>(20)</ns0:ref>. Differences such as these should be accommodated for when designing dietary assessment tools.</ns0:p><ns0:p>The usual assessment tools used in dietary research are <ns0:ref type='bibr'>24-hour</ns0:ref> dietary recall (openended, food consumed the previous day, conducted by trained interviewer), diet records (openended, participants trained to record own diet), and food frequency questionnaires (FFQs) (closed-ended, typically a food list and frequency of consumption in a given period). All have strength and limitations <ns0:ref type='bibr' target='#b21'>(21)</ns0:ref>, but due to low cost, low respondent burden and ease of use compared to other methods, FFQs are thought to be the best choice for measuring habitual diet in large populations. The usefulness and reliability of FFQs have been demonstrated with strong correlations with diet records <ns0:ref type='bibr' target='#b22'>(22,</ns0:ref><ns0:ref type='bibr' target='#b23'>23)</ns0:ref>, dietary recalls <ns0:ref type='bibr' target='#b24'>(24)</ns0:ref><ns0:ref type='bibr' target='#b26'>(25)</ns0:ref><ns0:ref type='bibr' target='#b27'>(26)</ns0:ref>, and objective biomarkers of diet <ns0:ref type='bibr' target='#b24'>(24,</ns0:ref><ns0:ref type='bibr' target='#b26'>25)</ns0:ref>. As an FFQ is a self-reported subjective tool, FFQs should be tested for validity alongside a reference tool.</ns0:p><ns0:p>The authors' aimed to conduct both a quantitative and qualitative review of all dietary studies conducted within each GCC country. To be as nationally representative as possible and to provide a current and more reflective picture of diet in the GCC, only studies carried out in multiple regions (must be a minimum of two regions) were included. Dietary research that used FFQs in individual GCC countries (Bahrain, Kuwait, Oman, Qatar, KSA, UAE) over the past ten years (2009-2019) were assessed. The characteristics of the studies were described and their quality was assessed using a widely accepted scoring tool <ns0:ref type='bibr' target='#b28'>(27,</ns0:ref><ns0:ref type='bibr' target='#b29'>28)</ns0:ref>. The objectives were to (1) identify multi-regional GCC dietary studies that used FFQs, (2) assess the quality of the studies, and (3) offer recommendations for future dietary assessments.</ns0:p></ns0:div>
<ns0:div><ns0:head>Method SEARCH STRATEGY AND INCLUSION CRITERIA</ns0:head><ns0:p>This review was conducted in May 2019. PubMed, Web of Science, MEDLINE, and Directory of Open Access Journals (DOAJ) databases were searched using the following terms: 'diet,' 'frequency questionnaire' in combination with each of the Gulf Cooperation Council countries ('Bahrain', 'Kuwait', 'Oman', 'Qatar', 'Saudi Arabia', 'UAE'). A total of 431 records were identified from PubMed (n = 241),Web of Science (n= 34), MEDLINE (n = 132) and DOAJ (n = 24). Duplicates (n=39) were removed, and the unique records (n = 392) were screened for the following inclusion criteria: (1) assessed diet using a food frequency questionnaire, (2) included data from multiple regions/cities (minimum two) of the Gulf country of focus, and (3) data were collected in the last ten years (that is, 2009 and later).</ns0:p></ns0:div>
<ns0:div><ns0:head>EXCLUSION OF STUDIES</ns0:head><ns0:p>Studies were excluded if they (1) examined data from only one specific region/city/population group and therefore were not necessarily nationally representative, (2) were multi-national studies that did not give Gulf-nation-specific results, (3) were not conducted in a GCC country, (4) were intervention studies where the diet had purposefully been changed, <ns0:ref type='bibr' target='#b4'>(5)</ns0:ref> were review or meta-analysis papers, (6) used an assessment tool other than a food frequency questionnaire, or <ns0:ref type='bibr' target='#b7'>(7)</ns0:ref> had no findings related to diet or did not report those findings. Therefore, the final analysis was limited to seven dietary studies (Fig. <ns0:ref type='figure' target='#fig_1'>1</ns0:ref>).</ns0:p></ns0:div>
<ns0:div><ns0:head>DATA CHARTING PROCESS</ns0:head><ns0:p>After an initial search and screening, the following data from each study were charted: publication year, author(s) name(s), geographical location, sample size, age range of participants, dietary assessment tool(s) used, units of measurement (for example, times/week, servings/day, etc.), total number of foods examined, number of Arab-specific foods (and where possible, the type and name of food), whether the questionnaire was validated, and dietary findings related to the most common foods studied. Any discrepancy was resolved through discussion and consensus among the authors.</ns0:p><ns0:p>CRITICAL APPRAISAL OF STUDIES Using a scoring system adapted from Newcastle-Ottawa Quality Assessment Scale for crosssectional studies <ns0:ref type='bibr' target='#b28'>(27)</ns0:ref>, each study was scored for (1) representativeness of the sample, (2) sample size, (3) non-respondents, (4) ascertainment of the exposure, (5) adaptability, (6) assessment of the outcome, and (7) statistical test (Appendix 1).</ns0:p><ns0:p>DATA ANALYSIS Study characteristics, along with main findings related to dietary intake/habits were tabulated. Additionally, indicators of study quality were assigned point values based on the quality assessment scoring scale and then summed. Each study was categorized as excellent (9-12 points), satisfactory (5-8 points), or unsatisfactory (0-4 points).</ns0:p></ns0:div>
<ns0:div><ns0:head>Results</ns0:head></ns0:div>
<ns0:div><ns0:head>STUDY CHARACTERISTICS</ns0:head><ns0:p>The search resulted in seven studies published between 2009 and 2019. Tables <ns0:ref type='table' target='#tab_3'>1 and 2</ns0:ref> show three studies were conducted in Saudi Arabia, one in Kuwait, one in Bahrain, and two in Qatar; there were no studies from Oman or the UAE. A majority of the studies (n=6) had sample sizes greater than 1000 participants, and all studies included a sample size justification. Almost all studies had a 1:1 male: female ratio (range 1: 0.9-1.4 male: female). Fifty-seven percent (n=4) of the studies were carried out with adolescents (12-19 years of age), whereas 33% (n=2) included both adolescents and adults. One study (14%) classified participants 18 years and older as adults, thus the study was considered to be carried out on an all-adult population <ns0:ref type='bibr' target='#b30'>(29)</ns0:ref>.</ns0:p><ns0:p>All studies used FFQs, but three administered the FFQ through face-to-face interviews; the rest were self-administered. One study <ns0:ref type='bibr' target='#b31'>(30)</ns0:ref> used pictures to deduce serving sizes.</ns0:p><ns0:p>The number of food items assessed ranged from two (non-specified fruits and vegetables) <ns0:ref type='bibr' target='#b32'>(31)</ns0:ref> to twenty items <ns0:ref type='bibr' target='#b30'>(29)</ns0:ref>. Only two of seven studies used validated questionnaires, adapted it for local cuisine, and had it pilot tested for suitability <ns0:ref type='bibr' target='#b30'>(29,</ns0:ref><ns0:ref type='bibr' target='#b33'>32)</ns0:ref> and from these, only one was conducted in Arabic; the other five studies did not use validated FFQs.</ns0:p><ns0:p>Key findings from each study varied based on the units of measurement. Frequency ranged from days per week, times per day, servings per day, to categories (such as, always, sometimes, never). Quantity options were servings per day, serving sizes, and serving sizes via selection of pictures.</ns0:p></ns0:div>
<ns0:div><ns0:head>QUALITY ASSESSMENT OF STUDIES</ns0:head><ns0:p>Only one of all included studies used a validated Arabic questionnaire (all were presented in English in the article) and none used any additional tools to measure diet. Donnelly et al. <ns0:ref type='bibr' target='#b30'>(29)</ns0:ref> translated the questionnaire to Arabic and back to English to ensure correct language usage. For relevance to local contexts, focus groups were conducted and the questionnaire pilot tested and thereafter further refined. Musaiger et al. <ns0:ref type='bibr' target='#b33'>(32)</ns0:ref> modified a previously validated questionnaire (Family Eating and Activity Habits Questionnaire) <ns0:ref type='bibr' target='#b34'>(33)</ns0:ref> and adapted it to ensure it reflected dietary habits of the target population. Contents of the FFQ were validated by experts in the field of nutrition, public health, and epidemiology and the questionnaire underwent pilot and test-retesting <ns0:ref type='bibr' target='#b33'>(32)</ns0:ref>.</ns0:p><ns0:p>Table <ns0:ref type='table'>3</ns0:ref> shows 57% (n =4) of the studies made an effort to include local foods, scoring a point for adaptability, whereas the other three studies either did not incorporate any local foods or did not mention it in their studies.</ns0:p><ns0:p>Five studies measured either frequency or quantity, whilst two studies scored the maximum three points for 'assessment of outcome' by having units of measurement that took into account both frequency and quantity (that is, times/week and servings/day).</ns0:p><ns0:p>All studies used appropriate statistical analysis and 86% (n =6) had an adequate response rate (≥60%). One study had 52.1% response rate ( <ns0:ref type='formula'>29</ns0:ref>) and one study did not compare between respondent and non-respondent characteristics or take non-responses into account (or did not mention it in their study) <ns0:ref type='bibr' target='#b35'>(34)</ns0:ref>.</ns0:p></ns0:div>
<ns0:div><ns0:head>Discussion</ns0:head><ns0:p>With such a high prevalence of diseases to which diet is a major contributor, it is surprising that there are so few multi-regional studies that investigated diet in the GCC in the past ten years. Five out of the seven studies included in this review did not use validated FFQs.</ns0:p><ns0:p>Dietary summaries show intake of fruit and vegetables being far below the recommended three servings of vegetables and two servings of fruit per day <ns0:ref type='bibr' target='#b36'>(35)</ns0:ref>. In Saudi Arabia, only 5.2% of individuals met the recommendation for fruit intake and 7.5% for vegetable intake. In contrast, consumption of sugary beverages was oversubscribed, with an average of 36% of adolescents (14-19 years old) reporting daily consumption <ns0:ref type='bibr' target='#b35'>(34)</ns0:ref> and 27% of 15-60 year olds <ns0:ref type='bibr' target='#b31'>(30)</ns0:ref>, exceeding local and global recommendations of sugary-drink consumption <ns0:ref type='bibr' target='#b37'>(36)</ns0:ref><ns0:ref type='bibr' target='#b38'>(37)</ns0:ref><ns0:ref type='bibr' target='#b39'>(38)</ns0:ref>. This low fruit and vegetable intake, combined with high sugary-beverage consumption, suggests a poor-quality diet across the GCC.</ns0:p><ns0:p>The varying methods of measuring diet made it difficult to compare consumption. For example, 57% of the studies assessed diet using frequency questions (how often), whilst 43% measured frequency and quantity (portions or serving size). At times, the response categories were too broad for in-depth analysis. For example, 'Do you regularly consume meals? Yes/No' <ns0:ref type='bibr' target='#b33'>(32)</ns0:ref> does not specify which meals, how many meals, or the content of the meals. Similarly, 'How often do you drink a glass of milk?' <ns0:ref type='bibr' target='#b40'>(39)</ns0:ref> does not quantify the size of the glass or the amount of milk consumed.</ns0:p><ns0:p>Adaptability was one of the main issues relating to study quality according to the quality assessment scoring scale. Studies need to make it explicit how they have categorized foods, for example, whether they have classified potatoes as starch, tuber, snack, fast food, etc. Four studies attempted to include local foods, with a maximum of two or three items added (and mentioned in the article) <ns0:ref type='bibr' target='#b30'>(29,</ns0:ref><ns0:ref type='bibr' target='#b31'>30,</ns0:ref><ns0:ref type='bibr' target='#b39'>38,</ns0:ref><ns0:ref type='bibr' target='#b41'>40)</ns0:ref>. It is concerning that the other three studies did not mention any native foods at all. In Tabacchi's review <ns0:ref type='bibr' target='#b42'>(41)</ns0:ref>, it is suggested that an FFQ with less than 70 food items reduces the quality of nutritional information that can be deduced. None of the studies included in this review had 70 items; the most was 20, the average being 11 items. Nutritional status and dietary patterns differ over time and from region to region; without the incorporation of local foods and without categorizing them under more common food groups, it is entirely possible to mask important epidemiological links between diet and disease.</ns0:p><ns0:p>An overall poor validity of FFQs was found in this review. Only one study used a validated Arabic FFQ and scored three points out of a possible four points on the quality assessment scale. Validation in large-scale studies is especially important as FFQs are prone to measurement errors and come with inherent self-bias. FFQs rely on an individual's memory and his/her own perception of food sizes, thus under-reporting remains a common problem <ns0:ref type='bibr' target='#b43'>(42)</ns0:ref><ns0:ref type='bibr' target='#b44'>(43)</ns0:ref><ns0:ref type='bibr' target='#b45'>(44)</ns0:ref><ns0:ref type='bibr' target='#b46'>(45)</ns0:ref>. Researchers have made extensive efforts in the last two decades to mitigate some of the errors with self-reporting data <ns0:ref type='bibr' target='#b47'>(46)</ns0:ref><ns0:ref type='bibr' target='#b48'>(47)</ns0:ref><ns0:ref type='bibr' target='#b49'>(48)</ns0:ref>, but diet and eating patterns are complex, and FFQs are still thought to have clear value and insight that solely objective measures cannot provide <ns0:ref type='bibr' target='#b50'>(49,</ns0:ref><ns0:ref type='bibr' target='#b51'>50)</ns0:ref>. One of the ways to minimize errors is to use a validated FFQ. FFQs are not one-size-fits-all, and it is integral that questionnaires be adapted/modified to suit the population with which they are being used. This includes first developing a good FFQ to standard procedure <ns0:ref type='bibr' target='#b53'>(51)</ns0:ref>, FFQs being in the native language, which for GCC is predominantly Arabic, and including as many local foods as possible.</ns0:p><ns0:p>Within obesity research, two areas are deficient: understanding the role of dietary habits in the obesity epidemic and sufficient intervention studies on weight loss via dietary change. Research on dietary habits in the obesity epidemic may be lacking due to a shortage of skilled researchers and research centers <ns0:ref type='bibr' target='#b12'>(12)</ns0:ref>. Obtaining accurate dietary data requires specialized nutritionists/dieticians and controlled research settings, but this is a problem across many Gulf states, where it is difficult to have sufficient numbers of well-trained staff to serve large populations and areas like Saudi Arabia (Table <ns0:ref type='table'>1</ns0:ref>). Investments should be made in specialized university health education and research courses and training in hospital departments; this will take time and resources but is a necessary step to produce expert personnel that can adequately face the challenges of regional obesity research <ns0:ref type='bibr' target='#b12'>(12)</ns0:ref>.</ns0:p><ns0:p>Research in this field may also be looking at risk factors found in Western countries and not necessarily exploring factors that are unique to the socio-cultural environment of the GCC. For example, women have been shown to be less active than men across Gulf countries <ns0:ref type='bibr' target='#b54'>(52,</ns0:ref><ns0:ref type='bibr' target='#b55'>53)</ns0:ref> and more sedentary than their British counterparts <ns0:ref type='bibr' target='#b56'>(54,</ns0:ref><ns0:ref type='bibr' target='#b59'>55)</ns0:ref>, but reasons for this behavior was poorly understood. Only by exploring the socioeconomic, environmental and cultural contexts further was it understood that the greatest barrier to physical activity for women was a lack of facilities rather than assumed low levels of knowledge, dress codes <ns0:ref type='bibr' target='#b60'>(56)</ns0:ref> or high obese-body acceptance <ns0:ref type='bibr' target='#b62'>(57)</ns0:ref>. Samara et al. suggested that future health strategies should focus on providing culturally sensitive exercise facilities for women <ns0:ref type='bibr' target='#b60'>(56)</ns0:ref>. A similar approach needs to be taken for nutrition and diet, where interventions, based on survey results, acknowledge and work with, not against, local culture and social norms <ns0:ref type='bibr' target='#b63'>(58)</ns0:ref>. Such intervention studies need to have tangible goals, clear action plans and sufficient follow-up to evaluate long-term effectiveness <ns0:ref type='bibr' target='#b64'>(59)</ns0:ref>.</ns0:p><ns0:p>Limitations of this review are that the search was carried out on four main databases; this may have missed studies published in other journals not found within these databases, and those that are currently underway or not yet published. However, additional cross-checking was performed with reference lists to ensure the maximum number of studies were screened. The small number of studies limited the generalizability of findings. To the authors' knowledge, there are no other reviews similar to the current study. There are studies that have looked at other methods for country-specific dietary assessment <ns0:ref type='bibr' target='#b50'>(49)</ns0:ref> and the Newcastle-Ottawa Assessment Scale, which was adapted for this study, has been used to assess study quality but not in the same context as the current study <ns0:ref type='bibr' target='#b28'>(27,</ns0:ref><ns0:ref type='bibr' target='#b65'>60)</ns0:ref>. Finally, although studies have looked at dietary research from other parts of the world, no study has quantified the number of dietary studies coming specifically from the GCC and assessed their quality. Our review is unique in these ways, so the results of this present study cannot be easily compared to other studies.</ns0:p><ns0:p>A particular strength is the quality assessment aspect of this review. Adapting a scoring system allowed for objective assessment of studies. It highlighted that most of the included studies were either satisfactory (n=4) or excellent (n=3), whilst making it clear that the greatest weaknesses were in the number of food items and the validity and adaptability of FFQs, which researchers should take into consideration when designing future studies. Another strength is that the review focused on large-scale, multi-regional studies, which are more representative of the respective GCC nations' populations.</ns0:p></ns0:div>
<ns0:div><ns0:head>RECOMMENDATIONS</ns0:head><ns0:p>As validity and adaptability were the lowest scoring categories, it is important to address this.</ns0:p><ns0:p>1. Validation can be assured by using a reference method. There are a variety of other methods used to measure diet, including self-reporting food records and 24-hour dietary recall (24-HDRs), but the most objective reference tool is food or nutrient biomarkers <ns0:ref type='bibr' target='#b21'>(21,</ns0:ref><ns0:ref type='bibr' target='#b66'>61)</ns0:ref>. In theory, biomarkers look like a promising method to remove the human error that comes with self-reported dietary data, but their widespread use is hindered because there are only a few known and validated biomarkers. One of the well-known biomarkers could be used as a reference measurement to validate FFQs and to assess their accuracy.</ns0:p><ns0:p>2. As KSA is the largest of the GCC countries, a quality assessment of all FFQs used in KSA should be undertaken. Comparisons should be made to see how similar they are, how inclusive they are of local cuisine and if the questionnaires are validated. This will be a labour-intensive task as the questionnaires are rarely attached to the articles or submitted as supplementary material; thus, authors will need to be contacted for original FFQs. This will give an overview of the versions of FFQs available and the Arabic food items included. By noting what foods are not represented in these questionnaires, additional foods can be added and attempts made to validate the FFQ.</ns0:p><ns0:p>A recent FFQ developed by Gosadi et al. ( <ns0:ref type='formula'>2017</ns0:ref>) is a promising start for KSA <ns0:ref type='bibr' target='#b67'>(62)</ns0:ref>. The Arabic FFQ had 140 food items and ensured it had a comprehensive food list by</ns0:p><ns0:p>comparing it with open-ended information from 24-hour dietary recalls to find that 85% of food items recalled were covered in the FFQ. The FFQ has been piloted and its reliability assessed (Cronbach's alpha test and test-retest) and it should now be used in other regions. This standard of FFQ development should be carried out with other GCC countries as well to better capture dietary habits. </ns0:p></ns0:div>
<ns0:div><ns0:head>Conclusions</ns0:head><ns0:p>This is the first review to collect, quantify and critique the quality of data from dietary studies conducted in GCC countries by using an objective scoring system approach. Study quality varied, and major weaknesses of FFQ validity and adaptability have been highlighted.</ns0:p><ns0:p>Findings consistently showed that the majority of GCC populations are not meeting the recommended fruit and vegetable intake, and sugary-beverage consumption is on the rise, implying a poor diet. However, interpretations are made with caution due to the low study sample included (n=7). In these GCC countries, where obesity levels are steadily rising, more dietary investigations are necessary. The use of validated FFQs in conjunction with other instruments like biomarkers, 24-hour recalls and/or food records is likely to provide more accurate dietary estimations.</ns0:p><ns0:p>In conclusion, it is essential that researchers develop well-designed, validated FFQs that are adapted for the GCC to standardise dietary assessments across studies.</ns0:p></ns0:div>
<ns0:div><ns0:head>Table 1 Background information and characteristics of Gulf Cooperation Countries (GCC)</ns0:head><ns0:p>(1). Manuscript to be reviewed Manuscript to be reviewed </ns0:p></ns0:div>
<ns0:div><ns0:head>Bahrain</ns0:head><ns0:note type='other'>Figure 1</ns0:note></ns0:div><ns0:figure xml:id='fig_0'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:06:49675:1:1:NEW 19 Aug 2020) Manuscript to be reviewed 3. The review only included cross-sectional studies because they give a current picture of diet (observations of diet at a given point in time). Carrying out a longitudinal study analysis (repeated observations of a population over time) would illuminate how diet has changed over time to make better-informed future predictions.</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_1'><ns0:head>Figure 1</ns0:head><ns0:label>1</ns0:label><ns0:figDesc>Figure 1 Flow chart of study eligibility of dietary studies conducted in GCC countries.</ns0:figDesc><ns0:graphic coords='27,42.52,178.87,525.00,393.75' type='bitmap' /></ns0:figure>
<ns0:figure type='table' xml:id='tab_1'><ns0:head>Table 2 (on next page)</ns0:head><ns0:label>2</ns0:label><ns0:figDesc /><ns0:table /></ns0:figure>
<ns0:figure type='table' xml:id='tab_2'><ns0:head>Table 2</ns0:head><ns0:label>2</ns0:label><ns0:figDesc>Study characteristics of national dietary assessment studies conducted in ArabGulf countries (n=7).</ns0:figDesc><ns0:table><ns0:row><ns0:cell>*where possible, names of Arab food have been included</ns0:cell></ns0:row><ns0:row><ns0:cell># number</ns0:cell></ns0:row><ns0:row><ns0:cell>SSB: sugar sweetened beverages</ns0:cell></ns0:row></ns0:table><ns0:note>PeerJ reviewing PDF | (2020:06:49675:1:1:NEW 19 Aug 2020)Manuscript to be reviewed 1</ns0:note></ns0:figure>
<ns0:figure type='table' xml:id='tab_3'><ns0:head>Table 2 Study characteristics of national dietary assessment studies conducted in Arab Gulf countries (n=7).</ns0:head><ns0:label>2</ns0:label><ns0:figDesc /><ns0:table><ns0:row><ns0:cell>2</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell cols='2'>Tool(s) used</ns0:cell><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell># of</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell>total</ns0:cell><ns0:cell># and type</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell>Age</ns0:cell><ns0:cell>Sample</ns0:cell><ns0:cell /><ns0:cell>food</ns0:cell><ns0:cell>of Arab</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>Author</ns0:cell><ns0:cell>Country</ns0:cell><ns0:cell>range</ns0:cell><ns0:cell>size</ns0:cell><ns0:cell>Type</ns0:cell><ns0:cell>items</ns0:cell><ns0:cell>food*</ns0:cell><ns0:cell cols='2'>Measurement Validated</ns0:cell><ns0:cell>Findings</ns0:cell></ns0:row><ns0:row><ns0:cell>Al Baho &</ns0:cell><ns0:cell>Kuwait</ns0:cell><ns0:cell>13 -15</ns0:cell><ns0:cell>2674</ns0:cell><ns0:cell>FFQ</ns0:cell><ns0:cell>6</ns0:cell><ns0:cell>2</ns0:cell><ns0:cell>times/day in</ns0:cell><ns0:cell>Not</ns0:cell><ns0:cell>Over 30 days, 36%</ns0:cell></ns0:row><ns0:row><ns0:cell>Badr,</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell>(1399 male;</ns0:cell><ns0:cell>(2011</ns0:cell><ns0:cell>(includes</ns0:cell><ns0:cell>Coriander</ns0:cell><ns0:cell>past 30 days</ns0:cell><ns0:cell>validated</ns0:cell><ns0:cell>of students usually</ns0:cell></ns0:row><ns0:row><ns0:cell>2011(40)</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell>1275</ns0:cell><ns0:cell>Kuwait</ns0:cell><ns0:cell>breakfast</ns0:cell><ns0:cell>(vegetable);</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell>ate fruits (≥2</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell>female)</ns0:cell><ns0:cell>GSHS)</ns0:cell><ns0:cell>meal)</ns0:cell><ns0:cell>KDD,</ns0:cell><ns0:cell>(except</ns0:cell><ns0:cell /><ns0:cell>times/day);</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell>KDcow,</ns0:cell><ns0:cell>breakfast:</ns0:cell><ns0:cell /><ns0:cell>19% ate vegetables</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell>Carnation</ns0:cell><ns0:cell>how often in</ns0:cell><ns0:cell /><ns0:cell>(≥3 times/day);</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell>(dairy)</ns0:cell><ns0:cell>last 30 days:</ns0:cell><ns0:cell /><ns0:cell>75% consumed</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell>Never, Rarely,</ns0:cell><ns0:cell /><ns0:cell>soft drink (≥1</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell>Sometimes,</ns0:cell><ns0:cell /><ns0:cell>times/day); 36%</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell>Mostly,</ns0:cell><ns0:cell /><ns0:cell>drank milk (≤2</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell>Always)</ns0:cell><ns0:cell /><ns0:cell>times /day); 48%</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell>had fast food (≥3</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell>times/week).</ns0:cell></ns0:row><ns0:row><ns0:cell>AlBuhairan</ns0:cell><ns0:cell>Saudi</ns0:cell><ns0:cell>12 -19</ns0:cell><ns0:cell>12575</ns0:cell><ns0:cell>FFQ</ns0:cell><ns0:cell>8</ns0:cell><ns0:cell>2</ns0:cell><ns0:cell>srvgs/day</ns0:cell><ns0:cell>Not</ns0:cell><ns0:cell>38% of</ns0:cell></ns0:row><ns0:row><ns0:cell>et al.,</ns0:cell><ns0:cell>Arabia</ns0:cell><ns0:cell /><ns0:cell>(6444 male;</ns0:cell><ns0:cell>(Global</ns0:cell><ns0:cell>(includes</ns0:cell><ns0:cell>Fatayer</ns0:cell><ns0:cell /><ns0:cell>validated</ns0:cell><ns0:cell>adolescents ate ≥1</ns0:cell></ns0:row><ns0:row><ns0:cell>2015(38)</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell>6131</ns0:cell><ns0:cell>School-</ns0:cell><ns0:cell>meals)</ns0:cell><ns0:cell>(snack);</ns0:cell><ns0:cell>breakfast: last</ns0:cell><ns0:cell /><ns0:cell>srvgs/day of fruit</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell>female)</ns0:cell><ns0:cell>based</ns0:cell><ns0:cell /><ns0:cell>molokhiya</ns0:cell><ns0:cell>30 days</ns0:cell><ns0:cell /><ns0:cell>and 54.3% ate ≥1</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell>Student</ns0:cell><ns0:cell /><ns0:cell>(vegetable)</ns0:cell><ns0:cell>(never, rarely,</ns0:cell><ns0:cell /><ns0:cell>srvgs/day of</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell>Health</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell>some, most,</ns0:cell><ns0:cell /><ns0:cell>vegetables. 38%</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell>Survey)</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell>daily)</ns0:cell><ns0:cell /><ns0:cell>drank ≥2</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell>Number of</ns0:cell><ns0:cell /><ns0:cell>carbonated</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell>main meals</ns0:cell><ns0:cell /><ns0:cell>beverages/day.</ns0:cell></ns0:row></ns0:table><ns0:note>PeerJ reviewing PDF | (2020:06:49675:1:1:NEW 19 Aug 2020) Manuscript to be reviewed PeerJ reviewing PDF | (2020:06:49675:1:1:NEW 19 Aug 2020) Manuscript to be reviewed day/week). PeerJ reviewing PDF | (2020:06:49675:1:1:NEW 19 Aug 2020) Manuscript to be reviewed PeerJ reviewing PDF | (2020:06:49675:1:1:NEW 19 Aug 2020)</ns0:note></ns0:figure>
<ns0:note place='foot'>PeerJ reviewing PDF | (2020:06:49675:1:1:NEW 19 Aug 2020)Manuscript to be reviewed</ns0:note>
<ns0:note place='foot'>PeerJ reviewing PDF | (2020:06:49675:1:1:NEW 19 Aug 2020)</ns0:note>
</ns0:body>
" | "Response to Reviewers’ Comments
Editor comment: Consider providing further details on the synthesis of contents/contexts of key cited literature. It is not only to cite literature, it is about the synthesis, the hows? the whys?
RESPONSE: Table 1 (cited on p. 5, line 143) has been created to give some background on the Gulf countries and context for the studies included in our review.
Reviewer 1 comments:
1. Authors could avoid the use of 'we' in the write up and rather use reporters' speech.
RESPONSE: This has been amended throughout the text as suggested.
2. Useful to extend the review including other search databases
RESPONSE: MEDLINE and Directory of Open Access Journals (DOAJ) were also searched as per the reviewer’s recommendations. After screening, one additional eligible study (Donnelly et al., 2018) was found and included in our results. The results and discussion sections and tables have been modified to reflect this.
Reviewer 2 comments:
1. Some words should be replaced and some sentences need to be rephrased or separated to ensure good flow. 1. Line 2 and 3: I recommend writing GCC in full in the title; 2. Line 39-42: should be divided into two complete sentences. Replace “i.e” with “that is”; 3. Line 48: replace “e.g” with “such as”; 4. Line 53-54: please rephrase this sentence “However, dietary studies have been limited; the research output from Arab countries constitutes ≈ 1% of global research (9).”; 5. Line 171: replace “sugared” with either “sweetened” or “sugary”
RESPONSE: These have been changed as the reviewer recommended.
2. Introduction would be more valuable to readers if more information are provided on the subject of obesity in the GCC, and some of the factors limiting research quantity and quality on this subject.
RESPONSE: More information has been provided in the introduction about obesity in the GCC. A paragraph (p.2-3, lines 54-65) was added to enumerate factors such as changes in social and cultural environments, physical activity, increased wealth (due to oil reserves), and diet and nutrition. Some possible factors limiting research on obesity have been added to the discussion section (p.7, lines 227-248).
3. The authors did not point out any research similar to what they have conducted.
RESPONSE: We acknowledge this and have now included this as a point in our discussion (p.7-8, lines 253-260)
4. The authors could include all the published articles they found in their search into this review (regardless of their strength and weaknesses) and pointing out how those studies could be improved
RESPONSE: We did not include all the articles from our search results in the main body of our review as they did not meet the inclusion criteria, and hence, were not assessed for quality. Only the studies that were eligible (total 7) were assessed for strengths and weaknesses. We did not include all articles found in our search as we felt this may overload readers with excessive and distracting information. However, we are happy to attach all search results and reasons for exclusion as an appendix at the reviewer’s request.
" | Here is a paper. Please give your review comments after reading it. |
9,811 | "<ns0:abstract xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:p>Background. COVID-19 pandemic is found to affect the mental health of the population.</ns0:p><ns0:p>Undergraduate medical students are especially prone to mental health disorders and hence could be more vulnerable to the impact of the pandemic. Methods. A prospective longitudinal study was conducted on 217 undergraduate medical students of a medical college in Chennai, India. Depression, anxiety & stress levels were recorded using Depression Anxiety Stress Scale 21 Items (DASS21) before & during COVID-19 outbreak in India in December 2019 & June 2020 respectively. In the follow-up survey, apart from DASS21, Pittsburgh Sleep Quality Index was used to assess sleep quality and a selfadministered questionnaire was used to assess the impact of COVID-19 related stressors including status of COVID-19 testing and interactions with COVID-19 patients, selfperceived levels of concerns & worries caused by COVID-19, related to academics [COVID-19-AA(academic apprehensions)] and worries for the self & family/friends [COVID-19-GA(general apprehensions)]. Cross sectional & longitudinal comparison of overall scores of depression, anxiety & stress and scores stratified by gender, year of study, place of residence and monthly family income were performed. Predictors for depression, anxiety & stress during COVID-19 were investigated using adjusted binary logistic regression analysis and results were expressed as adjusted odds ratio with 95% confidence interval (CI). P value<0.05 was considered statistically significant. Results. Average scores for depression (D), anxiety (A) & stress (S) during baseline survey were</ns0:p></ns0:div>
</ns0:abstract>
<ns0:body xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:head>Introduction</ns0:head><ns0:p>The World Health Organization (WHO) announced COVID-19 (SARS-CoV-2) outbreak initially, as a public health emergency of international concern (PHEIC) on January 30, 2020 and later on, owing to its rapid global spread, it was declared as a pandemic on <ns0:ref type='bibr'>March 11, 2020</ns0:ref><ns0:ref type='bibr'>(WHO 2020)</ns0:ref>. In India, the first COVID-19 case was reported in Kerala on January 30, 2020 and by May 19, the number of cases had crossed one hundred thousand and by July 7, India became the world's third worst hit nation with 7,19,665 confirmed cases and 20,160 deaths, following Brazil and United states of America. In India, Maharashtra is the worst hit state followed by Tamil Nadu with more than one hundred thousand cases and 1571 deaths <ns0:ref type='bibr' target='#b53'>(Newsdesk 2020)</ns0:ref>. The Government of India declared lockdown on 25th March 2020 as a measure to mitigate the spread of infection. However, prolonged lockdown is not only unfavorable to the individuals, it also significantly affects the nation's economy. As a way to revive and restore the affected economy, phase-wise upliftment of lockdown was announced from June 1 with easing of restrictions, while the lockdown was extended for the containment zones alone.</ns0:p><ns0:p>Public health emergencies such as epidemic/pandemic like SARS, MERS and Ebola outbreak is associated with increased psychological distress in the affected population <ns0:ref type='bibr' target='#b8'>(Batawi et al. 2019;</ns0:ref><ns0:ref type='bibr' target='#b38'>Lee et al. 2007;</ns0:ref><ns0:ref type='bibr'>Lötsch et al. 2017)</ns0:ref>. Maladaptive behaviors, emotional and defensive reactions are some of the psychological responses to pandemic <ns0:ref type='bibr' target='#b71'>(Taylor 2019)</ns0:ref>. Social isolation was found to be strongly associated with anxiety, depression, self-harm, and suicidal tendencies <ns0:ref type='bibr' target='#b47'>(Matthews et al. 2019)</ns0:ref>. Studies prove that social distancing for a longer duration could affect the mental health negatively <ns0:ref type='bibr' target='#b60'>(Reynolds et al. 2008)</ns0:ref>. Isolation, boredom, frustrations, worries about contracting the infection, lack of freedom, concerns for family/friends are some of the factors that could affect mental well-being during quarantine <ns0:ref type='bibr' target='#b11'>(Brooks et al. 2020)</ns0:ref>. Poor sleep quality and increased psychological distress were well documented during SARS pandemic <ns0:ref type='bibr' target='#b15'>(Chen et al. 2006;</ns0:ref><ns0:ref type='bibr' target='#b33'>Johal 2009)</ns0:ref>. Poor sleep is associated with negative emotions, depressive symptoms and increases the risk of mental illness <ns0:ref type='bibr' target='#b0'>(Agargun & Kara 1997;</ns0:ref><ns0:ref type='bibr' target='#b70'>Tao et al. 2017)</ns0:ref>.</ns0:p><ns0:p>In a recent study conducted during COVID-19 outbreak in India, one fifth of adult Indians were found to be suffering from depression and stress and one fourth from anxiety <ns0:ref type='bibr' target='#b62'>(Saikarthik et al. 2020)</ns0:ref>. Mental health of medical students is known to be even poorer, when compared to general population <ns0:ref type='bibr' target='#b9'>(Bergmann et al. 2019)</ns0:ref>. Medical education is the most demanding of all the other professional courses in terms of both academics and emotional component of the students <ns0:ref type='bibr' target='#b77'>(Wolf 1994</ns0:ref>). Globally, one in three medical students were found to have anxiety which is higher than the general population (Tian-Ci <ns0:ref type='bibr'>Quek et al. 2019)</ns0:ref>. Depression, suicidal ideation, suicide rates, substance abuse and mental health disorders were also found to be higher in medical students <ns0:ref type='bibr' target='#b28'>(Hays et al. 1996;</ns0:ref><ns0:ref type='bibr' target='#b50'>Molodynski et al. 2020;</ns0:ref><ns0:ref type='bibr'>Schwenk et al. 2010)</ns0:ref>. Though medical students have better access to mental health care, they were less likely to seek mental health help compared to general population mainly due to stigma surrounding mental health disorders <ns0:ref type='bibr' target='#b17'>(Chew-Graham et al. 2003)</ns0:ref>. This may lead to adaptation of untoward and harmful coping methods like excess alcohol consumption and substance abuse <ns0:ref type='bibr' target='#b61'>(Rosenthal & Okie 2005)</ns0:ref>.</ns0:p><ns0:p>Swine flu (H1N1) outbreak in the year 2009 was the previous outbreak of an infectious disease in the pandemic scale which India was exposed to <ns0:ref type='bibr'>(WHO 2009)</ns0:ref>. In general, undergraduate medical students in India are usually in the age group of late teens to mid-twenties, and hence the current COVID-19 infection is the first exposure to them as adults on a pandemic scale. In addition, medical students are facing challenges such as sudden changes in their training routine including teaching and assessment via online sessions which invariably increases screen time, decreased patient contact and interactions with peers, which could possibly hinder their training and increased risk of contracting the infection mainly among the students in clinical postings, to name a few. All these could eventually have a toll on the mental and emotional well-being of the medical students as they are on an unknown territory.</ns0:p><ns0:p>Earlier studies show that the negative impact of epidemic/pandemic on the mental health are higher in healthcare workers <ns0:ref type='bibr' target='#b39'>(Lee et al. 2018;</ns0:ref><ns0:ref type='bibr' target='#b45'>Lu et al. 2006;</ns0:ref><ns0:ref type='bibr' target='#b48'>McAlonan et al. 2007)</ns0:ref>. Unfortunately, there are limited studies done on the impact of epidemic/pandemic on the mental health of medical students. Studies on the impact of COVID-19 pandemic on medical students are limited to cross sectional surveys assessing attitude, awareness, knowledge, precautionary measures, concerns, risk perceptions, impact on education and confidence and fear of COVID-19 <ns0:ref type='bibr' target='#b1'>(Agarwal et al. 2020;</ns0:ref><ns0:ref type='bibr' target='#b3'>Ahmed et al. 2020b;</ns0:ref><ns0:ref type='bibr' target='#b18'>Choi et al. 2020;</ns0:ref><ns0:ref type='bibr' target='#b35'>Khasawneh et al. 2020;</ns0:ref><ns0:ref type='bibr' target='#b54'>Nguyen et al. 2020;</ns0:ref><ns0:ref type='bibr' target='#b69'>Taghrir et al. 2020)</ns0:ref>. Literature search showed only a single study about the psychological impact of COVID-19 on medical students, which cross sectionally assessed their anxiety levels <ns0:ref type='bibr' target='#b13'>(Cao et al. 2020</ns0:ref>).</ns0:p><ns0:p>Globally, few longitudinal studies, compared mental health before and during COVID-19, which found an increase in anxiety and depression symptoms in college students in china <ns0:ref type='bibr' target='#b40'>(Li et al. 2020a)</ns0:ref> , United States of America <ns0:ref type='bibr' target='#b31'>(Huckins et al. 2020</ns0:ref>) and deterioration of mental health in general population in United Kingdom <ns0:ref type='bibr' target='#b57'>(Pierce et al. 2020</ns0:ref>). To our knowledge, there are no studies analyzing the impact of COVID-19 on mental health of undergraduate medical students in a prospective manner to assess cause and relationship. We hypothesize that COVID-19 outbreak and quarantine would have a serious negative impact on the mental health of undergraduate medical students who are already a vulnerable population. Hence, we conducted a prospective longitudinal study in a medical college in Chennai, Tamil Nadu, which is a center for treating COVID-19, to investigate the mental health of undergraduate medical students over a duration of 6 months by analyzing data collected before and during COVID-19 outbreak in India. We did extensive investigation of possible confounders and predictors of mental health disorders including demographics, sleep quality, apprehensions related to and caused by COVID-19 in terms of academics and concerns for the self, family, friends and interpersonal relationships.</ns0:p></ns0:div>
<ns0:div><ns0:head>Materials & Methods</ns0:head></ns0:div>
<ns0:div><ns0:head>Participants and setting</ns0:head><ns0:p>The study was originally planned to be a cross sectional survey for assessing the mental health of the undergraduate medical students in the institution. There was a total of 300 medical students in the institution enrolled for undergraduate medical degree. Sample size for the study was calculated using Epi Info™ 7 (Centers for Disease Control and Prevention (CDC), Atlanta, Georgia, USA). Confidence interval was set at 95%, margin of error at 5%, and the reference prevalence level of depression of 48.4% obtained from thorough literature survey, <ns0:ref type='bibr' target='#b37'>(Kumar et al. 2017)</ns0:ref> which resulted in the minimum sample size of 168. After applying a 10% non-response rate, the final required minimal sample size was calculated to be 185.</ns0:p></ns0:div>
<ns0:div><ns0:head>Baseline (Before COVID-19) survey</ns0:head><ns0:p>The Bachelor of Medicine & Bachelor of Surgery (M.B.B.S) is a 5.5 years undergraduate medical course offered in India in which the first 2.5 years concentrate mostly on basic medical sciences (pre and paraclinical subjects) and the next 2 years on clinical subjects followed by 1 year of Compulsory Rotatory Resident Internship (CRRI). All 300 students studying in preclinical (1st year), paraclinical (1.5 years after preclinical year) and clinical years (pre-final year, final year and resident interns) were included for the study and convenience sampling method was used. The students were explained about the objective of the study and were informed that the participation was voluntary, and confidentiality will be maintained. 276 students agreed to take part in the study from whom written consent was obtained before the beginning of the study.</ns0:p><ns0:p>Basic sociodemographic details such as age, gender, year of study, area of current residence and family gross monthly income were collected, and the mental health status was assessed using Depression Anxiety Stress Scale 21 items (DASS21). Students below 18 years of age and those with self-reported history of any pre-existing chronic medical conditions including mental health disorders were excluded (5 were underage and 2 reported history of bronchial asthma). The remaining 269 participants who were included for the study were contacted during their free time, after classes and were encouraged to answer the survey sincerely and doubts, if any, were clarified. Email id and mobile number were collected from all the participants. This part of the study was conducted during the first two weeks of December 2019, which was before COVID-19 outbreak in India.</ns0:p></ns0:div>
<ns0:div><ns0:head>Follow-up (During COVID-19) survey</ns0:head><ns0:p>With the unexpected changes to normalcy caused by the COVID-19 outbreak and subsequent lockdown, the authors decided to prospectively study the mental health status of the medical students in order to assess the effects of COVID-19 on mental health of the study population. After obtaining permission from Institutional Ethics Committee (IEC), the original data from the cross-sectional study was decided to be taken as 'before COVID-19 data' (baseline) and another survey was conducted on June 2020 (June 10th to 20th) to collect 'during .</ns0:p><ns0:p>The follow-up survey was conducted via Google form whose link was sent through personal email IDs of the students which were collected during the baseline survey. This protocol was exercised in order to follow strict social distancing protocol and to avoid direct contact. The follow-up survey included 5 sections; first section had a detailed description of the purpose of the study, along with the informed consent. This section explained the importance and benefits of the survey in the current pandemic, highlighting the voluntary nature of participation and assurance of confidentiality of the collected data. Only after consenting to the study, the participants could access the remaining sections. The successive sections collected responses for demographic details, self-administered questionnaire, DASS21 and Pittsburgh Sleep Quality Index (PSQI). Out of the 269 participants from the baseline survey, 30 randomly selected students were included in a pilot study (described below) and were hence excluded from the follow-up survey. Out of the remaining 239 participants, 222 students responded, out of which, 5 responses were excluded because of being incomplete (response rate 90.8%). The final sample size of this prospective cohort study was 217. A flowchart illustrating the sample selection from baseline to follow-up survey is shown in figure <ns0:ref type='figure'>1</ns0:ref>.</ns0:p></ns0:div>
<ns0:div><ns0:head>Survey instruments</ns0:head><ns0:p>To fulfil the objective of our longitudinal study, 'during COVID-19' data included a selfadministered questionnaire and PSQI to assess the sleep quality of the students in addition to DASS21.</ns0:p></ns0:div>
<ns0:div><ns0:head>Estimation of impact of COVID-19 related stressors</ns0:head><ns0:p>A self-administered questionnaire was prepared by the authors after extensive literature search, discussion with peers and local experts <ns0:ref type='bibr' target='#b75'>(Wang et al. 2020;</ns0:ref><ns0:ref type='bibr' target='#b79'>Wong et al. 2007)</ns0:ref>. It included a total of 12 close ended questions out of which, Items 1-3 focused on the subjects' status of COVID-19 testing (Yes/No) and their interactions with COVID-19 positive patients (Yes/No/I don't know). The remaining nine items were designed to assess self-perceived levels of concerns and worries for the self (4-6) and family/friends (7-8) and those related to academics (9-12), due to COVID-19 outbreak and quarantine (Supplemental file-Other). The responses were measured on a Likert scale of score 1 to 5, with 1 being the least and 5 being the maximum. This questionnaire was first tested empirically on 30 students (15 each from pre/para clinical years and clinical years) as a pilot study <ns0:ref type='bibr'>(Hill & century 1998)</ns0:ref>. The collected feedback and responses were analyzed, and corrections were made in the form of changes in articulation and simplification of vocabulary with the help of experts in this field.</ns0:p></ns0:div>
<ns0:div><ns0:head>Identification of latent variables from the self-administered questionnaire</ns0:head><ns0:p>PeerJ reviewing PDF | (2020:08:51660:1:1:NEW 3 Sep 2020)</ns0:p><ns0:p>Manuscript to be reviewed An exploratory factor analysis (EFA) was performed on the 9 items (item 4-12) to determine the validity of the questionnaire and to identify latent variables that could enable the objective of the study. The EFA was conducted using principal component analysis with varimax rotation for factor extraction.</ns0:p><ns0:p>The extracted factors were analyzed for retention using Kaiser criterion (Eigen value >1), scree test and counter-validated using parallel analysis. The Eigen values obtained from parallel analysis, which are values generated randomly with the same number of variables and sample size, were compared with the factor solution generated by EFA with our data. The number of factors were decided by Eigen values of factors that were higher than the values obtained from parallel analysis. The two-factor solution thus obtained had 5 items in one (items 4-8) and 4 items in another (items 9-12). The factors were named COVID-19 related general apprehensions (COVID-19-GA) and COVID-19 related academic apprehensions (COVID-19-AA) respectively (Table <ns0:ref type='table' target='#tab_1'>S1</ns0:ref>). Reliability analysis was performed for each of the factors separately which presented with a high reliability, with the Cronbach's alpha score of 0.89 for COVID-19-GA and 0.91 for COVID-19-AA. COVID-19-GA was scored by totaling the scores of the five individual items with the total score ranging from 5 to 25 and higher scores denote higher general apprehension. Similarly, COVID-19-AA was scored by summing up the scores of the four individual items with the scores ranging from 4 to 20 and higher scores denote higher academic related apprehension.</ns0:p></ns0:div>
<ns0:div><ns0:head>Estimation of mental health status</ns0:head><ns0:p>Mental health status of the medical students was assessed using Lovibond and Lovibond's Depression Anxiety Stress Scale 21 items (DASS21) <ns0:ref type='bibr' target='#b44'>(Lovibond et al. 1995)</ns0:ref>. This scale consists of a total of 21 items with seven each for depression, anxiety and stress subscales. The total sub scores range from 0 to 42 and is categorized into normal, mild, moderate, severe and extremely severe. In this study, DASS21 sub scores were categorized dichotomously, with the participants being divided in to those who showed symptoms of depression, anxiety and stress and those who did not, based on the cut-off sub-scores of 9, 7 and 14 respectively <ns0:ref type='bibr' target='#b16'>(Cheung et al. 2015;</ns0:ref><ns0:ref type='bibr' target='#b44'>Lovibond et al. 1995)</ns0:ref>.</ns0:p></ns0:div>
<ns0:div><ns0:head>Estimation of sleep quality</ns0:head><ns0:p>Subjective sleep quality was assessed using Pittsburgh Sleep Quality Index (PSQI) which includes 21 items that assess seven components viz. subjective sleep quality, sleep duration, sleep latency, habitual sleep efficiency, use of sleep medications, sleep disturbance, and daytime dysfunction over the duration of two weeks prior to assessment. Global PSQI scores are obtained by summing up the seven individual sub scores and it ranges from 1 to 21 with higher scores (>5) denoting poor sleep quality <ns0:ref type='bibr' target='#b12'>(Buysse et al. 1989;</ns0:ref><ns0:ref type='bibr' target='#b58'>Rahe et al. 2015)</ns0:ref>. <ns0:ref type='table' target='#tab_4'>2020:08:51660:1:1:NEW 3 Sep 2020)</ns0:ref> Previous studies have shown high reliability of both DASS21 and PSQI among Indian undergraduate medical student population <ns0:ref type='bibr' target='#b65'>(Shad et al. 2015;</ns0:ref><ns0:ref type='bibr'>Yadav et al. 2016</ns0:ref>). In our study, both the scales showed good internal consistency and DASS21 scale demonstrated good testretest reliability. Cronbach α score for reliability for PSQI was 0.72 and for DASS21 scale 0.94 (depression subscale 0.85, anxiety subscale 0.84, stress subscale 0.87) and 0.94 (depression subscale 0.87, anxiety subscale 0.81, stress subscale 0.85) for baseline and follow-up survey respectively.</ns0:p></ns0:div>
<ns0:div><ns0:head>PeerJ reviewing PDF | (</ns0:head></ns0:div>
<ns0:div><ns0:head>Ethical consideration</ns0:head><ns0:p>Ethical approval was obtained from the IEC, Madha Medical College and Research Institute in Chennai (MMCRI/IEC/H/018/2020) and research was done in accordance with the Helsinki Declaration for research on human participants.</ns0:p></ns0:div>
<ns0:div><ns0:head>Statistical analysis</ns0:head><ns0:p>Descriptive statistics was performed for all the variables. The scores of depression, anxiety, stress and sleep quality were expressed as mean ± standard deviation (SD). Initially, unadjusted univariate association between the demographic variables and depression, anxiety and stress were performed. Mann Whitney U test and Kruskal Wallis test for continuous variables and Chi square test for categorical variables were used for cross sectional analysis. Wilcoxon signed rank test for continuous variables and McNemar's test for categorical variables were used for longitudinal analysis. Spearman's correlation test was performed to assess the correlation between the scores obtained from the survey instruments in both the surveys.</ns0:p><ns0:p>To explore the contributory factors associated with depression, anxiety and stress during COVID-19 outbreak (dependent variable), adjusted binary logistic regression analysis was performed. Independent variables included were scores of PSQI, COVID-19-GA, COVID-19-AA, dependent variable from baseline survey (depression, anxiety and stress sub-scores in respective regression models) as covariates (continuous variables) and responses for the items 1-3 from the self-administered questionnaire as independent factors (categorical variables). Crosssectional association between sleep quality and study parameters were analyzed using adjusted binary logistic regression. The effect of each of the independent variable was adjusted for sociodemographic variables which were considered to be potential confounders viz. age, gender, year of study, current residence and family monthly income, in separate binary regression models. The results were expressed as adjusted odds ratio (aOR), 95% confidence interval (95% CI) and P value (statistical significance set at two-tailed P<0.05). </ns0:p></ns0:div>
<ns0:div><ns0:head>Results</ns0:head><ns0:p>This longitudinal study includes 217 undergraduate medical students (78 males and 139 females), the average age was 20±1.6 years. 5.1% got tested for COVID-19 and they all tested negative. 14.3% had friends and family who tested positive for COVID-19 and 12% declared to have direct contact with COVID-19 patients (figure <ns0:ref type='figure'>2</ns0:ref>). The distribution of responses to the items of COVID-19-GA and COVID-19-AA is shown in figure <ns0:ref type='figure'>2</ns0:ref> and percentage distribution of depression, anxiety and stress in baseline and follow-up survey and PSQI in follow-up survey is shown in figure <ns0:ref type='figure' target='#fig_7'>3</ns0:ref>.</ns0:p></ns0:div>
<ns0:div><ns0:head>Cross sectional relationship between sociodemographic variables and depression, anxiety and stress</ns0:head><ns0:p>The cross-sectional and longitudinal relationship between sociodemographic variables and depression, anxiety and stress are shown in table 1. There was no significant cross-sectional relationship between the demographic variables and both baseline and follow-up depression, anxiety and stress scores except in the baseline survey where depression levels were higher in the students from rural sector than urban sector (P=0.039). The association between demographic variables and depression, anxiety and stress analyzed by binary logistic regression shows that age was a protective factor for depression in the follow-up survey (OR 0.737, 95%CI 0.565-0.961) (Table <ns0:ref type='table' target='#tab_3'>S2</ns0:ref>). Other than this, there were no significant associations between demographic variables and mental health in both baseline and follow-up survey (Table <ns0:ref type='table' target='#tab_3'>S2</ns0:ref>-4).</ns0:p></ns0:div>
<ns0:div><ns0:head>Comparison of baseline and follow-up depression, anxiety and stress stratified by sociodemographic variables</ns0:head><ns0:p>The overall prevalence (with 95% confidence interval) of depression, anxiety and stress before COVID-19 was 33.2% (27-39.9%), 21.2% (16-27.2%) and 20.7% (15.5-26.7%) and during COVID-19 outbreak was 35.5% (29.1-42.2%), 33.2% (27-39.9%) and 24.9% (19.3-31.2%) respectively. There was a significant increase in the prevalence and mean scores of anxiety and stress when compared to baseline scores (P<0.000) (table 1). In terms of prevalence, when compared to baseline values, the prevalence of anxiety has significantly increased during Manuscript to be reviewed COVID-19 in females (P=0.003), preclinical, paraclinical (P=0.014) and clinical years students (P=0.019), students from urban residence (P=0.005) and those with gross family income per month above 100,000INR (P=0.007). In terms of mean DASS21 scores, there was a statistically significant increase in anxiety with mild effect size and stress levels with moderate to strong effect size in medical students in the follow-up survey compared to baseline levels irrespective of gender, year of study, current residence and gross family income per month (below 50,000 INR and above 100,000 INR) (P<0.05). Out of the entire study population, the levels of depression have increased significantly in male students and students from urban residence when compared to before COVID-19 levels albeit with mild effect size (table 1).</ns0:p></ns0:div>
<ns0:div><ns0:head>Difference in ranks of DASS21 scores between baseline and follow-up survey</ns0:head><ns0:p>The difference in ranks of DASS21 scores between baseline and follow-up survey is shown in table 2. There was no significant change in depression (P=0.146) in medical students between the two surveys however anxiety and stress has increased (P<0.001) showing an increase in median scores in the follow-up survey comparatively (table 1,2). The incidence of depression, anxiety and stress based on DASS21 sub scores were found to be 2.3 (5 out of 217), 11.98 (26 out of 217) and 4.15 (9 out of 217) per 100 per 6 months respectively. 97 (44.7%), 89 (41.01%) and 142 (65.44%) students scored higher in depression, anxiety and stress in the follow-up survey when compared to their responses before COVID-19 outbreak. While 60 (28.04%), 69 (31.79) and 7 (3.22%) students scored lesser in depression, anxiety and stress in follow-up survey (table <ns0:ref type='table' target='#tab_2'>2</ns0:ref>).</ns0:p></ns0:div>
<ns0:div><ns0:head>Correlation between the scores of the survey instruments from baseline and follow-up survey</ns0:head><ns0:p>The results of Spearman correlation analysis of the scores of the survey instruments from baseline and follow-up survey is shown in table 3. There were significant positive correlations between PSQI and baseline and follow-up depression, anxiety and stress. The correlation between the baseline and follow-up depression, anxiety and stress scores indicate that a higher baseline score was associated with higher follow-up score and vice versa (P<0.001). There was a relatively weak but significant positive correlation between depression levels before COVID-19 and COVID-19 related general apprehensions (r=0.152, P=0.025) (table 3). Manuscript to be reviewed</ns0:p></ns0:div>
<ns0:div><ns0:head>Adjusted binary logistic regression analysis of sleep quality</ns0:head><ns0:p>Cross-sectional association between sleep quality and mental health in the follow-up survey is shown in figure <ns0:ref type='figure' target='#fig_8'>4</ns0:ref>. Students with higher depression, anxiety and stress scores during COVID-19 outbreak were found to be more likely to have poor sleep quality (P<0.001).</ns0:p></ns0:div>
<ns0:div><ns0:head>Adjusted binary logistic regression analysis of follow-up depression, anxiety and stress</ns0:head><ns0:p>The results of binary logistic regression analysis for follow-up depression, anxiety and stress are shown in figure <ns0:ref type='figure' target='#fig_9'>5-7</ns0:ref>. Poor sleep quality was found to be significantly associated with increase in depression, anxiety and stress (P<0.001). Higher baseline scores of depression, anxiety and stress were associated with higher levels of the same in follow-up survey (P<0.001). Higher COVID-19 related general apprehension was associated with higher levels of anxiety (P=0.016) and stress (P=0.005). Students who did not have any direct interactions with COVID-19 positive patients were found to be less likely to have symptoms of depression (P=0.017) and stress (P=0.004) when compared to those who did. Similarly, absence of COVID-19 positive cases in family and friends was found to be associated with decreased levels of stress (P=0.004).</ns0:p></ns0:div>
<ns0:div><ns0:head>Discussion</ns0:head><ns0:p>The present study investigates the mental health status of undergraduate medical students during COVID-19 outbreak in a medical college which is a government approved center for treating COVID-19. The medical college is located in Chennai, TamilNadu which is one of the top 5 worst COVID-19 hit metropolitan city in India. Longitudinal data analysis was used to test our hypothesis that COVID-19 outbreak and quarantine has a negative impact on the mental health of undergraduate medical students.</ns0:p><ns0:p>We found that 35.5% (95% CI 29.1-42.2%), 33.2% (95% CI 27-39.9%) and 24.9% (95% CI 19.3-31.2%) of the undergraduate medical students including resident interns showed symptoms of depression, anxiety and stress respectively during COVID-19 outbreak with the majority with moderate depression (15.2%) and anxiety (17.5%) and mild stress (13.4%). Based on the severity ranking, subjects with moderate and above ranking may present with a possible problem that may require intervention <ns0:ref type='bibr' target='#b21'>(Crawford & Henry 2003;</ns0:ref><ns0:ref type='bibr' target='#b44'>Lovibond et al. 1995;</ns0:ref><ns0:ref type='bibr' target='#b56'>Page et al. 2007;</ns0:ref><ns0:ref type='bibr' target='#b68'>Stormon et al. 2019)</ns0:ref>. In this study, 23.9%, 26.7% and 11.5% of the study population presented with moderate to extremely severe levels of depression, anxiety and stress respectively (figure <ns0:ref type='figure' target='#fig_7'>3</ns0:ref>). In our study, the six-month incidence of anxiety was found to be comparatively higher (11.98%) followed by stress (4.15%) and depression (2.3%). There were no other longitudinal studies conducted in medical students during COVID-19 assessing the incidence of mental health PeerJ reviewing PDF | (2020:08:51660:1:1:NEW 3 Sep 2020) disorders, however, a study conducted in general public in China found the one-month incidence of mental health disorders to be relatively low <ns0:ref type='bibr'>(Yali et al. 2020)</ns0:ref>. Though the incidence rate seemed to be relatively low in our study, compared to the baseline levels, 44.7%, 41.01% and 65.44% of the study population scored higher in depression, anxiety and stress sub scores during COVID-19 (table 2). When compared to baseline survey that was recorded before COVID-19 outbreak in India, there was a significant increase in prevalence and levels of anxiety and stress while depression levels remained unchanged during COVID-19. Similar studies conducted longitudinally in college students found significant increase in depression and anxiety during COVID-19 when compared to before COVID-19 levels. <ns0:ref type='bibr' target='#b40'>(Li et al. 2020a</ns0:ref><ns0:ref type='bibr' target='#b31'>, Huckins et al. 2020)</ns0:ref>. This negative impact of the pandemic could be attributed to sudden challenges faced by the medical students in terms of academics, uncertainties about future, fear of infection, news about shortage of personal protective equipment, quarantine induced boredom, frustrations, lack of freedom and fears caused by rumors and erroneous news in the media <ns0:ref type='bibr' target='#b7'>(Bao et al. 2020;</ns0:ref><ns0:ref type='bibr' target='#b23'>Ferrel & Ryan 2020)</ns0:ref>.</ns0:p><ns0:p>Investigation of the influence of demographics on mental health showed that, the increase in anxiety and stress levels in our study population was not affected by gender, year of study, current residence and family income per month. We also found no significant difference crosssectionally in depression, anxiety and stress between the groups of demographic variables during COVID-19. Binary logistic regression analysis of anxiety and stress and demographic variables as independent variables showed no significant association. In contrary, cross-sectional studies done on medical students in China and Brazil during this pandemic found significant associations between mental health disorders and place of residence, parental income/financial support and gender <ns0:ref type='bibr' target='#b13'>(Cao et al. 2020;</ns0:ref><ns0:ref type='bibr' target='#b63'>Sartorao Filho et al. 2020)</ns0:ref>. Our results show that the response of the medical students to COVID-19 epidemic in terms of the levels of anxiety and stress is similar, irrespective of gender, year of study, current residence and financial status of the family. Depression in the study population, on the other hand remained unchanged during COVID-19 in all the categories except in male gender and urban population. Binary logistic regression showed that increase in age could decrease the likelihood of depression (OR 0.737, 95%CI 0.565-0.961) which is consistent to previous studies during COVID-19 <ns0:ref type='bibr' target='#b2'>(Ahmed et al. 2020a;</ns0:ref><ns0:ref type='bibr' target='#b27'>González-Sanguino et al. 2020)</ns0:ref>. However, in our study, since the range of age is narrow (18-26 years) which mostly corresponds to the year of study, which had no significant association with depression, this observation becomes redundant. We found significant increase in depression levels in male students but not in female students. This is contradictory to the studies in medical students during COVID-19 which found female gender to be more at risk of developing depression symptoms <ns0:ref type='bibr' target='#b63'>(Sartorao Filho et al. 2020)</ns0:ref>. Females are more proactive in their response and awareness about the epidemic when compared to males and our results could be a possible implication of this (Brittni Frederiksen 2020). Despite higher levels of depression in the students from rural areas before COVID-19, their depression levels remained unchanged in the follow-up survey, while the students from the urban areas presented with an increase in depression when compared to before COVID-19 levels. Nearly 53% of India's cases are recorded in Mumbai, Delhi, Ahmedabad, Pune and Chennai alone, which are dubbed as top five COVID-19 cities <ns0:ref type='bibr' target='#b66'>(Shylendra 2020)</ns0:ref>. Urban areas are highly populated, and the epidemic is more active in urban areas than in rural. This led to the implementation of frequent lockdown measures by the respective State Governments in the COVID-19 hotspots which are mainly urban centers, which could have possibly increased the psychological distress in the students from urban locations.</ns0:p><ns0:p>Since we did not find significant differences in mental health among most of the groups of demographic variables in both baseline and follow-up survey, it is likely that the worsening of mental health status of the medical students found in our study is associated with COVID-19 related factors. To further elucidate this, we used adjusted binary logistic regression analysis to explore possible predictors. Our extensive literature survey showed all the collected demographic variables to have potential influence on the outcome. Hence, despite not finding significant associations between demographics and mental health, we adjusted the effects of each potential predictor for all the recorded demographic variables in the regression models <ns0:ref type='bibr' target='#b13'>(Cao et al. 2020;</ns0:ref><ns0:ref type='bibr' target='#b63'>Sartorao Filho et al. 2020;</ns0:ref><ns0:ref type='bibr' target='#b29'>Heinze & Dunkler 2017)</ns0:ref> An important finding in our study is the independent bidirectional association between poor sleep quality and mental health. In our study, 34.6% of the study population suffered from poor sleep quality which was found to be a significant independent predictor of depression (aOR 1.337, 95%CI 1.19-1.502), anxiety (aOR 1.227, 95%CI 1.106-1.363) and stress (aOR 1.371, 95%CI 1.209-1.555) during COVID-19 (figure <ns0:ref type='figure' target='#fig_9'>5-7</ns0:ref>), similar to a previous study by <ns0:ref type='bibr'>Cellini N et al., in Italy (Cellini et al. 2020)</ns0:ref>. Medical students are especially prone to have poor sleep quality because of the physically and emotionally challenging and intense training they undergo <ns0:ref type='bibr' target='#b78'>(Wong et al. 2005)</ns0:ref>. Poor sleep affects neurocognitive and psychomotor performance, emotional wellbeing, working capacity, academic performance, physical and mental health as well as quality of life <ns0:ref type='bibr' target='#b4'>(Al-Khani et al. 2019;</ns0:ref><ns0:ref type='bibr' target='#b24'>Flores 2009;</ns0:ref><ns0:ref type='bibr' target='#b26'>Giri et al. 2013;</ns0:ref><ns0:ref type='bibr' target='#b52'>Mume et al. 2011)</ns0:ref>. Due to lockdown measures and travel restrictions, students are facing decreased physical activity, lack of schedule, altered living conditions, increased screen time and time spent in social media, and altered sleep wake schedule including increased daytime nap duration <ns0:ref type='bibr' target='#b46'>(Majumdar et al. 2020)</ns0:ref>. All these factors in addition to higher demands of medical curriculum could lead to poor sleep which in turn affects mental wellbeing. Conversely, we also found that higher depression (aOR 1.114, 95%CI 1.07-1.159), anxiety (aOR 1.120, 95%CI 1.065-1.178) and stress levels (aOR 1.126, 95%CI 1.078-1.176) during COVID-19 were significant predictors of poor sleep quality (figure <ns0:ref type='figure' target='#fig_8'>4</ns0:ref>). Poor sleep is long since been considered an important symptom of mental health disorders. Sleep disturbance is a primary symptom of major depressive disorder <ns0:ref type='bibr' target='#b32'>(Jindal & Thase 2004)</ns0:ref>. Anxiety and stress negatively affect the body's ability to fall and stay asleep <ns0:ref type='bibr' target='#b20'>(Coplan et al. 2015;</ns0:ref><ns0:ref type='bibr' target='#b34'>Kalmbach et al. 2018</ns0:ref>). In addition, increased time spent in social media and digital devices is a way by which young adults cope with social isolation, and it is associated with increased tendencies to develop sleep disturbances <ns0:ref type='bibr' target='#b67'>(Sivertsen et al. 2019)</ns0:ref>. Thus, our findings show that poor sleep quality is both a cause and an effect of increased depression, anxiety and stress symptoms in medical students during this pandemic. Thus, worsening of one could exacerbate the other.</ns0:p><ns0:p>Students with higher baseline levels of depression, anxiety and stress were found to be more likely to have depression (aOR 1.27, 95%CI 1.185-1.361), anxiety (aOR 1.176, 95%CI 1.099-1.259) and stress (aOR 1.810, 95%CI 1.483-2.208) during COVID-19 outbreak (figure <ns0:ref type='figure' target='#fig_9'>5-7</ns0:ref>). There was also significant positive correlation between baseline and follow-up depression, anxiety and stress scores (table <ns0:ref type='table' target='#tab_4'>3</ns0:ref>). These results are consistent with a similar study conducted on college students in China <ns0:ref type='bibr' target='#b40'>(Li et al. 2020a)</ns0:ref>. Studies show associations between pre-existing mental health problems and mental health disorders in medical students <ns0:ref type='bibr'>(Yates et al. 2008)</ns0:ref>. Previous studies say that COVID-19 has a higher negative impact on people with mental health disorders when compared to those without any <ns0:ref type='bibr' target='#b6'>(Asmundson et al. 2020;</ns0:ref><ns0:ref type='bibr' target='#b72'>Taylor et al. 2020)</ns0:ref>. Medical students are proved to be at a higher risk of developing mental health disorders during the course of their training period <ns0:ref type='bibr' target='#b49'>(Moffat et al. 2004</ns0:ref>). With the added stressors related to the pandemic, it might be difficult to cope, especially for those students who had higher levels of depression, anxiety and stress to begin with, leading to exacerbation of symptoms.</ns0:p><ns0:p>A surprising finding in our study was that COVID-19 related academic apprehensions were not significantly associated with depression, anxiety and stress. This is in contrary to a recent study conducted in medical students in China, which found moderate positive correlation between worries about academic delay and anxiety <ns0:ref type='bibr' target='#b13'>(Cao et al. 2020)</ns0:ref>. Though social desirability response bias could be an attributing factor, our finding could be a reflection of the feel of assurance by the students due to drastic student-centered efforts taken by the medical college and universities in continuing medical education without disruption during this pandemic, which include classes and examinations conducted via online forums and constantly updating the students about any changes made and expected. Another reason could be that, since these changes that were instigated by the lockdown is in effect for around three months, the students would have adapted to this new normalcy in their training better than anticipated. On the other hand, higher COVID-19-GA scores were found to be a significant predictor for higher levels of anxiety (aOR 1.097, 95%CI 1.018-1.182) and stress (aOR 1.128, 95%CI 1.037-1.227) (figure <ns0:ref type='figure' target='#fig_10'>6, 7</ns0:ref>). In a study conducted on medical students in Pakistan, around 76% of the participants conveyed being worried about contracting COVID-19 during clinical postings and even more worried about insufficient care and improper treatment, if they contracted the infection <ns0:ref type='bibr' target='#b3'>(Ahmed et al. 2020b</ns0:ref>).</ns0:p><ns0:p>In a recent study conducted on adult Indian population, individuals with increased self-perceived risk of contracting COVID-19 were found to be more likely to have mental health disorders <ns0:ref type='bibr' target='#b62'>(Saikarthik et al. 2020)</ns0:ref>. A longitudinal study conducted in China in college students, found fear of infection to be significantly associated with anxiety and depression <ns0:ref type='bibr' target='#b40'>(Li et al. 2020a)</ns0:ref>. Our results show that medical students' fear of contracting and surviving if contracted with COVID-PeerJ reviewing PDF | (2020:08:51660:1:1:NEW 3 Sep 2020) 19 for the self, family and friends and worries about COVID-19 affecting interpersonal relationships have a negative impact on their mental health. Student population is highly active in social media which is filled with high amounts of misinformation which instils fear and affects mental well-being. Frequent use of social media was associated with higher prevalence of mental health problems during COVID-19 <ns0:ref type='bibr' target='#b25'>(Gao et al. 2020)</ns0:ref>. At the same time, social media could be used for communications among peers and family thereby offering much needed social support.</ns0:p><ns0:p>In this study, we found no significant association between being tested for COVID-19 and depression, anxiety and stress. The possible reason could be that all of those who got tested for COVID-19 were found negative. We also found that students without any COVID-19 positive patients in family and friends (aOR 0.259, 95%CI 0.069-0.968) and those who were not sure about having any direct interactions with COVID-19 positive patients (aOR 0.280 95%CI 0.092-0.856) were found to be less likely to have symptoms of stress (figure <ns0:ref type='figure'>7</ns0:ref>) and those without any direct interactions with COVID-19 positive patients were less likely to have symptoms of depression (aOR 0.31 95%CI 0.106-0.905) (figure <ns0:ref type='figure' target='#fig_9'>5</ns0:ref>) when compared to those who did, which was similar to the study by <ns0:ref type='bibr' target='#b13'>Cao et al. (Cao et al. 2020)</ns0:ref>. The R naught (R0) of SARS-CoV-2 is 2.2 (2-2.5), i.e each infected person spreads the infection to 2.2 individuals, in other words it is more contagious than seasonal flu <ns0:ref type='bibr' target='#b42'>(Li et al. 2020b)</ns0:ref>. This high contagious nature of the novel SARS-CoV-2 virus could be related to our findings.</ns0:p></ns0:div>
<ns0:div><ns0:head>Strengths and Limitations of the study</ns0:head><ns0:p>The longitudinal nature of this study outweighs the limitation of relatively small sample size and is its important strength, as the mental health status of the same medical students was investigated before and during COVID-19 enabling us to study the pattern, temporal order and predictors of changes in their mental status. To our knowledge, ours is the first study to analyze the effects of COVID-19 outbreak on the mental health of undergraduate medical students longitudinally. We elaborately studied the influence of demographic variables, stressors related to COVID-19 by categorizing them into general and academic apprehensions as well as sleep quality on mental health. With the known impact of negative mental health on medical students' career and life, we believe the results of our study is important due to the insights provided, that will help the medical educationists to address and devise strategies to overcome the COVID-19 induced negative impact on undergraduate medical students' mental health.</ns0:p><ns0:p>Our study is not without limitations. Though our study population includes students who were tested for possible COVID-19 infection, none turned out positive. Thus, our results cannot be extrapolated to medical students infected with COVID-19. Despite promising confidentiality, there could have been possible response bias by the students in answering the survey.</ns0:p></ns0:div>
<ns0:div><ns0:head>Conclusions</ns0:head><ns0:p>PeerJ reviewing PDF | (2020:08:51660:1:1:NEW 3 Sep 2020)</ns0:p><ns0:p>Manuscript to be reviewed COVID-19 has a negative impact on the mental health of undergraduate medical students. Prevalence and levels of anxiety and stress has increased due to COVID-19 outbreak and quarantine while the depression symptoms were unaltered. Poor sleep quality, higher levels of depression, anxiety and stress before COVID-19, increased worries about contracting and surviving, if contracted with COVID-19 for the self, family, and friends, and about COVID-19 affecting interpersonal relationships, presence of COVID-19 positive patients in family and friends and direct interactions with COVID-19 patients were found to be significant predictors of negative mental health in medical students. Neglecting the mental health of the medical students would lead to long term detrimental effects which not only will affect the quality of life of medical students and future physicians but also the overall performance of the healthcare system. An effective plan to safeguard the mental health of this already vulnerable population of undergraduate medical students is crucial. We strongly believe our findings would help the medical educationists in addressing and mitigating the rise in mental health disorders which could prove worse than the current pandemic itself. Further studies to analyze the temporal pattern of changes in mental health status of the medical students is warranted. Manuscript to be reviewed Difference in the ranks of DASS21 scores between baseline and follow-up surveys Manuscript to be reviewed Manuscript to be reviewed Correlation between scores of survey instruments from baseline and follow-up survey</ns0:p><ns0:p>The Manuscript to be reviewed Manuscript to be reviewed Manuscript to be reviewed Manuscript to be reviewed</ns0:p><ns0:note type='other'>Figure 7</ns0:note><ns0:p>Forest plot showing adjusted binary logistic regression analysis of follow-up stress scores aOR adjusted odds ratio; odds ratio adjusted for age, gender, year of study, urban/rural residential status, family's monthly financial status; 95%CI 95% confidence interval;</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:08:51660:1:1:NEW 3 Sep 2020)</ns0:p></ns0:div><ns0:figure xml:id='fig_0'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:08:51660:1:1:NEW 3 Sep 2020)Manuscript to be reviewed All the statistical analysis was performed using SPSS version 26 (IBM, New York, USA) and Parallel analysis was performed using scripts from O'Connor BP<ns0:ref type='bibr' target='#b55'>(O'Connor 2000)</ns0:ref> </ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_1'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:08:51660:1:1:NEW 3 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_2'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:08:51660:1:1:NEW 3 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_3'><ns0:head /><ns0:label /><ns0:figDesc>Yadav R, Gupta S, and Malhotra AK. 2016. A cross sectional study on depression, anxiety and their associated factors among medical students in Jhansi, Uttar Pradesh, India. Int J Community Med Public Health. 3(5):1209-1214. Yates J, James D, and Aston I. 2008. Pre-existing mental health problems in medical students: a retrospective survey. Medical Teacher. 30(3):319-321. 10.1080/01421590701797630</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_5'><ns0:head /><ns0:label /><ns0:figDesc>a follow-up < baseline, b follow-up > baseline, c follow-up = baseline PeerJ reviewing PDF | (2020:08:51660:1:1:NEW 3 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_6'><ns0:head /><ns0:label /><ns0:figDesc>Figure 1</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_7'><ns0:head>Figure 3</ns0:head><ns0:label>3</ns0:label><ns0:figDesc>Figure 3</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_8'><ns0:head>Figure 4 Forest</ns0:head><ns0:label>4</ns0:label><ns0:figDesc>Figure 4</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_9'><ns0:head>Figure 5 Forest</ns0:head><ns0:label>5</ns0:label><ns0:figDesc>Figure 5</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_10'><ns0:head>Figure 6 Forest</ns0:head><ns0:label>6</ns0:label><ns0:figDesc>Figure 6</ns0:figDesc></ns0:figure>
<ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='35,42.52,229.87,525.00,331.50' type='bitmap' /></ns0:figure>
<ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='40,42.52,250.12,525.00,217.50' type='bitmap' /></ns0:figure>
<ns0:figure type='table' xml:id='tab_0'><ns0:head>Table 1 (on next page)</ns0:head><ns0:label>1</ns0:label><ns0:figDesc>Relationship between demographic variables and baseline and follow-up depression,</ns0:figDesc><ns0:table><ns0:row><ns0:cell>anxiety and stress scores</ns0:cell><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>a square test, P value: Wilcoxon signed rank test, e P value: Kruskal Wallis test, r value: effect size of Wilcoxon signed rank test b P value: McNemar test, c P value: Mann Whitney U test,</ns0:cell><ns0:cell>d</ns0:cell><ns0:cell>P value: Chi</ns0:cell></ns0:row><ns0:row><ns0:cell>N (%): Number of subjects with depression, anxiety and stress</ns0:cell><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>Subjects with depression: depression sub-score > 9,</ns0:cell><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>Subjects with anxiety: anxiety sub-score > 7,</ns0:cell><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>Subjects with stress: stress sub-score > 14</ns0:cell><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>Significant P value in bold letters</ns0:cell><ns0:cell /><ns0:cell /></ns0:row></ns0:table></ns0:figure>
<ns0:figure type='table' xml:id='tab_1'><ns0:head>Table 1 :</ns0:head><ns0:label>1</ns0:label><ns0:figDesc>Relationship between demographic variables and baseline and follow-up depression, anxiety and stress 2 scores</ns0:figDesc><ns0:table><ns0:row><ns0:cell>Depression</ns0:cell><ns0:cell>Anxiety</ns0:cell></ns0:row></ns0:table><ns0:note>PeerJ reviewing PDF | (2020:08:51660:1:1:NEW 3 Sep 2020) Manuscript to be reviewed PeerJ reviewing PDF | (2020:08:51660:1:1:NEW 3 Sep 2020) Manuscript to be reviewed PeerJ reviewing PDF | (2020:08:51660:1:1:NEW 3 Sep 2020)</ns0:note></ns0:figure>
<ns0:figure type='table' xml:id='tab_2'><ns0:head>Table 2 (on next page)</ns0:head><ns0:label>2</ns0:label><ns0:figDesc /><ns0:table /></ns0:figure>
<ns0:figure type='table' xml:id='tab_3'><ns0:head>Table 2 :</ns0:head><ns0:label>2</ns0:label><ns0:figDesc>Difference in the ranks of DASS21 scores between baseline and follow-up surveys</ns0:figDesc><ns0:table><ns0:row><ns0:cell>Median score (50 th</ns0:cell></ns0:row><ns0:row><ns0:cell>percentile)</ns0:cell></ns0:row></ns0:table></ns0:figure>
<ns0:figure type='table' xml:id='tab_4'><ns0:head>Table 3 (on next page)</ns0:head><ns0:label>3</ns0:label><ns0:figDesc /><ns0:table /></ns0:figure>
<ns0:note place='foot'>PeerJ reviewing PDF | (2020:08:51660:1:1:NEW 3 Sep 2020)</ns0:note>
<ns0:note place='foot' n='2'>The results are expressed as ρ (Rho) value; (* P<0.05); (** P<0.001) PeerJ reviewing PDF | (2020:08:51660:1:1:NEW 3 Sep 2020)</ns0:note>
</ns0:body>
" | " Reviewer 1
Basic reporting
• The numbering unit of ‘lakh’ is not an internationally well-understood
concept, could the authors please consider revising this in Introduction.
The unit “lakh” has been changed to ‘hundred thousand” as per suggestion (Line no.
89, 92).
• It’s not very clear if there have been other longitudinal studies done before,
on the effect of COVID on depression, anxiety, and stress in other
populations/contexts. Please add something about this in the Introduction.
Agreed. Relevant points have been added in the introduction (Line no. 148).
‘Globally, few longitudinal studies compared mental health before and during
COVID-19, which found an increase in anxiety and depression symptoms in college
students in china (Li et al, 2020), US (Huckins et al., 2020) and deterioration of
mental health in general population in UK (Mathias Pierce et al., 2020)’
• Figures 1 and 2 appear to be in low-resolution.
A higher resolution of the figures 1 and 2 have been uploaded.
Results
• Please consider adding effect sizes where-ever possible, especially to the
changes in depression, anxiety and stress from baseline to follow-up survey,
regardless of whether the difference was p < .05 or not.
Effect sizes have been included in table 1 and in appropriate places in the results as
per suggestions
• Line 355 – I believe ‘a relatively weak but significant positive correlation’ is
more accurate than ‘a mild significant positive correlation’.
Change made as per suggestion (Line no 376).
• For Discussion, please talk about, whether there has been other longitudinal
studies conducted on the change of depression, anxiety, and stress from
before to during COVID, and if the magnitude of change found in the current
study is similar/different to that found for other populations in the other
studies.
Relevant points have been added as per suggestion (Line no. 420).
‘Similar studies conducted longitudinally in China and US found significant increase
in depression and anxiety during COVID-19 when compared to before COVID-19
levels in college students (Li et al, 2020), (Huckins et al., 2020).’
Comments for authors
Some typos noted:
• Line 221 – literature search (Line no. 228)
• Line 349 – the full stop should be placed after (table 2). (Line no. 366)
• Line 450 – missing “and” in sentence: challenging and intense training (Line
no. 478)
Corrected.
Reviewer 2
Experimental design
• Several details should be added, including the study site, sampling method at
baseline, and the dependent and independent variables in the adjusted binary
logistic regression. The limitations of small sample size should be addressed.
Study site – The study was conducted in Madha Medical College and Research
Institute, Chennai, India which is mentioned in line no. 290, under ethical
consideration and line no. 398, under discussion.
Sampling method – the sampling method used was convenience sampling (added
in line no.180)
“All 300 students studying in preclinical (1st year), paraclinical (1.5 years after
preclinical year) and clinical years (pre-final year, final year and resident interns)
were included for the study and convenience sampling method was employed.”
Dependent and independent variables in the adjusted binary logistic
regression
Depression, anxiety and stress during COVID-19 outbreak were the dependent
variables and the independent variables were scores of PSQI, COVID-19-General
Apprehension, COVID-19-Academic Apprehension, dependent variable from
baseline survey (depression, anxiety and stress sub-scores in respective regression
models) as covariates (continuous variables) and responses for the items 1-3 from
the self-administered questionnaire as independent factors (categorical variables).
For the regression analysis of sleep quality, PSQI score in follow-up survey was the
dependent variable and the independent variables were scores of COVID-19General Apprehension, COVID-19-Academic Apprehension, follow-up depression,
anxiety and stress sub-scores as covariates (continuous variables) and responses
for the items 1-3 from the self-administered questionnaire as independent factors
(categorical variables).
The effect of each of the independent variable was adjusted for sociodemographic
variables which were potential confounders viz. age, gender, year of study, current
residence and family monthly income, in separate binary regression models.
(Mentioned in line no 303-308, under statistical analysis)
The limitations of small sample size should be addressed.
Agreed. Relevant points added in line no. 555, under strengths and limitation of the
study.
The longitudinal nature of this study outweighs the limitation of relatively small
sample size and is its important strength, as the mental health status of the same
medical students was investigated before and during COVID-19 enabling us to study
the pattern, temporal order and predictors of changes in their mental status.
Validity of findings
• The results should be separated into sub-sectionals for better understanding.
Agreed. Sub-sections have been incorporated in the results section.
• The titles of all tables (including the supplementary tables) should be added.
• The statistic values should be added in both the results section and all
tables.
• The P=0.000 should be written as P<0.001
Changes incorporated in appropriate places
Reviewer 3
Comments for the authors
Abstract
1. Authors should report 95%CI of prevalence.
95% confidence interval of prevalence has been included
2. DASS 21 should report the full name when authors first mentioned it.
Corrected as per suggestions
3. Authors should report specific statistic values about predictors (e.g., OR,
95%CI.)
Corrected as per suggestions
Method
1. Authors should draw a flowchart to illustrate the sample selection from
baseline to follow-up.
As suggested, a flowchart has been included in the methodology as figure 1
2. For a prospective longitudinal study, why authors do not examine the
incidence of depression, anxiety and distress.
Agreed. Relevant points added in results and discussion
The incidence of depression, anxiety and stress based on DASS21 sub scores were
found to be 2.3 (5 out of 217), 11.98 (26 out of 217) and 4.15 (9 out of 217) per 100
per 6 months respectively (added in line no 361, 411).
3. For a prospective longitudinal study, why authors do not examine the RRs?
The objective of our study was to test the hypothesis that COVID-19 outbreak in
India has a negative impact on the mental health of undergraduate medical students
along with identifying the predictors of mental health disorders, while adjusting for
the effect of possible confounders viz. sociodemographic variables. After discussion
with the statistician and thorough literature search, the appropriate statistical tool for
our study was decided to be adjusted binary logistic regression analysis which
provided information regarding the predictors of mental health disorders while
adjusting for the effect of confounders. Since the test for our hypothesis was
answered by logistic regression analysis, odds ratio was preferred over risk ratio.
Reference:
Yu, H.Y.R., Ho, S.C., So, K.F.E. and Lo, Y.L. (2005), The psychological burden experienced
by Hong Kong midlife women during the SARS epidemic. Stress and Health, 21: 177-184.
doi:10.1002/smi.1051
Li HY, Cao H, Leung DY, Mak YWJIJoER, and Health P. 2020a. The Psychological Impacts
of a COVID-19 Outbreak on College Students in China: A Longitudinal Study. 17:3933.
4. What is the “Parallel analysis”? I did not found any results from this
analysis.
Parallel analysis was conducted to counter validate the factor solution obtained from
principal component analysis (PCA). Parallel analysis creates a random dataset with
the same numbers of observations and variables as the original data and computes
Eigen values of the correlation matrix. The Eigen values of the factors obtained from
PCA was compared with the Eigen values of the random data and only those factors
whose Eigen values were lesser than that obtained from parallel analysis were
retained.
Mentioned in line no. 248.
Results
1. Authors should add some subtitles in order to well understanding for
readers.
Agreed. Sub-sections have been incorporated in the results section.
2. Authors should re-design the table 1. A horizontal table is easy to
understand. Please turn the first column into the first row.
Agreed. Table 1 has been realigned as per instructions
3. All supplementary tables did not present the titles. Please add them.
Changes incorporated in appropriate places as per suggestion
4. In all tables. Please correct “p=0.000” to be “p<0.001”.
Changes incorporated in appropriate places as per suggestion
Discussion
1. As mentioned by authors in the limitation- “Majority of the study population
(69.1%) were in the preclinical and paraclinical years which could have
influenced the interpretation of results.” Why authors did not control this
variable in the association analysis.
Because of the unequal distribution of students based on the year of study and other
sociodemographic variables and since they were considered to be possible
confounders for the outcome variable, the effects of the sociodemographic variables
were controlled for in each regression model analysing the association between the
predictors and dependent variable. Hence “Majority of the study population (69.1%)
were in the preclinical and paraclinical years which could have influenced the
interpretation of results” is removed from the limitations section.
" | Here is a paper. Please give your review comments after reading it. |
9,812 | "<ns0:abstract xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:p>Background. COVID-19 pandemic is found to affect the mental health of the population. Undergraduate medical students are especially prone to mental health disorders and hence could be more vulnerable to the impact of the pandemic.</ns0:p></ns0:div>
<ns0:div><ns0:head>Methods.</ns0:head><ns0:p>A prospective longitudinal study was conducted on 217 undergraduate medical students in a medical college at Chennai, India. Depression, anxiety, and stress levels were recorded using Depression Anxiety Stress Scale 21 Items (DASS21) before and during COVID-19 outbreak in India in December 2019 and June 2020 respectively. In the follow-up survey, in addition to DASS21, Pittsburgh Sleep Quality Index to assess sleep quality and a self-administered questionnaire to assess the impact of COVID-19 related stressors were used. The self administered questionnaire assessed the status of COVID-19 testing, interactions with COVID-19 patients, self-perceived levels of concerns and worries related to academics [COVID-19-AA (academic apprehensions)] and those pertaining to the self and family/friends [COVID-19-GA (general apprehensions)]. Cross-sectional and longitudinal comparison of overall scores of depression, anxiety, and stress and scores stratified by gender, year of study, place of residence and monthly family income were performed. Predictors for depression, anxiety, and stress during COVID-19 were investigated using adjusted binary logistic regression analysis and results were expressed as adjusted odds ratio with 95% confidence interval (CI). A P value <0.05 was considered statistically significant.</ns0:p></ns0:div>
</ns0:abstract>
<ns0:body xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:head>Introduction</ns0:head><ns0:p>The World Health Organization (WHO) announced COVID-19 outbreak initially, as a public health emergency of international concern (PHEIC) on January 30, 2020 and later declared as a pandemic on <ns0:ref type='bibr'>March 11, 2020</ns0:ref><ns0:ref type='bibr'>(WHO 2020)</ns0:ref>. In India, the first COVID-19 case was reported in Kerala on January 30, 2020, and by May 19, the number of cases had crossed one hundred thousand. By September 7, India became the world's second worst hit nation with 4.2 million confirmed COVID-19 cases following United States of America and has recorded 71,642 deaths <ns0:ref type='bibr'>(Times, 2020)</ns0:ref>. Within India, the state of Maharashtra was the worst hit state followed by Andhra Pradesh and Tamil Nadu, contributing to 21.6%, 11.8% and 11% of the total cases respectively <ns0:ref type='bibr' target='#b51'>(Newsdesk, 2020)</ns0:ref>. The Government of India declared a nationwide lockdown on 25th March 2020, as a measure to mitigate the spread of infection. However, prolonged lockdown is not only unfavorable to the individuals, it also significantly affects the nation's economy. As a way to revive and restore the affected economy, a phase-wise upliftment of lockdown was announced from June 1, easing of some restrictions, while the lockdown was maintained for the containment zones alone.</ns0:p><ns0:p>Public health emergencies during epidemic/pandemic like SARS, MERS and Ebola outbreak were associated with increased psychological distress in the affected population <ns0:ref type='bibr' target='#b7'>(Batawi et al. 2019;</ns0:ref><ns0:ref type='bibr' target='#b36'>Lee et al. 2007;</ns0:ref><ns0:ref type='bibr' target='#b42'>Lotsch et al. 2017)</ns0:ref>. Maladaptive behaviors, emotional and defensive reactions were some of the psychological responses to pandemic (Taylor 2019). Social isolation was found to be strongly associated with anxiety, depression, self-harm, and suicidal tendencies <ns0:ref type='bibr' target='#b46'>(Matthews et al. 2019)</ns0:ref>. Studies indicated that social distancing for a longer duration could affect the mental health negatively <ns0:ref type='bibr' target='#b58'>(Reynolds et al. 2008)</ns0:ref>. Isolation, boredom, frustrations, worries about contracting the infection, lack of freedom, concerns for family/friends are some of the factors that could affect mental well-being during quarantine <ns0:ref type='bibr' target='#b11'>(Brooks et al. 2020)</ns0:ref>. Poor sleep quality and increased psychological distress were also well-documented during earlier pandemics <ns0:ref type='bibr' target='#b15'>(Chen et al. 2006;</ns0:ref><ns0:ref type='bibr' target='#b32'>Johal 2009)</ns0:ref>. In particular, poor sleep was associated with negative emotions, depressive symptoms and increased the risk of mental illness <ns0:ref type='bibr' target='#b0'>(Agargun & Kara 1997;</ns0:ref><ns0:ref type='bibr'>Tao et al. 2017)</ns0:ref>.</ns0:p><ns0:p>In a recent study conducted during COVID-19 outbreak in India, one fifth of adults were found to suffer from depression and stress and one fourth from anxiety <ns0:ref type='bibr' target='#b60'>(Saikarthik et al. 2020)</ns0:ref>. Mental health of medical students was found to be even poorer, when compared to general population <ns0:ref type='bibr' target='#b9'>(Bergmann et al. 2019)</ns0:ref>. Medical education is the most demanding of all the other professional programs in terms of both academics and emotional component of the students <ns0:ref type='bibr'>(Wolf 1994)</ns0:ref>. Globally, one in three medical students were found to have anxiety, which was higher than the general population <ns0:ref type='bibr'>(Tian-Ci Quek et al. 2019)</ns0:ref>. Level of depression, suicidal ideation, suicide rates, substance abuse and mental health disorders were also found to be higher among medical students <ns0:ref type='bibr' target='#b27'>(Hays et al. 1996;</ns0:ref><ns0:ref type='bibr' target='#b49'>Molodynski et al. 2020;</ns0:ref><ns0:ref type='bibr'>Schwenk et al. 2010)</ns0:ref>. Although medical students have better access to mental health care, they were less likely to seek mental health help compared to general population, mainly due to stigma surrounding mental health disorders. This may lead to untoward and harmful coping methods like excess alcohol consumption and substance abuse <ns0:ref type='bibr' target='#b17'>(Chew Graham et al. 2003;</ns0:ref><ns0:ref type='bibr' target='#b59'>Rosenthal & Okie 2005)</ns0:ref>.</ns0:p><ns0:p>Swine flu (H1N1) outbreak in 2009 was the last outbreak of an infectious disease in a pandemic scale to which India was exposed <ns0:ref type='bibr'>(WHO 2009)</ns0:ref>. Undergraduate medical students in India are usually in the age group of late teens to mid-twenties, and hence the current COVID-19 infection is the first exposure to them as adults on a pandemic level. In addition, medical students are facing challenges such as sudden changes in their training routine, including teaching and assessment via online sessions, decreased patient contact and interactions with peers to name a few. These changes result in increased screen time, possible hinderance to their training and increased risk of contracting the infection mainly among the students in clinical postings. All these factors could eventually exert a toll on the mental and emotional well-being of the medical students as they are on an unknown territory.</ns0:p><ns0:p>Earlier studies show that the negative impact of epidemic/pandemic on the mental health are higher in healthcare workers <ns0:ref type='bibr' target='#b37'>(Lee et al. 2018;</ns0:ref><ns0:ref type='bibr' target='#b44'>Lu et al. 2006;</ns0:ref><ns0:ref type='bibr' target='#b47'>McAlonan et al. 2007</ns0:ref>). Unfortunately, only limited studies were done on the impact of epidemic/pandemic on the mental health of medical students. Studies on the impact of COVID-19 pandemic on medical students are limited to cross-sectional surveys assessing attitude, awareness, knowledge, precautionary measures, concerns, risk perceptions, impact on education and confidence, and fear of COVID-19 <ns0:ref type='bibr' target='#b1'>(Agarwal et al. 2020;</ns0:ref><ns0:ref type='bibr' target='#b3'>Ahmed et al. 2020b;</ns0:ref><ns0:ref type='bibr' target='#b18'>Choi et al. 2020;</ns0:ref><ns0:ref type='bibr' target='#b34'>Khasawneh et al. 2020;</ns0:ref><ns0:ref type='bibr' target='#b52'>Nguyen et al. 2020;</ns0:ref><ns0:ref type='bibr'>Taghrir et al. 2020)</ns0:ref>. Literature search showed only a single study about the psychological impact of COVID-19 on medical students, which cross-sectionally assessed their anxiety levels <ns0:ref type='bibr' target='#b13'>(Cao et al. 2020</ns0:ref>).</ns0:p><ns0:p>Globally, few longitudinal studies compared mental health before and during COVID-19, and found an increase in anxiety and depression symptoms in college students in China <ns0:ref type='bibr' target='#b39'>(Li et al. 2020a)</ns0:ref>, the United States <ns0:ref type='bibr' target='#b30'>(Huckins et al. 2020</ns0:ref>) and a deterioration of mental health in the general population in the United Kingdom <ns0:ref type='bibr' target='#b55'>(Pierce et al. 2020</ns0:ref>). To our knowledge, there are no studies analyzing the impact of COVID-19 on mental health of undergraduate medical students prospectively to assess cause and relationship. From these observations, we hypothesized that COVID-19 outbreak and quarantine would have a serious negative impact on the mental health of undergraduate medical students. Hence, we conducted a prospective longitudinal study to investigate the mental health of undergraduate medical students over a duration of 6 months by analyzing data collected before and during COVID-19 outbreak in India. The study was conducted in a medical college in Chennai, Tamil Nadu, which is a center for treating COVID-19 patients. We did an extensive investigation of possible confounders and predictors of mental health disorders including demographics, sleep quality, apprehensions related to and caused by COVID-19 in terms of academics and concerns for the self, family, friends, and interpersonal relationships.</ns0:p></ns0:div>
<ns0:div><ns0:head>Materials & Methods</ns0:head></ns0:div>
<ns0:div><ns0:head>Participants and setting</ns0:head><ns0:p>The study was originally planned to be a cross-sectional survey for assessing the mental health of the undergraduate medical students in the institution. There were 300 medical students in the institution enrolled for undergraduate medical degree. The Bachelor of Medicine and Bachelor of Surgery (M.B.B.S) is a 5.5 years undergraduate medical course offered in India in which the first 2.5 years concentrate mostly on basic medical sciences (pre and para-clinical subjects) and the next 2 years on clinical subjects followed by 1 year of Compulsory Rotatory Resident Internship (CRRI). All 300 students studying in pre-clinical (1st year), para-clinical (1.5 years after preclinical year) and clinical years (pre-final year, final year, and resident interns) were included for the study and convenience sampling method was used. The students were explained about the objective of the study and were informed that the participation was voluntary, and confidentiality will be maintained. 276 students out of the total 300 students agreed to take part in the study from whom written consent was obtained before the beginning of the study.</ns0:p></ns0:div>
<ns0:div><ns0:head>Baseline (Before COVID-19) survey</ns0:head><ns0:p>Basic sociodemographic details such as age, gender, year of study, area of current residence and family gross monthly income were collected, and the mental health status was assessed using Depression Anxiety Stress Scale 21 items (DASS21). Students below 18 years of age and those with self-reported history of any pre-existing chronic medical conditions including mental health disorders were excluded (5 were underage and 2 reported history of bronchial asthma). The remaining 269 participants who were included for the study were contacted during their free time, after classes and were encouraged to answer the survey sincerely and doubts were clarified. Email id and mobile number were collected from all the participants. This part of the study was conducted during the first two weeks of December 2019, which was before COVID-19 outbreak in India.</ns0:p></ns0:div>
<ns0:div><ns0:head>Follow-up (During COVID-19) survey</ns0:head><ns0:p>With the unexpected changes to normalcy caused by the COVID-19 outbreak and subsequent lockdown, the authors decided to prospectively study the mental health status of the medical students to assess the effects of COVID-19 on mental health of the study population. After obtaining permission from Institutional Ethics Committee (IEC), the original data from the crosssectional study was decided to be taken as 'before COVID-19 data' (baseline) and another survey was conducted on June 2020 (June 10 to 20) to collect 'during .</ns0:p><ns0:p>The follow-up survey was conducted via Google form whose link was sent through personal email IDs of the students which were collected during the baseline survey. This protocol was exercised in order to follow strict social distancing protocol and to avoid direct contact. The follow-up survey included 5 sections; first section had a detailed description of the purpose of the study, along with the informed consent. This section explained the importance and benefits of the survey in the current pandemic, highlighting the voluntary nature of participation and assurance of confidentiality of the collected data. Only after consenting to the study, the participants could access the remaining sections. The successive sections collected responses for demographic details, self-administered questionnaire, DASS21 and Pittsburgh Sleep Quality Index (PSQI). Out of the 269 participants from the baseline survey, 30 randomly selected students were included in a pilot study (described below) and were hence excluded from the follow-up survey. Out of the remaining 239 participants, 222 students responded, from which 5 responses were excluded because of being incomplete (response rate 90.8%). The final sample size of this prospective longitudinal study was 217. A flowchart illustrating the sample selection from baseline to follow-up survey is shown in Figure <ns0:ref type='figure'>1</ns0:ref>.</ns0:p></ns0:div>
<ns0:div><ns0:head>Survey instruments</ns0:head><ns0:p>To fulfil the objective of our longitudinal study, besides DASS21, 'during COVID-19' data also included a self-administered questionnaire to assess the impact of COVID-19 related stressors and PSQI to assess the sleep quality of the students.</ns0:p></ns0:div>
<ns0:div><ns0:head>Assessment of the impact of COVID-19 related stressors</ns0:head><ns0:p>A self-administered questionnaire was prepared by the authors after an extensive literature search, discussion with peers and local experts <ns0:ref type='bibr'>(Wang et al. 2020;</ns0:ref><ns0:ref type='bibr'>Wong et al. 2007</ns0:ref>). It included 12 close-ended questions out of which, Items 1-3 focused on the subjects' status of COVID-19 testing (Yes/No) and their interactions with COVID-19 patients (Yes/No/I don't know). The remaining nine items were designed to assess self-perceived levels of concerns and worries for the self (4-6) and family/friends (7-8) and those related to academics (9-12), due to COVID-19 outbreak and quarantine (Supplemental file-Other). The responses were measured on a Likert scale of score 1 to 5, with 1 being the least and 5 being the maximum. This questionnaire was first tested empirically on 30 students (15 each from pre/para clinical years and clinical years) as a pilot study <ns0:ref type='bibr'>(Hill & century 1998)</ns0:ref>. The collected feedback and responses were analyzed, and corrections were made in the form of changes in articulation and simplification of vocabulary with the help of experts in this field.</ns0:p></ns0:div>
<ns0:div><ns0:head>Identification of latent variables from the self-administered questionnaire</ns0:head><ns0:p>An exploratory factor analysis (EFA) was performed on the 9 items (item 4-12) to determine the validity of the questionnaire and to identify latent variables that could enable the objective of the study. The EFA was conducted using principal component analysis with varimax rotation for factor extraction.</ns0:p><ns0:p>The extracted factors were analyzed for retention using Kaiser criterion (Eigen value >1), Scree test and counter-validated using parallel analysis. The Eigen values obtained from parallel analysis, which are values generated randomly with the same number of variables and sample size, were compared with the factor solution generated by EFA with our data. Eigen values of the factors that were higher than the values obtained from parallel analysis decided the number of factors. The two-factor solution thus obtained had 5 items in one (items 4-8) and 4 items in another (items 9-12). The factors were named COVID-19-related general apprehensions (COVID-19-GA) and COVID-19-related academic apprehensions (COVID-19-AA) respectively (Table <ns0:ref type='table' target='#tab_1'>S1</ns0:ref>). Reliability analysis was performed for each of the factors separately which presented with a high reliability, with the Cronbach's alpha score of 0.89 for COVID-19-GA and 0.91 for COVID-19-AA. COVID-19-GA was scored by totaling the scores of the five individual items with the total score ranging from 5 to 25 and higher scores denote higher general apprehension. Similarly, COVID-19-AA was scored by summing up the scores of the four individual items with the scores ranging from 4 to 20 and higher scores denote higher academic related apprehension.</ns0:p></ns0:div>
<ns0:div><ns0:head>Estimation of mental health status</ns0:head><ns0:p>Mental health status of the medical students was assessed using Lovibond and Lovibond's Depression Anxiety Stress Scale 21 items (DASS21) <ns0:ref type='bibr' target='#b43'>(Lovibond et al. 1995)</ns0:ref>. This scale comprises of 21 items with seven each for depression, anxiety, and stress subscales. The total sub scores range from 0 to 42 and is categorized into normal, mild, moderate, severe, and extremely severe. In this study, DASS21 sub scores were categorized dichotomously, with the participants being divided in to those who showed symptoms of depression, anxiety and stress and those who did not, based on the cut-off sub-scores of 9, 7 and 14 respectively <ns0:ref type='bibr' target='#b16'>(Cheung et al. 2015;</ns0:ref><ns0:ref type='bibr' target='#b43'>Lovibond et al. 1995)</ns0:ref>.</ns0:p></ns0:div>
<ns0:div><ns0:head>Estimation of sleep quality</ns0:head><ns0:p>Subjective sleep quality was assessed using Pittsburgh Sleep Quality Index (PSQI) which includes 21 items that assess seven components viz. subjective sleep quality, sleep duration, sleep latency, habitual sleep efficiency, use of sleep medications, sleep disturbance, and daytime dysfunction over the duration of two weeks prior to assessment. Global PSQI scores are obtained by summing up the seven individual sub scores and it ranges from 1 to 21 with higher scores (>5) denoting poor sleep quality <ns0:ref type='bibr' target='#b12'>(Buysse et al. 1989;</ns0:ref><ns0:ref type='bibr' target='#b56'>Rahe et al. 2015)</ns0:ref>.</ns0:p><ns0:p>Previous studies have shown high reliability of both DASS21 and PSQI among Indian undergraduate medical student population <ns0:ref type='bibr' target='#b62'>(Shad et al. 2015;</ns0:ref><ns0:ref type='bibr'>Yadav et al. 2016</ns0:ref>). In our study, both the scales showed good internal consistency and DASS21 scale demonstrated good testretest reliability. Cronbach alpha score for reliability for PSQI was 0.72 and for DASS21 scale 0.94 (depression subscale 0.85, anxiety subscale 0.84, stress subscale 0.87) and 0.94 (depression <ns0:ref type='table' target='#tab_3'>PeerJ reviewing PDF | (2020:08:51660:2:0:NEW 18 Sep 2020)</ns0:ref> Manuscript to be reviewed subscale 0.87, anxiety subscale 0.81, stress subscale 0.85) for baseline and follow-up survey, respectively.</ns0:p></ns0:div>
<ns0:div><ns0:head>Ethical consideration</ns0:head><ns0:p>Ethical approval was obtained from the IEC, Madha Medical College and Research Institute in Chennai (MMCRI/IEC/H/018/2020) and research was done in accordance with the Helsinki Declaration for research on human participants.</ns0:p></ns0:div>
<ns0:div><ns0:head>Statistical analysis</ns0:head><ns0:p>Descriptive statistics was performed for all the variables. The scores of depression, anxiety, stress, and sleep quality were expressed as mean ± standard deviation (SD). Initially, unadjusted univariate association between the demographic variables and depression, anxiety and stress were performed. Mann Whitney U test and Kruskal Wallis test for continuous variables and Chisquare test for categorical variables were used for cross-sectional analysis. Wilcoxon signed rank test for continuous variables and McNemar's test for categorical variables were used for longitudinal analysis. Spearman's correlation test was performed to assess the correlation between the scores obtained from the survey instruments in both the surveys.</ns0:p><ns0:p>To explore the contributory factors associated with depression, anxiety, and stress during COVID-19 outbreak (dependent variable), adjusted binary logistic regression analysis was performed. Independent variables included were scores of PSQI, COVID-19-GA, COVID-19-AA, dependent variable from baseline survey (depression, anxiety and stress sub-scores in respective regression models) as covariates (continuous variables) and responses for the items 1-3 from the self-administered questionnaire as independent factors (categorical variables). Crosssectional association between sleep quality and study parameters were analyzed using adjusted binary logistic regression. The effect of each of the independent variable was adjusted for sociodemographic variables which were considered to be potential confounders viz. age, gender, year of study, current residence, and family monthly income, in separate binary regression models. The results were expressed as adjusted odds ratio (aOR), 95% confidence interval (95% CI) and P value (statistical significance set at two-tailed P<0.05).</ns0:p><ns0:p>All the statistical analysis was performed using SPSS version 26 (IBM, New York, USA) and Parallel analysis was performed using scripts from O'Connor BP (O'Connor 2000)</ns0:p></ns0:div>
<ns0:div><ns0:head>Results</ns0:head><ns0:p>This longitudinal study includes 217 undergraduate medical students (78 males and 139 females); the average age was 20±1.6 years. 5.1% got tested for COVID-19, and they all tested negative. 14.3% had friends and family who tested positive for COVID-19 and 12% declared to have had direct contact with COVID-19 patients (Figure <ns0:ref type='figure'>2</ns0:ref>). The distribution of responses to the items of COVID-19-GA and COVID-19-AA is shown in Figure <ns0:ref type='figure'>2</ns0:ref> and percentage distribution of depression, anxiety and stress in baseline and follow-up survey and PSQI in follow-up survey is shown in Figure <ns0:ref type='figure' target='#fig_3'>3</ns0:ref>.</ns0:p></ns0:div>
<ns0:div><ns0:head>Cross-sectional relationship between sociodemographic variables and depression, anxiety, and stress</ns0:head><ns0:p>The cross-sectional and longitudinal relationship between sociodemographic variables and depression, anxiety, and stress are shown in Table <ns0:ref type='table' target='#tab_0'>1</ns0:ref>. There was no significant cross-sectional relationship between the demographic variables and both baseline and follow-up depression, anxiety, and stress scores except in the baseline survey where depression levels were higher in the students from rural sector than urban sector (P=0.039). The association between demographic variables and depression, anxiety, and stress analyzed by binary logistic regression showed that age was a protective factor for depression in the follow-up survey (OR 0.737, 95% CI 0.565-0.961) (Table <ns0:ref type='table' target='#tab_3'>S2</ns0:ref>). Other than this, there were no significant associations between demographic variables and mental health in both baseline and follow-up survey (Tables <ns0:ref type='table' target='#tab_3'>S2-4</ns0:ref>).</ns0:p></ns0:div>
<ns0:div><ns0:head>Comparison of baseline and follow-up depression, anxiety and stress stratified by sociodemographic variables</ns0:head><ns0:p>The overall prevalence (with 95% confidence interval) of depression, anxiety and stress before COVID-19 was 33.2% (27-39.9%), 21.2% (16-27.2%) and 20.7% (15.5-26.7%) and during COVID-19 outbreak was 35.5% (29.1-42.2%), 33.2% (27-39.9%) and 24.9% (19.3-31.2%) respectively. There was a significant increase in the prevalence and mean scores of anxiety and stress when compared to baseline scores (P<0.001) (Table <ns0:ref type='table' target='#tab_0'>1</ns0:ref>). In terms of prevalence, when compared to baseline values, the prevalence of anxiety was significantly increased during COVID-19 in females (P=0.003), pre-clinical, para-clinical (P=0.014) and clinical year students (P=0.019), students from urban residence (P=0.005) and those with gross family income above 100,000 INR per month (P=0.007). In terms of mean DASS21 scores, there was a statistically significant increase in anxiety with mild effect size and stress levels with moderate to strong effect size in medical students in the follow-up survey compared to baseline levels irrespective of gender, year of study, current residence and gross family income per month (below 50,000 INR and above 100,000 INR) (P<0.05). Out of the entire study population, the levels of depression have increased significantly in male students and students from urban residence when compared to before COVID-19 levels albeit with mild effect size (Table <ns0:ref type='table' target='#tab_0'>1</ns0:ref>).</ns0:p></ns0:div>
<ns0:div><ns0:head>Difference in ranks of DASS21 scores between baseline and follow-up survey</ns0:head><ns0:p>The difference in ranks of DASS21 scores between baseline and follow-up survey is shown in Table <ns0:ref type='table' target='#tab_2'>2</ns0:ref>. There was no significant change in depression (P=0.146) in medical students between the two surveys; however, anxiety and stress has increased (P<0.001), showing an increase in median scores in the follow-up survey (Tables <ns0:ref type='table' target='#tab_3'>1, 2</ns0:ref>). The incidence of depression, anxiety, and stress based on DASS21 sub-scores were found to be 2.3 (5 out of 217), 11.98 (26 out of 217) and 4.15 (9 out of 217) per 100 per 6 months, respectively. 97 (44.7%), 89 (41.01%) and 142 (65.44%) students scored higher in depression, anxiety, and stress in the follow-up survey when compared to their responses before COVID-19 outbreak. While 60 (28.04%), 69 (31.79) and 7</ns0:p><ns0:p>(3.22%) students scored lesser in depression, anxiety, and stress in follow-up survey (Table <ns0:ref type='table' target='#tab_2'>2</ns0:ref>).</ns0:p></ns0:div>
<ns0:div><ns0:head>Correlation between the scores of the survey instruments from baseline and follow-up survey</ns0:head><ns0:p>The results of Spearman correlation analysis of the scores of the survey instruments from baseline and follow-up survey is shown in Table <ns0:ref type='table' target='#tab_4'>3</ns0:ref>. There were significant positive correlations between PSQI and baseline and follow-up depression, anxiety, and stress. The correlation between the baseline and follow-up depression, anxiety, and stress scores indicates that a higher baseline score was associated with higher follow-up score and vice versa (P<0.001). There was a relatively weak, but significant positive correlation between depression levels before COVID-19 and COVID-19-related general apprehensions (r=0.152, P=0.025) (Table <ns0:ref type='table' target='#tab_4'>3</ns0:ref>).</ns0:p></ns0:div>
<ns0:div><ns0:head>Adjusted binary logistic regression analysis of sleep quality</ns0:head><ns0:p>PeerJ reviewing PDF | (2020:08:51660:2:0:NEW 18 Sep 2020)</ns0:p><ns0:p>Manuscript to be reviewed Cross-sectional association between sleep quality and mental health in the follow-up survey is shown in Figure <ns0:ref type='figure' target='#fig_4'>4</ns0:ref>. Students with higher depression, anxiety, and stress scores during COVID-19 outbreak were found to be more likely to have poor sleep quality (P<0.001).</ns0:p></ns0:div>
<ns0:div><ns0:head>Adjusted binary logistic regression analysis of follow-up depression, anxiety, and stress</ns0:head><ns0:p>The results of binary logistic regression analysis for follow-up depression, anxiety, and stress are shown in Figures <ns0:ref type='figure' target='#fig_5'>5-7</ns0:ref>. Poor sleep quality was found to be significantly associated with an increase in depression, anxiety, and stress (P<0.001). Higher baseline scores of depression, anxiety, and stress were associated with higher levels of the same in follow-up survey (P<0.001).</ns0:p><ns0:p>Higher COVID-19-related general apprehension was associated with higher levels of anxiety (P=0.016), and stress (P=0.005). Students who did not have any direct interactions with COVID-19 patients were found to be less likely to have symptoms of depression (P=0.017) and stress (P=0.004) when compared to those who did. Similarly, absence of COVID-19 patients in family and friends was found to be associated with decreased levels of stress (P=0.004).</ns0:p></ns0:div>
<ns0:div><ns0:head>Discussion</ns0:head><ns0:p>The present study investigated the mental health status of undergraduate medical students in a medical college, which is a government-approved center for treating COVID-19 patients. The medical college is in Chennai, TamilNadu, which is one of the top 5 COVID-19 affected metropolitan cities in India. Longitudinal data analysis was used to test our hypothesis that COVID-19 outbreak and quarantine has a negative impact on the mental health of undergraduate medical students.</ns0:p><ns0:p>We found that 35.5% (95% CI 29.1-42.2%), 33.2% (95% CI 27-39.9%) and 24.9% (95% CI 19.3-31.2%) of the undergraduate medical students, including resident interns showed symptoms of depression, anxiety, and stress respectively during COVID-19 outbreak with the majority with moderate depression (15.2%), moderate anxiety (17.5%), and mild stress (13.4%). Based on the severity ranking, subjects with moderate and above ranking may present with a possible problem that may require intervention <ns0:ref type='bibr' target='#b20'>(Crawford & Henry 2003;</ns0:ref><ns0:ref type='bibr' target='#b43'>Lovibond et al. 1995;</ns0:ref><ns0:ref type='bibr' target='#b54'>Page et al. 2007;</ns0:ref><ns0:ref type='bibr' target='#b65'>Stormon et al. 2019)</ns0:ref>. In this study, 23.9%, 26.7% and 11.5% of the study population presented with moderate to extremely severe levels of depression, anxiety, and stress respectively (Figure <ns0:ref type='figure' target='#fig_3'>3</ns0:ref>). The six-month incidence of anxiety was found to be comparatively higher (11.98%) followed by stress (4.15%) and depression (2.3%). There were no other longitudinal studies conducted in medical students during COVID-19 pandemic assessing the incidence of mental health disorders; however, a study conducted in the general public in China found the one-month incidence of mental health disorders to be relatively low <ns0:ref type='bibr'>(Yali et al. 2020)</ns0:ref>. Though the incidence rate seemed to be relatively low in our study, 44.7%, 41.01% and 65.44% of the study population scored higher in depression, anxiety, and stress sub scores during COVID-19 (Table <ns0:ref type='table' target='#tab_2'>2</ns0:ref>). When compared to baseline survey that was recorded before COVID-19 outbreak in India, there was a significant increase in prevalence and levels of anxiety and stress, while that of the depression remained unchanged during COVID-19. Similar studies conducted longitudinally in college students found a significant increase in depression and anxiety when compared to before COVID-19 levels. <ns0:ref type='bibr' target='#b39'>(Li et al. 2020a</ns0:ref><ns0:ref type='bibr' target='#b30'>, Huckins et al. 2020)</ns0:ref>. This negative impact of the pandemic could be attributed to sudden challenges faced by the medical students in terms of academics, uncertainties about future, fear of infection, news about shortage of personal protective equipment, quarantine induced boredom, frustrations, lack of freedom, and fears caused by rumors and misleading news in the media <ns0:ref type='bibr' target='#b6'>(Bao et al. 2020;</ns0:ref><ns0:ref type='bibr' target='#b22'>Ferrel & Ryan 2020)</ns0:ref>.</ns0:p><ns0:p>Investigation of the influence of demographics on mental health showed that the increase in anxiety and stress levels in our study population was not affected by gender, year of study, current residence, or family income. We also found no significant cross-sectional difference in depression, anxiety, and stress between the groups of demographic variables during COVID-19. Binary logistic regression analysis of anxiety and stress and demographic variables as independent variables showed no significant association. In contrary, cross-sectional studies done on medical students in China and Brazil during this pandemic found significant associations between mental health disorders and place of residence, parental income/financial support, and gender <ns0:ref type='bibr' target='#b13'>(Cao et al. 2020;</ns0:ref><ns0:ref type='bibr' target='#b61'>Sartorao Filho et al. 2020)</ns0:ref>. Our results show that the response of the medical students to COVID-19 pandemic in terms of the levels of anxiety and stress is similar, irrespective of gender, year of study, current residence, and financial status of the family. Depression in the study population, on the other hand, remained unchanged during COVID-19 in all the categories except in male and urban population. Binary logistic regression showed that increase in age could decrease the likelihood of depression (OR 0.737, 95% CI 0.565-0.961) which is consistent to previous studies during COVID-19 <ns0:ref type='bibr' target='#b2'>(Ahmed et al. 2020a;</ns0:ref><ns0:ref type='bibr' target='#b26'>González-Sanguino et al. 2020)</ns0:ref>. However, in our study, the range of age was narrow (18-26 years) and mostly corresponded to the year of study, which had no significant association with depression. Hence, this observation becomes redundant. We found a significant increase in depression levels in male students but not in female students. This finding is contradictory to the study by Sartorao Filho et al., which found female medical students to be more at risk of developing depression symptoms during this pandemic <ns0:ref type='bibr' target='#b61'>(Sartorao Filho et al. 2020)</ns0:ref>. Females are more proactive in their response and awareness about the epidemic when compared to males, and our results could be a possible implication of this (Brittni Frederiksen 2020). Despite higher levels of depression in the students from rural areas before COVID-19, their depression levels remained unchanged in the follow-up survey. However, the students from the urban areas presented with an increase in depression when compared to before COVID-19 levels. Nearly 53% of India's cases were recorded in Mumbai, Delhi, Ahmedabad, Pune, and Chennai alone, which were listed as top five COVID-19 cities <ns0:ref type='bibr' target='#b63'>(Shylendra 2020)</ns0:ref>. Urban areas are highly populated, and the epidemic is more active in these areas than in rural areas. This led to the implementation of frequent lockdown measures by the respective State Governments in the COVID-19 hotspots, which are mainly urban centers. This could have possibly increased the psychological distress in the students from urban locations.</ns0:p><ns0:p>Since we did not find significant differences in mental health among most of the groups of demographic variables in both baseline and follow-up survey, it is likely that the worsening of mental health status of the medical students found in our study is associated with COVID-19related factors. To further elucidate this, we used adjusted binary logistic regression analysis to explore possible predictors. Our extensive literature survey showed all the collected demographic variables to have a potential influence on the outcome. Hence, despite not finding significant associations between demographics and mental health, we adjusted the effects of each potential predictor for all the recorded demographic variables in the regression models <ns0:ref type='bibr' target='#b13'>(Cao et al. 2020;</ns0:ref><ns0:ref type='bibr' target='#b61'>Sartorao Filho et al. 2020;</ns0:ref><ns0:ref type='bibr' target='#b28'>Heinze & Dunkler 2017)</ns0:ref> An important finding in our study is the independent bidirectional association between poor sleep quality and mental health. In our study, 34.6% of the study population suffered from poor sleep quality which was found to be a significant independent predictor of depression (aOR 1.337, 95% CI 1.19-1.502), anxiety (aOR 1.227, 95% CI 1.106-1.363) and stress (aOR 1.371, 95% CI 1.209-1.555) during COVID-19 (Figures <ns0:ref type='figure' target='#fig_5'>5-7</ns0:ref>), similar to a previous study by <ns0:ref type='bibr'>Cellini N et al., in Italy (Cellini et al. 2020)</ns0:ref>. Medical students are especially prone to poor sleep quality because of the physically and emotionally challenging and intense training they undertake <ns0:ref type='bibr'>(Wong et al. 2005)</ns0:ref>. Poor sleep affects neurocognitive and psychomotor performance, emotional wellbeing, working capacity, academic performance, physical and mental health as well as quality of life <ns0:ref type='bibr' target='#b4'>(Al-Khani et al. 2019;</ns0:ref><ns0:ref type='bibr' target='#b23'>Flores 2009;</ns0:ref><ns0:ref type='bibr' target='#b25'>Giri et al. 2013;</ns0:ref><ns0:ref type='bibr' target='#b50'>Mume et al. 2011)</ns0:ref>. Due to lockdown measures and travel restrictions, students are facing decreased physical activity, lack of schedule, altered living conditions, increased screen time and time spent in social media, and altered sleep wake schedule including increased daytime nap duration <ns0:ref type='bibr' target='#b45'>(Majumdar et al. 2020)</ns0:ref>. All these factors in addition to higher demands of medical curriculum could lead to poor sleep, which in turn affects mental wellbeing. Conversely, we also found that higher depression (aOR 1.114, 95% CI 1.07-1.159), anxiety (aOR 1.120, 95% CI 1.065-1.178) and stress levels (aOR 1.126, 95% CI 1.078-1.176) during COVID-19 were significant predictors of poor sleep quality (Figure <ns0:ref type='figure' target='#fig_4'>4</ns0:ref>). Poor sleep is long since been considered an important symptom of mental health disorders. Sleep disturbance is a primary symptom of major depressive disorder <ns0:ref type='bibr' target='#b31'>(Jindal & Thase 2004)</ns0:ref>. Anxiety and stress negatively affect the body's ability to fall and stay asleep <ns0:ref type='bibr' target='#b19'>(Coplan et al. 2015;</ns0:ref><ns0:ref type='bibr' target='#b33'>Kalmbach et al. 2018</ns0:ref>). In addition, increased time spent in social media and digital devices is a way by which young adults cope with social isolation, and it is associated with increased tendencies to develop sleep disturbances <ns0:ref type='bibr' target='#b64'>(Sivertsen et al. 2019</ns0:ref>). Thus, our findings show that poor sleep quality is both a cause and an effect of increased depression, anxiety, and stress symptoms in medical students during this pandemic. Thus, worsening of one could exacerbate the other.</ns0:p><ns0:p>Students with higher baseline levels of depression, anxiety, and stress were found to be more likely to have depression (aOR 1.27, 95% CI 1.185-1.361), anxiety (aOR 1.176, 95% CI 1.099-1.259) and stress (aOR 1.810, 95% CI 1.483-2.208) during COVID-19 outbreak (Figures <ns0:ref type='figure' target='#fig_5'>5-7</ns0:ref>). There was also significant positive correlation between baseline and follow-up depression, anxiety, and stress scores (Table <ns0:ref type='table' target='#tab_4'>3</ns0:ref>). These results are consistent with a similar study conducted on college students in China <ns0:ref type='bibr' target='#b39'>(Li et al. 2020a)</ns0:ref>. Studies show associations between pre-existing mental health problems and mental health disorders in medical students <ns0:ref type='bibr'>(Yates et al. 2008)</ns0:ref>.</ns0:p><ns0:p>Previous studies indicate that COVID-19 has a higher negative impact on people with mental health disorders when compared to those without any <ns0:ref type='bibr' target='#b5'>(Asmundson et al. 2020;</ns0:ref><ns0:ref type='bibr' target='#b5'>Taylor et al. 2020)</ns0:ref>. Medical students are proven to be at a higher risk of developing mental health disorders during the course of their training period <ns0:ref type='bibr' target='#b48'>(Moffat et al. 2004</ns0:ref>). With the added stressors related to the pandemic, it might be difficult to cope, especially for those students who had higher levels of depression, anxiety, and stress to begin with, leading to exacerbation of symptoms.</ns0:p><ns0:p>A surprising finding in our study was that COVID-19-related academic apprehensions were not significantly associated with depression, anxiety, and stress. This is in contrary to a recent study conducted in medical students in China, which found moderate positive correlation between worries about academic delay and anxiety <ns0:ref type='bibr' target='#b13'>(Cao et al. 2020)</ns0:ref>. Although social desirability response bias could be an attributing factor, our finding could be a reflection of the feel of assurance by the students because of drastic student-centered efforts taken by the medical college and universities in continuing medical education. Another reason could be that, since these lockdown instigated changes were in effect for around three months, the students would have adapted to this new normalcy in their training better than anticipated. On the other hand, higher COVID-19-GA scores were found to be a significant predictor for higher levels of anxiety (aOR 1.097, 95% CI 1.018-1.182) and stress (aOR 1.128, 95% CI 1.037-1.227) (Figures <ns0:ref type='figure' target='#fig_6'>6, 7</ns0:ref>). In a study conducted on medical students in Pakistan, around 76% of the participants conveyed being worried about contracting COVID-19 during clinical postings and even more worried about insufficient care and improper treatment, if they contracted the infection <ns0:ref type='bibr' target='#b3'>(Ahmed et al. 2020b</ns0:ref>). In a recent study conducted on adult Indian population, individuals with increased self-perceived risk of contracting COVID-19 were found to be more likely to have mental health disorders <ns0:ref type='bibr' target='#b60'>(Saikarthik et al. 2020)</ns0:ref>. A longitudinal study conducted in China in college students, found fear of infection to be significantly associated with anxiety and depression <ns0:ref type='bibr' target='#b39'>(Li et al. 2020a)</ns0:ref>. Our results demonstrated that medical students' self-perceived levels of worries for the self, family, and friends about contracting COVID-19; surviving if contracted with COVID-19; and COVID-19 affecting interpersonal relationships have a negative impact on their mental health. Student population is highly active in social media, which is filled with high amounts of misinformation, adding fear, and affecting mental well-being. Frequent use of social media was associated with higher prevalence of mental health problems during COVID-19 <ns0:ref type='bibr' target='#b24'>(Gao et al. 2020)</ns0:ref>. At the same time, social media could be used for communications among peers and family, thereby offering much needed social support.</ns0:p><ns0:p>In this study, we found no significant association between being tested for COVID-19 and depression, anxiety, and stress. The possible reason could be that all of those who got tested for COVID-19 were found negative. We also found that students without any COVID-19 patients among their family and friends (aOR 0.259, 95% CI 0.069-0.968) and those who were not sure about having any direct interactions with COVID-19 patients (aOR 0.280 95% CI 0.092-0.856) were found to be less likely to have symptoms of stress (Figure <ns0:ref type='figure'>7</ns0:ref>). Students without any direct interactions with COVID-19 patients were less likely to have symptoms of depression (aOR 0.31 95% CI 0.106-0.905) (Figure <ns0:ref type='figure' target='#fig_5'>5</ns0:ref>) when compared to those who did. These findings were in line with the study by <ns0:ref type='bibr'>Cao et al. in China (Cao et al. 2020)</ns0:ref>. The R naught (R0) of SARS-CoV-2 is 2.2 (2-2.5), i.e. each infected person spreads the infection to 2.2 individuals, in other words it is more contagious than seasonal flu <ns0:ref type='bibr' target='#b41'>(Li et al. 2020b)</ns0:ref>. This high contagious nature of the novel SARS-CoV-2 virus could be related to our findings.</ns0:p></ns0:div>
<ns0:div><ns0:head>Strengths and Limitations of the study</ns0:head><ns0:p>The longitudinal nature of this study is its major strength which outweighs the limitation of relatively small sample size. The study investigated the mental health status of the same medical students before and during COVID-19 pandemic, which enabled in studying the pattern, temporal order and predictors of changes in their mental health status. To our knowledge, this is the first study to analyze the effects of COVID-19 outbreak on the mental health of undergraduate medical students longitudinally. We elaborately studied the influence of demographic variables, stressors related to COVID-19 by categorizing them into general and academic apprehensions as well as sleep quality, on mental health. With the known impact of negative mental health on medical students' career and life, we believe the results of our study is important due to the insights provided, that will help the medical educators to address and devise strategies to overcome the pandemic-induced negative impact on undergraduate medical students' mental health.</ns0:p><ns0:p>Our study is not without limitations. Though our study population includes students who were tested for COVID-19, none turned out positive. Thus, our results cannot be extrapolated to medical students infected with the virus. Despite promising confidentiality, there could have been possible response bias by the students in answering the survey.</ns0:p></ns0:div>
<ns0:div><ns0:head>Conclusions</ns0:head><ns0:p>Prevalence and levels of anxiety and stress increased, and depression symptoms remained unaltered during COVID-19 outbreak and quarantine. Poor sleep quality, higher levels of depression, anxiety, and stress before COVID-19, increased worries for the self, family, and friends about contracting and surviving, if contracted with COVID-19, and about COVID-19 affecting interpersonal relationships, presence of COVID-19 patients in family and friends and direct interactions with COVID-19 patients were found to be significant predictors of negative mental health in medical students.</ns0:p><ns0:p>Neglecting the mental health of the medical students would lead to long-term detrimental effects, which not only will affect the quality of life of medical students and future physicians, but also the overall performance of the healthcare system. An effective plan to safeguard the mental health of this already vulnerable population of undergraduate medical students is crucial. We strongly believe our findings would help the medical educators in addressing and mitigating the rise in mental health disorders, which could prove worse than the current pandemic itself. Further studies to analyze the temporal pattern of changes in mental health status of the medical students are warranted. Manuscript to be reviewed Difference in the ranks of DASS21 scores between baseline and follow-up surveys Manuscript to be reviewed Manuscript to be reviewed Correlation between scores of survey instruments from baseline and follow-up survey</ns0:p><ns0:p>The Manuscript to be reviewed Manuscript to be reviewed Manuscript to be reviewed Manuscript to be reviewed</ns0:p><ns0:note type='other'>Figure 7</ns0:note><ns0:p>Forest plot showing adjusted binary logistic regression analysis of follow-up stress scores aOR adjusted odds ratio; odds ratio adjusted for age, gender, year of study, urban/rural residential status, family's monthly financial status; 95%CI 95% confidence interval;</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:08:51660:2:0:NEW 18 Sep 2020)</ns0:p></ns0:div><ns0:figure xml:id='fig_1'><ns0:head /><ns0:label /><ns0:figDesc>a follow-up < baseline, b follow-up > baseline, c follow-up = baseline PeerJ reviewing PDF | (2020:08:51660:2:0:NEW 18 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_2'><ns0:head /><ns0:label /><ns0:figDesc>Figure 1</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_3'><ns0:head>Figure 3</ns0:head><ns0:label>3</ns0:label><ns0:figDesc>Figure 3</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_4'><ns0:head>Figure 4 Forest</ns0:head><ns0:label>4</ns0:label><ns0:figDesc>Figure 4</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_5'><ns0:head>Figure 5 Forest</ns0:head><ns0:label>5</ns0:label><ns0:figDesc>Figure 5</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_6'><ns0:head>Figure 6 Forest</ns0:head><ns0:label>6</ns0:label><ns0:figDesc>Figure 6</ns0:figDesc></ns0:figure>
<ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='34,42.52,229.87,525.00,331.50' type='bitmap' /></ns0:figure>
<ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='39,42.52,250.12,525.00,217.50' type='bitmap' /></ns0:figure>
<ns0:figure type='table' xml:id='tab_0'><ns0:head>Table 1 (on next page)</ns0:head><ns0:label>1</ns0:label><ns0:figDesc>Relationship between demographic variables and baseline and follow-up depression,</ns0:figDesc><ns0:table><ns0:row><ns0:cell>anxiety and stress scores</ns0:cell><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>a square test, P value: Wilcoxon signed rank test, e P value: Kruskal Wallis test, r value: effect size of Wilcoxon signed rank test b P value: McNemar test, c P value: Mann Whitney U test,</ns0:cell><ns0:cell>d</ns0:cell><ns0:cell>P value: Chi</ns0:cell></ns0:row><ns0:row><ns0:cell>N (%): Number of subjects with depression, anxiety and stress</ns0:cell><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>Subjects with depression: depression sub-score > 9,</ns0:cell><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>Subjects with anxiety: anxiety sub-score > 7,</ns0:cell><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>Subjects with stress: stress sub-score > 14</ns0:cell><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>Significant P value in bold letters</ns0:cell><ns0:cell /><ns0:cell /></ns0:row></ns0:table></ns0:figure>
<ns0:figure type='table' xml:id='tab_1'><ns0:head>Table 1 :</ns0:head><ns0:label>1</ns0:label><ns0:figDesc>Relationship between demographic variables and baseline and follow-up depression, anxiety and stress 2 scores</ns0:figDesc><ns0:table><ns0:row><ns0:cell>Depression</ns0:cell><ns0:cell>Anxiety</ns0:cell></ns0:row></ns0:table><ns0:note>PeerJ reviewing PDF | (2020:08:51660:2:0:NEW 18 Sep 2020) Manuscript to be reviewed PeerJ reviewing PDF | (2020:08:51660:2:0:NEW 18 Sep 2020) Manuscript to be reviewed PeerJ reviewing PDF | (2020:08:51660:2:0:NEW 18 Sep 2020)</ns0:note></ns0:figure>
<ns0:figure type='table' xml:id='tab_2'><ns0:head>Table 2 (on next page)</ns0:head><ns0:label>2</ns0:label><ns0:figDesc /><ns0:table /></ns0:figure>
<ns0:figure type='table' xml:id='tab_3'><ns0:head>Table 2 :</ns0:head><ns0:label>2</ns0:label><ns0:figDesc>Difference in the ranks of DASS21 scores between baseline and follow-up surveys</ns0:figDesc><ns0:table><ns0:row><ns0:cell>Median score (50 th</ns0:cell></ns0:row><ns0:row><ns0:cell>percentile)</ns0:cell></ns0:row></ns0:table></ns0:figure>
<ns0:figure type='table' xml:id='tab_4'><ns0:head>Table 3 (on next page)</ns0:head><ns0:label>3</ns0:label><ns0:figDesc /><ns0:table /></ns0:figure>
<ns0:note place='foot'>PeerJ reviewing PDF | (2020:08:51660:2:0:NEW 18 Sep 2020)</ns0:note>
<ns0:note place='foot' n='2'>The results are expressed as ρ (Rho) value; (* P<0.05); (** P<0.001) PeerJ reviewing PDF | (2020:08:51660:2:0:NEW 18 Sep 2020)</ns0:note>
</ns0:body>
" | " Editor’s comments
My first concern is the English language is not written in native American or
British English. Further, the paper also has some language issues, for
example, line 44 and 52, '&' should not be used, 'and' is appropriate. Line
116-117, it is inappropriate to say 'depression is found to be higher', which
should be the level of depression. I strongly suggest the authors to have their
paper polished by native speakers.
As suggested, the manuscript was edited for language correction by native English
speakers.
My second concern is the unnecessarily long abstract. I think it can be written
in a more concise way. For example, there is no need to provide so many
statistical details here. In addition, it is not necessary to abbreviate depression
as 'D', as well as anxiety and stress.
Corrections made as suggested.
My third concern is the inaccurate conclusion. Because one of the three
mental health indicators, depression, did not change significantly after the
outbreak, it seems not strict to say 'has a negative impact on mental health'.
The term 'mental health' is broad, which should be specific to anxiety and
stress based on findings of the present study.
Third, line 86-89, please update the figures accordingly. India has been the
second in the world.
Corrections made as suggested.
Fourth, line 167-170, sample size estimation is described in a confused way.
First, it should not be limited to cross-sectional design, cohort design should
also be considered. Second, the sample is recruited from an institution, not an
infinite population. Please consult a bio-statistician to address this issue.
The methodology section has been corrected as per suggestions from a
biostatistician.
" | Here is a paper. Please give your review comments after reading it. |
9,813 | "<ns0:abstract xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:p>Ecdysis was a vital process during the lives of trilobites, in addition to preserving the morphological changes in trilobite ontogeny, the preservation of its action often captured interesting behavioral information. Abundant exuviae of Ovalocephalus tetrasulcatus are preserved in the Ordovician strata in central Hubei, China, and some of them are arranged with two or three together end to end or superimposed. The preserved patterns and burial conditions indicate that these specimens were caused by the active behavior of trilobites.</ns0:p><ns0:p>It is speculated that these exuvial clusters were formed by two or three trilobites in line to molt; that is, after one trilobite finished molting, other trilobites molted in front of, behind, or overlying the previously molted shells. This ecdysis strategy is interpreted as related to the postulated herding behavior of some trilobites, representing a behavioral response of the trilobites to choose a nearby safe zone during some risky life activities.</ns0:p></ns0:div>
</ns0:abstract>
<ns0:body xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:head>Introduction</ns0:head><ns0:p>The biomineralized or chitinous exoskeleton of arthropods hinders the growth of their bodies; therefore, individuals must shed their old shells many times during growth and development <ns0:ref type='bibr' target='#b19'>(Moussian, 2013;</ns0:ref><ns0:ref type='bibr' target='#b8'>Daley and Drage, 2016)</ns0:ref>. Trilobites, as an extinct group of arthropods, also needed to shed their shells as they grew <ns0:ref type='bibr' target='#b10'>(Fortey, 2014)</ns0:ref>. Different trilobites exhibited different molting techniques <ns0:ref type='bibr' target='#b16'>(Henningsmoen, 1975)</ns0:ref>; most trilobites shed their shells through separating the librigenae from the cranidium <ns0:ref type='bibr' target='#b18'>(McNamara and Rudkin, 1984;</ns0:ref><ns0:ref type='bibr' target='#b30'>Whittington, 1990)</ns0:ref>, whereas for trilobites with librigenae fused with the cranidium, separation of the cephalon from the thoracopygon usually occurred during molting <ns0:ref type='bibr' target='#b25'>(Speyer, 1985;</ns0:ref><ns0:ref type='bibr' target='#b29'>Wang and Han, 1997)</ns0:ref>, and some genera had multiple exuvial modes <ns0:ref type='bibr' target='#b2'>(Brandt, 1993;</ns0:ref><ns0:ref type='bibr' target='#b3'>Budil and Bruthansová, 2005)</ns0:ref>. In addition to ecdysis reflecting the ontogenetic development process of trilobites, some exceptionally preserved trilobite specimens also contain behavioral information. For example, some trilobites shed their shells by hiding in empty shells or burrows of other animals <ns0:ref type='bibr' target='#b9'>(Davis et al., 2001;</ns0:ref><ns0:ref type='bibr' target='#b5'>Chatterton et al., 2003;</ns0:ref><ns0:ref type='bibr' target='#b6'>Chatterton and Fortey, 2008;</ns0:ref><ns0:ref type='bibr' target='#b37'>Zong et al., 2016)</ns0:ref>, and even molted infaunally <ns0:ref type='bibr' target='#b24'>(Rustán et al., 2011)</ns0:ref>, reflecting the hiding behavior of trilobites. Some phacopids may also have exhibited asymmetric behaviors during molting <ns0:ref type='bibr' target='#b38'>(Zong and Gong, 2017)</ns0:ref>. Other trilobites collectively shed their shells, which may have been related to molting-mating behavior <ns0:ref type='bibr' target='#b27'>(Speyer and Brett, 1985;</ns0:ref><ns0:ref type='bibr' target='#b26'>Speyer, 1990)</ns0:ref>.</ns0:p><ns0:p>South China is an important area for trilobite fossils, and abundant Ordovician trilobites have been collected from Hubei Province <ns0:ref type='bibr' target='#b17'>(Lu, 1975;</ns0:ref><ns0:ref type='bibr' target='#b35'>Zhou and Zhen, 2008)</ns0:ref>. However, there are few reports on exuvial specimens and correlation research. Only photographs of exuvial specimens were attached to the identification of genera and species in the paleontological literature; these exuvial specimens have not been systematically described, their patterns have not been classified and explained, and the behavioral strategy of these trilobites during molting has not been analyzed <ns0:ref type='bibr' target='#b15'>(Han and Wang, 2000)</ns0:ref>. Here, I collected many exuviae of Ovalocephalus tetrasulcatus (Phacopida, Pliomeridae) from the Upper Ordovician in central Hubei. Some of the specimens were found with two or three in a line, or preserved partly or even completely overlapping; these patterns are regarded as representing the active behavior of trilobites during molting, and may be related to the herding behavior of some trilobites. They provide new material for understanding exuvial techniques of trilobites, and the behavior of trilobites when molting.</ns0:p></ns0:div>
<ns0:div><ns0:head>Materials & Methods</ns0:head><ns0:p>All specimens were collected from the Upper Ordovician Linhsiang Formation of the Daozimiao section in Jinshan County, central Hubei (GPS: N 30°59′51.49″, E113°06′26.04″) (Fig. <ns0:ref type='figure' target='#fig_0'>1</ns0:ref>). The Linhsiang Formation is widely distributed in the middle Yangtze region, and is mainly composed of yellow-green calcareous mudstones with a few siliceous mudstones in the Daozimiao section, but differs from the nodular muddy limestones of the Linhsiang Formation in the type section in Linxiang, Hunan Province. In the Daozimiao section, the Linhsiang Formation is conformably underlain by the muddy limestones of the Upper Ordovician Pagoda Formation and overlain by the graptolite shales of the Upper Ordovician Wufeng Formation. The 1.5-m-thick calcareous mudstones from the top of the Linhsiang Formation (Fig. <ns0:ref type='figure' target='#fig_0'>1d</ns0:ref>) yield an abundant and diverse trilobite fauna, including members of the Metagnostidae, Cyclopygidae, Phillipsinellidae, Encrinuridae, Telephinidae, Raphiophoridae, Cheiruridae, Dionididae, Trinucleidae, Asaphidae, Pliomeridae, and Remopleurididae. In addition, the Foliomena fauna (brachiopods) <ns0:ref type='bibr' target='#b31'>(Zhan and Jin, 2005)</ns0:ref>, ostracods, echinoderms, machaeridians, and trace fossils are also found in the same horizon. Based on the trilobite assemblage, the age of the top of the Linhsiang Formation is constrained to the middle Kaitian (early Ashgill) <ns0:ref type='bibr' target='#b32'>(Zhou et al., 2016)</ns0:ref>.</ns0:p><ns0:p>The 1.5-m-thick trilobite-bearing calcareous mudstones are homogeneous without bedding structures. There are no cross-bedding and graded bedding in the longitudinal sections of the rocks, but a small amount of authigenic pyrites preserved in that (Fig. <ns0:ref type='figure' target='#fig_1'>2a-d</ns0:ref>). These evidence suggest the calcareous mudstones formed in a relatively calm and deepwater environment, which is consistent with the results of the characteristics of the brachiopod association in the same interval <ns0:ref type='bibr' target='#b31'>(Zhan and Jin, 2005)</ns0:ref>. Ovalocephalus tetrasulcatus (Kielan 1960) is the only pliomerid trilobite in the Linhsiang Formation of Jingshan. Ovalocephalus is largely restricted to peri-Gondwana, but is widely distributed in the Ordovician of China <ns0:ref type='bibr' target='#b34'>(Zhou et al., 2010)</ns0:ref>. In the Linhsiang Formation of Jingshan, most specimens of Ovalocephalus tetrasulcatus are articulated or nearly articulated exoskeletons, as well as enrolled specimens. This pattern reflects that the trilobites were not transported by current before burial, and were buried in situ. Although some biotic burrows can be observed in the calcareous mudstones (Fig. <ns0:ref type='figure' target='#fig_1'>2e, f</ns0:ref>), no trace of burrows can be found near the trilobites (Fig. <ns0:ref type='figure' target='#fig_1'>2b, c</ns0:ref>), which can ruled out the possibility that trilobites were affected by biotic disturbance before burial or preserved in burrows.</ns0:p><ns0:p>I collected more than 100 specimens containing articulated or nearly articulated exoskeletons from the calcareous mudstones at the top of the Linhsiang Formation of Jingshan, including 13 specimens that showed two exoskeletons end to end or overlapping one another, and three specimens that showed three exoskeletons end to end or two of them overlapping one another, which obviously differed from isolated single exoskeletons preserved in the same interval. The fossils in Fig. <ns0:ref type='figure' target='#fig_2'>3</ns0:ref> were whitened with magnesium oxide powder, and all photographs were captured using a Nikon D5100 camera with a Micro-Nikkor 55 mm F3.5 lens. The axial azimuth measurements of the trilobites and the rose chart were completed in CorelDRAW X7.</ns0:p></ns0:div>
<ns0:div><ns0:head>Results</ns0:head><ns0:p>In the present work, complete exoskeletons include all specimens with articulated cephala, thoraces, and pygidia or enrolled exoskeletons, indicating that they are fossilized carapaces of Ovalocephalus tetrasulcatus. Among these nearly complete exoskeletons, a portion of them are articulated thoraces and pygidia, but with the cephala separated and preserved nearby (Fig. <ns0:ref type='figure' target='#fig_2'>3a</ns0:ref>), which are considered exuviae <ns0:ref type='bibr' target='#b15'>(Han and Wang, 2000)</ns0:ref>. Another kind has librigenae separated from the cranidia, but the cranidia and thoracopyga are still articulated, or the cranidia are separated from the thoraces and slightly rotated. These specimens include those with two librigenae separated from cranidia, but still preserved nearby (Fig. <ns0:ref type='figure' target='#fig_2'>3b</ns0:ref>), and a portion of these librigenae were inverted or rotated (Fig. <ns0:ref type='figure' target='#fig_2'>3d</ns0:ref>); in addition, some specimens have one librigena separated from the cranidium and inverted, but another still in situ (Fig. <ns0:ref type='figure' target='#fig_2'>3c</ns0:ref>). Both types of inverted or rotated librigenae separated from the cranidium are similar to the exuvial mode of many trilobites <ns0:ref type='bibr'>(Henningsmoen,1975;</ns0:ref><ns0:ref type='bibr' target='#b18'>McNamara and Rudkin, 1984)</ns0:ref>, indicating that they are most likely exuviae of Ovalocephalus tetrasulcatus, rather than corpses that were broken up by bottom current or organism, and the natural decomposition of the corpses. This indicates that there is more than one exuvial mode in Ovalocephalus tetrasulcatus; i.e., separation of the cephalon from the thoracopygon, separation of the librigenae from the cranidium, or both existing at the same time. A similar diversity of exuvial patterns has been found in other phacopid trilobites <ns0:ref type='bibr' target='#b2'>(Brandt, 1993;</ns0:ref><ns0:ref type='bibr' target='#b3'>Budil and Bruthansová, 2005;</ns0:ref><ns0:ref type='bibr' target='#b7'>Chen, 2011)</ns0:ref>.</ns0:p><ns0:p>In the sixteen specimens with two or three exoskeletons partly superimposed or end to end, all the specimens show separation of the cephalon or librigenae, and the separated cephala or librigenae are still preserved near the thoracopyga (Fig. <ns0:ref type='figure' target='#fig_3'>4</ns0:ref>), indicating that all of these specimens are exuviae of Ovalocephalus tetrasulcatus. The central axis of the most frontal Ovalocephalus tetrasulcatus in the exuvial clusters was used as the directrix to calculate the axial azimuth of the posterior trilobites. The results showed that there was an obvious dominant orientation; that is, the axial azimuth of the trilobites in these clusters had obvious consistency (Fig. <ns0:ref type='figure' target='#fig_4'>5</ns0:ref>), indicating they were not the result of accidental burial events. The separated librigenae preserved near the cranidia may also preclude them from being the result of postmortem transport. Therefore, these exuviae are inferred to have been caused by the active behavior of trilobites.</ns0:p></ns0:div>
<ns0:div><ns0:head>Discussion</ns0:head><ns0:p>Possible causes of coupled exuviae Preservation of two Ovalocephalus tetrasulcatus exoskeletons preserved together is similar to the preservation of an arthropod in the middle of the act of molting (García-Bellido and Collins, <ns0:ref type='formula'>2004</ns0:ref>), but the latter case includes one corpse and one exuvia, whereas all these specimens from Jingshan are exuviae, without corpses; thus, the possibility that the Ovalocephalus tetrasulcatus individuals were buried when molting can be excluded. Several cases of queuing of trilobites have been reported in the past, but most of these case are corpse fossils, and the number of trilobites is typically more than three; thus, they are regarded as representing unexpected burial during the migration of trilobites <ns0:ref type='bibr' target='#b23'>(Radwański et al., 2009;</ns0:ref><ns0:ref type='bibr' target='#b0'>Błażejowski et al., 2016;</ns0:ref><ns0:ref type='bibr' target='#b28'>Vannier et al., 2019)</ns0:ref>. <ns0:ref type='bibr' target='#b6'>Chatterton and Fortey (2008)</ns0:ref> reported trilobites aligned for molting, but they were preserved in burrows. In Jingshan, there are no biological burrow near the Ovalocephalus tetrasulcatus specimens (Fig. <ns0:ref type='figure' target='#fig_1'>2b, c</ns0:ref>), and the number of exuviae per specimens is only two or three, in contrast with the former. Since <ns0:ref type='bibr' target='#b27'>Speyer and Brett (1985)</ns0:ref> first reported synchronous ecdysis in Middle Devonian phacopids in New York, this behavior has been identified in some other trilobites <ns0:ref type='bibr' target='#b20'>(Paterson et al. 2007)</ns0:ref>, usually with many trilobites concentrated in a certain place to shed their shells, which is thought to be related to gregarious behavior and is probably associated with the copulation and reproduction of trilobites <ns0:ref type='bibr' target='#b26'>(Speyer, 1990)</ns0:ref>. A similar pattern may exist in Ovalocephalus tetrasulcatus, and pairs of exuviae are relatively easy to understand for mating behavior after molting. However, there are obvious differences of sizes in some exuviae in the same cluster (Fig. <ns0:ref type='figure' target='#fig_3'>4d</ns0:ref>), or in different clusters (Fig. <ns0:ref type='figure' target='#fig_3'>4b, e</ns0:ref>). Moreover, there are also clusters with three exuviae (Fig. <ns0:ref type='figure' target='#fig_3'>4g-i</ns0:ref>); thus it is difficult to explain these exuvial clusters as the result of molting-mating behavior.</ns0:p><ns0:p>I speculate that these specimens may reflect a particular exuvial behavior of Ovalocephalus tetrasulcatus; because superposition (Fig. <ns0:ref type='figure' target='#fig_3'>4b</ns0:ref>) and even almost complete overlapping (Fig. <ns0:ref type='figure' target='#fig_3'>4c, hi</ns0:ref>) exist in the exuvial clusters, it is unlikely that two or three trilobites shed their shells synchronously. Instead, they may have shed their shells in lines; that is, after one trilobite finished molting, the others shed their shells as well. Alternatively, perhaps the first trilobite finished molting and left an empty shell, and then later, other trilobites came to the same place to molt. The consistency of the long axes of the exuviae may be related to the seabed topography at that time; for example, molting could have occurred in a narrow shallow gully or on a gentle slope. This is may have been somewhat similar to the molting of extant cicadas on trees, where they occasionally molt in line or overlapping <ns0:ref type='bibr' target='#b1'>(Bobo, 2016;</ns0:ref><ns0:ref type='bibr'>Muchen, 2016;</ns0:ref><ns0:ref type='bibr' target='#b36'>Zhu and Wang, 2017)</ns0:ref>. Implications for the behavioral strategy of trilobites Arthropods are weak during and after molting, which leaves them vulnerable to predators and even other members of their species. Living shrimps and crabs usually hide in rock crevices or water plants for molting. Trilobites would also have needed a quiet and undisturbed environment when they shed their shells <ns0:ref type='bibr' target='#b16'>(Henningsmoen, 1975;</ns0:ref><ns0:ref type='bibr' target='#b13'>Han, 2006)</ns0:ref>. For example, some trilobites used the empty shells of cephalopods and gastropods as shelter for molting <ns0:ref type='bibr' target='#b4'>(Chatterton, 1971;</ns0:ref><ns0:ref type='bibr' target='#b2'>Brandt, 1993;</ns0:ref><ns0:ref type='bibr' target='#b9'>Davis et al., 2001;</ns0:ref><ns0:ref type='bibr' target='#b37'>Zong et al., 2016)</ns0:ref>, some even shed their shells under the empty shells of larger trilobites <ns0:ref type='bibr' target='#b12'>(Gutiérrez-Marco et al., 2009)</ns0:ref>, and others molted in burrows of other animals <ns0:ref type='bibr' target='#b5'>(Chatterton et al., 2003;</ns0:ref><ns0:ref type='bibr' target='#b6'>Chatterton & Fortey, 2008)</ns0:ref>. The Upper Ordovician strata in Jingshan have yielded a large number of cephalopods, and I also found nautiloid fossils in the Linhsiang Formation. These predators would have threatened great harm to molting trilobites, and the trilobites therefore would have needed to find safe places for ecdysis. However, when there was no shelter on the seafloor, or insufficient space to hide, possibly, they may have chosen to follow congeneric trilobites nearby to molt; alternatively, the remaining exuviae of other trilobites might have suggested that the location was suitable or safe for molting, thus attracting the later trilobites to molt in the same position, thus forming exuvial clusters. It is worth noting that there are only two or three exuviae in each cluster, which may be because the posterior trilobites adopted the principle of proximity when choosing the molting site. This is similar to the herding behavior of animals in crisis situations, indicating that Ovalocephalus tetrasulcatus would take the initiative to choose the nearest safe area to carry out vulnerable life activities.</ns0:p></ns0:div>
<ns0:div><ns0:head>Conclusions</ns0:head><ns0:p>The exuviae of Ovalocephalus tetrasulcatus from the Upper Ordovician Linxiang Formation presented two types: separation of the librigenae from the cranidium, and separation of the cephalon from the thoracopygon, reflecting the diversity of the exuvial modes. Some specimens have two or three exuviae arranged end to end, and some have partly and even completely superimposed exuviae together in clusters. The preserved patterns and burial conditions indicate that these specimens are products of the active behavior of trilobites, rather than mechanical transport by currents, unexpected burial in the middle of the act of molting, or collective molting before mating. The preserved patterns and overlapping phenomena of the exuviae indicate that these clusters were formed by two or three trilobites lining up to shed their shells in a long and narrow feature of seafloor topography. They likely represent the behavioral response that the trilobites chose to follow in certain risky life activities, indicating that herding behavior existed in these Ordovician trilobites. </ns0:p></ns0:div><ns0:figure xml:id='fig_0'><ns0:head>Figure 1</ns0:head><ns0:label>1</ns0:label><ns0:figDesc>Figure 1</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_1'><ns0:head>Figure 2</ns0:head><ns0:label>2</ns0:label><ns0:figDesc>Figure 2</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_2'><ns0:head>Figure 3</ns0:head><ns0:label>3</ns0:label><ns0:figDesc>Figure 3</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_3'><ns0:head>Figure 4</ns0:head><ns0:label>4</ns0:label><ns0:figDesc>Figure 4</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_4'><ns0:head>Figure 5</ns0:head><ns0:label>5</ns0:label><ns0:figDesc>Figure 5</ns0:figDesc></ns0:figure>
</ns0:body>
" | "Reviewer 1 (Nigel Hughes)
Basic reporting
Review of “Coupled exuviae of the Ordovician Ovalocephalus (Pliomeridae, Trilobita) in south China and its behavioral implications (#48931)”
By: Ruiwen Zong
Reviewer: Nigel Hughes
The paper presents data that in the associations of trilobites studied there is a clear tendency for alignment of long axes in clusters of differently sized mature individuals that include several specimens, but rarely more than 4. Many of these individuals show patterns of partial disarticulation, mostly either at the “neck suture” or at the facial sutures. As far as I am aware this particular style of trilobite sclerite association is new and therefore of some interest: it joins a growing literature of case study descriptions that collectively build the descriptive basis of trilobite sclerite disposition. So, in those terms it certainly has merit as a study.
Where I am less keen is the interpretative part. PeerJ is a short format journal, and the paper does not appear to indicate a store of supplementary information that documents and supports the contentions made herein. Such information needs to be available. The paper does refer to some other studies, but it’s unclear whether these provide the specific documentation needed. For example:
1. More information of sedimentology (perhaps on an online supplement?) is necessary. Sedimentary evidence is insufficient to assess represent events taking place in burrows either with or without a roof) or on the seafloor at sediment water interface, and no data is presented on the range of directions that the various clusters pointed towards – only that of the internal variance in orientation within clusters. The sedimentary log provides is fine for orientating us stratigraphically but tells us almost nothing of the details of the depositional regime. The bed yielding the clusters is ~1.5 m thick. What is it comprised of? Are there multiple separate depositional events? What is the relationship between episodes of deposition and the clusters? Here are they located specifically within the bed? Do different clusters on the same bedding surface align? The paper speculates that clusters accumulated on the sediment surface over a period of time. Is there any evidence from sedimentology to corroborate this?
No. 1: Yes, I added a new Figure and some sentences to discuss the sedimentary environment and taphonomy of coupled Ovalocephalus, i.e., ‘The 1.5-m-thick trilobite-bearing calcareous mudstones are homogeneous without bedding structures. There are no cross-bedding and graded bedding in the longitudinal sections of the rocks, but a small amount of authigenic pyrites preserved in that (Fig. 2a–d). These evidence suggest the calcareous mudstones formed in a relatively calm and deepwater environment, which is consistent with the results of the characteristics of the brachiopod association in the same interval (Zhan and Jin, 2005). In other words, these trilobites were not transported by the current before burial. Although some biotic burrows can be observed in the calcareous mudstones (Fig. 2e, f), no trace of burrows can be found near the trilobites (Fig. 2b, c), which can ruled out the possibility that trilobites were affected by biotic disturbance before burial or preserved in burrows.’.
I speculate that clusters accumulated on the sediment surface over a period of time, because superposition (Fig. 4b) and even almost complete overlapping (Fig. 4c, h–i) exist in the exuvial clusters, it is unlikely that two or three trilobites shed their shells synchronously. However, this time interval may be very short, such as a few hours or a few days, so there may be no sedimentary record left.
2. As Whittington (1990) pointed out, making the case for trilobite sclerite associations being the product of ecdysis is a challenge. In my view, there is a recent tendency towards tolerating relatively lax criteria for the confident recognition of exuvae. For example, this paper treats the study of Brandt (1993) as an accepted authority – I view that study as flawed in several ways. In my view it’s possible to argue that almost any association of sclerites derived from the same individual is the result of ecdysis – but that doesn’t mean that all actually are the result of that process. As Whittington (1990) made clear, there have to be specific arguments from the disposition of sclerites that make not only ecdysis a candidate process for the disposition, but also one that is more likely than the sum of alternative explanations.
What’s novel (and thus genuinely interesting) here is the association of sclerites derived from individuals and the alignment of multiple individuals, because these particular associations place limits on the processes that can reasonably explain them. What I was hoping for in this paper was, as in Whittington 1990, constraints provided by the alignment pattern itself that independently supports ecdysis as the explanation for the arrangement of the sclerites. I don’t see that here. What if these were animals clustered in burrows that then decayed or were scavenged (see for example Hughes and Cooper, 1999 Journal of Paleontology for an example of scavenging of carcasses)? The pattern of sclerite disposition reported are quite diverse – not the stereotyped associations whose posture can only be reasonably be explained by movements associated with ecdysis. For example, Fig 2 shows a whole range of associations, including a partly enrolled specimen. Of course, the diversity of postures could be because Brandt (and Daley and followers) are right and trilobites did show an unusual diversity of ecdysial strategies atypical compared to living arthropods. But other explanations – such as the disturbance of carcasses (or exuvae for that matter) by various agents (physical processes or biological agents) are possible. How can it be argued that the weight of evidence favors a diversity of modes of ecdysis?
No. 2: Yes, more than one exuvial mode in some trilobites maybe controversial. For the disarticulated exoskeletons of Ovalocephalus in the Linhsiang Formation (Figure 2, now Figure 3 in revised MS), based on the evidence presented in the MS, as well as the additional evidence added in the revised MS, other possible causes for the disarticulated exoskeletons can be ruled out. More specifically, there are no cross-bedding and graded bedding in the longitudinal sections of the Ovalocephalus-bearing calcareous mudstones, but a small amount of authigenic pyrites preserved in that (Fig. 2a–d), which suggest these calcareous mudstones formed in a relatively calm and deepwater environment, i.e., these trilobites were not transported by the current before burial. Moreover, no trace of burrows can be found near the trilobites (Fig. 2b, c), which can ruled out the possibility that trilobites were affected by biotic disturbance before burial or preserved in burrows. Some exoskeletons with inverted librigenae preserved in the environment without the physical and/or biological interference, there are not consistent with the natural decomposition of the corpses. So, I think these disarticulated exoskeletons all belong the exuviae of Ovalocephalus, and that, a similar diversity of exuvial patterns has been found in other phacopid trilobite.
What to say overall? I am very not familiar with PeerJ, but I certainly see a useful publication emerging from this work. It will need careful documentation of sedimentology and taphonomy of the associations, and I’d like to see at least a good portion of that work in the main text of the publication, rather than relegated to the supplementary materials because I think the documentation is critical to interpretation. If PeerJ supports that format then I think a revised version should be encouraged. If not, I would suggest a longer manuscript submitted to a discipline specific journal.
Experimental design
see above
Validity of the findings
see above
Comments for the Author
see above
Annotated manuscript
The reviewer has also provided an annotated manuscript as part of their review:
Comments in PDF:
Line 17: ecdysis is a process - it can't 'capture' anything. The preservation of exuvae might capture information. You mean 'and preservation of its action often captured
No. 3: Yes, I revised this sentence.
Line 20-21:This sounds like you assume these sclerite dispositions are exuvae and then interpret behavior. But this is not how an argument can be made - a recurrent pattern has explainable by a unique and particular process.
No. 4: Yes, I revised this sentence.
Line 22: I don't understand what you mean here by this phase
No. 5: This sentence was deleted according to the review of Reviewer 2.
Line 26: trilobites molted multiple times in their lives. It was surely a process that involved risk but to call is a 'crisis' is overly dramatic
No. 6: Yes, I revised this sentence, i.e., ‘representing a behavioral response of the trilobites to choose a nearby safe zone during some risky life activities.’.
Line 28: delete ‘will’, replace ‘hinder’ with ‘hinders’
No. 7: Yes, done.
Line 36: This is a point-of-view but not universally accepted
No. 8: As I answered in the No. 2, more than one exuvial mode in some trilobites maybe controversial. However, for the disarticulated exoskeletons of Ovalocephalus in the Linhsiang Formation (Figure 2, now Figure 3 in revised MS), based on the evidence presented in the MS, as well as the additional evidence added in the revised MS, other possible causes for the disarticulated exoskeletons can be ruled out.
Line 50: this sentence is not complete
No. 9: Yes, I revised it.
Line 125: Collins I believe
No. 10: Yes, done.
Reviewer 2 (Brian Chatterton)
Basic reporting
Review of “Coupled exuviae of the Ordovician Include” (Pliomeridae, Trilobita) in South China and its behavioural implications by Ruiwen Zong.
1) English language, while clearly not written by an anglophone, is on the whole easy to follow (but see detailed suggested improvements below).
2) The author has done a fairly thorough job of researching the topic, and that is shown by his knowledge of the relevant literature.
3) The figures are of reasonable quality and necessary for the paper.
4) I have a number of suggestions that I hope would serve to improve the paper slightly, and perhaps change the conclusions slightly.
5) The paper is interesting and worthy of publication.
6) Raw data are supplied.
7) The paper includes original research, and improves our understanding of ancient events and biology.
8) Methods are described with sufficient detail and information.
9) On whole, findings are clear, original, but I would suggest that some minor but significant changes should be made to them (see detailed comments below).
10) Manuscript provides interesting new data and ideas.
Detailed comments on MS:
Title: Ovalocaphalus should be in a different font to rest of title to show that it is a binomial name.
No. 11: Yes, done.
Line 16: change “reflecting” to “preserving”.
No. 12: Yes, done.
Lines 21-22: remove “,combined with the overlying phenomenon,” as it is confusing.
No. 13: Yes, done.
Line 24: change “the herding behavior of trilobites” to “the postulated herding behavior of some trilobites”.
No. 14: Yes, done.
Line 25: I do not think that this work has demonstrated what is staled here: “representing a behavioral response of the trilobites to choose a nearby safe zone in a crisis situation”. See more detailed comments below.
No. 15: Yes, I revised this sentence according to the review of Reviewer 1, i.e., ‘representing a behavioral response of the trilobites to choose a nearby safe zone during some risky life activities.’.
Line 27, trilobites do not have an exoskeleton of just chitin, it is mainly calcite with various amounts of Mg, probably with a chitinous outer layer (that is true ion many other arthropods (crabs, lobsters, etc.), so perhaps
No. 16: Yes, I replaced it with ‘biomineralized or chitinous exoskeleton’.
Line 35: the thoracopygon is the term used for the thorax and the pygidium (see “morphological terms” in the latest Treatise on trilobites, Whittington et al., 1997, p. 329).
No. 17: Yes, I replaced all ‘thoracopygidium (thoracopygidia)’ with ‘thoracopygon (thoracopyga)’ in the MS.
Lines 45-49. Not really relevant to the argument, could be left out.
No. 18: Yes, I deleted them.
Line 52: change “literatures” to ‘literature’. Literature is usually used as a collective noun.
No. 19: Yes, done.
Liner 58: change “”related to the herding behavior of trilobites” to :’herding behavior of some trilobites” - since we do not know that all trilobites exhibited this behavior (some probably did not).
No. 20: Yes, done.
Lines 59-60” Change to read “They provide new material for understanding exuvial techniques of trilobites, and the behavior of trilobites when melting”.
No. 21: Yes, done.
Line 70: change “trilobites with the high abundance” to “ an abundant and diverse trilobite fauna”
No. 22: Yes, done.
Line 74: change “machaerids” to “machaeridians”
No. 23: Yes, done.
Line 78-80: Change to read “Ovalocephalus tetrasulcatus (Kielan, 1960) is the only pliomerid trilobite in the Linhsiang Formation. Ovalocephalus is largely restricted to peri-Gondwana, but is widely distributed in the Ordovician of China (Zhou et al., 2010).
No. 24: Yes, done.
Line 87 and 88: Change “together” to “one another”.
No. 25: Yes, done.
Line 89: Change “same horizon” to “ same interval”.
No. 26: Yes, done.
Line 94: Change “In the above materials” to “In the present work”, and change “included” to “include”
No. 27: Yes, done.
Line 95: change “corpses” to “carapaces” or “exoskeletons”. If they are molts it is not appropriate to call them “corpses” as if they were carcasses or cadavers.
No. 28: Yes, done.
Line 102: change “specimens has” to “specimens have”
No. 29: Yes, done.
Line 113” change “thoracopygidia” to “thoracopyga” (see reference above).
No. 30: Yes, done.
Line 124-125: “Remove first “together” and change to read “Preservation of two Ovalocephalus tetrasulcatus exoskeletons preserved together is similar to the preservation of an arthropod in …”
No. 31: Yes, done.
Line 133: either “are no burrows” or “is no burrow”. How about “no evidence of burrows”, as burrows tree often very difficult to see or not preserved in strata that are of a single composition (particularly pure mudstones).
No. 32: I added a new Figure (Figure 2) in the MS. There are some borrows in the trilobite-bearing calcareous mudstones from the top of the Linhsiang Formation, and they are easily recognized in the bedding surface or the longitudinal sections of rocks (Fig. 2e, f). However, the burrows can not be found near the coupled Ovalocephalus, or in the longitudinal sections of coupled Ovalocephalus-bearing calcareous mudstones (Fig. 2b, c). So I ruled out the possibility that these coupled Ovalocephalus were preserved in burrows.
Line 143: change “thus, so, it” to “thus it”
No. 33: Yes, done.
Lines 145-147. I agree that it makes no sense to regard them as the products of a single animal returning to the same place to molt. We do not know how many times trilobites were able to reproduce or whether larger, older trilobites were able to reproduce with younger animals (as occurs in many species today). We do not think that they were like insects and copulated once and then died (although that is a possible scenario for some forms). We also know that trilobites, like any other species probably had a bell curve relative to size (with large and small individuals of the same age and degree of sexual maturity - we certainly have some examples that show a size range in most molt clusters/instars - see my chapter in the treatise and a number of later papers, often with Catherine Cronier as an author).
No. 34: Yes, I deleted this part.
Line 153: change “latter: to “later”
No. 35: Yes, done.
Lines 156-157: I wonder why you have not considered sea floor currents as a reason for the trilobite molts pointing in the same direction (your effective rose diagram). It makes much more sense that the trilobites were orienting themselves relative to prevailing sea floor currents than a need to invoke linear depressions in an environment where the currents were not strong enough to move the exuviae. The currents could still have been strong enough that a trilobite which was about to lose its exoskeleton (so it would have been slow and weak, and subject to predation) might wish to minimize its exposure to the current (by facing into it to reduce drag forces). I know that at the substrate surface (the base of the hydrodynamic boundary layer) there was practically no current, but unless the trilobite burrowed mostly into the sediment before molting it must have been high enough to have been affected by sea floor currents. We also know that sea floor currents strong enough to form ripple marks can occur at depths of thousands of meters in some areas of the ocean floor (probably not in regions that deposit mudstones?).
No. 36: I added a new Figure and some sentences to discuss the sedimentary environment and taphonomy of coupled Ovalocephalus, i.e., ‘The 1.5-m-thick trilobite-bearing calcareous mudstones are homogeneous without bedding structures. There are no cross-bedding and graded bedding in the longitudinal sections of the rocks, but a small amount of authigenic pyrites preserved in that (Fig. 2a–d). These evidence suggest the calcareous mudstones formed in a relatively calm and deepwater environment, which is consistent with the results of the characteristics of the brachiopod association in the same interval (Zhan and Jin, 2005). In other words, these trilobites were not transported by the current before burial. Although some biotic burrows can be observed in the calcareous mudstones (Fig. 2e, f), no trace of burrows can be found near the trilobites (Fig. 2b, c), which can ruled out the possibility that trilobites were affected by biotic disturbance before burial or preserved in burrows.’.
In addition, the separated librigenae preserved near the cranidia may also preclude them from being the result of postmortem transport.
Line 163. I suggested that phacopids probably selected nautiloid shells as melting sites as far back as 1971 in Palaeontographica A (for the Devonian of Australia).
No. 37: Yes, I added this reference, i.e., Chatterton BDE. 1971. Taxonomy and ontogeny of Siluro–Devonian trilobites from near Yass, New South Wales. Palaeontographica, Abteilung A 137: 1–108.
Line 167: change “nautilus: to “nautiloid” (could not be Nautilus in the Ordovician!)
No. 38: Yes, done.
Line 186: Why do you think that the trilobites reproduced before molting. If it was at all an active behavior, it would be easier to accomplish while they still had an exoskeleton, and then they could have melted afterwards. Limulus and related forms certainly moulted with the exoskeleton in place, as shown in modern studies and Jurassic trace fossil studies.
No. 39: Molting before mating? It is ‘molting before mating’ in Line 186. The view of ‘molting before mating’ was raised by Speyer and Brett (1985) and Speyer (1990). I don’t think these coupled Ovalocephalus is the product of molting–mating behavior, and this view has been discussed in the ‘Possible causes of coupled exuviae’ section.
Lines 188-190. I am not sure that I like these closing comments. You have not demonstrated that there were any “ crisis situations” unless of course you are referring to attacks by nautiloids. Also, getting together in twos or threes (quite possibly for reproduction” does not really, to my mind, mean “herding behavior”
No. 40: Yes, I deleted the ‘crisis situations’, and replaced it with ‘certain risky life activities’.
In these coupled Ovalocephalus, there are obvious differences of sizes in some exuviae in the same cluster (Fig. 4d), or in different clusters (Fig. 4b, e). Moreover, there are also clusters with three exuviae (Fig. 4g–i). More importantly, superposition (Fig. 4b) and even almost complete overlapping (Fig. 4c, h–i) exist in the exuvial clusters, it is unlikely that two or three trilobites shed their shells synchronously. So, I think it is difficult to explain these exuvial clusters as the result of molting before reproduction.
I have not read through the references carefully, but would point out that on line 249, it is Radwanski and Radwanska, 2009) - I looked at the original paper.
No. 41: Yes, I revised this reference, and checked other references.
I enjoyed reading this paper.
Experimental design
Ok
Validity of the findings
Slight changes suggested
Comments for the Author
I enjoyed reading this. I hope that you take my suggested improvements seriously. They are meant to be constructive and to improve the work.
" | Here is a paper. Please give your review comments after reading it. |
9,814 | "<ns0:abstract xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:p>Ecdysis was a vital process during the lives of trilobites, in addition to preserving the morphological changes in trilobite ontogeny, the preservation of its action often captured interesting behavioral information. Abundant exuviae of Ovalocephalus tetrasulcatus are preserved in the Ordovician strata in central Hubei, China, and some of them are arranged with two or three together end to end or superimposed. The preserved patterns and burial conditions indicate that these specimens were caused by the active behavior of trilobites.</ns0:p><ns0:p>It is speculated that these exuvial clusters were formed by two or three trilobites in line to molt; that is, after one trilobite finished molting, other trilobites molted in front of, behind, or overlying the previously molted shells. This ecdysis strategy is interpreted as related to the postulated herding behavior of some trilobites, representing a behavioral response of the trilobites to choose a nearby safe zone during some risky life activities.</ns0:p></ns0:div>
</ns0:abstract>
<ns0:body xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:head>Introduction</ns0:head><ns0:p>The biomineralized or chitinous exoskeleton of arthropods hinders the growth of their bodies; therefore, individuals must shed their old shells many times during growth and development <ns0:ref type='bibr' target='#b19'>(Moussian, 2013;</ns0:ref><ns0:ref type='bibr' target='#b8'>Daley and Drage, 2016)</ns0:ref>. Trilobites, as an extinct group of arthropods, also needed to shed their shells as they grew <ns0:ref type='bibr' target='#b10'>(Fortey, 2014)</ns0:ref>. Different trilobites exhibited different molting techniques <ns0:ref type='bibr' target='#b16'>(Henningsmoen, 1975)</ns0:ref>; most trilobites shed their shells through separating the librigenae from the cranidium <ns0:ref type='bibr' target='#b18'>(McNamara and Rudkin, 1984;</ns0:ref><ns0:ref type='bibr' target='#b30'>Whittington, 1990)</ns0:ref>, whereas for trilobites with librigenae fused with the cranidium, separation of the cephalon from the thoracopygon usually occurred during molting <ns0:ref type='bibr' target='#b25'>(Speyer, 1985;</ns0:ref><ns0:ref type='bibr' target='#b29'>Wang and Han, 1997)</ns0:ref>, and some genera had multiple exuvial modes <ns0:ref type='bibr' target='#b2'>(Brandt, 1993;</ns0:ref><ns0:ref type='bibr' target='#b3'>Budil and Bruthansová, 2005)</ns0:ref>. In addition to ecdysis reflecting the ontogenetic development process of trilobites, some exceptionally preserved trilobite specimens also contain behavioral information. For example, some trilobites shed their shells by hiding in empty shells or burrows of other animals <ns0:ref type='bibr' target='#b9'>(Davis et al., 2001;</ns0:ref><ns0:ref type='bibr' target='#b5'>Chatterton et al., 2003;</ns0:ref><ns0:ref type='bibr' target='#b6'>Chatterton and Fortey, 2008;</ns0:ref><ns0:ref type='bibr' target='#b37'>Zong et al., 2016)</ns0:ref>, and even molted infaunally <ns0:ref type='bibr' target='#b24'>(Rustán et al., 2011)</ns0:ref>, reflecting the hiding behavior of trilobites. Some phacopids may also have exhibited asymmetric behaviors during molting <ns0:ref type='bibr' target='#b38'>(Zong and Gong, 2017)</ns0:ref>. Other trilobites collectively shed their shells, which may have been related to molting-mating behavior <ns0:ref type='bibr' target='#b27'>(Speyer and Brett, 1985;</ns0:ref><ns0:ref type='bibr' target='#b26'>Speyer, 1990)</ns0:ref>.</ns0:p><ns0:p>South China is an important area for trilobite fossils, and abundant Ordovician trilobites have been collected from Hubei Province <ns0:ref type='bibr' target='#b17'>(Lu, 1975;</ns0:ref><ns0:ref type='bibr' target='#b35'>Zhou and Zhen, 2008)</ns0:ref>. However, there are few reports on exuvial specimens and correlation research. Only photographs of exuvial specimens were attached to the identification of genera and species in the paleontological literature; these exuvial specimens have not been systematically described, their patterns have not been classified and explained, and the behavioral strategy of these trilobites during molting has not been analyzed <ns0:ref type='bibr' target='#b15'>(Han and Wang, 2000)</ns0:ref>. Here, I collected many exuviae of Ovalocephalus tetrasulcatus (Phacopida, Pliomeridae) from the Upper Ordovician in central Hubei. Some of the specimens were found with two or three in a line, or preserved partly or even completely overlapping; these patterns are regarded as representing the active behavior of trilobites during molting, and may be related to the herding behavior of some trilobites. They provide new material for understanding exuvial techniques of trilobites, and the behavior of trilobites when molting.</ns0:p></ns0:div>
<ns0:div><ns0:head>Materials & Methods</ns0:head><ns0:p>All specimens were collected from the Upper Ordovician Linhsiang Formation of the Daozimiao section in Jinshan County, central Hubei (GPS: N 30°59′51.49″, E113°06′26.04″) (Fig. <ns0:ref type='figure'>1</ns0:ref>). The Linhsiang Formation is widely distributed in the middle Yangtze region, and is mainly composed of yellow-green calcareous mudstones with a few siliceous mudstones in the Daozimiao section, but differs from the nodular muddy limestones of the Linhsiang Formation in the type section in Linxiang, Hunan Province. In the Daozimiao section, the Linhsiang Formation is conformably underlain by the muddy limestones of the Upper Ordovician Pagoda Formation and overlain by the graptolite shales of the Upper Ordovician Wufeng Formation. The 1.5-m-thick calcareous mudstones from the top of the Linhsiang Formation (Fig. <ns0:ref type='figure'>1d</ns0:ref>) yield an abundant and diverse trilobite fauna, including members of the Metagnostidae, Cyclopygidae, Phillipsinellidae, Encrinuridae, Telephinidae, Raphiophoridae, Cheiruridae, Dionididae, Trinucleidae, Asaphidae, Pliomeridae, and Remopleurididae. In addition, the Foliomena fauna (brachiopods) <ns0:ref type='bibr' target='#b32'>(Zhan and Jin, 2005)</ns0:ref>, ostracods, echinoderms, machaeridians, and trace fossils are also found in the same horizon. Based on the trilobite assemblage, the age of the top of the Linhsiang Formation is constrained to the middle Katian (early Ashgill) <ns0:ref type='bibr' target='#b33'>(Zhou et al., 2016)</ns0:ref>.</ns0:p><ns0:p>The 1.5-m-thick trilobite-bearing calcareous mudstones are homogeneous, without bedding structures. There are no cross-bedding and graded bedding visible in the longitudinal sections of the rocks, but a small amount of authigenic pyrites is preserved in this interval (Fig. <ns0:ref type='figure' target='#fig_1'>2a-d</ns0:ref>). This evidence suggests the calcareous mudstones formed in a relatively calm and deepwater environment, which is consistent with the results of the characteristics of the brachiopod association found in the same interval <ns0:ref type='bibr' target='#b32'>(Zhan and Jin, 2005)</ns0:ref>. Ovalocephalus tetrasulcatus (Kielan 1960) is the only pliomerid trilobite in the Linhsiang Formation of Jingshan. Ovalocephalus is largely restricted to peri-Gondwana, but is widely distributed in the Ordovician of China <ns0:ref type='bibr' target='#b34'>(Zhou et al., 2010)</ns0:ref>. In the Linhsiang Formation of Jingshan, most specimens of Ovalocephalus tetrasulcatus are articulated or nearly articulated exoskeletons, as well as enrolled specimens. This pattern suggests that the trilobites were not transported by current before burial, and were buried in situ. Although some biotic burrows can be observed in the calcareous mudstones (Fig. <ns0:ref type='figure' target='#fig_1'>2e, f</ns0:ref>), no traces of burrows have been found near the trilobites (Fig. <ns0:ref type='figure' target='#fig_1'>2b, c</ns0:ref>), which rules out the likelihood that trilobites were affected by biotic disturbance before burial or preserved in burrows.</ns0:p><ns0:p>I collected more than 100 specimens containing articulated or nearly articulated exoskeletons from the calcareous mudstones at the top of the Linhsiang Formation of Jingshan, including 13 specimens that showed two exoskeletons end to end or overlapping one another, and three specimens that show three exoskeletons end to end or two of them overlapping one another, which obviously differed from isolated single exoskeletons preserved in the same interval. The fossils in Fig. <ns0:ref type='figure' target='#fig_2'>3</ns0:ref> were whitened with magnesium oxide powder, and all photographs were captured using a Nikon D5100 camera with a Micro-Nikkor 55 mm F3.5 lens. The axial azimuth measurements of the trilobites and the rose chart were completed in CorelDRAW X7.</ns0:p></ns0:div>
<ns0:div><ns0:head>Results</ns0:head><ns0:p>In the present work, complete exoskeletons include all specimens with articulated cephala, thoraces, and pygidia or enrolled exoskeletons, indicating that they are fossilized carapaces of Ovalocephalus tetrasulcatus. Among these nearly complete exoskeletons, a portion of them are articulated thoraces and pygidia, but with the cephala separated and preserved nearby (Fig. <ns0:ref type='figure' target='#fig_2'>3a</ns0:ref>), which are considered exuviae <ns0:ref type='bibr' target='#b15'>(Han and Wang, 2000)</ns0:ref>. Another kind has librigenae separated from the cranidia, but the cranidia and thoracopyga are still articulated, or the cranidia are separated from the thoraces and slightly rotated. These specimens include those with two librigenae separated from cranidia, but still preserved nearby (Fig. <ns0:ref type='figure' target='#fig_2'>3b</ns0:ref>), and a portion of these librigenae were inverted or rotated (Fig. <ns0:ref type='figure' target='#fig_2'>3d</ns0:ref>); in addition, some specimens have one librigena separated from the cranidium and inverted, but another still in situ (Fig. <ns0:ref type='figure' target='#fig_2'>3c</ns0:ref>). Both types of inverted or rotated librigenae separated from the cranidium are similar to the exuvial mode of many trilobites <ns0:ref type='bibr'>(Henningsmoen,1975;</ns0:ref><ns0:ref type='bibr' target='#b18'>McNamara and Rudkin, 1984)</ns0:ref>, indicating that they are most likely exuviae of Ovalocephalus tetrasulcatus, rather than corpses that were broken up by bottom currents and/or organisms, and the natural decomposition of the corpses. This indicates that there is more than one exuvial mode in Ovalocephalus tetrasulcatus; i.e., separation of the cephalon from the thoracopygon, separation of the librigenae from the cranidium, or both existing at the same time. A similar diversity of exuvial patterns has been found in other phacopid trilobites <ns0:ref type='bibr' target='#b2'>(Brandt, 1993;</ns0:ref><ns0:ref type='bibr' target='#b3'>Budil and Bruthansová, 2005;</ns0:ref><ns0:ref type='bibr' target='#b7'>Chen, 2011)</ns0:ref>.</ns0:p><ns0:p>In the sixteen specimens with two or three exoskeletons partly superimposed or end to end, all the specimens show separation of the cephalon or librigenae, and the separated cephala or librigenae are still preserved near the thoracopyga (Fig. <ns0:ref type='figure' target='#fig_3'>4</ns0:ref>), indicating that all of these specimens are exuviae of Ovalocephalus tetrasulcatus. The central axis of the most frontal Ovalocephalus tetrasulcatus in the exuvial clusters was used as the directrix to calculate the axial azimuth of the posterior trilobites. The results showed that there was an obvious dominant orientation; that is, the axial azimuth of the trilobites in these clusters had obvious consistency (Fig. <ns0:ref type='figure' target='#fig_4'>5</ns0:ref>), indicating they were not the result of accidental burial events. The separated librigenae preserved near the cranidia may also preclude them from being the result of postmortem transport. Therefore, these exuviae are inferred to have been caused by the active behavior of trilobites.</ns0:p></ns0:div>
<ns0:div><ns0:head>Discussion</ns0:head><ns0:p>Possible causes of coupled exuviae Preservation of two Ovalocephalus tetrasulcatus exoskeletons preserved together is similar to the preservation of an arthropod in the middle of the act of molting <ns0:ref type='bibr' target='#b11'>(García-Bellido and Collins, 2004)</ns0:ref>, but the latter case includes one corpse and one exuvia, whereas all these specimens from Jingshan are exuviae, without corpses; thus, the possibility that the Ovalocephalus tetrasulcatus individuals were buried when molting can be excluded. Several cases of queuing of trilobites have been reported in the past, but most of these cases involve corpse fossils, and the number of trilobites is typically more than three; thus, they are regarded as representing unexpected burial during the migration of trilobites <ns0:ref type='bibr' target='#b23'>(Radwański et al., 2009;</ns0:ref><ns0:ref type='bibr' target='#b0'>Błażejowski et al., 2016;</ns0:ref><ns0:ref type='bibr' target='#b28'>Vannier et al., 2019)</ns0:ref>. <ns0:ref type='bibr' target='#b6'>Chatterton and Fortey (2008)</ns0:ref> reported trilobites aligned for molting, but they were preserved in burrows. In Jingshan, there are no biological burrows near the Ovalocephalus tetrasulcatus specimens (Fig. <ns0:ref type='figure' target='#fig_1'>2b, c</ns0:ref>), and the number of exuviae per specimens is only two or three, in contrast with the former. Since <ns0:ref type='bibr' target='#b27'>Speyer and Brett (1985)</ns0:ref> first reported synchronous ecdysis in Middle Devonian phacopids in New York, this behavior has been identified in some other trilobites <ns0:ref type='bibr' target='#b20'>(Paterson et al. 2007)</ns0:ref>, usually with many trilobites concentrated in a certain place to shed their shells, which is thought to be related to gregarious behavior and is probably associated with the copulation and reproduction of trilobites <ns0:ref type='bibr' target='#b26'>(Speyer, 1990)</ns0:ref>. A similar pattern may exist in Ovalocephalus tetrasulcatus, and pairs of exuviae are relatively easy to understand for mating behavior after molting. However, there are obvious differences of sizes in some exuviae in the same cluster (Fig. <ns0:ref type='figure' target='#fig_3'>4d</ns0:ref>), or in different clusters (Fig. <ns0:ref type='figure' target='#fig_3'>4b, e</ns0:ref>). More importantly, there are also clusters with three exuviae (Fig. <ns0:ref type='figure' target='#fig_3'>4g-i</ns0:ref>), and the superposition (Fig. <ns0:ref type='figure' target='#fig_3'>4b</ns0:ref>) even almost complete overlapping (Fig. <ns0:ref type='figure' target='#fig_3'>4c, h-i</ns0:ref>) is visible in the exuvial clusters, it is unlikely that two or three trilobites shed their shells synchronously, the time interval (such as a few hours or a few days) may be exist in two or three exuvial process. Thus, it is difficult to explain these exuvial clusters as the result of molting-mating behavior.</ns0:p><ns0:p>I speculate that these specimens may reflect a particular exuvial behavior of Ovalocephalus tetrasulcatus; because superposition (Fig. <ns0:ref type='figure' target='#fig_3'>4b</ns0:ref>) and even almost complete overlapping (Fig. <ns0:ref type='figure' target='#fig_3'>4c, hi</ns0:ref>) exist in the exuvial clusters, it is unlikely that two or three trilobites shed their shells synchronously. Instead, they may have shed their shells in lines; that is, after one trilobite finished molting, the others shed their shells as well. Alternatively, perhaps the first trilobite finished molting and left an empty shell, and then later, other trilobites came to the same place to molt. The consistency of the long axes of the exuviae may be related to the seabed topography at that time; for example, molting could have occurred in a narrow shallow gully or on a gentle slope. This is may have been somewhat similar to the molting of extant cicadas on trees, where they occasionally molt in line or overlapping <ns0:ref type='bibr' target='#b1'>(Bobo, 2016;</ns0:ref><ns0:ref type='bibr'>Muchen, 2016;</ns0:ref><ns0:ref type='bibr' target='#b36'>Zhu and Wang, 2017)</ns0:ref>. Implications for the behavioral strategy of trilobites Arthropods are weak during and after molting, which leaves them vulnerable to predators and even other members of their species. Living shrimps and crabs usually hide in rock crevices or water plants for molting. Trilobites would also have needed a quiet and undisturbed environment when they shed their shells <ns0:ref type='bibr' target='#b16'>(Henningsmoen, 1975;</ns0:ref><ns0:ref type='bibr' target='#b14'>Han, 2006)</ns0:ref>. For example, some trilobites used the empty shells of cephalopods and gastropods as shelter for molting <ns0:ref type='bibr' target='#b4'>(Chatterton, 1971;</ns0:ref><ns0:ref type='bibr' target='#b2'>Brandt, 1993;</ns0:ref><ns0:ref type='bibr' target='#b9'>Davis et al., 2001;</ns0:ref><ns0:ref type='bibr' target='#b37'>Zong et al., 2016)</ns0:ref>, some even shed their shells under the empty shells of larger trilobites <ns0:ref type='bibr' target='#b13'>(Gutiérrez-Marco et al., 2009)</ns0:ref>, and others molted in burrows of other animals <ns0:ref type='bibr' target='#b5'>(Chatterton et al., 2003;</ns0:ref><ns0:ref type='bibr' target='#b6'>Chatterton & Fortey, 2008)</ns0:ref>. The Upper Ordovician strata in Jingshan have yielded a large number of cephalopods, and I also found nautiloid fossils in the Linhsiang Formation. These predators would have threatened great harm to molting trilobites, and the trilobites therefore would have needed to find safe places for ecdysis. However, when there was no shelter on the seafloor, or insufficient space to hide, possibly, they may have chosen to follow congeneric trilobites nearby to molt; alternatively, the remaining exuviae of other trilobites might have suggested that the location was suitable or safe for molting, thus attracting the later trilobites to molt in the same position, thus forming exuvial clusters. It is worth noting that there are only two or three exuviae in each cluster, which may be because the posterior trilobites adopted the principle of proximity when choosing the molting site. This is similar to the herding behavior of animals in crisis situations, indicating that Ovalocephalus tetrasulcatus would take the initiative to choose the nearest safe area to carry out vulnerable life activities.</ns0:p></ns0:div>
<ns0:div><ns0:head>Conclusions</ns0:head><ns0:p>The exuviae of Ovalocephalus tetrasulcatus from the Upper Ordovician Linxiang Formation presented two types: separation of the librigenae from the cranidium, and separation of the cephalon from the thoracopygon, reflecting the diversity of the exuvial modes. Some specimens have two or three exuviae arranged end to end, and some have partly and even completely superimposed exuviae together in clusters. The preserved patterns and burial conditions indicate that these specimens are products of the active behavior of trilobites, rather than mechanical transport by currents, unexpected burial in the middle of the act of molting, or collective molting before mating. The preserved patterns and overlapping phenomena of the exuviae indicate that these clusters were formed by two or three trilobites lining up to shed their shells in a long and narrow feature of seafloor topography. They likely represent the behavioral response that the trilobites chose to follow in certain risky life activities, indicating that herding behavior existed in these Ordovician trilobites. </ns0:p></ns0:div><ns0:figure xml:id='fig_0'><ns0:head /><ns0:label /><ns0:figDesc>Figure 1</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_1'><ns0:head>Figure 2</ns0:head><ns0:label>2</ns0:label><ns0:figDesc>Figure 2</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_2'><ns0:head>Figure 3</ns0:head><ns0:label>3</ns0:label><ns0:figDesc>Figure 3</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_3'><ns0:head>Figure 4</ns0:head><ns0:label>4</ns0:label><ns0:figDesc>Figure 4</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_4'><ns0:head>Figure 5</ns0:head><ns0:label>5</ns0:label><ns0:figDesc>Figure 5</ns0:figDesc></ns0:figure>
</ns0:body>
" | "Reviewer 2 (Brian Chatterton)
Basic reporting
no comment
Experimental design
no comment
Validity of the findings
The author has done a good job making most of the changes suggested. There are some improvements needed to the English language in the MS (see suggestions in the revised MS sent back. It would have been easier to have made a list of changes that need to be made had there been line numbers in the revised MS.
Yes, I revised the MS according to comments in the PDF.
I am still not convinced why sex could not or would not have been accomplished before molting (which seems to me to be more likely).
The only other general problem that I have with this work is that the author seems to think that only one male is or can be involved in impregnating or fertilizing the eggs of a female, whereas there are numerous examples in the animal kingdom, involving many classes and orders, of more than one male attempting to do this job (ranging from our own species to fishes, squids, bees and probably many other invertebrates, if one was to research this area thoroughly). So the fact that there are three moult carapaces in some of the specimens described in this paper, if they are directly associated with mating behaviour, could simply be the results of there being 2 males and one female in the group.
I added some sentences to exclude the possibility of ‘molting–mating behavior’. In addition to the obvious differences of sizes in some exuviae in the clusters, and three exuviae in some clusters. More importantly, it is unlikely that two or three trilobites shed their shells synchronously, based on the superposition (Fig. 4b) even almost complete overlapping (Fig. 4c, h–i) is visible in some exuvial clusters, i.e., the time interval (such as a few hours or a few days or even longer) may be exist in two or three exuvial process. Thus, it is difficult to explain these exuvial clusters as the result of molting–mating behavior.
Just a note that it is the Katian Stage, not the Kaitian.
Yes, I replaced it.
Comments for the Author
Please check minor changes to the MS to improve the English language and note the comments above under validity of the findings. Otherwise, I am pleased that you have clearly carefully considered the improvements suggested by the reviewers.
Yes, done.
" | Here is a paper. Please give your review comments after reading it. |
9,815 | "<ns0:abstract xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:p>Ecdysis was a vital process during the lives of trilobites, in addition to preserving the morphological changes in trilobite ontogeny, the preservation of its action often captured interesting behavioral information. Abundant exuviae of Ovalocephalus tetrasulcatus are preserved in the Ordovician strata in central Hubei, China, and some of them are arranged with two or three together end to end or superimposed. The preserved patterns and burial conditions indicate that these specimens were caused by the active behavior of trilobites.</ns0:p><ns0:p>It is speculated that these exuvial clusters were formed by two or three trilobites in line to molt; that is, after one trilobite finished molting, other trilobites molted in front of, behind, or overlying the previously molted shells. This ecdysis strategy is interpreted as related to the postulated herding behavior of some trilobites, representing a behavioral response of the trilobites to choose a nearby safe zone during some risky life activities.</ns0:p></ns0:div>
</ns0:abstract>
<ns0:body xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:head>Introduction</ns0:head><ns0:p>The biomineralized or chitinous exoskeleton of arthropods hinders the growth of their bodies; therefore, individuals must shed their old shells many times during growth and development <ns0:ref type='bibr' target='#b19'>(Moussian, 2013;</ns0:ref><ns0:ref type='bibr' target='#b8'>Daley and Drage, 2016)</ns0:ref>. Trilobites, as an extinct group of arthropods, also needed to shed their shells as they grew <ns0:ref type='bibr' target='#b10'>(Fortey, 2014)</ns0:ref>. Different trilobites exhibited different molting techniques <ns0:ref type='bibr' target='#b16'>(Henningsmoen, 1975)</ns0:ref>; most trilobites shed their shells through separating the librigenae from the cranidium <ns0:ref type='bibr' target='#b18'>(McNamara and Rudkin, 1984;</ns0:ref><ns0:ref type='bibr' target='#b30'>Whittington, 1990)</ns0:ref>, whereas for trilobites with librigenae fused with the cranidium, separation of the cephalon from the thoracopygon usually occurred during molting <ns0:ref type='bibr' target='#b25'>(Speyer, 1985;</ns0:ref><ns0:ref type='bibr' target='#b29'>Wang and Han, 1997)</ns0:ref>, and some genera had multiple exuvial modes <ns0:ref type='bibr' target='#b2'>(Brandt, 1993;</ns0:ref><ns0:ref type='bibr' target='#b3'>Budil and Bruthansová, 2005)</ns0:ref>. In addition to ecdysis reflecting the ontogenetic development process of trilobites, some exceptionally preserved trilobite specimens also contain behavioral information. For example, some trilobites shed their shells by hiding in empty shells or burrows of other animals <ns0:ref type='bibr' target='#b9'>(Davis et al., 2001;</ns0:ref><ns0:ref type='bibr' target='#b5'>Chatterton et al., 2003;</ns0:ref><ns0:ref type='bibr' target='#b6'>Chatterton and Fortey, 2008;</ns0:ref><ns0:ref type='bibr' target='#b37'>Zong et al., 2016)</ns0:ref>, and even molted infaunally <ns0:ref type='bibr' target='#b24'>(Rustán et al., 2011)</ns0:ref>, reflecting the hiding behavior of trilobites. Some phacopids may also have exhibited asymmetric behaviors during molting <ns0:ref type='bibr' target='#b38'>(Zong and Gong, 2017)</ns0:ref>. Other trilobites collectively shed their shells, which may have been related to molting-mating behavior <ns0:ref type='bibr' target='#b27'>(Speyer and Brett, 1985;</ns0:ref><ns0:ref type='bibr' target='#b26'>Speyer, 1990)</ns0:ref>.</ns0:p><ns0:p>South China is an important area for trilobite fossils, and abundant Ordovician trilobites have been collected from Hubei Province <ns0:ref type='bibr' target='#b17'>(Lu, 1975;</ns0:ref><ns0:ref type='bibr' target='#b35'>Zhou and Zhen, 2008)</ns0:ref>. However, there are few reports on exuvial specimens and correlation research. Only photographs of exuvial specimens were attached to the identification of genera and species in the paleontological literature; these exuvial specimens have not been systematically described, their patterns have not been classified and explained, and the behavioral strategy of these trilobites during molting has not been analyzed <ns0:ref type='bibr' target='#b15'>(Han and Wang, 2000)</ns0:ref>. Here, I collected many exuviae of Ovalocephalus tetrasulcatus (Phacopida, Pliomeridae) from the Upper Ordovician in central Hubei. Some of the specimens were found with two or three in a line, or preserved partly or even completely overlapping; these patterns are regarded as representing the active behavior of trilobites during molting, and may be related to the herding behavior of some trilobites. They provide new material for understanding exuvial techniques of trilobites, and the behavior of trilobites when molting.</ns0:p></ns0:div>
<ns0:div><ns0:head>Materials & Methods</ns0:head><ns0:p>All specimens were collected from the Upper Ordovician Linhsiang Formation of the Daozimiao section in Jinshan County, central Hubei (GPS: N 30°59′51.49″, E113°06′26.04″) (Fig. <ns0:ref type='figure'>1</ns0:ref>). The Linhsiang Formation is widely distributed in the middle Yangtze region, and is mainly composed of yellow-green calcareous mudstones with a few siliceous mudstones in the Daozimiao section, but differs from the nodular muddy limestones of the Linhsiang Formation in the type section in Linxiang, Hunan Province. In the Daozimiao section, the Linhsiang Formation is conformably underlain by the muddy limestones of the Upper Ordovician Pagoda Formation and overlain by the graptolite shales of the Upper Ordovician Wufeng Formation. The 1.5-m-thick calcareous mudstones from the top of the Linhsiang Formation (Fig. <ns0:ref type='figure'>1d</ns0:ref>) yield an abundant and diverse trilobite fauna, including members of the Metagnostidae, Cyclopygidae, Phillipsinellidae, Encrinuridae, Telephinidae, Raphiophoridae, Cheiruridae, Dionididae, Trinucleidae, Asaphidae, Pliomeridae, and Remopleurididae. In addition, the Foliomena fauna (brachiopods) <ns0:ref type='bibr' target='#b32'>(Zhan and Jin, 2005)</ns0:ref>, ostracods, echinoderms, machaeridians, and trace fossils are also found in the same horizon. Based on the trilobite assemblage, the age of the top of the Linhsiang Formation is constrained to the middle Katian (early Ashgill) <ns0:ref type='bibr' target='#b33'>(Zhou et al., 2016)</ns0:ref>.</ns0:p><ns0:p>The 1.5-m-thick trilobite-bearing calcareous mudstones are homogeneous, without bedding structures. There are no cross-bedding and graded bedding visible in the longitudinal sections of the rocks, but a small amount of authigenic pyrites is preserved in this interval (Fig. <ns0:ref type='figure' target='#fig_1'>2a-d</ns0:ref>). This evidence suggests the calcareous mudstones formed in a relatively calm and deepwater environment, which is consistent with the results of the characteristics of the brachiopod association found in the same interval <ns0:ref type='bibr' target='#b32'>(Zhan and Jin, 2005)</ns0:ref>. Ovalocephalus tetrasulcatus (Kielan 1960) is the only pliomerid trilobite in the Linhsiang Formation of Jingshan. Ovalocephalus is largely restricted to peri-Gondwana, but is widely distributed in the Ordovician of China <ns0:ref type='bibr' target='#b34'>(Zhou et al., 2010)</ns0:ref>. In the Linhsiang Formation of Jingshan, most specimens of Ovalocephalus tetrasulcatus are articulated or nearly articulated exoskeletons, as well as enrolled specimens. This pattern suggests that the trilobites were not transported by current before burial, and were buried in situ. Although some biotic burrows can be observed in the calcareous mudstones (Fig. <ns0:ref type='figure' target='#fig_1'>2e, f</ns0:ref>), no traces of burrows have been found near the trilobites (Fig. <ns0:ref type='figure' target='#fig_1'>2b, c</ns0:ref>), which rules out the likelihood that trilobites were affected by biotic disturbance before burial or preserved in burrows.</ns0:p><ns0:p>I collected more than 100 specimens containing articulated or nearly articulated exoskeletons from the calcareous mudstones at the top of the Linhsiang Formation of Jingshan, including 13 specimens that showed two exoskeletons end to end or overlapping one another, and three specimens that show three exoskeletons end to end or two of them overlapping one another, which obviously differed from isolated single exoskeletons preserved in the same interval. The fossils in Fig. <ns0:ref type='figure' target='#fig_2'>3</ns0:ref> were whitened with magnesium oxide powder, and all photographs were captured using a Nikon D5100 camera with a Micro-Nikkor 55 mm F3.5 lens. The axial azimuth measurements of the trilobites and the rose chart were completed in CorelDRAW X7.</ns0:p></ns0:div>
<ns0:div><ns0:head>Results</ns0:head><ns0:p>In the present work, complete exoskeletons include all specimens with articulated cephala, thoraces, and pygidia or enrolled exoskeletons, indicating that they are fossilized carapaces of Ovalocephalus tetrasulcatus. Among these nearly complete exoskeletons, a portion of them are articulated thoraces and pygidia, but with the cephala separated and preserved nearby (Fig. <ns0:ref type='figure' target='#fig_2'>3a</ns0:ref>), which are considered exuviae <ns0:ref type='bibr' target='#b15'>(Han and Wang, 2000)</ns0:ref>. Another kind has librigenae separated from the cranidia, but the cranidia and thoracopyga are still articulated, or the cranidia are separated from the thoraces and slightly rotated. These specimens include those with two librigenae separated from cranidia, but still preserved nearby (Fig. <ns0:ref type='figure' target='#fig_2'>3b</ns0:ref>), and a portion of these librigenae were inverted or rotated (Fig. <ns0:ref type='figure' target='#fig_2'>3d</ns0:ref>); in addition, some specimens have one librigena separated from the cranidium and inverted, but another still in situ (Fig. <ns0:ref type='figure' target='#fig_2'>3c</ns0:ref>). Both types of inverted or rotated librigenae separated from the cranidium are similar to the exuvial mode of many trilobites <ns0:ref type='bibr'>(Henningsmoen,1975;</ns0:ref><ns0:ref type='bibr' target='#b18'>McNamara and Rudkin, 1984)</ns0:ref>, indicating that they are most likely exuviae of Ovalocephalus tetrasulcatus, rather than corpses that were broken up by bottom currents and/or organisms, and the natural decomposition of the corpses. This indicates that there is more than one exuvial mode in Ovalocephalus tetrasulcatus; i.e., separation of the cephalon from the thoracopygon, separation of the librigenae from the cranidium, or both existing at the same time. A similar diversity of exuvial patterns has been found in other phacopid trilobites <ns0:ref type='bibr' target='#b2'>(Brandt, 1993;</ns0:ref><ns0:ref type='bibr' target='#b3'>Budil and Bruthansová, 2005;</ns0:ref><ns0:ref type='bibr' target='#b7'>Chen, 2011)</ns0:ref>.</ns0:p><ns0:p>In the sixteen specimens with two or three exoskeletons partly superimposed or end to end, all the specimens show separation of the cephalon or librigenae, and the separated cephala or librigenae are still preserved near the thoracopyga (Fig. <ns0:ref type='figure' target='#fig_3'>4</ns0:ref>), indicating that all of these specimens are exuviae of Ovalocephalus tetrasulcatus. The central axis of the most frontal Ovalocephalus tetrasulcatus in the exuvial clusters was used as the directrix to calculate the axial azimuth of the posterior trilobites. The results showed that there was an obvious dominant orientation; that is, the axial azimuth of the trilobites in these clusters had obvious consistency (Fig. <ns0:ref type='figure' target='#fig_4'>5</ns0:ref>), indicating they were not the result of accidental burial events. The separated librigenae preserved near the cranidia may also preclude them from being the result of postmortem transport. Therefore, these exuviae are inferred to have been caused by the active behavior of trilobites.</ns0:p></ns0:div>
<ns0:div><ns0:head>Discussion</ns0:head><ns0:p>Possible causes of coupled exuviae Preservation of two Ovalocephalus tetrasulcatus exoskeletons preserved together is similar to the preservation of an arthropod in the middle of the act of molting <ns0:ref type='bibr' target='#b11'>(García-Bellido and Collins, 2004)</ns0:ref>, but the latter case includes one corpse and one exuvia, whereas all these specimens from Jingshan are exuviae, without corpses; thus, the possibility that the Ovalocephalus tetrasulcatus individuals were buried when molting can be excluded. Several cases of queuing of trilobites have been reported in the past, but most of these cases involve corpse fossils, and the number of trilobites is typically more than three; thus, they are regarded as representing unexpected burial during the migration of trilobites <ns0:ref type='bibr' target='#b23'>(Radwański et al., 2009;</ns0:ref><ns0:ref type='bibr' target='#b0'>Błażejowski et al., 2016;</ns0:ref><ns0:ref type='bibr' target='#b28'>Vannier et al., 2019)</ns0:ref>. <ns0:ref type='bibr' target='#b6'>Chatterton and Fortey (2008)</ns0:ref> reported trilobites aligned for molting, but they were preserved in burrows. In Jingshan, there are no biological burrows near the Ovalocephalus tetrasulcatus specimens (Fig. <ns0:ref type='figure' target='#fig_1'>2b, c</ns0:ref>), and the number of exuviae per specimens is only two or three, in contrast with the former. Since <ns0:ref type='bibr' target='#b27'>Speyer and Brett (1985)</ns0:ref> first reported synchronous ecdysis in Middle Devonian phacopids in New York, this behavior has been identified in some other trilobites <ns0:ref type='bibr' target='#b20'>(Paterson et al. 2007)</ns0:ref>, usually with many trilobites concentrated in a certain place to shed their shells, which is thought to be related to gregarious behavior and is probably associated with the copulation and reproduction of trilobites <ns0:ref type='bibr' target='#b26'>(Speyer, 1990)</ns0:ref>. A similar pattern may exist in Ovalocephalus tetrasulcatus, and pairs of exuviae are relatively easy to understand for mating behavior after molting. However, there are obvious differences of sizes in some exuviae in the same cluster (Fig. <ns0:ref type='figure' target='#fig_3'>4d</ns0:ref>), or in different clusters (Fig. <ns0:ref type='figure' target='#fig_3'>4b, e</ns0:ref>). More importantly, there are also clusters with three exuviae (Fig. <ns0:ref type='figure' target='#fig_3'>4g-i</ns0:ref>), and the superposition (Fig. <ns0:ref type='figure' target='#fig_3'>4b</ns0:ref>) even almost complete overlapping (Fig. <ns0:ref type='figure' target='#fig_3'>4c, h-i</ns0:ref>) is visible in the exuvial clusters, it is unlikely that two or three trilobites shed their shells synchronously, given the time interval likely represented (a few hours or days). Thus, it is difficult to explain these exuvial clusters as the result of moltingmating behavior.</ns0:p><ns0:p>I speculate that these specimens may reflect a particular exuvial behavior of Ovalocephalus tetrasulcatus; because superposition (Fig. <ns0:ref type='figure' target='#fig_3'>4b</ns0:ref>) and even almost complete overlapping (Fig. <ns0:ref type='figure' target='#fig_3'>4c, hi</ns0:ref>) exist in the exuvial clusters, it is unlikely that two or three trilobites shed their shells synchronously. Instead, they may have shed their shells in lines; that is, after one trilobite finished molting, the others shed their shells as well. Alternatively, perhaps the first trilobite finished molting and left an empty shell, and then later, other trilobites came to the same place to molt. The consistency of the long axes of the exuviae may be related to the seabed topography at that time; for example, molting could have occurred in a narrow shallow gully or on a gentle slope. This is may have been somewhat similar to the molting of extant cicadas on trees, where they occasionally molt in line or overlapping <ns0:ref type='bibr' target='#b1'>(Bobo, 2016;</ns0:ref><ns0:ref type='bibr'>Muchen, 2016;</ns0:ref><ns0:ref type='bibr' target='#b36'>Zhu and Wang, 2017)</ns0:ref>. Implications for the behavioral strategy of trilobites Arthropods are weak during and after molting, which leaves them vulnerable to predators and even other members of their species. Living shrimps and crabs usually hide in rock crevices or water plants for molting. Trilobites would also have needed a quiet and undisturbed environment when they shed their shells <ns0:ref type='bibr' target='#b16'>(Henningsmoen, 1975;</ns0:ref><ns0:ref type='bibr' target='#b14'>Han, 2006)</ns0:ref>. For example, some trilobites used the empty shells of cephalopods and gastropods as shelter for molting <ns0:ref type='bibr' target='#b4'>(Chatterton, 1971;</ns0:ref><ns0:ref type='bibr' target='#b2'>Brandt, 1993;</ns0:ref><ns0:ref type='bibr' target='#b9'>Davis et al., 2001;</ns0:ref><ns0:ref type='bibr' target='#b37'>Zong et al., 2016)</ns0:ref>, some even shed their shells under the empty shells of larger trilobites <ns0:ref type='bibr' target='#b13'>(Gutiérrez-Marco et al., 2009)</ns0:ref>, and others molted in burrows of other animals <ns0:ref type='bibr' target='#b5'>(Chatterton et al., 2003;</ns0:ref><ns0:ref type='bibr' target='#b6'>Chatterton & Fortey, 2008)</ns0:ref>. The Upper Ordovician strata in Jingshan have yielded a large number of cephalopods, and I also found nautiloid fossils in the Linhsiang Formation. These predators would have threatened great harm to molting trilobites, and the trilobites therefore would have needed to find safe places for ecdysis. However, when there was no shelter on the seafloor, or insufficient space to hide, possibly, they may have chosen to follow congeneric trilobites nearby to molt; alternatively, the remaining exuviae of other trilobites might have suggested that the location was suitable or safe for molting, thus attracting the later trilobites to molt in the same position, thus forming exuvial clusters. It is worth noting that there are only two or three exuviae in each cluster, which may be because the posterior trilobites adopted the principle of proximity when choosing the molting site. This is similar to the herding behavior of animals in crisis situations, indicating that Ovalocephalus tetrasulcatus would take the initiative to choose the nearest safe area to carry out vulnerable life activities.</ns0:p></ns0:div>
<ns0:div><ns0:head>Conclusions</ns0:head><ns0:p>The exuviae of Ovalocephalus tetrasulcatus from the Upper Ordovician Linxiang Formation presented two types: separation of the librigenae from the cranidium, and separation of the cephalon from the thoracopygon, reflecting the diversity of the exuvial modes. Some specimens have two or three exuviae arranged end to end, and some have partly and even completely superimposed exuviae together in clusters. The preserved patterns and burial conditions indicate that these specimens are products of the active behavior of trilobites, rather than mechanical transport by currents, unexpected burial in the middle of the act of molting, or collective molting before mating. The preserved patterns and overlapping phenomena of the exuviae indicate that these clusters were formed by two or three trilobites lining up to shed their shells in a long and narrow feature of seafloor topography. They likely represent the behavioral response that the trilobites chose to follow in certain risky life activities, indicating that herding behavior existed in these Ordovician trilobites. </ns0:p></ns0:div><ns0:figure xml:id='fig_0'><ns0:head /><ns0:label /><ns0:figDesc>Figure 1</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_1'><ns0:head>Figure 2</ns0:head><ns0:label>2</ns0:label><ns0:figDesc>Figure 2</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_2'><ns0:head>Figure 3</ns0:head><ns0:label>3</ns0:label><ns0:figDesc>Figure 3</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_3'><ns0:head>Figure 4</ns0:head><ns0:label>4</ns0:label><ns0:figDesc>Figure 4</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_4'><ns0:head>Figure 5</ns0:head><ns0:label>5</ns0:label><ns0:figDesc>Figure 5</ns0:figDesc></ns0:figure>
</ns0:body>
" | "Editor comments (Bruce Lieberman)
MINOR REVISIONS
The paper requires only one minor revision for copy editing/grammatical purposes. With this change it will be ready to go for publication. The necessary change is related to the new text the author inserted to address one of reviewer Chatterton's comments. In particular, for current lines 168-170 it reads 'it is unlikely that two or three trilobites shed their shells synchronously, the time interval (such as a few hours or a few days) may be exist in two or three exuvial process. ' To correct this, please change the text to 'it is unlikely that two or three trilobites shed their shells synchronously, given the time interval likely represented (a few hours or days). '
All of the other changes the author made look good and these addressed all of reviewer Chatterton's comments.
Please make this change and resubmit. Thank you.
Yes, done.
" | Here is a paper. Please give your review comments after reading it. |
9,816 | "<ns0:abstract xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:p>Background: Wearable activity trackers are regarded as a new opportunity to deliver health promotion interventions. The algorithms typically contain two stages: 1) activity classification and 2) quantification of energy expenditures by applying activity-specific regression equations. Although the predictions of traditional activity trackers rely primarily on the processing of accelerometer sensor data, the emergence of smart clothes with multi-sensing capacities could contribute to a more accurate evaluation of daily physical behaviors. This study aims to 1) develop an activity recognition algorithm based on the processing of plantar pressure information provided by a smart-shoe prototype and 2) to determine the optimal hardware and software configurations. Method: Seventeen subjects wore a pair of smart-shoe prototypes composed of plantar pressure measurement insoles, and they performed the following nine activities: sitting, standing, walking on a flat surface, walking upstairs, walking downstairs, walking up a slope, running, cycling, and completing office work. The insole featured seven pressure sensors. For each activity, four minutes of plantar pressure data were collected. The plantar pressure data were cut in overlapping windows of different lengths and 167 features were extracted for each window. A random forest model was trained to recognize activity using some selected sensor configurations and different numbers of data features. The resulting activity recognition algorithms were evaluated using data samples specifically directed to validation trials. Results: Using all the sensors and all 167 features, the smart shoes predicted the activities with an average success of 89%. 'Running' demonstrated the highest percentage of good predictions (100%). 'Walking up a slope' was linked with the lowest performance (63%), with the majority of the false positives being 'walking on a flat surface' and 'walking upstairs.' Some 2-and 3-sensor configurations were linked with an</ns0:p></ns0:div>
</ns0:abstract>
<ns0:body xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:head>Introduction</ns0:head><ns0:p>The promotion of an active lifestyle among our populations remains an on-going problem <ns0:ref type='bibr' target='#b1'>(Barreto, 2013)</ns0:ref>. Fortunately, the recent boom in the marketing of activity trackers provides new tools to address this issue. The term 'activity tracker' is defined as a category of wearable devices, which aims to provide users with feedback on their physical behaviors, physical fitness, and physical activity. This feedback can be provided through a wide variety of parameters, including 'stepcount,' time spent in activities of selected intensities (sedentary, light, moderate, or vigorous activities), number of floors climbed, and daily energy expenditures (expressed in kilocalories). This type of device has been demonstrated to be effective in supporting active lifestyles and is now widely considered in the development of health promotion policies <ns0:ref type='bibr' target='#b4'>(Bravata et al., 2007;</ns0:ref><ns0:ref type='bibr' target='#b3'>Bonomi & Westerterp, 2012;</ns0:ref><ns0:ref type='bibr' target='#b13'>Gal et al., 2018</ns0:ref><ns0:ref type='bibr' target='#b16'>, Jennings et al., 2017)</ns0:ref>.</ns0:p><ns0:p>From a technological perspective, the majority of contemporary activity trackers integrate one MEMS 3-axis accelerometer chip that allows the sensing of the user's body motion. They are typically worn at the hip or wrist and provide feedback on the amount of daily physical activity <ns0:ref type='bibr' target='#b32'>(Romanzini, Petroski & Reichert, 2012;</ns0:ref><ns0:ref type='bibr' target='#b18'>Kamada et al., 2016)</ns0:ref>. State-of-the-art algorithms directed at evaluating physical behaviors typically feature an activity classification method <ns0:ref type='bibr' target='#b36'>(Staudenmayer et al., 2009</ns0:ref><ns0:ref type='bibr' target='#b26'>, Ohkawara et al., 2011</ns0:ref><ns0:ref type='bibr' target='#b2'>, Bassett, Rowlands & Trost, 2012)</ns0:ref>. The following are examples of activity classes that are frequently proposed when using the information of one single accelerometer: locomotive vs. non-locomotive vs. mixed activity and sedentary vs. light intensity vs. moderate intensity vs. vigorous intensity activities <ns0:ref type='bibr' target='#b19'>(Karabulut, Crouter & Bassett, 2005</ns0:ref><ns0:ref type='bibr' target='#b28'>, Oshima et al., 2010)</ns0:ref>. However, latest trackers now feature multi-sensing technologies (e.g., gyroscope, altimeter, light reflectance, thermal resistor), increasing the amount of available information (e.g., inclination, altitude, heart rate, skin temperature) for physical behavior evaluation, and calling for the development of algorithmic suites able to handle the full wealth of available information <ns0:ref type='bibr' target='#b6'>(Chen & Bassett, 2005</ns0:ref><ns0:ref type='bibr' target='#b29'>, Park et al., 2011)</ns0:ref>. This trend toward multi-sensing evaluation is expected to benefit from current innovations in the field of wearable technologies and smart clothes, which are designed to work in an interconnected network of 5G devices. In such a fast-evolving context, the current methods could rapidly become outdated, and smart clothes able to collect physiological or mechanical information could assume an ever more central role in the evaluation of physical activity <ns0:ref type='bibr' target='#b15'>(Intille, 2012;</ns0:ref><ns0:ref type='bibr' target='#b7'>Chen et al., 2012;</ns0:ref><ns0:ref type='bibr' target='#b8'>Chen et al., 2016)</ns0:ref>. In the near future, higher-level algorithms will function in the cloud and be capable of collecting available data from a large number of connected devices, and they can select the most relevant information depending on the context to proceed to a continuous and ever more accurate evaluation on physical behaviors. Among these, smart shoes or smart insoles oriented toward the assessment of physical behaviors could be used to evaluate the interaction with the ground and to help refine activity classification.</ns0:p><ns0:p>Medical insoles capable of measuring plantar pressures have already been commercialized as transportable alternatives to force platforms (Fscan, Tekscan, Inc., USA; ParoTec, Paromed GmbH & Co. KG; PedoSmart). These devices provide a reliable analysis of the center of pressure to assess posture, gait stability, mobility disorders, fall risk, and some other physical considerations. Recently, smart-shoe systems intended for athletes have also been proposed (Nike + Sensor, Nike, Inc., USA; SportProfiler, Digitsol, France; Torin IQ, Altra Running, USA; Mijia, Xiaomi, China). They typically provide feedback on plantar pressure distribution, foot landing type, cadence, and contact duration with the ground, among other measurements. To date, smart-shoe systems aimed at monitoring physical behaviors in daily life have only been presented in the scientific literature <ns0:ref type='bibr' target='#b9'>(De Pinho André, Diniz & Fuks, 2017;</ns0:ref><ns0:ref type='bibr' target='#b25'>Ngueleu et al., 2019)</ns0:ref>. Devices mentioning a high rate of activity recognition typically have multi-sensing abilities, including accelerometer sensors, gyroscopes, temperature sensors, and GPS antennas, providing a large amount of information to the prediction algorithm. However, the inclusion of several in-shoe sensors would likely induce higher production costs as well as challenges for product designers. Furthermore, the high rates of behavior recognition presented in the literature are at times inherent to the study protocols, which may only include a limited number of activities or focus on specific clinical populations, thus preventing the generalization of the results <ns0:ref type='bibr' target='#b9'>(De Pinho André, Diniz & Fuks, 2017;</ns0:ref><ns0:ref type='bibr' target='#b25'>Ngueleu et al., 2019)</ns0:ref>.</ns0:p><ns0:p>Hence, a smart-insole or smart-shoe system that only uses plantar pressure information and that could recognize multiple human daily life activities has yet to be developed. The present research aims to develop efficient and effective activity recognition algorithms for smart-insole devices featuring 1-7 plantar pressure sensors. Nine daily life activities are considered. The smart-insole prototype used in the present study is equipped with the 7-sensor plantar pressure measurement insole, described elsewhere <ns0:ref type='bibr'>(Saito et al., 2011;</ns0:ref><ns0:ref type='bibr' target='#b23'>Nakajima et al., 2014)</ns0:ref>. The data analysis is conducted using machine learning methods. The identification of the best hardware and software configurations is conducted through a data processing logical frame, which may be re-used by designers willing to develop smart-shoes devices.</ns0:p></ns0:div>
<ns0:div><ns0:head>Materials & Methods</ns0:head></ns0:div>
<ns0:div><ns0:head>7-sensor plantar pressure measurement insole</ns0:head><ns0:p>The shoe hygienic insoles, which are 2 mm thick, were equipped with seven force-sensing resistors (FSR400, Interlink Electronics, Inc., CA). The sensors respond to stimulation ranging from 0.2-20 N (8.13-813 kPa), allowing the measurement of human peak plantar pressure <ns0:ref type='bibr' target='#b24'>(Nandikolla et al., 2017)</ns0:ref>. The sensors were placed on the heel, lateral midfoot, center of the midfoot, lateral forefoot, center of the forefoot, medial forefoot, and big toe (Fig. <ns0:ref type='figure'>1</ns0:ref>). They were connected to a 12-bit resolution data acquisition unit with a wireless data transmission sampling rate capacity of 100 Hz, allowing real-time recording during normal ambulatory activities. Insoles with a similar configuration have proven to be valid for the evaluation of posture and gait in previous studies <ns0:ref type='bibr'>(Saito et al., 2011;</ns0:ref><ns0:ref type='bibr' target='#b23'>Nakajima et al., 2014;</ns0:ref><ns0:ref type='bibr' target='#b0'>Anzai et al. 2020)</ns0:ref>. Multiple pairs of the insole in different sizes were available.</ns0:p></ns0:div>
<ns0:div><ns0:head>Data collection</ns0:head><ns0:p>The experimental protocol was approved by the Ochanomizu University research ethics committee <ns0:ref type='bibr'>(#2018-01)</ns0:ref>. A total of 17 female subjects signed written consents and participated in the trial (age: 26 ± 9 years old, weight: 49 ± 3 kg). All the participants were healthy and did not present mobility disorders. The 7-sensor plantar pressure measurement insoles were inserted in a pair of commercial sneakers (Vans Fable 2, VF Corporation) with stiff and flat midsoles. The insoles and shoes were available from size 22 cm to 27 cm. The participants wore shoes and insoles that best matched their foot size. They performed the following nine activities: sitting, standing, walking on a flat surface, walking upstairs, walking downstairs, walking up a slope, running, cycling, and office work (Table <ns0:ref type='table'>1</ns0:ref>). The duration of each activity was approximately 4 min, except 'walking on a flat surface' and 'running,' which was approximately 8 min. The order in which the 9 activities were completed was randomly selected for each subject. Eleven subjects completed the nine activities. During the course of the experiment, certain subjects expressed a desire to shorten their participation mainly owing to upcoming agenda conflicts, discomfort, or tiredness. Two subjects completed eight activities, two subjects completed seven activities, one subject completed six activities, and one subject completed five activities. For each subject, data for 'walking on a flat surface', 'walking upstairs', 'walking downstairs' and 'running', respectively, may have been stored in two files. The final dataset consisted of 196 files corresponding to the 140 activities completed by the 17 subjects. Each file contained 14 independent plantar pressure time series (seven sensors for each of the left and right feet).</ns0:p></ns0:div>
<ns0:div><ns0:head>Data preprocessing</ns0:head><ns0:p>The raw data were converted into Newtons (N) and smoothed using a second-order low-pass Butterworth filter with a cutoff frequency of 5 Hz. For each data file, the time series were cut in windows of <ns0:ref type='bibr'>1, 5, 10, 15, 20, 25, 30, 35, 40, 45, 50, 55, and 60</ns0:ref> s, with an overlap of 50%.</ns0:p></ns0:div>
<ns0:div><ns0:head>Feature extraction</ns0:head><ns0:p>For each window, 167 data features were extracted using the information from seven sensors on each foot. The features were extracted from the different types of analysis presented in Table <ns0:ref type='table'>2</ns0:ref>.</ns0:p><ns0:p>The features were grouped into the five following categories:</ns0:p><ns0:p>• General statistics analysis: The mean, maximum, standard deviation, and median were calculated for each time series. This category included 56 extracted features.</ns0:p><ns0:p>• Peak analysis: The peak number, average and standard deviation (SD) of the interval between peaks, average and SD of the peak magnitudes, and average and SD of the peak widths were calculated for each time series using the SciPy library <ns0:ref type='bibr' target='#b17'>(Jones et al., 2001)</ns0:ref>. The peak widths were calculated at 30% of the peak height. The default parameters of the library were used for the computation of all other features extracted from the peak analysis. This category included 98 extracted features.</ns0:p><ns0:p>• Gait phase analysis: The envelope of the signal of the seven sensors was calculated for each foot. For each identified full stance phase, the difference in the force peak yield between the foot contact on the ground (early stance phase) and the foot lift (late stance phase) was calculated and the values averaged over the window. The average duration of the double float phase was also calculated or was set to the null value when such phase does not exist. Two features were extracted in this category.</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:05:48920:1:1:NEW 4 Sep 2020)</ns0:p><ns0:p>• Frequency domain analysis: The signal of the 14 sensors was summed up and a fast Fourier transform (FFT) was conducted. Preliminary FFT analyses were conducted. The following 5 features were extracted from the AC component of the discrete frequency component series (0.05-50 Hz) and included in the final analysis: 1) power density, 2) frequency signal weighted average from 1.67-10 Hz, 3) skewness of the frequency components below 10 Hz, 4) mean of the AC components from 2-10 Hz, and 5) standard deviation of the same segment. Events with a frequency lesser than 2 Hz were assumed to be related to the gait cycle. Gait cycle-related behaviors were expected to be described by the features extracted from the above described peak analysis. Moreover, human movements are assumed to not exceed a frequency greater than 10 Hz. Therefore, only the spectral signals at frequencies less than 10 Hz were considered in the present analysis. Five features were extracted in this category.</ns0:p><ns0:p>• Pressure distribution analysis: The envelope of the signal of sensors 4, 5, 6, and 7, located in the forefoot area (Fig. <ns0:ref type='figure'>1A</ns0:ref>), was computed. The difference between the mean of this new series of data and the plantar pressures detected by sensor 1 (heel, Fig. <ns0:ref type='figure'>1A</ns0:ref>) was calculated for the left and right feet. The difference was averaged to express the anterior-posterior distribution of the plantar pressures. The difference between the mean of the plantar pressures detected by sensor 6 (head of the first metatarsal) and mean of the plantar pressures detected by sensor 4 (lateral forefoot) were calculated for the left and right feet. The values were averaged to express the medial-lateral distribution of the plantar pressures. Moreover, a Pearson correlation test was used to test the 1) agreement between the envelope of sensor 4, 5, 6, and 7 signals and sensor 1 signal and 2) agreement between the signal of sensor 4 and that of sensor 6. These correlation coefficients were calculated for both the left and right feet. Six features were extracted in this category.</ns0:p><ns0:p>The final number of extracted features depended on the number of sensors included in the processing (cf. paragraphs 'Window length,' 'Number and location of sensors,' 'Number of features').</ns0:p></ns0:div>
<ns0:div><ns0:head>Design of activity recognition algorithms</ns0:head><ns0:p>In the present study, the smart-shoe activity prediction algorithms were developed using machinelearning techniques. Data used as input included as many dimensions as the number of features extracted, i.e., 167 when using the information from the seven sensors for each foot. Preliminary processing including different machine-learning methods (e.g., k-means clustering, support vector machine) indicated higher performances for the random forest models (results not provided). The analysis presented in this manuscript focuses on the development of random forest models able to process plantar pressure information for activity recognition. The machine-learning analysis was completed using the Python scikit-learn module <ns0:ref type='bibr' target='#b31'>(Pedregosa et al., 2011)</ns0:ref>. 'Forests' were made of 100 decision trees. Each tree in the forest produced an independent prediction (here, an activity), and the mode of the predictions was chosen as the forest decision. Each tree was constructed using a random subset of the dataset, according to the bagging method described elsewhere <ns0:ref type='bibr' target='#b5'>(Breiman, 2001)</ns0:ref>. During the construction process, the nodes were successively split until all data points corresponded to the same activity; that is, until the tree's gini impurity score was equal to zero. This configuration enabled each tree in the forest to output one single prediction (also called a pure decision). Highly informative features could appear in several trees and tended to appear in the nodes that were closer to the root of the trees. Conversely, <ns0:ref type='table'>PeerJ reviewing PDF | (2020:05:48920:1:1:NEW 4 Sep 2020)</ns0:ref> Manuscript to be reviewed features with poor discriminating capacities appeared in less nodes across the entire forest (Fig. <ns0:ref type='figure'>2</ns0:ref> and Supplementary Material 1).</ns0:p><ns0:p>For the training, the data of six subjects (i.e., approximately 33% of the dataset) were used under five different subject-wise assignments (Fig. <ns0:ref type='figure' target='#fig_5'>3</ns0:ref>). At the training stage, only the data of the subjects who completed the nine activities were used. A total of 20 training-test runs were performed for each assignment, with each run using different random subsets of the dataset (hereafter called 'random states'). For the testing process, the generated random forest modules evaluated the data of the remaining 11 subjects. The results averaged across all five assignments (i.e., across 100 forests), were presented as confusion matrices of the predictions vs. actual activities. The results were also presented as mean (minimum, maximum) when summarizing the overall performance across all activities. All the results presented in this manuscript correspond to the outcome of the evaluation of the random forest modules using the test samples only. None of the reported scores are related to the training phases.</ns0:p></ns0:div>
<ns0:div><ns0:head>Data analysis framework</ns0:head><ns0:p>As illustrated in Figure <ns0:ref type='figure' target='#fig_6'>4</ns0:ref>, the analyses are integrated in a 3-stage logical flow. 1) The window length analysis aims at identifying the optimum analytic window length.</ns0:p><ns0:p>2) The analysis of a pre-selected set of 25 sensor configurations, i.e. configurations using the information of different numbers of sensors and/or the information of sensors placed at different locations, aims at identifying the best hardware combination for each possible number of sensors ranging from 1 to 6 (the 7-sensor configuration only has 1 possible combination). This analysis was conducted using the optimum window length identified in 1).</ns0:p><ns0:p>3) A final analysis exploring the contribution of each feature to the forest outputs aims at finding the most efficient number of features to be used for each of the seven best sensor configurations identified in 2). Again, this analysis was conducted using the optimum window length identified in 1). Further details related to each of the three stages are given in the three following subsections.</ns0:p></ns0:div>
<ns0:div><ns0:head>Stage 1: window length</ns0:head><ns0:p>The described analysis was performed for different window lengths <ns0:ref type='bibr'>(1, 5, 10, 15, 20, 25, 30, 35, 40, 45, 50, 55, and 60 s)</ns0:ref> using the data of the seven sensors per shoe with each data point having 167 dimensions corresponding to the maximal number of data features that were possible to extract. The optimum window length was defined at the point where the slope of the function describing the prediction rate vs. window length began to decrease. This optimum window length was used for all subsequent analysis.</ns0:p></ns0:div>
<ns0:div><ns0:head>Stage 2: number and location of sensors</ns0:head><ns0:p>The processing was repeated from scratch for different sensor configurations, i.e., different location and/or number of sensors, for the optimal window length only. Twenty-five configurations were selected among the 127 possible combinations. The selection was performed using subjective criteria: 1) reproduction of a selection of the configurations found in the literature or in the industrial sector of running shoes, 2) promotion of combinations allowing the collection of relevant information for the prediction of gait and postural behaviors, and 3) selection of combinations that are believed to perform poorly (Fig. <ns0:ref type='figure' target='#fig_6'>4</ns0:ref>). The number of dimensions of the data points decreased in accordance with the decreased number of sensors. For certain configurations with the same number of sensors, the data points present different numbers of dimensions. Indeed, as indicated in Table <ns0:ref type='table'>2</ns0:ref>, some features may need specific sensor locations to be computed. All the tested configurations are noted in Figure <ns0:ref type='figure'>2</ns0:ref>.</ns0:p></ns0:div>
<ns0:div><ns0:head>Stage 3: number of features</ns0:head><ns0:p>The processing was again repeated from scratch with the best configurations only and for a decreasing number of features, which were removed one-by-one based on their discriminating capacities (Fig. <ns0:ref type='figure' target='#fig_6'>4</ns0:ref>). For each sensor configuration, the analysis was performed with the maximum number of available features (similar to what was performed for the previous process, cf. paragraph 'Number and location of sensors'). Features were ranked relative to their discriminating capacities, i.e., from the highest to lowest informative feature, across the 100 runs of the analysis (five assignment × 20 random states, cf. 'Prediction algorithm: training and test'). The lowest informative feature was removed from the dataset, and a new repetition of training-test runs was performed. The entire process was repeated until only one feature remained. The minimum number of features corresponding to the inflection point for the prediction rate vs. number of the feature was considered to be the optimum number of inputs. A total of 686 combinations of sensor configurations and number of features were tested (i.e. best 1-sensor: 29, best 2-sensor: 54, best 3sensor: 76, best 4-sensor: 98, best 5-sensor: 120, best 6-sensor: 142, 7-sensor: 167).</ns0:p></ns0:div>
<ns0:div><ns0:head>Results</ns0:head><ns0:p>The 'prediction rates' and 'rates of good predictions' presented in the text and figures refer to the accuracy, calculated as follows: correctly predicted sample/total number of samples. When reporting statistical results, the terms 'average' and 'mean' point to the average accuracy across the 100 forests of one round of evaluations (see Fig. <ns0:ref type='figure' target='#fig_5'>3</ns0:ref>). The expressions 'best single forest' and 'best performer' refer to the one single forest that showed the best accuracy score among the 100 forests produced for one round of evaluations. Conversely, the term 'worst single forest' points to the one single forest that showed the worst accuracy score among the 100 forests produced for one round of evaluations. Logical links between the 3 stages of the analysis are shown in Fig. <ns0:ref type='figure' target='#fig_6'>4</ns0:ref>. The values indicated at the intersections of 'true label' and 'prediction' in the confusion matrices refer to sensitivity, calculated as follows: true positives/(true positives + false negatives).</ns0:p></ns0:div>
<ns0:div><ns0:head>Stage 1: window length</ns0:head><ns0:p>The average performances of the 7-sensor configuration tested at different window lengths <ns0:ref type='bibr'>(1, 5, 10, 15, 20, 25, 30, 35, 40, 45, 50, 55, and 60 s)</ns0:ref> are presented in the Fig. <ns0:ref type='figure'>5</ns0:ref>. The full set of 167 features was used for all tests. The best prediction rate was obtained with a 45-s window length: 0.90 (min: 0.86, max: 0.91). A 20-s length was associated with an average of 0.89 (min: 0.82, max: 0.91). The average prediction rates for window lengths between 20-60 s showed marginal variations within the 0.89-0.90 range. To preserve the highest possible temporal resolution for future applications, 20 s was selected as the optimum length. The rest of the analyses were conducted using a 20-second window length.</ns0:p><ns0:p>As indicated in Fig. <ns0:ref type='figure'>6</ns0:ref>, 'walking up a slope' could be confused with 'walking on a flat surface' or 'walking upstairs.' Confusions between 'walking upstairs' and 'walking downstairs' and between 'standing' and 'office work' were noted to a certain extent depending on the window length.</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:05:48920:1:1:NEW 4 Sep 2020)</ns0:p><ns0:p>Manuscript to be reviewed</ns0:p></ns0:div>
<ns0:div><ns0:head>Stage 2: number and location of sensors</ns0:head><ns0:p>The average performances of a subset of 25 selected sensor configurations are presented in Fig. <ns0:ref type='figure'>7</ns0:ref>. For each configuration, the analyses were performed using all the available features. The best average prediction rate was 0.89. In addition to the 7-sensor configuration, this rate was observed for the four following configurations: • 6 sensors, 145 (heel, lateral midfoot, lateral forefoot, medial forefoot, center of the midfoot, center of the forefoot): 0.89 (min: 0.82, max: 0.92) • 6 sensors, 142, (heel, lateral midfoot, lateral forefoot, big toe, center of the midfoot, center of the forefoot): 0.89 (min: 0.83, max: 0.91) • 5 sensors, 120 features (heel, lateral midfoot, lateral forefoot, center of the midfoot, and center of the forefoot): 0.89 (min: 0.85, max: 0.92) • 4 sensors, 98 features (heel, lateral midfoot, lateral forefoot, center of the forefoot): 0.89 (min: 0.85, max: 0.92) Regarding the best performers, selected forests achieved a prediction rate of 0.92. This result was obtained with a 3-sensor configuration only (heel, lateral midfoot, center of the forefoot). All selected configurations with at least two sensors produced an average rate of good predictions of 0.80 or more. All the configurations with at least five sensors produced an average rate of good predictions of 0.85 or more. All the configurations with at least two sensors, which were expected to perform well, produced an average rate of good predictions of 0.87 or more. Certain forests with one sensor located at the center of the forefoot could compute prediction rates as high as 0.86. The mean and maximum rates of good predictions of a larger panel of 67 selected configurations are presented in Supplementary Material 2.</ns0:p><ns0:p>The confusion matrices presented in Fig. <ns0:ref type='figure'>8 and 9</ns0:ref> indicate the sensitivity score of each activity, for the best and worst 1-, 2-, 3-, 4-, 5-, and 6-sensor configurations. Among the best sensor configurations, the decrease in the average prediction rate observed when reducing the number of sensors from two to one might be explained mainly by the higher levels of confusion between 'sitting' and 'standing' and between 'walking downstairs' and 'walking upstairs.' For example, the best 1-sensor configuration wrongly predicted 'walking upstairs' instead of 'walking downstairs' in 36% of the cases. Among the worst configurations, the decrease in the average prediction rate observed when reducing the number of sensors from four to three could be explained mainly by a decrease in sensitivity for 'cycling' (0.91 and 0.80).</ns0:p></ns0:div>
<ns0:div><ns0:head>Stage 3: number of features</ns0:head><ns0:p>The changes in performance of the seven selected configurations when decreasing, one-by-one, the number of features used for the prediction are displayed in Fig. <ns0:ref type='figure'>10</ns0:ref>. For these configurations, the mean rate of good predictions increased from an average 0.46 ± 0.03 when using one feature to 0.87 ± 0.04 when using a set of 20 high performance features. Using 20 features only, all the selected configurations demonstrated a mean rate of good predictions greater than 0.85, with the worst single forest scoring at 0.81 (2-sensor configuration), except for the 1-sensor configuration, which demonstrated a rate of 0.78 (min: 0.72, max: 0.83). The data are presented in Supplementary Material 2. The overall performance remained constant when the predictions were computed with more features. The mean rate of good predictions exhibited an average of 0.87 ± 0.04 when considering computations performed using the maximum number of available features, i.e., 29, PeerJ reviewing PDF | (2020:05:48920:1:1:NEW 4 Sep 2020)</ns0:p><ns0:p>Manuscript to be reviewed <ns0:ref type='bibr'>54, 76, 98, 120, 142, and 167</ns0:ref>, respectively, for the selected 1-, 2-, 3-, 4-, 5-, 6-, and 7-sensor configurations.</ns0:p><ns0:p>Considering a 20-feature cut-off below which features became increasingly important, 44 important features were identified over the 20-140 alternatives enabled by the seven selected configurations (Fig. <ns0:ref type='figure' target='#fig_3'>11</ns0:ref>). Seven features systematically ranked among the 20 most important features of the seven selected configurations: average peak interval of the left foot heel sensor (peak analysis), average peak magnitude of the right foot heel sensor (peak analysis), mean of the AC component (frequency domain), number of peaks for the right foot heel sensor (peak analysis), number of peaks for the left foot heel sensor (peak analysis), standard deviation of the left foot heel sensor plantar pressures (general statistics), and standard deviation of the AC component (frequency domain). Among the 44 important features, 24 belong to the 'peak analysis' category, 16 to the 'general statistics' category, 3 to the 'frequency domain' category, 1 to the 'gait phase' category, and 0 to the 'pressure distribution' category. Regarding the 7-sensor configuration only, the features related to the heel and central forefoot were identified five times. No feature directly extracted from the analysis of the big toe pressure ranked among the set of important features.</ns0:p><ns0:p>The best 1-sensor configuration (heel sensor) using the single most informative feature demonstrated a mean rate of good predictions of 0.43 (min: 0.41, max: 0.44). Only the 'running' and 'sitting' activities demonstrated a sensitivity score greater than 50% (Fig. <ns0:ref type='figure' target='#fig_17'>12</ns0:ref>). As indicated in Fig. <ns0:ref type='figure' target='#fig_17'>12</ns0:ref>, the other selected sensor configurations were associated with sensitivity scores of 82% or more for all activities except 'office work,' 'walking up a slope,' and 'walking upstairs,' when using a limited number (i.e., 9-23) of features. Regarding the 7-sensor configuration specifically, the confusions noted when using the 23 most informative features (Fig. <ns0:ref type='figure' target='#fig_17'>12</ns0:ref>) were similar to the ones noted when using the full number of available features (Fig. <ns0:ref type='figure'>6D</ns0:ref>), except for 'walking downstairs,' which had a better sensitivity (0.92 vs. 0.84) when using 23 features only. That phenomenon can be explained by the greater difficulty to fit a classifier with a higher number of dimensions. In theory, the same performance should be attainable with more features, at the risk of overfitting the system and decreasing its generality (performance on unknown data) <ns0:ref type='bibr' target='#b21'>(Lever, Krzywinski & Altman, 2016)</ns0:ref>. Confusion matrices of some selected single forests produced with the best 4-sensor configuration are presented in Fig. <ns0:ref type='figure' target='#fig_19'>13</ns0:ref>. A low performance single forest with a relatively high number of features (49 over a maximum of 98 available) was associated with low sensitivity scores for the 'office work,' 'walking downstairs,' and 'walking up the slope' activities (0.74, 0.61, and 0.56, respectively), consistent with the pattern that has been frequently found on confusion matrices, as displayed in Fig. <ns0:ref type='figure'>6, 8, and 9</ns0:ref>. Interestingly, the worst single forest among the ones built with 29 features only had the highest sensitivity for the 'walking up the slope' activity (0.70). Finally, confusion matrices of the best single forests built with 86 and 22 features demonstrated a similar pattern of missed predictions, with 'office work' and 'walking up the slope' being relatively poorly recognized (<0.85 and <0.65, respectively).</ns0:p></ns0:div>
<ns0:div><ns0:head>Supplementary results</ns0:head><ns0:p>A more comprehensive analysis has been conducted considering a larger panel of 67 sensor configurations. Random forest modules were systematically created and tested for each window length candidates (1-60 s) and each possible number of features (maximum to one), without any selection of the best sensor configurations like in the 3-stage data processing flow presented in Fig. <ns0:ref type='figure' target='#fig_6'>4</ns0:ref> (the results of which have been presented in the above three subsections). The machine Manuscript to be reviewed learning procedure was the same as the one detailed in the method section. Therefore, 75,172 additional analyses have been completed, resulting in the computation of 7,517,200 forests. These supplementary analyses were associated with higher prediction scores, highlighting the whole potential of using plantar pressure data for the recognition of physical behaviors. As shown in the Supplementary Material 3, 297 sensor configurations were associated with at least one forest presenting a prediction score of 0.92 or more. Regarding the highest scores, at least 12 forests presented a rate of good predictions of 0.94. The best average scores ranged from 0.54 to 0.91, a scale similar to the one of the results of the 3-stage analysis (Supplementary Material 2). Eightyseven sensor configurations were associated with average rates of good predictions of 0.90 or more. As shown in Supplementary Material 4B, the best performances observed in these analyses are systematically associated with window lengths of 30 s or longer. Regarding the identification of an optimum analytic window length, the results still points to a period of 20 s (Supplementary Material 4A). The question of the time resolution is discussed later in the manuscript. The results of these supplementary analyses are summarized in Supplementary Material <ns0:ref type='figure' target='#fig_11'>3 and 4</ns0:ref>.</ns0:p></ns0:div>
<ns0:div><ns0:head>Discussion</ns0:head><ns0:p>In the present study, homemade smart shoes mounted with seven pressure sensors were used to collect plantar pressures during nine daily life activities. From the plantar pressure data, 167 features bearing a potential interest for the characterization of gait and posture were extracted. Random forest models using subject-wise training-test assignments were utilized to develop smartshoe activity recognition algorithms. A 20-s window length was identified as the optimal period for the extraction of the features. Forests could recognize activities at an average rate of good predictions of 0.89, with certain single forests demonstrating a rate as high as 0.92. Reducing the number of sensors to two (heel and lateral forefoot) and selecting 20 high performance features maintained the average rate of good predictions above 0.85.</ns0:p></ns0:div>
<ns0:div><ns0:head>Performances</ns0:head><ns0:p>Smart shoes in their maximal configuration (i.e., 7 sensors per foot and 167 features extracted from the collected plantar pressures) allow random forest modules to recognize activities at a rate of good predictions of 0.89 (min: 0.82, max: 0.91). Each single activity was associated with a sensitivity score of at least 0.87, except 'office work' and 'walking up a slope,' which presented lower scores (0.80 and 0.63, respectively) (Fig. <ns0:ref type='figure'>6B</ns0:ref>). 'Office work' was confused with 'standing' in 18% of cases. The latter is not surprising considering the content of the 'office work' activity, which includes a considerable number of tasks realized in the standing posture. Numerous subjects consumed a significant amount of time writing and erasing notes on a white table board while performing the 'office work'-labelled activity. This could have created this confusion with the 'standing' activity. Moreover, poor predictions involving the 'walking up a slope' activity being confused with 'walking upstairs' or 'walking on a flat surface' was a recurrent issue of the present analysis. This type of confusion occurred regardless of the sensor configuration or the number of features used as input. Depending on the field of application, several of the above-mentioned confusions could have marginal or significant consequences on the final evaluation of physical behaviors. Future smart-shoe studies should also consider extracting data features that are more likely to report on slope-related gait alterations. Manuscript to be reviewed Conversely, random forest module outcomes indicated only a small number of confusions for the 'cycling,' 'running,' or 'sitting' activities. Although 'running' and 'sitting' are typically well recognized in research protocols that use accelerometer sensors, which remain the current primary hardware choice for activity trackers <ns0:ref type='bibr' target='#b30'>(Pavey et al., 2017;</ns0:ref><ns0:ref type='bibr' target='#b39'>Trost, Zheng & Wong, 2014;</ns0:ref><ns0:ref type='bibr' target='#b40'>Voicu et al., 2019)</ns0:ref>, the recognition of 'sitting' behaviors has actually proven technically challenging in real-life conditions <ns0:ref type='bibr' target='#b20'>(Kerr et al., 2018)</ns0:ref>. Extrinsic behavioral factors, such as people leaving their tracking device to charge when they are resting or sitting, render the assessment of sedentary behaviors even more difficult. In the present study, smart shoes demonstrated high level of sensitivity for 'sitting' (0.96 or more for any of the selected configurations with at least two sensors) and low level of confusion with the other sedentary activity (i.e., 'standing,' (0.00-0.01), except for some 1-and 2-sensor configurations). Such outcomes should be considered as promising for the monitoring of sedentary behaviors outside the house. Finally, differences were noted among the single forests for the performance in each activity. For example, one forest tagged with a low overall performance displayed in Fig. <ns0:ref type='figure' target='#fig_19'>13</ns0:ref> performed surprisingly well for the recognition of the 'walking up a slope' activity. However, this enhanced performance would appear to be possible at the expense of an altered sensitivity for other activities. This may reflect the capacity of random forest modules to specialize for one given type of activity. Further analyses, which are beyond the scope of the present report, would be necessary to identify the 'ins and outs' of forest specialization and determine if the random forest method could be adapted to the specific case of smart shoes to obtain more homogenous recognition rates across activities. For example, forests with a higher number of trees or hierarchical models assigning data points to sub-classes before proceeding to the final evaluation could be considered for future studies.</ns0:p></ns0:div>
<ns0:div><ns0:head>Comparison with previous studies and originality</ns0:head><ns0:p>Several reviews have summarized the outcomes of studies interested in the validity of instrumented insoles developed for activity recognition <ns0:ref type='bibr' target='#b9'>(De Pinho André, Diniz & Fuks, 2017</ns0:ref><ns0:ref type='bibr' target='#b25'>, Ngueleu et al., 2019)</ns0:ref>. Similar to the observations in the present research, specific studies have reported excellent performances, with rates of good predictions scoring frequently over 0.90. However, they can also be linked with experimental limitations, altering the generalization of the results, such as a small number of tested activities, small number of subjects, special groups of individuals, and trainingtest procedures completed separately for each subject. Moreover, the majority of these studies have used hardware with multi-sensing capabilities. Hegde et al. ( <ns0:ref type='formula'>2017</ns0:ref>) developed the SmartStep system, which featured three pressure sensors, one 3-axis accelerometer, and a gyroscope. The pressure sensors were placed at the heel, first metatarsal head (i.e., equivalent to the medial forefoot), and big toe. They tested the activity recognition capabilities of the SmartStep system for a wide range of daily life activities. Similar to the present report, they observed an average rate of good prediction of approximatively 0.90. They also reported recurrent mis-predictions for 'walking downstairs' (0.62), which is frequently confused with 'walking on a flat surface' and 'walking upstairs' and for 'shelving items' (0.61), the description of which resembles the 'office work' activity of the present study, and which is frequently confused with 'standing.' Smart shoebased activity recognition projects appear to be associated with redundant challenges related to ascending and/or descending locomotive activities and activities combining locomotive and nonlocomotive behaviors. Moreover, in another recent study, el Achkar et al. ( <ns0:ref type='formula'>2016</ns0:ref>) used a simple decision tree classifier to achieve excellent rates of good predictions for nine activities, including Manuscript to be reviewed 'walking downstairs' (0.98), 'walking upstairs' (0.99), and 'walking uphill' (0.96). However, their smart-insole system featured a barometer in addition to eight pressure sensors, one 3-axis accelerometer, one 3-axis gyroscope, and one 3-axis magnetometer, which surely helped the assessment of ascending and/or descending locomotive behaviors.</ns0:p><ns0:p>According to <ns0:ref type='bibr' target='#b25'>Ngueleu et al. (2019)</ns0:ref>, smart shoe-based activity recognition studies that only use plantar pressure information are limited. Although some of these studies reported acceptable performance, protocols were typically limited to a small number of locomotive behaviors <ns0:ref type='bibr' target='#b42'>(Zhang et al., 2005;</ns0:ref><ns0:ref type='bibr' target='#b44'>Zhang, Poslad, 2014)</ns0:ref>, small number of subjects, or training-test procedures completed separately for each individual <ns0:ref type='bibr' target='#b37'>(Sugimoto et al., 2010)</ns0:ref>. Therefore, the present research provides important findings to the relatively small corpus of knowledge on plantar pressure-based activity recognition. Other originalities of the present research include the use of a random forest modeling method to develop different activity classifiers and a comparison of different sensor configurations (number and location) within one single experimental protocol.</ns0:p></ns0:div>
<ns0:div><ns0:head>Temporal resolution, sensor configuration, number of features, manufacturing, and algorithmic considerations</ns0:head><ns0:p>Although windows of 30 and 45 s were linked with better performances for the recognition of 'office work' and 'walking up a slope', overall, the performances were consistent across all analyses performed with a window size of 20 s or longer (Fig. <ns0:ref type='figure'>6 B, C, and D</ns0:ref>). In real-life situations, a short window length reduces the probability of overlapping activities over the span of one analytic period. Therefore, a 20-s length with a 50% overlap between windows was selected as the optimum window length. It allowed computing predictions every 10 s. Considering future applications, this relatively high temporal resolution would allow applying a second statistical algorithmic layer consisting of comparing the prediction of one given window with the ones of its neighbors <ns0:ref type='bibr' target='#b41'>(Witowski et al., 2014)</ns0:ref>. This would provide the opportunity to have a set of six 'instant' predictions to determine the dominant behavior every minute. Further explorations that include free-living experiments are necessary to elaborate further on the issue of temporal resolution.</ns0:p><ns0:p>One interesting finding of the present study is the marginal alteration of the overall performance obtained with a reduced number of sensors. Although configurations without the heel sensor systematically present lower performances, other configurations that include at least two sensors demonstrate average rates of good predictions of 0.87 or more (Fig. <ns0:ref type='figure'>7</ns0:ref>). The absence of a heel sensor appears to worsen confusions between ascending and descending activities and between 'office work' and 'standing' (Fig. <ns0:ref type='figure'>8 and 9</ns0:ref>). Using one sensor only, the average rates of good predictions declined below 0.80. Furthermore, marginal variations of the overall performance were noted when reducing the number of features down to approximately 20 (Fig. <ns0:ref type='figure'>10</ns0:ref>). The reduction of the number of features given to the forests was accomplished in a manner that favored the most contributive features. Extracts from the FFT and peak analyses were redundant in the lists of 20 important features (Fig. <ns0:ref type='figure' target='#fig_3'>11</ns0:ref>). However, this result could also be the mere reflection of the higher number of gait activities included in the present protocol, which all present cyclic plantar pressure patterns. Therefore, future studies should include a more balanced number of locomotive, nonlocomotive, and mixed activities to determine whether this trend is confirmed or not. Moreover, no feature extracted directly from the big toe sensor ever scored among the 20 most important features. This location may not be relevant for smart-shoe prototypes aimed at behavior recognition. Although the 167 data features were selected to be as comprehensive as possible and to accommodate the analysis on the reduction of the number of sensors, the list of potentially informative features is not closed. Future studies could propose extracting different features to boost the performance on a similar or different subset of activities. With respect to the abovediscussed results, shoe manufacturers willing to develop activity recognition devices should probably consider the opportunity to implement a minimalist sensor configuration instead of the full 7-sensor configuration. They should also consider the relevance of using an exhaustive number of features, whereas a subset of 20 features has been demonstrated to perform equally well. All these considerations will influence shoe design (relative to the location of sensors and other hardware), microprocessor selection (relative to the computational needs), and, ultimately, the financial cost of the device <ns0:ref type='bibr' target='#b11'>(Eskofier et al., 2017)</ns0:ref>.</ns0:p></ns0:div>
<ns0:div><ns0:head>Limitations and strengths</ns0:head><ns0:p>Some characteristics of the present protocol could limit the interpretation of the results presented in this report and should be mentioned clearly for the readers. First, the current protocol only includes nine different daily life activities. This number puts the study among smart-shoe protocols testing a large sample of activities <ns0:ref type='bibr' target='#b25'>(Ngueleu et al. 2019)</ns0:ref>. The challenges related to the recognition of activities that potentially present closed plantar pressure patterns are addressed in an adequate manner. However, a larger number of activities should be studied in the future to reflect more exhaustively physical behaviors of the daily life, e.g., sport activities and a wider panel of activities combining locomotive and non-locomotive behaviors. Second, the experimental design does not include further validation of forest performances in real-life situations. Similarly, no comparison with commercial activity monitors was performed. Future protocols should include a free-living validation to increase the generalization of the results to real-life situations. Third, the present protocol includes a relatively homogenous population. Subjects were all healthy women. To address this potential issue, five different subject-wise training-test assignments were used to develop and test the forests. In addition, a conservative 6-11 training-test assignment ratio has been used to limit the wealth of the available information during the training phase and create more challenging conditions relying on inter-individual differences. However, a more heterogeneous sample of the population must be tested before generalizing further the results of the present study. A more heterogeneous population would indeed provide a more diverse information to the training algorithms, which could also result in increased good prediction scores. Overall, given the homogeneity of the population used in the present study, one should exercise caution when interpreting the results. The best configurations identified in the present study could differ from one population to the other. Designers are therefore encouraged to select a subject sample large enough to be representative of the targeted population and provide the wealthiest possible information to the machine learning algorithms. Finally, the present work does not address the question of a multi-sensing environment. Given that alternative sensing options could already be embedded in other type of devices (e.g., activity trackers, smartphones), one could consider that smart shoes should primarily specialize in the collection of information on the foot-ground interaction. The present protocol allows focusing on the sole performance of plantar pressurebased activity recognition to assess the relevance of including smart shoes in a network of devices dedicated to physical activity evaluation <ns0:ref type='bibr' target='#b8'>(Chen et al., 2016</ns0:ref><ns0:ref type='bibr' target='#b11'>, Eskofier et al., 2017)</ns0:ref>.</ns0:p></ns0:div>
<ns0:div><ns0:head>Conclusions</ns0:head><ns0:p>PeerJ reviewing PDF | (2020:05:48920:1:1:NEW 4 Sep 2020)</ns0:p></ns0:div>
<ns0:div><ns0:head>Manuscript to be reviewed</ns0:head><ns0:p>In this work, random forest modules as behavior recognition algorithms for plantar pressure measurement using smart shoes were explored and proved relevant. Indeed, smart shoes mounted with seven pressure sensors and extracting 167 plantar pressure data features could recognize nine different daily life activities with an average of good prediction of 0.89. Interestingly, the results suggest a marginal reduction of performance for configurations downgraded to two, three, four, five, or six sensors and the computation of approximately 20 plantar pressure data features, which could ease the design and manufacturing of smart-shoe products. Future studies are necessary to generalize the present findings to a larger sample of the population and larger number of behaviors. Considering the trend toward the development of wearable devices with 5G capacities, smart shoes could become a crucial element of systems allowing self-monitoring of physical activity, thus having an important role in promoting active and healthy lifestyles. Zoom view on a selected branch of one regression tree of one selected forest. During the training process, the nodes (diamonds) are split until all data points correspond to one activity. At each node, the decision is based on the parameter that best discriminates the sample into two sub-samples. The process is repeated until the generation of a pure offspring, i.e., leaves (rounded corner rectangles) containing the data points of one given activity only. The full tree is available in the Supplementary Material 1 (window length: 20 s, configuration: seven sensors, assignment: 1, run: 1). The number of features depended on the number and location of the sensors. The number of features from left to right, one sensor: 29, 29; two sensors: <ns0:ref type='bibr'>54, 54, 54, 51; three sensors: 76, 76, 76, 73; four sensors: 98, 101, 98, 98; five sensors: 120, 123, 123, 123, 120, 120; six sensors: 142, 145, 145, 142</ns0:ref>; and seven sensors: 167. Green boxes: expected to perform well <ns0:ref type='bibr'>(positions 1, 2, 3, 4, 7, 8, 11, 12, 13, 15, 16, 17, 18, 21, 22, 23, 25</ns0:ref> from left to right). Pink boxes: expected to perform poorly <ns0:ref type='bibr'>(positions 5,6,9,10,14,19,20,24</ns0:ref> from left to right) (cf. Manuscript to be reviewed Values refers to the normalized sensitivity of the selected forest for each activity. (A) One selected bad performer with a relatively low number of used features. (B) One selected good performer with a relatively low number of used features. (C) One selected bad performer with a relatively high number of used features. (D) One selected good performer with a relatively high number of used features. 'Best' refers to the results obtained from the forest with the highest prediction rate. 'Worst' refers to the results obtained from the forest with the lowest prediction rate. 'Performer' here refers to one single forest. N: number of features.</ns0:p><ns0:note type='other'>Figure 3</ns0:note><ns0:note type='other'>Figure 13</ns0:note></ns0:div><ns0:figure xml:id='fig_0'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:05:48920:1:1:NEW 4 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_1'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:05:48920:1:1:NEW 4 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_2'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:05:48920:1:1:NEW 4 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_3'><ns0:head>Figure 1 Figure 1 .</ns0:head><ns0:label>11</ns0:label><ns0:figDesc>Figure 1</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_4'><ns0:head>Figure 2 Figure 2 .</ns0:head><ns0:label>22</ns0:label><ns0:figDesc>Figure 2</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_5'><ns0:head>Figure 3 .</ns0:head><ns0:label>3</ns0:label><ns0:figDesc>Figure 3. Overview of the machine-learning procedure.</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_6'><ns0:head>Figure 4</ns0:head><ns0:label>4</ns0:label><ns0:figDesc>Figure 4</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_7'><ns0:head>Figure 4 .</ns0:head><ns0:label>4</ns0:label><ns0:figDesc>Figure 4. Chart of the 3-stage data processing flow.</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_8'><ns0:head>Figure 5 Figure 5 .</ns0:head><ns0:label>55</ns0:label><ns0:figDesc>Figure 5</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_9'><ns0:head>Figure 6 Figure 6 .</ns0:head><ns0:label>66</ns0:label><ns0:figDesc>Figure 6</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_10'><ns0:head>Figure 7 Figure 7 .</ns0:head><ns0:label>77</ns0:label><ns0:figDesc>Figure 7</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_11'><ns0:head>Figure 4 )</ns0:head><ns0:label>4</ns0:label><ns0:figDesc>Figure4). Red diamonds: mean values.</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_12'><ns0:head>Figure 8 Figure 8 .</ns0:head><ns0:label>88</ns0:label><ns0:figDesc>Figure 8</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_13'><ns0:head>Figure 9 Figure 9 .</ns0:head><ns0:label>99</ns0:label><ns0:figDesc>Figure 9</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_14'><ns0:head>Figure 10 Figure 10 .</ns0:head><ns0:label>1010</ns0:label><ns0:figDesc>Figure 10</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_15'><ns0:head>Figure 11</ns0:head><ns0:label>11</ns0:label><ns0:figDesc>Figure 11</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_16'><ns0:head>Figure 11 .</ns0:head><ns0:label>11</ns0:label><ns0:figDesc>Figure 11. Identification of most important features.</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_17'><ns0:head>Figure 12</ns0:head><ns0:label>12</ns0:label><ns0:figDesc>Figure 12</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_18'><ns0:head>Figure 12 .</ns0:head><ns0:label>12</ns0:label><ns0:figDesc>Figure 12. Confusion matrices for four selected configurations using different numbers of features.</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_19'><ns0:head>Figure 13 .</ns0:head><ns0:label>13</ns0:label><ns0:figDesc>Figure 13. Confusion matrices for selected forests captured from the best 4-sensor configuration.</ns0:figDesc></ns0:figure>
<ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='23,42.52,70.87,525.00,396.00' type='bitmap' /></ns0:figure>
<ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='30,42.52,70.87,525.00,393.75' type='bitmap' /></ns0:figure>
<ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='41,42.52,70.87,525.00,402.00' type='bitmap' /></ns0:figure>
<ns0:note place='foot'>PeerJ reviewing PDF | (2020:05:48920:1:1:NEW 4 Sep 2020)</ns0:note>
<ns0:note place='foot' n='6'>PeerJ reviewing PDF | (2020:05:48920:1:1:NEW 4 Sep 2020)</ns0:note>
</ns0:body>
" | "Dear Dr. Justin Keogh,
Thank you very much for giving us the opportunity to revise our manuscript.
We would like to apologize for the delay of our answer. Additional computations have been necessary to address some of the comments. We now have a full picture of the results, which comforts our prior interpretations. We have made several modifications to address all comments of both reviewers. Reviewer 1 wanted some changes in the results section, which were incompatible with our method, while reviewer 2 found our method “meaningful”. We believe that the choices made to revise the manuscript will content both reviewers. The method used in the previous version was preserved but better described (as recommended by the reviewer 1), while the results of some supplementary analyses using a different methodology (more comprehensive) have been added in the supplementary material section and briefly described in the core of the manuscript.
We are grateful for the improvements to the manuscript that have resulted from the requested revisions.
Please find our detailed response to reviewers with our alterations appear in green.
Kind regards (on behalf of all coauthors),
Julien Tripette
Reviewer 1
Basic reporting
no comments
Experimental design
no comments
Validity of the findings
no comments
Comments for the Author
The paper is well structured and has systematic exploration of using plantar pressure for activity classification. However, due to excessive details about sensor and feature selection, the paper is a little difficult to understand specifically for results.
Thank you for the positive comment and for having pointed out this issue in clarity.
Regarding this specific issue, a chart describing the data processing flow has been included in Figure 4 (and the Table 3 of the previous version has been removed).
Figure 4. Chart of the 3-stage data processing flow. Stage 1: “window length”. Stage 2: “number and location of sensors”. Stage 3: “number of features”. Green bars: sensor configurations that were expected to perform well. Pink bars: sensor configurations that were expected to perform poorly. The orange/dotted connectors indicate the logical links between each stage of the analysis.
In addition, the sub-section “Data analysis framework” has been added in the method section:
Line 244-255: “As illustrated in Figure 4, the analyses are integrated in a 3-stage logical flow.
1) The window length analysis aims at identifying the optimum analytic window length.
2) The analysis of a pre-selected set of 25 sensor configurations, i.e. configurations using the information of different numbers of sensors and/or the information of sensors placed at different locations, aims at identifying the best hardware combination for each possible number of sensors ranging from 1 to 6 (the 7-sensor configuration only has 1 possible combination). This analysis was conducted using the optimum window length identified in 1).
3) A final analysis exploring the contribution of each feature to the forest outputs aims at finding the most efficient number of features to be used for each of the seven best sensor configurations identified in 2). Again, this analysis was conducted using the optimum window length identified in 1).
Further details related to each of the three stages are given in the three following subsections.”
The subsection titles have been changed as follows in both the method and results sections.
Line 257, 306: “window length” “Stage 1: window length”
Line 265, 321: “Number and location of sensors” “Stage 2: number and location of sensors”
Line 278, 355: “Number of features” “Stage 3: number of features”
Finally, we have made the following addition at the end of the introduction to inform the reader about our intention to lead the analysis within one predetermined logical frame, rather than brute-forcing the analysis for all possible (tens of thousands) hardware/software combinations:
Line 117-120: “The data analysis is conducted using machine learning methods. The identification of the best hardware and software configurations is conducted through a data processing logical frame, which may be re-used by designers willing to develop smart-shoes devices.”
We hope these changes make the paper less difficult to read.
Several other changes have been done throughout the manuscript in accordance with the reviewer’s comments. As recommended, the manuscript (including the results section) underwent some major modifications, which are described in line with our answer to comments 1-10. We sincerely thank the reviewer for the improvements that have resulted from the requested revisions.
Here are some comments:
1) I generally recommend the authors to remove the descriptions about energy expenditure in the paper, considering the paper does not work on algorithm related to energy expenditure and the paper focuses on activity classification only.
We agree. We have removed most of the instances referring to energy expenditure predictions. Several modifications (withdrawing mainly) have been made in the abstract (line 23-27), in the introduction (line 55, 69-70), in the discussion (line 451-453, 561), and sporadically, elsewhere to align the content, when we referred to the “energy expenditure” or 'quantification of physical activity' in the previous version.
2) I recommend the authors to rewrite result section. The authors used all sensors and features configuration to determine optimal window size. And then use the determined window size for the following sections. This raised a question that maybe in other configuration of sensor and features, this window size is not optimal. Please give more explanations about this. Why do the authors simplify the selection of window sizes while selecting features and sensors?
As recommended, several modifications have been done in the results section. Before listing these changes, we would like to make some clarifications. The selection of the optimum (rather than best) window length is not only a machine learning question and relates more widely to the issue of the temporal resolution. This question is discussed in the first paragraph of the discussion sub-section entitled “Temporal resolution, sensor configuration, number of features, manufacturing, and algorithmic considerations” (line 518-530). In short, the aim of the “window length” analysis was to identify the shortest possible analytic window allowing the computation of an “acceptable” rate of good predictions. The aim was not to find the window length with the best absolute score. In the method section, we mention:
Line 261-263: “The optimum window length was defined at the point where the slope of the function describing the prediction rate vs. window length began to decrease. This optimum window length was used for all subsequent analysis.”
Moreover, the data used in the present analysis were acquired using a structured experimental design. No pollution of one experimental activity by another activity could have happened in our protocol. This would be very different in real-life settings. In short, the longer the window length, the higher the chance to have a combination of activities within the span of one analytic window, a situation that the current experimental design is not covering. The application of our method to real-life situations would only be done at window lengths small enough to avoid as much as possible a mixture of activities during one analytic period. Further elaborations are made in the discussion section. Again, these considerations appear in the sub-section entitled “Temporal resolution, sensor configuration, number of features, manufacturing, and algorithmic considerations” (line 518-530).
In order to address the reviewer’s concern, and clarify our aim to the readers, we have made the following changes:
In the method section (line 245): “The window length analysis aims at identifying the optimum analytic window length.”
In the results section (line 311-314): “The average prediction rates for window lengths between 20-60 s showed marginal variations within the 0.89-0.90 range. To preserve the highest possible temporal resolution for future applications, 20 s was selected as the optimum length. The rest of the analyses were conducted using a 20-second window length.”
In addition, the specific aim of the “window length” analysis is appearing in the new Figure 4 (left panel: “stage 1”): “Selecting the optimum window length”.
We hope that these modifications, taken together with the considerations that already appeared in the discussion (see above), are sufficient to orient the reader toward this idea of “optimum” rather than “highest performance” window length.
This being said, the intuition of the reviewer was correct. We have run our main code, not only for a 20-second window length, but for all window lengths suggested in the manuscript (1, 5, 10, 15, 20, 25, 30, 35, 40, 45, 50, 55, 60 seconds) and observed better results at longer window lengths. This exhaustive analysis has been performed on an extended panel of 67 sensor configurations. However, only the results of the 25 sensor configurations selected in the manuscript are summarized in Supplementary Material 4 (duplicated below). Panel A follows the same format as Fig. 5 shown in the manuscript. The bars describe the best result possible for each of the selected 25 sensor configurations trained and tested with all possible numbers of features configurations repeated at all window lengths. The relation between the window length and the rate of good predictions shows a similar pattern in both Supplementary Material 4A and Fig.5. However, the new analysis reveals higher scores for the window lengths between 5 and 15 seconds. For window lengths between 20 and 60 seconds, the average scores seem marginally higher (about + 0.01). With respect to our selection criterion detailed in the method section (line 261-262), the optimum window length would still be 20 seconds when considering this new set of results. In addition, as shown in Supplementary Material 4B (which follows the same format as Fig. 7 shown in the manuscript), the sensor configuration expected to perform well (green) indeed performed better when tuning the number of features and the window length. Some average rates are ~0.02 higher compared to the results presented in the manuscript. Some single forests may reach a rate of 0.94 of good predictions. The higher scores observed in Supplementary Material 4B indeed are associated with longer window lengths. The Supplementary Material 4C is the same as the Supplementary Material 4B but with the window size locked at 20 seconds. The results are very similar to the one shown in the manuscript in Fig. 7. This new analysis shows that the window length variable impacts significantly the sensitivity of our activity recognition algorithm.
While these new results may look more promising than the ones included in the core of the manuscript, the elaborations done earlier on the selection of the optimum rather than the best window length still prevail. Selecting a longer window length impacts negatively the time resolution and would make real-life applications more challenging.
In addition, one may not omit one current limitation of our study. As noted in the discussion (“limitations and strength,” line 570-576): “the present protocol includes a relatively homogenous population. […] a more heterogeneous sample of the population must be tested before generalizing further the results of the present study”. Therefore, chasing the very best forests over a pool of millions with no logical framework leading the data processing flow might not make any sense to us.
Supplementary Material 4. Summary of the supplementary analyses following the formats of Figure 5 and 7. A: window length effect on activity recognition rate (best combination for “number and location of sensors” and “number of features”). Pink boxes: 1, 5, 10 and 15 seconds. Green box: 20 seconds (considered optimum). Yellow boxes: 25, 30, 35, 40, 45, 50, 55 and 60 seconds. Red diamonds: mean values. B: Performance of activity recognition of random forest algorithms for 25 sensor configurations (best combination for “window length” and “number of features”). Green boxes: sensor configurations that were expected to perform well. Pink boxes: sensor configurations that were expected to perform poorly. Red diamonds: mean values. C: Performance of activity recognition of random forest algorithms for 25 sensor configurations (“window length”: 20 seconds, “number of features”: best average rate of good prediction). Green and pink boxes: same chart as for panel B.
However, we agree with the reviewer, it is important to show the reader the whole potentiality of plantar pressure data for the recognition of physical behaviors. Therefore, to comply with the reviewer request, Supplementary Material 3 and 4 have been added to the Supplementary Material section. A major amendment to the results section has been made (line 406-425).
“Supplementary results
A more comprehensive analysis has been conducted considering a larger panel of 67 sensor configurations. Random forest modules were systematically created and tested for each window length candidates (1-60 s) and each possible number of features (maximum to one), without any selection of the best sensor configurations like in the 3-stage data processing flow presented in Fig. 4 (the results of which have been presented in the above three subsections). The machine learning procedure was the same as the one detailed in the method section. Therefore, 75,172 additional analyses have been completed, resulting in the computation of 7,517,200 forests.
These supplementary analyses were associated with higher prediction scores, highlighting the whole potential of using plantar pressure data for the recognition of physical behaviors. As shown in the Supplementary Material 3, 297 sensor configurations were associated with at least one forest presenting a prediction score of 0.92 or more. Regarding the highest scores, at least 12 forests presented a rate of good predictions of 0.94. The best average scores ranged from 0.54 to 0.91, a scale similar to the one of the results of the 3-stage analysis (Supplementary Material 2). Eighty-seven sensor configurations were associated with average rates of good predictions of 0.90 or more. As shown in Supplementary Material 4B, the best performances observed in these analyses are systematically associated with window lengths of 30 s or longer. Regarding the identification of an optimum analytic window length, the results still points to a period of 20 s (Supplementary Material 4A). The question of the time resolution is discussed later in the manuscript. The results of these supplementary analyses are summarized in Supplementary Material 3 and 4.”
Please note that this analysis was possible using the code shared with the previous version of the manuscript. However, the ReadMe.md of the github repository has been slightly updated to includes clear directions on how to run the code at different window lengths and for a sample of 25, 67 or 127 sensor configurations. Running the main code for all window lengths (1 to 60 seconds) for an extended sample of 67 sensor configurations at all possible number of features, needed 7 days of calculation. Running the code for all the 127 configurations would have double this time. Here are our hardware specs: Ubuntu 18.04 running on an Intel Xeon CPU with 32 cores (64 threads) at 2.40GHz and 64GB of RAM. We believed that such a long computation time gives more credit to the 3-stage approach that we opted for in the present study.
3) In abstract, please give a concise description for validation method? Like cross validation or hold-out validation?
Among usual terminologies, the single expression that would best describe our protocol is “subject-wise validation”. This terminology is grasping the most important feature of the machine learning design, i.e., one given subject label cannot appear in both the training and testing samples in one analytic run (see our motivations line 572-574). The predictions obtained using such a protocol are robust to inter-individual differences.
We did not use a cross validation design. What may best describe our protocol is the term “multi-hold-out”. The hold-out validation procedure was repeated 5 times using different assignments to training and test samples each time.
We made the following changes in the abstract
Line 35-39: “Data were split into training and test samples using a subject-wise assignment method. A random forest model was trained to recognize activity. The resulting activity recognition algorithms were evaluated on the test sample. A multi hold-out procedure allowed repeating the operation with 5 different assignments.”
In addition, the expression “subject-wise” is now used in the method section (line 232) and throughout the manuscript where the term “subject-independent” previously appeared. Finally, we added the Figure 3 to illustrate our training-test design.
Figure 3. Overview of the machine-learning procedure. Training-test ratio: 6-11 (subject-wise). Testing method: multi- hold-out (data assigned to 5 different training-test combinations). For each assignment, 20 runs are conducted using different random subsets of the dataset (or “random states”). Blue: training samples. Salmon: test samples.
4) In Line 40-43, all prediction accuracy percentage is obtained under window size of 20 seconds?
Yes, it was obtained with a 20-seconds window size. The following changes were made in the abstract.
Line 39-40: “The analytic conditions were modulated to test 1) different window lengths (1-60 seconds) […].”
Line 41-42: “A window length of 20 s was found to be optimum and therefore used for the rest of the analysis.”
Related to the abstract section, we have also made the following addition to detail a little more the description of the results.
Line 46-47: “Reducing the number of features down to 20 does not alter significantly the performance of the algorithm.”
5) For one category of activity, sitting, there could be some confounding factors which influence the measurements? Could the author release information about height of the chair? As we know, if the subject's lower leg is shorted than the height, the foot will be not on the ground while sitting and plantar pressure is 0. In some cases, the foot could be on the ground.
Thank you for this remark. The height of the chair was 45cm. All subjects were able to touch the ground with their feet when they were sitting. There was no further instruction related to the sitting posture, in order to favor real-life sitting behavior. The information about chair height is now included in Table 1: “Chatting and browsing the internet with smartphone while sitting on an office chair (height: 45cm, all subjects able to touch the ground with their feet when sitting), indoors.”
6) In line 74-75, nature of acceleration data does not allow refining of the classes until all daily life activities are identified. It is unclear and why?
We agree, our point was unclear. This section had been modified in order to address the first comment of the reviewer (line 69-74). This statement does not appear anymore in the revised manuscript.
7) In 232-233, the authors mentioned 20 train-validation runs? so all runs is on training dataset (6 subjects)? Is it cross validation? If the reviewer's understanding correct, the validation dataset (11 subjects) might be better called testing dataset to avoid confusion?
The reviewer's understanding is correct. We did not use the right terminology. The term “validation” has been changed for “test” or “testing” throughout the manuscript.
The training was stochastic. For one given assignment, the data of the training sample (6 subjects) are used for building twenty different original forests. To reach this number, as shown in the shared code, the “random_state” (or “seed” in the computer science terminology) parameter is given 20 different values. The “random_state” influence the composition of the data subsets used for building predictors, resulting in the creation of one single forest each time the “random_state” parameter takes a different value (over a pool of 20 possible values). In the present study, the “bagging” method was used to build the predictors (Breiman, 2001). Thus, twenty original forest are created, i.e. one for each configuration of the “random_state” parameter (see. scikit-learn module: https://scikit-learn.org/stable/modules/generated/sklearn.ensemble.RandomForestClassifier.html).
These descriptions are detailed in the manuscript (line 220-229), and one reference points to the scikit-learn library (line 218: “Pedregosa et al., 2011”)
No cross-validation process was performed during the training process. Then, each single forest (over the pool of 20 generated forests) is evaluated using the data of the test sample (the 11 remaining subject not assigned to the training sample). Consequently, running the “RandomForestClassifyer” function 20 times, i.e with the “random_state” parameter taking 20 different values, results in the training and the testing of 20 original random forest modules. The operation was repeated for the 5 different assignments (“multi hold-out).
We believe that the additional Figure 3 (shown above) makes the whole machine-learning design clearer in the revised version.
8) From 240-274, the window size, number of sensors and features are all determined by training dataset? […]
Thank you for this question. No, of course they were determined through the actual performance evaluated using the data of the test sample only. No score related to the processing of training data were/are reported in the manuscript. We have made the following addition in the method section to make it clear to the reader.
Line 239-241: “All the results presented in this manuscript correspond to the outcome of the evaluation of the random forest modules using the test samples only. None of the reported scores are related to the training phases.”
[…] Please ensure not to touch the validation dataset while doing sensor and feature selection and hyper-parameters configuration. The dataset deriving reported accuracy percentage should not be used in model configuration.
The modules evaluated at one stage of the study, e.g., “window length” or “number and location of sensors” are not touched or reused for any of the subsequent analyses, e.g., “number and location of sensor” or “number of features”. We restart the training from scratch at each stage. In the method section, the following sentences have been modified to ease the understanding of the reader.
Line 266-267: “The processing was repeated from scratch for different sensor configurations, i.e., different location and/or number of sensors, for the optimal window length only.”
Line 279-281: “The processing was again repeated from scratch with the best configurations only and for a decreasing number of features, which were removed one-by-one based on their discriminating capacities (Fig. 4).”
We believe that the details above paired with the information included in the new subsection “Data analysis framework” (line 243-254 in the method section, see previous answers), and the new Figures 3 and 4 (see. our answer to the previous comments) give a more accurate picture of our protocol. We are not conducting any cross-validation procedure. Consequently, no overlap between a cross-validation sample and the test samples could occur. This confusion may be the result of the wrong terminology we used in the previous version of the paper (“validation” instead of “test”), which has been corrected. We apologize.
9) From line 282-284, the authors first mention the best accuracy is achieved at 45s, and then what does best forest mean at 30 seconds? […]
We apologize that we did not explain clearly the terminology used for the description of the results. We added the following details in the top paragraph of the result section.
Line 295-304: “The “prediction rates” and “rates of good predictions” presented in the text and figures refer to the accuracy, calculated as follows: correctly predicted sample/total number of samples. When reporting statistical results, the terms “average” and “mean” point to the average accuracy across the 100 forests of one round of evaluations (see Fig. 3). The expressions “best single forest” and “best performer” refer to the one single forest that showed the best accuracy score among the 100 forests produced for one round of evaluations. Conversely, the term “worst single forest” points to the one single forest that showed the worst accuracy score among the 100 forests produced for one round of evaluations. Logical links between the 3 stages of the analysis are shown in Fig. 4. The values indicated at the intersections of “true label” and “prediction” in the confusion matrices refer to sensitivity, calculated as follows: true positives/(true positives + false negatives).”
We aligned the terminology accordingly, in the results section (line 345, 353, 385, 386, 391, 398, 401). We changed a number of instances where using the term “performer” was somewhat confusing (line 347, 351, “performer” changed for “configuration”). Finally, to avoid any confusion, we have withdrawn the description of results for the 30-second window length (line 310).
[…] And then why the author select 20 seconds as the optimal window size. Please give clear explanations of rules about selection of window size.
We thank the reviewer for this comment. We have addressed the issue of window length selection in our answer to the comment #2.
10) In number of sensors and features subsection, the protocol of evaluation is not clear. Did the authors first determined sensor number while using all features? And then fixing the optimal sensor number and then work on feature selection? Is this understanding correct? […]
We would like to thank the reviewer for this question, but we are afraid that her/his understanding may not be correct for this particular issue. We apologize for the inaccuracies in our protocol description that lead to this misunderstanding. We hope that the changes made to address the previous comments, will help the reader:
• data processing flow chart (Figure 4)
• new subsection “Data analysis framework” (line 243-254 in the method section, or see above)
As shown in the new Figure 4, only the 7-sensor configuration was used to create random forest modules and identify the optimal window length (Stage 1). At this stage, the computations were performed using all the maximum number of features only, i.e. all the 167 there were possible to compute. In stage 2 (“number and location of sensors”), the “sensor number” (here we read “the sensor configurations, including both the number and location of sensors”) has been preselected before any computation was performed. This selection was done using a number of criteria that are described in the method section.
Line 267-272: “Twenty-five configurations were selected among the 127 possible combinations. The selection was performed using subjective criteria: 1) reproduction of a selection of the configurations found in the literature or in the industrial sector of running shoes, 2) promotion of combinations allowing the collection of relevant information for the prediction of gait and postural behaviors, and 3) selection of combinations that are believed to perform poorly (Fig. 4).”
We hope that the above small syntax change now makes the whole point clearer.
[…] It seems there are a lot of combinations of sensors and features? For 7 sensors, the total number of combinations is 7*6*5....*1. If considering the feature number, the combination is more various. […]
We thank the reviewer for this remark. The number of sensor combinations can actually be calculated with the following equation:
Number of combinations = 27–1
Where “2” refers to the two possible conditions, i.e. included-not included, “7” is the number of sensors, and “-1” removes the non-existent 0-sensor combination.
The total number of combinations is 127 as described in the method (see above or line 268).
For the 25 pre-selected sensor configurations tested at one given window length, the total number of configurations that includes variation on the number of features is 2437. For the 7 “best” selected sensor configurations that were assigned to the “number of features” analysis, the total number of configurations that includes variation on the number of features is 686 for one given window length (20 s in the paper). These numbers now appear in the method section too.
Line 289-291: “A total of 686 combinations of sensor configurations and number of features were tested (i.e. best 1-sensor: 29, best 2-sensor: 54, best 3-sensor: 76, best 4-sensor: 98, best 5-sensor: 120, best 6-sensor: 142, 7-sensor: 167).”
They also remain in the caption of Figure 7 in the revised version.
[…] The reviewer accept some degree of simplification, but please give more clear explanations.
We thank the reviewer and share the same opinion. We also believe that some degree of simplification is necessary to ease readers’ understanding. This is why we had initially chosen the 3-stage data processing flow (1) “window length”, 2) “number and location of sensors”, 3) “number of features”). We apologize that the first description of this simplification was not clear enough. We have added Figure 4 that provides a clear chart of the data processing flow and emphasize the logical link between each stage of the analysis. We have made a significant amount of amendments and modifications to remove the sources of misunderstanding. The code shared with the previous version of the manuscript already allowed further exploration of the dataset. However, we have now included a more exhaustive version of our results in the supplementary material section so that the reader interested in a more detailed version of the results does not need to go through time-consuming computations. The code has also been modified so that any reader can easily launch 3 types of analysis with either 25, 67 or 127 sensor configurations.
It was our great pleasure to address the reviewer’s comments. We believe the manuscript has been improved in many aspects. We sincerely thank the reviewer for her/his significant contribution to our report.
Reviewer 2
Basic reporting
The rational for this study is clearly presented, with appropriate links to literature in the field. Authors provided sufficient field background, relying on adequate references.
Thank you for the positive comment.
Experimental design
Authors used meaningful methods, which are well described. I only have two questions:
1. Did participants perform the 9 tasks in a pre-defined order or randomly-selected order? This can possibly influence the performance of algorithms to recognize activities.
We thank the reviewer for this comment related to the experimental protocol and the method. No, we did not set any predefined order for the 9 activities. The following sentence has been added in the method section to emphasize this feature of the protocol.
Line 147-148: “The order in which the 9 activities were completed was randomly selected for each subject.”
2. To train algorithms, authors used a five-different subject-independent training–validation assignments. Would a higher number of subject for training have changed the performance of algorithms regarding good predictions?
We thank the reviewer for this comment. This remark is absolutely important. Our preliminary analyses were conducted using either a 30-70 or a 50-50 training-test assignment rate. Slightly better results were obtained with a larger training set. However, we opted for a more conservative approach to balance one limitation of our paper, which is the relatively homogeneous population. The following modification has been done in the discussion (subsection: “limitations and strengths”) to mention this strength:
Line 571-584: “[…] the present protocol includes a relatively homogenous population. Subjects were all healthy women. To address this potential issue, five different subject-wise training–test assignments were used to develop and test the forests. In addition, a conservative 6-11 training-test assignment ratio has been used to limit the wealth of the available information during the training phase and create more challenging conditions relying on inter-individual differences. However, a more heterogeneous sample of the population must be tested before generalizing further the results of the present study. A more heterogeneous population would indeed provide a more diverse information to the training algorithms, which could also result in increased good prediction scores. Overall, given the homogeneity of the population used in the present study, one should exercise caution when interpreting the results. The best configurations identified in the present study could differ from one population to the other. Designers are therefore encouraged to select a subject sample large enough to be representative of the targeted population and provide the wealthiest possible information to the machine learning algorithms.”
Validity of the findings
Results are presented with details; text, figures and tables complement each other. Even the algorithms showed acceptable performance, there are room for improvement. This inspire my following question:
The average good prediction was 0.89, which is good but not perfect. Would a more heterogenous sample have contributed to increase the performance of the algorithms?
Thank you for this comment. Yes, as discussed in the manuscript, the lack of heterogeneity of our population is one important limitation of the study. First, it prevents generalizing too much the results. Conceptually, in a subject-wise machine learning protocol, a homogenous population makes the evaluation easier since the subjects assigned to the test sample share the same characteristics as the subjects assigned to the training sample. Second, it curbs the learning process since the corpus of information available to the training algorithm to describe the targeted phenomena (physical behaviors in the present study) is limited to the few characteristics shared by the individual in the homogenous population (healthy Asian women in the present study). Related to this second point, and as suggested by the reviewer, we indeed believe that higher scores could have been produced with a more heterogeneous sample.
We have seen a link between this comment and the previous one. Consequently, these two comments were addressed together in the manuscript. We hope that the modifications made in the subsection “limitation and strength” (line 568-581 or see above) are sufficient to address this point as well.
We also included the following remark in the abstract.
Line 51-52: “Future experiments must include a more heterogeneous population.”
Comments for the Author
This study reports results on the performance of random forest algorithms for the recognition of daily life activities using plantar pressure information collected with a smart-shoe. For these goals, authors enrolled 17 female healthy subject with mean age of 26 ± 9 years old who took par in the experimental study. Their results confirmed the possibility of high-performance human behavior recognition using plantar pressure data only. The study is interesting. The manuscript is well written and conclusions are supported by reported results. There are only few aspects I would like authors to comment and possibly consider the above questions to improve the quality of their manuscript.
We thank the reviewer for the positive comments overall. It was our great pleasure to address the reviewer’s comments. The questions allowed us to back our approach and underly some strengths of our protocol. In particular, we believe that the question of limitations vs. strengths will be clearer for the readers. We sincerely thank the reviewer for her/his significant contribution to our report.
" | Here is a paper. Please give your review comments after reading it. |
9,817 | "<ns0:abstract xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:p>Background: Wearable activity trackers are regarded as a new opportunity to deliver health promotion interventions. The algorithms typically contain two stages: 1) activity classification and 2) quantification of energy expenditures by applying activity-specific regression equations. Although the predictions of traditional activity trackers rely primarily on the processing of accelerometer sensor data, the emergence of smart clothes with multi-sensing capacities could contribute to a more accurate evaluation of daily physical behaviors. This study aims to 1) develop an activity recognition algorithm based on the processing of plantar pressure information provided by a smart-shoe prototype and 2) to determine the optimal hardware and software configurations. Method: Seventeen subjects wore a pair of smart-shoe prototypes composed of plantar pressure measurement insoles, and they performed the following nine activities: sitting, standing, walking on a flat surface, walking upstairs, walking downstairs, walking up a slope, running, cycling, and completing office work. The insole featured seven pressure sensors. For each activity, four minutes of plantar pressure data were collected. The plantar pressure data were cut in overlapping windows of different lengths and 167 features were extracted for each window. A random forest model was trained to recognize activity using some selected sensor configurations and different numbers of data features. The resulting activity recognition algorithms were evaluated using data samples specifically directed to validation trials. Results: Using all the sensors and all 167 features, the smart shoes predicted the activities with an average success of 89%. 'Running' demonstrated the highest percentage of good predictions (100%). 'Walking up a slope' was linked with the lowest performance (63%), with the majority of the false positives being 'walking on a flat surface' and 'walking upstairs.' Some 2-and 3-sensor configurations were linked with an</ns0:p></ns0:div>
</ns0:abstract>
<ns0:body xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:head>Introduction</ns0:head><ns0:p>The promotion of an active lifestyle among our populations remains an on-going problem <ns0:ref type='bibr' target='#b1'>(Barreto, 2013)</ns0:ref>. Fortunately, the recent boom in the marketing of activity trackers provides new tools to address this issue. The term 'activity tracker' is defined as a category of wearable devices, which aims to provide users with feedback on their physical behaviors, physical fitness, and physical activity. This feedback can be provided through a wide variety of parameters, including 'stepcount,' time spent in activities of selected intensities (sedentary, light, moderate, or vigorous activities), number of floors climbed, and daily energy expenditures (expressed in kilocalories). This type of device has been demonstrated to be effective in supporting active lifestyles and is now widely considered in the development of health promotion policies <ns0:ref type='bibr' target='#b4'>(Bravata et al., 2007;</ns0:ref><ns0:ref type='bibr' target='#b3'>Bonomi & Westerterp, 2012;</ns0:ref><ns0:ref type='bibr' target='#b13'>Gal et al., 2018</ns0:ref><ns0:ref type='bibr' target='#b16'>, Jennings et al., 2017)</ns0:ref>.</ns0:p><ns0:p>From a technological perspective, the majority of contemporary activity trackers integrate one MEMS 3-axis accelerometer chip that allows the sensing of the user's body motion. They are typically worn at the hip or wrist and provide feedback on the amount of daily physical activity <ns0:ref type='bibr' target='#b31'>(Romanzini, Petroski & Reichert, 2012;</ns0:ref><ns0:ref type='bibr' target='#b18'>Kamada et al., 2016)</ns0:ref>. State-of-the-art algorithms directed at evaluating physical behaviors typically feature an activity classification method <ns0:ref type='bibr' target='#b35'>(Staudenmayer et al., 2009</ns0:ref><ns0:ref type='bibr' target='#b25'>, Ohkawara et al., 2011</ns0:ref><ns0:ref type='bibr' target='#b2'>, Bassett, Rowlands & Trost, 2012)</ns0:ref>. The following are examples of activity classes that are frequently proposed when using the information of one single accelerometer: locomotive vs. non-locomotive vs. mixed activity and sedentary vs. light intensity vs. moderate intensity vs. vigorous intensity activities <ns0:ref type='bibr' target='#b19'>(Karabulut, Crouter & Bassett, 2005</ns0:ref><ns0:ref type='bibr' target='#b27'>, Oshima et al., 2010)</ns0:ref>. However, latest trackers now feature multi-sensing technologies (e.g., gyroscope, altimeter, light reflectance, thermal resistor), increasing the amount of available information (e.g., inclination, altitude, heart rate, skin temperature) for physical behavior evaluation, and calling for the development of algorithmic suites able to handle the full wealth of available information <ns0:ref type='bibr' target='#b6'>(Chen & Bassett, 2005</ns0:ref><ns0:ref type='bibr' target='#b28'>, Park et al., 2011)</ns0:ref>. This trend toward multi-sensing evaluation is expected to benefit from current innovations in the field of wearable technologies and smart clothes, which are designed to work in an interconnected network of 5G devices. In such a fast-evolving context, the current methods could rapidly become outdated, and smart clothes able to collect physiological or mechanical information could assume an ever more central role in the evaluation of physical activity <ns0:ref type='bibr' target='#b15'>(Intille, 2012;</ns0:ref><ns0:ref type='bibr' target='#b7'>Chen et al., 2012;</ns0:ref><ns0:ref type='bibr' target='#b8'>Chen et al., 2016)</ns0:ref>. In the near future, higher-level algorithms will function in the cloud and be capable of collecting available data from a large number of connected devices, and they can select the most relevant information depending on the context to proceed to a continuous and ever more accurate evaluation on physical behaviors. Among these, smart shoes or smart insoles oriented toward the assessment of physical behaviors could be used to evaluate the interaction with the ground and to help refine activity classification.</ns0:p><ns0:p>Medical insoles capable of measuring plantar pressures have already been commercialized as transportable alternatives to force platforms (Fscan, Tekscan, Inc., USA; ParoTec, Paromed GmbH & Co. KG; PedoSmart). These devices provide a reliable analysis of the center of pressure to assess posture, gait stability, mobility disorders, fall risk, and some other physical considerations. Recently, smart-shoe systems intended for athletes have also been proposed (Nike + Sensor, Nike, Inc., USA; SportProfiler, Digitsol, France; Torin IQ, Altra Running, USA; Mijia, Xiaomi, China). They typically provide feedback on plantar pressure distribution, foot landing type, cadence, and contact duration with the ground, among other measurements. To date, smart-shoe systems aimed at monitoring physical behaviors in daily life have only been presented in the scientific literature <ns0:ref type='bibr' target='#b9'>(De Pinho André, Diniz & Fuks, 2017;</ns0:ref><ns0:ref type='bibr' target='#b24'>Ngueleu et al., 2019)</ns0:ref>. Devices mentioning a high rate of activity recognition typically have multi-sensing abilities, including accelerometer sensors, gyroscopes, temperature sensors, and GPS antennas, providing a large amount of information to the prediction algorithm. However, the inclusion of several in-shoe sensors would likely induce higher production costs as well as challenges for product designers. Furthermore, the high rates of behavior recognition presented in the literature are at times inherent to the study protocols, which may only include a limited number of activities or focus on specific clinical populations, thus preventing the generalization of the results <ns0:ref type='bibr' target='#b9'>(De Pinho André, Diniz & Fuks, 2017;</ns0:ref><ns0:ref type='bibr' target='#b24'>Ngueleu et al., 2019)</ns0:ref>.</ns0:p><ns0:p>Hence, a smart-insole or smart-shoe system that only uses plantar pressure information and that could recognize multiple human daily life activities has yet to be developed. The present research aims to develop efficient and effective activity recognition algorithms for smart-insole devices featuring 1-7 plantar pressure sensors. Nine daily life activities are considered. The smart-insole prototype used in the present study is equipped with the 7-sensor plantar pressure measurement insole, described elsewhere <ns0:ref type='bibr'>(Saito et al., 2011;</ns0:ref><ns0:ref type='bibr' target='#b22'>Nakajima et al., 2014)</ns0:ref>. The data analysis is conducted using machine learning methods. The identification of the best hardware and software configurations is conducted through a data processing logical frame, which may be re-used by designers willing to develop smart-shoes devices.</ns0:p></ns0:div>
<ns0:div><ns0:head>Materials & Methods</ns0:head></ns0:div>
<ns0:div><ns0:head>7-sensor plantar pressure measurement insole</ns0:head><ns0:p>The shoe hygienic insoles, which are 2 mm thick, were equipped with seven force-sensing resistors (FSR400, Interlink Electronics, Inc., CA). The sensors respond to stimulation ranging from 0.2-20 N (8.13-813 kPa), allowing the measurement of human peak plantar pressure <ns0:ref type='bibr' target='#b23'>(Nandikolla et al., 2017)</ns0:ref>. The sensors were placed on the heel, lateral midfoot, center of the midfoot, lateral forefoot, center of the forefoot, medial forefoot, and big toe (Fig. <ns0:ref type='figure'>1</ns0:ref>). They were connected to a 12-bit resolution data acquisition unit with a wireless data transmission sampling rate capacity of 100 Hz, allowing real-time recording during normal ambulatory activities. Insoles with a similar configuration have proven to be valid for the evaluation of posture and gait in previous studies <ns0:ref type='bibr'>(Saito et al., 2011;</ns0:ref><ns0:ref type='bibr' target='#b22'>Nakajima et al., 2014;</ns0:ref><ns0:ref type='bibr' target='#b0'>Anzai et al. 2020)</ns0:ref>. Multiple pairs of the insole in different sizes were available.</ns0:p></ns0:div>
<ns0:div><ns0:head>Data collection</ns0:head><ns0:p>The experimental protocol was approved by the Ochanomizu University research ethics committee <ns0:ref type='bibr'>(#2018-01)</ns0:ref>. A total of 17 female subjects signed written consents and participated in the trial (age: 26 ± 9 years old, weight: 49 ± 3 kg). All the participants were healthy and did not present mobility disorders. The 7-sensor plantar pressure measurement insoles were inserted in a pair of commercial sneakers (Vans Fable 2, VF Corporation) with stiff and flat midsoles. The insoles and shoes were available from size 22 cm to 27 cm. The participants wore shoes and insoles that best matched their foot size. They performed the following nine activities: sitting, standing, walking on a flat surface, walking upstairs, walking downstairs, walking up a slope, running, cycling, and office work (Table <ns0:ref type='table'>1</ns0:ref>). The duration of each activity was approximately 4 min, except 'walking on a flat surface' and 'running,' which was approximately 8 min. The order in which the 9 activities were completed was randomly selected for each subject. Eleven subjects completed the nine activities. During the course of the experiment, certain subjects expressed a desire to shorten their participation mainly owing to upcoming agenda conflicts, discomfort, or tiredness. Two subjects completed eight activities, two subjects completed seven activities, one subject completed six activities, and one subject completed five activities. For each subject, data for 'walking on a flat surface', 'walking upstairs', 'walking downstairs' and 'running', respectively, may have been stored in two files. The final dataset consisted of 196 files corresponding to the 140 activities completed by the 17 subjects. Each file contained 14 independent plantar pressure time series (seven sensors for each of the left and right feet).</ns0:p></ns0:div>
<ns0:div><ns0:head>Data preprocessing</ns0:head><ns0:p>The raw data were converted into Newtons (N) and smoothed using a second-order low-pass Butterworth filter with a cutoff frequency of 5 Hz. For each data file, the time series were cut in windows of <ns0:ref type='bibr'>1, 5, 10, 15, 20, 25, 30, 35, 40, 45, 50, 55, and 60</ns0:ref> s, with an overlap of 50%.</ns0:p></ns0:div>
<ns0:div><ns0:head>Feature extraction</ns0:head><ns0:p>For each window, 167 data features were extracted using the information from seven sensors on each foot. The features were extracted from the different types of analysis presented in Table <ns0:ref type='table'>2</ns0:ref>.</ns0:p><ns0:p>The features were grouped into the five following categories:</ns0:p><ns0:p>• General statistics analysis: The mean, maximum, standard deviation, and median were calculated for each time series. This category included 56 extracted features.</ns0:p><ns0:p>• Peak analysis: The peak number, average and standard deviation (SD) of the interval between peaks, average and SD of the peak magnitudes, and average and SD of the peak widths were calculated for each time series using the SciPy library <ns0:ref type='bibr' target='#b17'>(Jones et al., 2001)</ns0:ref>. The peak widths were calculated at 30% of the peak height. The default parameters of the library were used for the computation of all other features extracted from the peak analysis. This category included 98 extracted features.</ns0:p><ns0:p>• Gait phase analysis: The envelope of the signal of the seven sensors was calculated for each foot. For each identified full stance phase, the difference in the force peak yield between the foot contact on the ground (early stance phase) and the foot lift (late stance phase) was calculated and the values averaged over the window. The average duration of the double float phase was also calculated or was set to the null value when such phase does not exist. Two features were extracted in this category.</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:05:48920:2:0:NEW 17 Sep 2020)</ns0:p><ns0:p>• Frequency domain analysis: The signal of the 14 sensors was summed up and a fast Fourier transform (FFT) was conducted. Preliminary FFT analyses were conducted. The following 5 features were extracted from the AC component of the discrete frequency component series (0.05-50 Hz) and included in the final analysis: 1) power density, 2) frequency signal weighted average from 1.67-10 Hz, 3) skewness of the frequency components below 10 Hz, 4) mean of the AC components from 2-10 Hz, and 5) standard deviation of the same segment. Events with a frequency lesser than 2 Hz were assumed to be related to the gait cycle. Gait cycle-related behaviors were expected to be described by the features extracted from the above described peak analysis. Moreover, human movements are assumed to not exceed a frequency greater than 10 Hz. Therefore, only the spectral signals at frequencies less than 10 Hz were considered in the present analysis. Five features were extracted in this category.</ns0:p><ns0:p>• Pressure distribution analysis: The envelope of the signal of sensors 4, 5, 6, and 7, located in the forefoot area (Fig. <ns0:ref type='figure'>1A</ns0:ref>), was computed. The difference between the mean of this new series of data and the plantar pressures detected by sensor 1 (heel, Fig. <ns0:ref type='figure'>1A</ns0:ref>) was calculated for the left and right feet. The difference was averaged to express the anterior-posterior distribution of the plantar pressures. The difference between the mean of the plantar pressures detected by sensor 6 (head of the first metatarsal) and mean of the plantar pressures detected by sensor 4 (lateral forefoot) were calculated for the left and right feet. The values were averaged to express the medial-lateral distribution of the plantar pressures. Moreover, a Pearson correlation test was used to test the 1) agreement between the envelope of sensor 4, 5, 6, and 7 signals and sensor 1 signal and 2) agreement between the signal of sensor 4 and that of sensor 6. These correlation coefficients were calculated for both the left and right feet. Six features were extracted in this category.</ns0:p><ns0:p>The final number of extracted features depended on the number of sensors included in the processing (cf. paragraphs 'Window length,' 'Number and location of sensors,' 'Number of features').</ns0:p></ns0:div>
<ns0:div><ns0:head>Design of activity recognition algorithms</ns0:head><ns0:p>In the present study, the smart-shoe activity prediction algorithms were developed using machinelearning techniques. Data used as input included as many dimensions as the number of features extracted, i.e., 167 when using the information from the seven sensors for each foot. Preliminary processing including different machine-learning methods (e.g., k-means clustering, support vector machine) indicated higher performances for the random forest models (results not provided). The analysis presented in this manuscript focuses on the development of random forest models able to process plantar pressure information for activity recognition. The machine-learning analysis was completed using the Python scikit-learn module <ns0:ref type='bibr' target='#b30'>(Pedregosa et al., 2011)</ns0:ref>. 'Forests' were made of 100 decision trees. Each tree in the forest produced an independent prediction (here, an activity), and the mode of the predictions was chosen as the forest decision. Each tree was constructed using a random subset of the dataset, according to the bagging method described elsewhere <ns0:ref type='bibr' target='#b5'>(Breiman, 2001)</ns0:ref>. During the construction process, the nodes were successively split until all data points corresponded to the same activity; that is, until the tree's gini impurity score was equal to zero. This configuration enabled each tree in the forest to output one single prediction (also called a pure decision). Highly informative features could appear in several trees and tended to appear in the nodes that were closer to the root of the trees. Conversely, <ns0:ref type='table'>PeerJ reviewing PDF | (2020:05:48920:2:0:NEW 17 Sep 2020)</ns0:ref> Manuscript to be reviewed features with poor discriminating capacities appeared in less nodes across the entire forest (Fig. <ns0:ref type='figure'>2</ns0:ref> and Supplementary Material 1).</ns0:p><ns0:p>For the training, the data of six subjects (i.e., approximately 33% of the dataset) were used under five different subject-wise assignments (Fig. <ns0:ref type='figure' target='#fig_9'>3</ns0:ref>). At the training stage, only the data of the subjects who completed the nine activities were used. A total of 20 training-test runs were performed for each assignment, with each run using different random subsets of the dataset (hereafter called 'random states'). For the testing process, the generated random forest modules evaluated the data of the remaining 11 subjects. The results averaged across all five assignments (i.e., across 100 forests), were presented as confusion matrices of the predictions vs. actual activities. The results were also presented as mean (minimum, maximum) when summarizing the overall performance across all activities. All the results presented in this manuscript correspond to the outcome of the evaluation of the random forest modules using the test samples only. None of the reported scores are related to the training phases.</ns0:p></ns0:div>
<ns0:div><ns0:head>Data analysis framework</ns0:head><ns0:p>As illustrated in Figure <ns0:ref type='figure' target='#fig_10'>4</ns0:ref>, the analyses are integrated in a 3-stage logical flow. 1) The window length analysis aims at identifying the optimum analytic window length.</ns0:p><ns0:p>2) The analysis of a pre-selected set of 25 sensor configurations, i.e. configurations using the information of different numbers of sensors and/or the information of sensors placed at different locations, aims at identifying the best hardware combination for each possible number of sensors ranging from 1 to 6 (the 7-sensor configuration only has 1 possible combination). This analysis was conducted using the optimum window length identified in 1).</ns0:p><ns0:p>3) A final analysis exploring the contribution of each feature to the forest outputs aims at finding the most efficient number of features to be used for each of the seven best sensor configurations identified in 2). Again, this analysis was conducted using the optimum window length identified in 1). Further details related to each of the three stages are given in the three following subsections.</ns0:p></ns0:div>
<ns0:div><ns0:head>Stage 1: window length</ns0:head><ns0:p>The described analysis was performed for different window lengths <ns0:ref type='bibr'>(1, 5, 10, 15, 20, 25, 30, 35, 40, 45, 50, 55, and 60 s)</ns0:ref> using the data of the seven sensors per shoe with each data point having 167 dimensions corresponding to the maximal number of data features that were possible to extract. The optimum window length was defined at the point where the slope of the function describing the prediction rate vs. window length began to decrease. This optimum window length was used for all subsequent analysis.</ns0:p></ns0:div>
<ns0:div><ns0:head>Stage 2: number and location of sensors</ns0:head><ns0:p>The processing was repeated from scratch for different sensor configurations, i.e., different location and/or number of sensors, for the optimal window length only. Twenty-five configurations were selected among the 127 possible combinations of sensors. The selection was performed using subjective criteria: 1) reproduction of a selection of the configurations found in the literature or in the industrial sector of running shoes, 2) selection of combinations allowing the collection of relevant information for the prediction of gait and postural behaviors, and 3) selection of combinations that are believed to miss some pieces of relevant information for the prediction of gait and postural behaviors (Fig. <ns0:ref type='figure' target='#fig_10'>4</ns0:ref>). Seventeen configurations were selected with respect to criteria Manuscript to be reviewed 1) and 2). These configurations were expected to perform well. Eight configurations were selected with respect to the criterion 3). These configurations were expected to perform poorly. The number of dimensions of the data points decreased in accordance with the decreased number of sensors. For certain configurations with the same number of sensors, the data points present different numbers of dimensions. Indeed, as indicated in Table <ns0:ref type='table'>2</ns0:ref>, some features may need specific sensor locations to be computed. All the tested configurations are noted in Figure <ns0:ref type='figure' target='#fig_10'>4</ns0:ref>.</ns0:p></ns0:div>
<ns0:div><ns0:head>Stage 3: number of features</ns0:head><ns0:p>The processing was again repeated from scratch with the best configurations only and for a decreasing number of features, which were removed one-by-one based on their discriminating capacities (Fig. <ns0:ref type='figure' target='#fig_10'>4</ns0:ref>). For each sensor configuration, the analysis was performed with the maximum number of available features (similar to what was performed for the previous process, cf. paragraph 'Number and location of sensors'). Features were ranked relative to their discriminating capacities, i.e., from the highest to lowest informative feature, across the 100 runs of the analysis (five assignment × 20 random states, cf. 'Prediction algorithm: training and test'). The lowest informative feature was removed from the dataset, and a new repetition of training-test runs was performed. The entire process was repeated until only one feature remained. The minimum number of features corresponding to the inflection point for the prediction rate vs. number of the feature was considered to be the optimum number of inputs. A total of 686 combinations of sensor configurations and number of features were tested (i.e. best 1-sensor: 29, best 2-sensor: 54, best 3sensor: 76, best 4-sensor: 98, best 5-sensor: 120, best 6-sensor: 142, 7-sensor: 167).</ns0:p></ns0:div>
<ns0:div><ns0:head>Results</ns0:head><ns0:p>The 'prediction rates' and 'rates of good predictions' presented in the text and figures refer to the accuracy, calculated as follows: correctly predicted sample/total number of samples. When reporting statistical results, the terms 'average' and 'mean' point to the average accuracy across the 100 forests of one round of evaluations (see Fig. <ns0:ref type='figure' target='#fig_9'>3</ns0:ref>). The expressions 'best single forest' and 'best performer' refer to the one single forest that showed the best accuracy score among the 100 forests produced for one round of evaluations. Conversely, the term 'worst single forest' points to the one single forest that showed the worst accuracy score among the 100 forests produced for one round of evaluations. Logical links between the 3 stages of the analysis are shown in Fig. <ns0:ref type='figure' target='#fig_10'>4</ns0:ref>. The values indicated at the intersections of 'true label' and 'prediction' in the confusion matrices refer to sensitivity, calculated as follows: true positives/(true positives + false negatives).</ns0:p></ns0:div>
<ns0:div><ns0:head>Stage 1: window length</ns0:head><ns0:p>The average performances of the 7-sensor configuration tested at different window lengths <ns0:ref type='bibr'>(1, 5, 10, 15, 20, 25, 30, 35, 40, 45, 50, 55, and 60 s)</ns0:ref> are presented in the Fig. <ns0:ref type='figure'>5</ns0:ref>. The full set of 167 features was used for all tests. The best prediction rate was obtained with a 45-s window length: 0.90 (min: 0.86, max: 0.91). A 20-s length was associated with an average of 0.89 (min: 0.82, max: 0.91). The average prediction rates for window lengths between 20-60 s showed marginal variations within the 0.89-0.90 range. To preserve the highest possible temporal resolution for future applications, 20 s was selected as the optimum length. The rest of the analyses were conducted using a 20-second window length. Manuscript to be reviewed As indicated in Fig. <ns0:ref type='figure'>6</ns0:ref>, 'walking up a slope' could be confused with 'walking on a flat surface' or 'walking upstairs.' Confusions between 'walking upstairs' and 'walking downstairs' and between 'standing' and 'office work' were noted to a certain extent depending on the window length.</ns0:p></ns0:div>
<ns0:div><ns0:head>Stage 2: number and location of sensors</ns0:head><ns0:p>The average performances of a subset of 25 selected sensor configurations are presented in Fig. <ns0:ref type='figure'>7</ns0:ref>. For each configuration, the analyses were performed using all the available features. The best average prediction rate was 0.89. In addition to the 7-sensor configuration, this rate was observed for the four following configurations:</ns0:p><ns0:p>• 6 sensors, 145 (heel, lateral midfoot, lateral forefoot, medial forefoot, center of the midfoot, center of the forefoot): 0.89 (min: 0.82, max: 0.92) • 6 sensors, 142, (heel, lateral midfoot, lateral forefoot, big toe, center of the midfoot, center of the forefoot): 0.89 (min: 0.83, max: 0.91) • 5 sensors, 120 features (heel, lateral midfoot, lateral forefoot, center of the midfoot, and center of the forefoot): 0.89 (min: 0.85, max: 0.92) • 4 sensors, 98 features (heel, lateral midfoot, lateral forefoot, center of the forefoot): 0.89 (min: 0.85, max: 0.92) Regarding the best performers, selected forests achieved a prediction rate of 0.92. This result was obtained with a 3-sensor configuration only (heel, lateral midfoot, center of the forefoot). All selected configurations with at least two sensors produced an average rate of good predictions of 0.80 or more. All the configurations with at least five sensors produced an average rate of good predictions of 0.85 or more. All the configurations with at least two sensors, which were expected to perform well, produced an average rate of good predictions of 0.87 or more. Certain forests with one sensor located at the center of the forefoot could compute prediction rates as high as 0.86. The mean and maximum rates of good predictions of a larger panel of 67 selected configurations are presented in Supplementary Material 2.</ns0:p><ns0:p>The confusion matrices presented in Fig. <ns0:ref type='figure'>8 and 9</ns0:ref> indicate the sensitivity score of each activity, for the best and worst 1-, 2-, 3-, 4-, 5-, and 6-sensor configurations. Among the best sensor configurations, the decrease in the average prediction rate observed when reducing the number of sensors from two to one might be explained mainly by the higher levels of confusion between 'sitting' and 'standing' and between 'walking downstairs' and 'walking upstairs.' For example, the best 1-sensor configuration wrongly predicted 'walking upstairs' instead of 'walking downstairs' in 36% of the cases. Among the worst configurations, the decrease in the average prediction rate observed when reducing the number of sensors from four to three could be explained mainly by a decrease in sensitivity for 'cycling' (0.91 and 0.80).</ns0:p></ns0:div>
<ns0:div><ns0:head>Stage 3: number of features</ns0:head><ns0:p>The changes in performance of the seven selected configurations when decreasing, one-by-one, the number of features used for the prediction are displayed in Fig. <ns0:ref type='figure'>10</ns0:ref>. For these configurations, the mean rate of good predictions increased from an average 0.46 ± 0.03 when using one feature to 0.87 ± 0.04 when using a set of 20 high performance features. Using 20 features only, all the selected configurations demonstrated a mean rate of good predictions greater than 0.85, with the worst single forest scoring at 0.81 (2-sensor configuration), except for the 1-sensor configuration, Manuscript to be reviewed which demonstrated a rate of 0.78 (min: 0.72, max: 0.83). The data are presented in Supplementary Material 2. The overall performance remained constant when the predictions were computed with more features. The mean rate of good predictions exhibited an average of 0.87 ± 0.04 when considering computations performed using the maximum number of available features, i.e., <ns0:ref type='bibr'>29, 54, 76, 98, 120, 142, and 167</ns0:ref>, respectively, for the selected 1-, 2-, 3-, 4-, 5-, 6-, and 7-sensor configurations.</ns0:p><ns0:p>Considering a 20-feature cut-off below which features became increasingly important, 44 important features were identified over the 20-140 alternatives enabled by the seven selected configurations (Fig. <ns0:ref type='figure' target='#fig_7'>11</ns0:ref>). Seven features systematically ranked among the 20 most important features of the seven selected configurations: average peak interval of the left foot heel sensor (peak analysis), average peak magnitude of the right foot heel sensor (peak analysis), mean of the AC component (frequency domain), number of peaks for the right foot heel sensor (peak analysis), number of peaks for the left foot heel sensor (peak analysis), standard deviation of the left foot heel sensor plantar pressures (general statistics), and standard deviation of the AC component (frequency domain). Among the 44 important features, 24 belong to the 'peak analysis' category, 16 to the 'general statistics' category, 3 to the 'frequency domain' category, 1 to the 'gait phase' category, and 0 to the 'pressure distribution' category. Regarding the 7-sensor configuration only, the features related to the heel and central forefoot were identified five times. No feature directly extracted from the analysis of the big toe pressure ranked among the set of important features.</ns0:p><ns0:p>The best 1-sensor configuration (heel sensor) using the single most informative feature demonstrated a mean rate of good predictions of 0.43 (min: 0.41, max: 0.44). Only the 'running' and 'sitting' activities demonstrated a sensitivity score greater than 50% (Fig. <ns0:ref type='figure' target='#fig_18'>12</ns0:ref>). As indicated in Fig. <ns0:ref type='figure' target='#fig_18'>12</ns0:ref>, the other selected sensor configurations were associated with sensitivity scores of 82% or more for all activities except 'office work,' 'walking up a slope,' and 'walking upstairs,' when using a limited number (i.e., 9-23) of features. Regarding the 7-sensor configuration specifically, the confusions noted when using the 23 most informative features (Fig. <ns0:ref type='figure' target='#fig_18'>12</ns0:ref>) were similar to the ones noted when using the full number of available features (Fig. <ns0:ref type='figure'>6D</ns0:ref>), except for 'walking downstairs,' which had a better sensitivity (0.92 vs. 0.84) when using 23 features only. That phenomenon can be explained by the greater difficulty to fit a classifier with a higher number of dimensions. In theory, the same performance should be attainable with more features, at the risk of overfitting the system and decreasing its generality (performance on unknown data) <ns0:ref type='bibr' target='#b21'>(Lever, Krzywinski & Altman, 2016)</ns0:ref>. Confusion matrices of some selected single forests produced with the best 4-sensor configuration are presented in Fig. <ns0:ref type='figure' target='#fig_20'>13</ns0:ref>. A low performance single forest with a relatively high number of features (49 over a maximum of 98 available) was associated with low sensitivity scores for the 'office work,' 'walking downstairs,' and 'walking up the slope' activities (0.74, 0.61, and 0.56, respectively), consistent with the pattern that has been frequently found on confusion matrices, as displayed in Fig. <ns0:ref type='figure'>6</ns0:ref>, 8, and 9. Interestingly, the worst single forest among the ones built with 29 features only had the highest sensitivity for the 'walking up the slope' activity (0.70). Finally, confusion matrices of the best single forests built with 86 and 22 features demonstrated a similar pattern of missed predictions, with 'office work' and 'walking up the slope' being relatively poorly recognized (<0.85 and <0.65, respectively).</ns0:p></ns0:div>
<ns0:div><ns0:head>Supplementary results</ns0:head><ns0:p>PeerJ reviewing PDF | (2020:05:48920:2:0:NEW 17 Sep 2020)</ns0:p></ns0:div>
<ns0:div><ns0:head>Manuscript to be reviewed</ns0:head><ns0:p>A more comprehensive analysis has been conducted considering a larger panel of 67 sensor configurations. Random forest modules were systematically created and tested for each window length candidates (1-60 s) and each possible number of features (maximum to one), without any selection of the best sensor configurations like in the 3-stage data processing flow presented in Fig. <ns0:ref type='figure' target='#fig_10'>4</ns0:ref> (the results of which have been presented in the above three subsections). The machine learning procedure was the same as the one detailed in the method section. Therefore, 75,172 additional analyses have been completed, resulting in the computation of 7,517,200 forests. These supplementary analyses were associated with higher prediction scores, highlighting the whole potential of using plantar pressure data for the recognition of physical behaviors. As shown in the Supplementary Material 3, 297 sensor configurations were associated with at least one forest presenting a prediction score of 0.92 or more. Regarding the highest scores, at least 12 forests presented a rate of good predictions of 0.94. The best average scores ranged from 0.54 to 0.91, a scale similar to the one of the results of the 3-stage analysis (Supplementary Material 2). Eightyseven sensor configurations were associated with average rates of good predictions of 0.90 or more. As shown in Supplementary Material 4B, the best performances observed in these analyses are systematically associated with window lengths of 30 s or longer. Regarding the identification of an optimum analytic window length, the results still points to a period of 20 s (Supplementary Material 4A). The question of the time resolution is discussed later in the manuscript. The results of these supplementary analyses are summarized in Supplementary Material <ns0:ref type='figure' target='#fig_11'>3 and 4</ns0:ref>.</ns0:p></ns0:div>
<ns0:div><ns0:head>Discussion</ns0:head><ns0:p>In the present study, homemade smart shoes mounted with seven pressure sensors were used to collect plantar pressures during nine daily life activities. From the plantar pressure data, 167 features bearing a potential interest for the characterization of gait and posture were extracted. Random forest models using subject-wise training-test assignments were utilized to develop smartshoe activity recognition algorithms. A 20-s window length was identified as the optimal period for the extraction of the features. Forests could recognize activities at an average rate of good predictions of 0.89, with certain single forests demonstrating a rate as high as 0.92. Reducing the number of sensors to two (heel and lateral forefoot) and selecting 20 high performance features maintained the average rate of good predictions above 0.85.</ns0:p></ns0:div>
<ns0:div><ns0:head>Performances</ns0:head><ns0:p>Smart shoes in their maximal configuration (i.e., 7 sensors per foot and 167 features extracted from the collected plantar pressures) allow random forest modules to recognize activities at a rate of good predictions of 0.89 (min: 0.82, max: 0.91). Each single activity was associated with a sensitivity score of at least 0.87, except 'office work' and 'walking up a slope,' which presented lower scores (0.80 and 0.63, respectively) (Fig. <ns0:ref type='figure'>6B</ns0:ref>). 'Office work' was confused with 'standing' in 18% of cases. The latter is not surprising considering the content of the 'office work' activity, which includes a considerable number of tasks realized in the standing posture. Numerous subjects consumed a significant amount of time writing and erasing notes on a white table board while performing the 'office work'-labelled activity. This could have created this confusion with the 'standing' activity. Moreover, poor predictions involving the 'walking up a slope' activity being confused with 'walking upstairs' or 'walking on a flat surface' was a recurrent issue of the present analysis. This type of confusion occurred regardless of the sensor configuration or the number of features used as input. Depending on the field of application, several of the above-mentioned Manuscript to be reviewed confusions could have marginal or significant consequences on the final evaluation of physical behaviors. Future smart-shoe studies should also consider extracting data features that are more likely to report on slope-related gait alterations.</ns0:p><ns0:p>Conversely, random forest module outcomes indicated only a small number of confusions for the 'cycling,' 'running,' or 'sitting' activities. Although 'running' and 'sitting' are typically well recognized in research protocols that use accelerometer sensors, which remain the current primary hardware choice for activity trackers <ns0:ref type='bibr' target='#b29'>(Pavey et al., 2017;</ns0:ref><ns0:ref type='bibr' target='#b38'>Trost, Zheng & Wong, 2014;</ns0:ref><ns0:ref type='bibr' target='#b39'>Voicu et al., 2019)</ns0:ref>, the recognition of 'sitting' behaviors has actually proven technically challenging in real-life conditions <ns0:ref type='bibr' target='#b20'>(Kerr et al., 2018)</ns0:ref>. Extrinsic behavioral factors, such as people leaving their tracking device to charge when they are resting or sitting, render the assessment of sedentary behaviors even more difficult. In the present study, smart shoes demonstrated high level of sensitivity for 'sitting' (0.96 or more for any of the selected configurations with at least two sensors) and low level of confusion with the other sedentary activity (i.e., 'standing,' (0.00-0.01), except for some 1-and 2-sensor configurations). Such outcomes should be considered as promising for the monitoring of sedentary behaviors outside the house. Finally, differences were noted among the single forests for the performance in each activity. For example, one forest tagged with a low overall performance displayed in Fig. <ns0:ref type='figure' target='#fig_20'>13</ns0:ref> performed surprisingly well for the recognition of the 'walking up a slope' activity. However, this enhanced performance would appear to be possible at the expense of an altered sensitivity for other activities. This may reflect the capacity of random forest modules to specialize for one given type of activity. Further analyses, which are beyond the scope of the present report, would be necessary to identify the 'ins and outs' of forest specialization and determine if the random forest method could be adapted to the specific case of smart shoes to obtain more homogenous recognition rates across activities. For example, forests with a higher number of trees or hierarchical models assigning data points to sub-classes before proceeding to the final evaluation could be considered for future studies.</ns0:p></ns0:div>
<ns0:div><ns0:head>Comparison with previous studies and originality</ns0:head><ns0:p>Several reviews have summarized the outcomes of studies interested in the validity of instrumented insoles developed for activity recognition <ns0:ref type='bibr' target='#b9'>(De Pinho André, Diniz & Fuks, 2017</ns0:ref><ns0:ref type='bibr' target='#b24'>, Ngueleu et al., 2019)</ns0:ref>. Similar to the observations in the present research, specific studies have reported excellent performances, with rates of good predictions scoring frequently over 0.90. However, they can also be linked with experimental limitations, altering the generalization of the results, such as a small number of tested activities, small number of subjects, special groups of individuals, and trainingtest procedures completed separately for each subject. Moreover, the majority of these studies have used hardware with multi-sensing capabilities. Hegde et al. ( <ns0:ref type='formula'>2017</ns0:ref>) developed the SmartStep system, which featured three pressure sensors, one 3-axis accelerometer, and a gyroscope. The pressure sensors were placed at the heel, first metatarsal head (i.e., equivalent to the medial forefoot), and big toe. They tested the activity recognition capabilities of the SmartStep system for a wide range of daily life activities. Similar to the present report, they observed an average rate of good prediction of approximatively 0.90. They also reported recurrent mis-predictions for 'walking downstairs' (0.62), which is frequently confused with 'walking on a flat surface' and 'walking upstairs' and for 'shelving items' (0.61), the description of which resembles the 'office work' activity of the present study, and which is frequently confused with 'standing.' Smart shoe- Manuscript to be reviewed based activity recognition projects appear to be associated with redundant challenges related to ascending and/or descending locomotive activities and activities combining locomotive and nonlocomotive behaviors. Moreover, in another recent study, el Achkar et al. ( <ns0:ref type='formula'>2016</ns0:ref>) used a simple decision tree classifier to achieve excellent rates of good predictions for nine activities, including 'walking downstairs' (0.98), 'walking upstairs' (0.99), and 'walking uphill' (0.96). However, their smart-insole system featured a barometer in addition to eight pressure sensors, one 3-axis accelerometer, one 3-axis gyroscope, and one 3-axis magnetometer, which surely helped the assessment of ascending and/or descending locomotive behaviors.</ns0:p><ns0:p>According to <ns0:ref type='bibr' target='#b24'>Ngueleu et al. (2019)</ns0:ref>, smart shoe-based activity recognition studies that only use plantar pressure information are limited. Although some of these studies reported acceptable performance, protocols were typically limited to a small number of locomotive behaviors <ns0:ref type='bibr' target='#b41'>(Zhang et al., 2005;</ns0:ref><ns0:ref type='bibr' target='#b43'>Zhang, Poslad, 2014)</ns0:ref>, small number of subjects, or training-test procedures completed separately for each individual <ns0:ref type='bibr' target='#b37'>(Sugimoto et al., 2010)</ns0:ref>. Therefore, the present research provides important findings to the relatively small corpus of knowledge on plantar pressure-based activity recognition. Other originalities of the present research include the use of a random forest modeling method to develop different activity classifiers and a comparison of different sensor configurations (number and location) within one single experimental protocol.</ns0:p></ns0:div>
<ns0:div><ns0:head>Temporal resolution, sensor configuration, number of features, manufacturing, and algorithmic considerations</ns0:head><ns0:p>Although windows of 30 and 45 s were linked with better performances for the recognition of 'office work' and 'walking up a slope', overall, the performances were consistent across all analyses performed with a window size of 20 s or longer (Fig. <ns0:ref type='figure'>6 B, C, and D</ns0:ref>). In real-life situations, a short window length reduces the probability of overlapping activities over the span of one analytic period. Therefore, a 20-s length with a 50% overlap between windows was selected as the optimum window length. It allowed computing predictions every 10 s. Considering future applications, this relatively high temporal resolution would allow applying a second statistical algorithmic layer consisting of comparing the prediction of one given window with the ones of its neighbors <ns0:ref type='bibr' target='#b40'>(Witowski et al., 2014)</ns0:ref>. This would provide the opportunity to have a set of six 'instant' predictions to determine the dominant behavior every minute. Further explorations that include free-living experiments are necessary to elaborate further on the issue of temporal resolution.</ns0:p><ns0:p>One interesting finding of the present study is the marginal alteration of the overall performance obtained with a reduced number of sensors. Although configurations without the heel sensor systematically present lower performances, other configurations that include at least two sensors demonstrate average rates of good predictions of 0.87 or more (Fig. <ns0:ref type='figure'>7</ns0:ref>). The absence of a heel sensor appears to worsen confusions between ascending and descending activities and between 'office work' and 'standing' (Fig. <ns0:ref type='figure'>8 and 9</ns0:ref>). Using one sensor only, the average rates of good predictions declined below 0.80. Furthermore, marginal variations of the overall performance were noted when reducing the number of features down to approximately 20 (Fig. <ns0:ref type='figure'>10</ns0:ref>). The reduction of the number of features given to the forests was accomplished in a manner that favored the most contributive features. Extracts from the FFT and peak analyses were redundant in the lists of 20 important features (Fig. <ns0:ref type='figure' target='#fig_7'>11</ns0:ref>). However, this result could also be the mere reflection of the higher number of gait activities included in the present protocol, which all present cyclic plantar pressure patterns. Therefore, future studies should include a more balanced number of locomotive, non- Manuscript to be reviewed locomotive, and mixed activities to determine whether this trend is confirmed or not. Moreover, no feature extracted directly from the big toe sensor ever scored among the 20 most important features. This location may not be relevant for smart-shoe prototypes aimed at behavior recognition. Although the 167 data features were selected to be as comprehensive as possible and to accommodate the analysis on the reduction of the number of sensors, the list of potentially informative features is not closed. Future studies could propose extracting different features to boost the performance on a similar or different subset of activities. With respect to the abovediscussed results, shoe manufacturers willing to develop activity recognition devices should probably consider the opportunity to implement a minimalist sensor configuration instead of the full 7-sensor configuration. They should also consider the relevance of using an exhaustive number of features, whereas a subset of 20 features has been demonstrated to perform equally well. All these considerations will influence shoe design (relative to the location of sensors and other hardware), microprocessor selection (relative to the computational needs), and, ultimately, the financial cost of the device <ns0:ref type='bibr' target='#b11'>(Eskofier et al., 2017)</ns0:ref>.</ns0:p></ns0:div>
<ns0:div><ns0:head>Limitations and strengths</ns0:head><ns0:p>Some characteristics of the present protocol could limit the interpretation of the results presented in this report and should be mentioned clearly for the readers. First, the current protocol only includes nine different daily life activities. This number puts the study among smart-shoe protocols testing a large sample of activities <ns0:ref type='bibr' target='#b24'>(Ngueleu et al. 2019)</ns0:ref>. The challenges related to the recognition of activities that potentially present closed plantar pressure patterns are addressed in an adequate manner. However, a larger number of activities should be studied in the future to reflect more exhaustively physical behaviors of the daily life, e.g., sport activities and a wider panel of activities combining locomotive and non-locomotive behaviors. Second, the experimental design does not include further validation of forest performances in real-life situations. Similarly, no comparison with commercial activity monitors was performed. Future protocols should include a free-living validation to increase the generalization of the results to real-life situations. Third, the present protocol includes a relatively homogenous population. Subjects were all healthy women. To address this potential issue, five different subject-wise training-test assignments were used to develop and test the forests. In addition, a conservative 6-11 training-test assignment ratio has been used to limit the wealth of the available information during the training phase and create more challenging conditions relying on inter-individual differences. However, a more heterogeneous sample of the population must be tested before generalizing further the results of the present study. A more heterogeneous population would indeed provide a more diverse information to the training algorithms, which could also result in increased good prediction scores. Overall, given the homogeneity of the population used in the present study, one should exercise caution when interpreting the results. The best configurations identified in the present study could differ from one population to the other. Designers are therefore encouraged to select a subject sample large enough to be representative of the targeted population and provide the wealthiest possible information to the machine learning algorithms. Finally, the present work does not address the question of a multi-sensing environment. Given that alternative sensing options could already be embedded in other type of devices (e.g., activity trackers, smartphones), one could consider that smart shoes should primarily specialize in the collection of information on the foot-ground interaction. The present protocol allows focusing on the sole performance of plantar pressurebased activity recognition to assess the relevance of including smart shoes in a network of devices dedicated to physical activity evaluation <ns0:ref type='bibr' target='#b8'>(Chen et al., 2016</ns0:ref><ns0:ref type='bibr' target='#b11'>, Eskofier et al., 2017)</ns0:ref>. Manuscript to be reviewed</ns0:p></ns0:div>
<ns0:div><ns0:head>Conclusions</ns0:head><ns0:p>In this work, random forest modules as behavior recognition algorithms for plantar pressure measurement using smart shoes were explored and proved relevant. Indeed, smart shoes mounted with seven pressure sensors and extracting 167 plantar pressure data features could recognize nine different daily life activities with an average of good prediction of 0.89. Interestingly, the results suggest a marginal reduction of performance for configurations downgraded to two, three, four, five, or six sensors and the computation of approximately 20 plantar pressure data features, which could ease the design and manufacturing of smart-shoe products. Future studies are necessary to generalize the present findings to a larger sample of the population and larger number of behaviors. Considering the trend toward the development of wearable devices with 5G capacities, smart shoes could become a crucial element of systems allowing self-monitoring of physical activity, thus having an important role in promoting active and healthy lifestyles. Zoom view on a selected branch of one regression tree of one selected forest. During the training process, the nodes (diamonds) are split until all data points correspond to one activity. At each node, the decision is based on the parameter that best discriminates the sample into two sub-samples. The process is repeated until the generation of a pure offspring, i.e., leaves (rounded corner rectangles) containing the data points of one given activity only. The full tree is available in the Supplementary Material 1 (window length: 20 s, configuration: seven sensors, assignment: 1, run: 1). The number of features depended on the number and location of the sensors. The number of features from left to right, one sensor: 29, 29; two sensors: <ns0:ref type='bibr'>54, 54, 54, 51; three sensors: 76, 76, 76, 73; four sensors: 98, 101, 98, 98; five sensors: 120, 123, 123, 123, 120, 120; six sensors: 142, 145, 145, 142</ns0:ref>; and seven sensors: 167. Green boxes: sensor configurations that were expected to perform well <ns0:ref type='bibr'>(positions 1, 2, 3, 4, 7, 8, 11, 12, 13, 15, 16, 17, 18, 21, 22, 23, 25</ns0:ref> from left to right). Pink boxes: sensor configurations that were expected to perform poorly <ns0:ref type='bibr'>(positions 5, 6, 9, 10, 14, 19, 20, 24</ns0:ref> from left to right) (cf. Figure <ns0:ref type='figure' target='#fig_10'>4</ns0:ref>). Red diamonds: mean values. Manuscript to be reviewed Values refers to the normalized sensitivity of the selected forest for each activity. (A) One selected bad performer with a relatively low number of used features. (B) One selected good performer with a relatively low number of used features. (C) One selected bad performer with a relatively high number of used features. (D) One selected good performer with a relatively high number of used features. 'Best' refers to the results obtained from the forest with the highest prediction rate. 'Worst' refers to the results obtained from the forest with the lowest prediction rate. 'Performer' here refers to one single forest. N: number of features.</ns0:p><ns0:note type='other'>Figure 3</ns0:note><ns0:note type='other'>Figure 13</ns0:note></ns0:div><ns0:figure xml:id='fig_0'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:05:48920:2:0:NEW 17 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_1'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:05:48920:2:0:NEW 17 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_2'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:05:48920:2:0:NEW 17 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_3'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:05:48920:2:0:NEW 17 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_4'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:05:48920:2:0:NEW 17 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_5'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:05:48920:2:0:NEW 17 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_6'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:05:48920:2:0:NEW 17 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_7'><ns0:head>Figure 1 Figure 1 .</ns0:head><ns0:label>11</ns0:label><ns0:figDesc>Figure 1</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_8'><ns0:head>Figure 2 Figure 2 .</ns0:head><ns0:label>22</ns0:label><ns0:figDesc>Figure 2</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_9'><ns0:head>Figure 3 .</ns0:head><ns0:label>3</ns0:label><ns0:figDesc>Figure 3. Overview of the machine-learning procedure.</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_10'><ns0:head>Figure 4</ns0:head><ns0:label>4</ns0:label><ns0:figDesc>Figure 4</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_11'><ns0:head>Figure 4 .</ns0:head><ns0:label>4</ns0:label><ns0:figDesc>Figure 4. Chart of the 3-stage data processing flow.</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_12'><ns0:head>Figure 7 Figure 7 .</ns0:head><ns0:label>77</ns0:label><ns0:figDesc>Figure 7</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_13'><ns0:head>Figure 8 Figure 8 .</ns0:head><ns0:label>88</ns0:label><ns0:figDesc>Figure 8</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_14'><ns0:head>Figure 9 Figure 9 .</ns0:head><ns0:label>99</ns0:label><ns0:figDesc>Figure 9</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_15'><ns0:head>Figure 10 Figure 10 .</ns0:head><ns0:label>1010</ns0:label><ns0:figDesc>Figure 10</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_16'><ns0:head>Figure 11</ns0:head><ns0:label>11</ns0:label><ns0:figDesc>Figure 11</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_17'><ns0:head>Figure 11 .</ns0:head><ns0:label>11</ns0:label><ns0:figDesc>Figure 11. Identification of most important features.</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_18'><ns0:head>Figure 12</ns0:head><ns0:label>12</ns0:label><ns0:figDesc>Figure 12</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_19'><ns0:head>Figure 12 .</ns0:head><ns0:label>12</ns0:label><ns0:figDesc>Figure 12. Confusion matrices for four selected configurations using different numbers of features.</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_20'><ns0:head>Figure 13 .</ns0:head><ns0:label>13</ns0:label><ns0:figDesc>Figure 13. Confusion matrices for selected forests captured from the best 4-sensor configuration.</ns0:figDesc></ns0:figure>
<ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='23,42.52,70.87,525.00,396.00' type='bitmap' /></ns0:figure>
<ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='26,42.52,70.87,525.00,447.75' type='bitmap' /></ns0:figure>
<ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='27,42.52,280.87,525.00,324.75' type='bitmap' /></ns0:figure>
<ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='31,42.52,70.87,525.00,393.75' type='bitmap' /></ns0:figure>
<ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='42,42.52,70.87,525.00,402.00' type='bitmap' /></ns0:figure>
<ns0:note place='foot' n='6'>PeerJ reviewing PDF | (2020:05:48920:2:0:NEW 17 Sep 2020)</ns0:note>
</ns0:body>
" | "Dear Dr. Justin Keogh,
Thank you very much for your positive feedback.
We have made some small alterations in the text and in the Figure 4 to address the remaining comment of the Reviewer #1. We are grateful for the resulting improvements.
Please find our detailed response. Our alterations appear in green.
Here is a summarize of our changes at the attention of the editorial staff:
• 2 minor alterations in Figure 4 (no change in color chart). The caption has been slightly edited too.
• 1 minor alteration of Figure 7 caption.
• 3 minor alterations of the text: line 268-275.
• Correction of 1 typo: line 279.
Kind regards (on behalf of all coauthors),
Julien Tripette
Reviewer 1
Basic reporting
NA
Experimental design
1) In Figure 4, the color bar in stage 2 is unclear. What does 'expected to perform' mean? It should belong to 'RESULTS'
Thank you for having pointed out this remaining issue in clarity.
We would like to apologize; we have found one mistake in the Figure 4. This mistake might have been the source of the misunderstanding. In the previous version, the sliders separating the green and pink colors were not at the correct location for the '4-sensor configurations' and “6-sensor configurations” panels. Because of this error 10 configurations were marked has “expected to perform poorly”, instead of 8 configurations as indicated in Figure 7.
We have adjusted the position of the sliders. Thank you.
Now, the configurations marked in green in Figure 4 (i.e. configurations that are expected to perform well) are also marked in green in Figure 7. Similarly, the configurations marked in pink (i.e. configurations that are expected to perform well) in Figure 4 are marked in pink in Figure 7.
Figure 4. Chart of the 3-stage data processing flow. Stage 1: “window length”. Stage 2: “number and location of sensors”. Twenty-five configurations were selected among the 127 possible combinations of sensors (see. “method”, “stage 2: number and location of sensors”). Some of these configurations were expected to perform well (green bars). Some of these configurations were expected to perform poorly (pink bars). Stage 3: “number of features”. The orange/dotted connectors indicate the logical links between each stage of the analysis.
In both Figure 4 and 7, the green and pink colors are not pointing to any result. These colors are used to point to some sensor configurations that are expected to perform well (green) or expected to perform poorly (pink). These assumptions have been made before running any analysis as explained in the method section. We have made some alterations in the text to make our point clearer.
Line 267-275 (“Materials and Methods”, “Stage 2: number and location of sensors”): “Twenty-five configurations were selected among the 127 possible combinations of sensors. The selection was performed using subjective criteria: 1) reproduction of a selection of the configurations found in the literature or in the industrial sector of running shoes, 2) selection of combinations allowing the collection of relevant information for the prediction of gait and postural behaviors, and 3) selection of combinations that are believed to miss some pieces of relevant information for the prediction of gait and postural behaviors (Fig. 4). Seventeen configurations were selected with respect to criteria 1) and 2). These configurations were expected to perform well. Eight configurations were selected with respect to the criterion 3). These configurations were expected to perform poorly.”
The caption of Figure 4 has been altered to increase the clarity (see Figure 4 above).
The caption of Figure 7 has also been slightly altered:
Figure 7. Performance of activity recognition of random forest algorithms for 25 sensor configurations. The number of features depended on the number and location of the sensors. The number of features from left to right, one sensor: 29, 29; two sensors: 54, 54, 54, 51; three sensors: 76, 76, 76, 73; four sensors: 98, 101, 98, 98; five sensors: 120, 123, 123, 123, 120, 120; six sensors: 142, 145, 145, 142; and seven sensors: 167. Green boxes: sensor configurations that were expected to perform well (positions 1, 2, 3, 4, 7, 8, 11, 12, 13, 15, 16, 17, 18, 21, 22, 23, 25 from left to right). Pink boxes: sensor configurations that were expected to perform poorly (positions 5, 6, 9, 10, 14, 19, 20, 24 from left to right) (cf. Figure 4). Red diamonds: mean values.
This being said the results presented in Figure 7 indeed confirmed our assumptions. The configurations that were expected to perform well (marked in green) indeed produced the highest prediction scores (relatively to the number of sensors). The configurations that were expected to perform poorly (pink) indeed produced the lowest prediction scores.
Related to this issue we have also corrected the following typo mistake.
Line 279: “All the tested configurations are noted in Figure 4.” (Figure 2 Figure 4).
Validity of the findings
NA
It was our great pleasure to address the reviewer’s comments during these two rounds of revision. We hope that this detailed answer fully addresses her/his remaining comment. We sincerely thank the reviewer for her/his constructive comments and significant contribution to the improvement of the manuscript.
" | Here is a paper. Please give your review comments after reading it. |
9,818 | "<ns0:abstract xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:p>Jatropha curcas L. belongs to Euphorbiaceae family, it synthesizes flavonoid and diterpene compounds that have showed antioxidant, anti-inflammatory, anticancer, antiviral, antimicrobial, antifungal and insecticide activity. Seeds of this plant accumulate phorbol esters, which are tigliane type diterpenes, reported as toxic and, depending on its concentration, toxic and non-toxic varieties has been identified. The aim of this work was to characterize the chemical profile of the extracts from seeds, leaves and callus of both varieties (toxic and non-toxic) of Jatropha curcas, to verify the presence of important compounds in dedifferentiated cells and consider the possibility of using these cultures for the massive production of metabolites. Callus induction was obtained using NAA (1.5 mg.L -1 ) and BAP (1.5 mg.L -1 ) after 21 d for both varieties. Thin layer chromatography analysis showed differences in compounds accumulation in callus from non-toxic variety throughout the time of culture, diterpenes showed an increase along the time, in contrast with flavonoids which decreased. Based on the results obtained through microQTOF-QII spectrometer it is suggested a higher accumulation of phorbol esters, derived from 12deoxy-16-hydroxy-phorbol (m/z 365 [M+H] + ), in callus of 38 d than those of 14 d culture, from both varieties. Unlike flavonoids accumulation, the MS chromatograms analysis allowed to suggest lower accumulation of flavonoids as the culture time progresses, in callus from both varieties. The presence of 6 glycosylated flavonoids is also suggested in leaf and callus extracts derived from both varieties (toxic and non-toxic), including: apigenin 6-C-α-L-arabinopyranosyl-8-C-β-D-xylopyranoside (m/z 535 [M+H] + ), apigenin 4'-O-rhamnoside (m/z 417 [M+H] + ), vitexin (m/z 433 [M+H] + ), vitexin 4'-O-glucoside-2''-Orhamnoside (m/z 741 [M+H] + ), vicenin-2 (m/z 595 [M+H] + ), and vicenin-2,6''-O-glucoside (m/z 757 [M+H] + ).</ns0:p></ns0:div>
</ns0:abstract>
<ns0:body xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:head>Introduction</ns0:head><ns0:p>Jatropha curcas L. (Euphorbiaceae) is a multipurpose plant native to Mesoamerica, it is important because of its usefulness as raw material in biofuels production <ns0:ref type='bibr' target='#b57'>(Salvador-Figueroa et al., 2015)</ns0:ref>, as well as, in veterinary and human traditional medicine <ns0:ref type='bibr' target='#b71'>(Zhang et al., 2017)</ns0:ref>. Several compounds with different biological activities have been isolated from different species of Jatropha <ns0:ref type='bibr'>(Ferreira-Rodriguez et al., 2016;</ns0:ref><ns0:ref type='bibr' target='#b23'>Katagi et al., 2016)</ns0:ref>. The identification of biologically active compounds extracted from different organs of this plant has been reported <ns0:ref type='bibr' target='#b49'>(Prasad, Izam & Khan, 2012;</ns0:ref><ns0:ref type='bibr' target='#b59'>Sharma, Dhamija & Parashar, 2012)</ns0:ref>. Isolated compounds or whole plant extracts have been studied because of their potential pharmacological activity <ns0:ref type='bibr' target='#b7'>(Cocan et al., 2018)</ns0:ref>. Biological effect of J. curcas includes antibacterial <ns0:ref type='bibr' target='#b50'>(Rampadarath et al., 2016)</ns0:ref>, cytotoxic <ns0:ref type='bibr' target='#b23'>(Katagi et al., 2016)</ns0:ref>, antiinflammatory <ns0:ref type='bibr' target='#b56'>(Salim et al., 2018)</ns0:ref>, and antifungal <ns0:ref type='bibr' target='#b1'>(Abdelgader, Suleiman & Ali, 2019;</ns0:ref><ns0:ref type='bibr' target='#b61'>Srinivasan, Palanisamy & Mulpuri, 2019)</ns0:ref>. Most research on J. curcas have been done with toxic varieties; toxicity is referred to phorbol esters content in seeds.</ns0:p><ns0:p>In Mexico, Brazil and India, it have been identified non-toxic varieties of this species with very low or non-detectable levels of phorbol esters (PEs) in seeds <ns0:ref type='bibr' target='#b29'>(Laviola et al., 2010;</ns0:ref><ns0:ref type='bibr' target='#b34'>Martínez-Herrera, Chel-Guerrero & Martínez-Ayala, 2004;</ns0:ref><ns0:ref type='bibr' target='#b26'>Kumar, Anand & Reddy, 2011)</ns0:ref>. PEs are known as Jatropha factors because each one of them has the same nucleus diterpene moiety, namely, 12deoxy-16-hydroxy-phorbol (DHP) which is coupled to unstables intramolecular diterpenes (named C 1 -C 6 factors) <ns0:ref type='bibr' target='#b20'>(Hirota et al., 2017)</ns0:ref>.</ns0:p><ns0:p>Plants are the most successful source of chemical compounds, which potential mode of action makes them an alternative phytomedicinal drug, since several natural products have shown benefits against human diseases <ns0:ref type='bibr' target='#b5'>(Aye et al., 2019)</ns0:ref>. Several compounds are tissue-specific accumulated, and are usually structurally complex <ns0:ref type='bibr' target='#b4'>(Armaly et al., 2015)</ns0:ref>. Therefore it is necessary the use of chemical analysis techniques to isolate and identify the extracted plant metabolites <ns0:ref type='bibr' target='#b19'>(Hernandez & Sarlah, 2019)</ns0:ref>. There are a few cases where the use of plant cell culture of Jatropha curcas has allowed the production of bioactive compounds <ns0:ref type='bibr' target='#b3'>(Alvero-Bascos & Ungson, 2012;</ns0:ref><ns0:ref type='bibr' target='#b31'>Mahalakshmi, Eganathan & Parida, 2013;</ns0:ref><ns0:ref type='bibr'>Nassar, El-Ahmay & Al-Azizi, 2013;</ns0:ref><ns0:ref type='bibr' target='#b70'>Zaragoza-Martínez et al., 2016)</ns0:ref>, the study of the culture at different stages of toxic and non-toxic varieties, generate the opportunity to design biotechnological models for production of bioactive compounds i.e. terpenoids, alkaloids, flavonoids <ns0:ref type='bibr'>(Abdelgavir & Van Staden, 2013;</ns0:ref><ns0:ref type='bibr' target='#b54'>Sabandar et al., 2013)</ns0:ref> providing opportunities for new drugs discovery. Secondary metabolites are generally in complex matrices at very low concentrations in plant organs, and lower in dedifferentiated cells. These compounds have a wide range of polarities, therefore it is necessary the use of solvents with different polarity to obtain the extracts <ns0:ref type='bibr' target='#b6'>(Chemat et al., 2019)</ns0:ref>. The aim of this work was to characterize the chemical profile of the extracts from callus of both varieties (toxic and non-toxic) of Jatropha curcas, through the cell culture, to verify the presence of important compounds in dedifferentiated cells and consider the possibility of using these cultures for the massive production of bioactive compounds.</ns0:p></ns0:div>
<ns0:div><ns0:head>Materials & Methods</ns0:head></ns0:div>
<ns0:div><ns0:head>Plant material</ns0:head><ns0:p>Seeds and young leaves of Jatropha curcas were collected. Non-toxic variety samples from Centro de Desarrollo de Productos Bióticos- <ns0:ref type='bibr'>IPN,</ns0:ref><ns0:ref type='bibr'>Yautepec,</ns0:ref><ns0:ref type='bibr'>Morelos,</ns0:ref><ns0:ref type='bibr'>México (18º53'09''N,</ns0:ref><ns0:ref type='bibr'>99º03'38''W)</ns0:ref>. The toxic variety samples were collected from Campo Experimental Zacatepec, Instituto Nacional de Investigaciones Agrícolas y Pecuarias (INIFAP), <ns0:ref type='bibr'>Zacatepec,</ns0:ref><ns0:ref type='bibr'>Morelos,</ns0:ref><ns0:ref type='bibr'>México (18º39'23''N,</ns0:ref><ns0:ref type='bibr'>99º11'28''W)</ns0:ref>.</ns0:p><ns0:p>To induce cell dedifferentiation, two different explants were surface-sterilized according to <ns0:ref type='bibr' target='#b64'>Vanegas et al., 2002.</ns0:ref> Leaf blade of approximately 0.25 cm 2 and petiole of approximately 3 mm in length were cultured in MS medium <ns0:ref type='bibr' target='#b38'>(Murashige & Skoog, 1962)</ns0:ref> supplemented with sucrose (30 g.L -1 ), phytagel (3 g.L -1 ) (Sigma-Aldrich®). Since there are no reports of the induction of dedifferentiated cells in the varieties analyzed in this study, the combinations of three concentrations (0.0, 1.5 and 3.0 mg.L -1 ) of naphthaleneacetic acid (NAA) and 6-benzylaminopurine (BAP) were evaluated according to <ns0:ref type='bibr' target='#b67'>Verma (2013)</ns0:ref>, pH was adjusted to 5.7, media were sterilized at 121 °C for 15 min. Ten explants per Petri dish with 3 repetitions per treatment were incubated at 25 ± 2 °C, photoperiod of 16 h light/8 h darkness for 35 d <ns0:ref type='bibr' target='#b28'>(Kumar et al., 2015)</ns0:ref>. Explants dedifferentiation was recorded every seven days using a stereoscopic microscope <ns0:ref type='bibr'>(Nikon, model SMZ 1500, Japan)</ns0:ref>. In order to observe differences in accumulation of compounds during callus development, completely dedifferentiated cells were cultured under the above described conditions for 38 d, samples were taken on days 0, 2 and every 4 d thereafter.</ns0:p><ns0:p>Fresh washed leaves were indoors dried at 25 ± 2 °C during 3 weeks. Seeds without tegument and callus, were oven dried at 50 °C for 48 h, dried samples were ground with a mortar and sieved through a mesh size 53 µm.</ns0:p></ns0:div>
<ns0:div><ns0:head>Ultrasound assisted extraction (UAE)</ns0:head><ns0:p>UAE was performed with an ultrasound bath Branson (2510R-MTH, CT, USA) with automatic control of time and temperature and ultrasound frequency of 40 kHz. 500 mg of grounded biomass dry weight (dw) were placed into a 50 mL borosilicate glass conical Erlenmeyer flask, then 20 mL of ethanol 80% (v/v) were added, and sonicated at 40 ± 5 °C during 30 min <ns0:ref type='bibr'>(Bazaldúa et al., 2019;</ns0:ref><ns0:ref type='bibr' target='#b44'>Pandey et al., 2018;</ns0:ref><ns0:ref type='bibr' target='#b13'>Dumitraşcu et al., 2019)</ns0:ref>. During sonication flasks were suspended in the water without contact with the bottom of the ultrasonic bath, subsequently they were vortexed. Supernatant was filtered, concentrated to dryness at 25 ± 2 °C, and solubilized in 500 μL of HPLC grade MeOH (Sigma-Aldrich®) for chromatographic analysis <ns0:ref type='bibr' target='#b55'>(Saeed et al., 2006;</ns0:ref><ns0:ref type='bibr'>Liu et al, 2013)</ns0:ref>.</ns0:p><ns0:p>Phorbol esters (PEs) rich defatted extract 500 mg of dried sample were packed in a filter paper cartridge and defatted in a Soxhlet equipment with petroleum ether (60-80 °C) (Sigma-Aldrich®) for 4 h. Petroleum ether (Fermont®) extract was concentrated using rotary evaporator at 40 °C, 90 rpm, and 900 mbar. The methyl esters in the resulting oil, were extracted with MeOH, later filtered and concentrated to dryness at 25 ± 2 °C, then solubilized in 500 μL of HPLC grade MeOH for chromatographic analysis <ns0:ref type='bibr' target='#b9'>(Demissie & Lele, 2010)</ns0:ref>.</ns0:p></ns0:div>
<ns0:div><ns0:head>Thin layer chromatography</ns0:head><ns0:p>Extracts were applied on normal phase silica plates <ns0:ref type='bibr'>(Merck Millipore,</ns0:ref><ns0:ref type='bibr'>60 F 254 ,</ns0:ref><ns0:ref type='bibr'>Germany)</ns0:ref>. were used as mobile phase, reference standards were phorbol-12-myristate 13-acetate (PMA, Sigma-Aldrich®), quercetin, and vitexin (Sigma-Aldrich®), both plates were revealed with anisaldehyde <ns0:ref type='bibr' target='#b24'>(Kathiravan & Raman, 2010)</ns0:ref>.</ns0:p><ns0:p>To analyze extracts obtained by sonication-ethanol 80% and Soxhlet-methanol a mobile phase consisting of chloroform-methanol (97:3) was used. The reference standard was PMA, and the plates were cerium sulfate-revealed, then observed at 366 nm, and white light. Retention factor (Rf) and color from the spots were compared with chromatographic terpenes profiles described by <ns0:ref type='bibr' target='#b53'>Reich & Schibli (2007)</ns0:ref>.</ns0:p></ns0:div>
<ns0:div><ns0:head>MicrOTOF Q-II analysis</ns0:head><ns0:p>Electrospray ionization analysis (ESI) was performed using a micrOTOF-Q II mass spectrometer (Bruker Daltonics, Bremen, Germany) according to <ns0:ref type='bibr' target='#b30'>León-López et al. (2015)</ns0:ref>. Samples were solubilized in 500 μL of HPLC grade MeOH and filtered with a syringe filter (nylon membrane, 0.45 μm, Agilent Technologies, Santa Clara, CA, USA). The molecular ions related to the extracts were analyzed in positive ion mode (ESI + ). 20 μL of sample were directly injected into the evaporation chamber, capillary potential was -4.5 kV, gas temperature of 200 °C, drying gas flow of 4 L min -1 and nebulizer gas pressure of 0.4 Bar. Detection was performed at 50-3000 m/z. The predictive structures of the MS/MS partitioning profile were established utilizing the Competitive Fragmentation Modeling for Metabolite Identification (CFM-ID. Version 3.0, 2019) platform from Wishart-lab (http://cfmid3.wishartlab.com), which is referred to in the PubChem-NCBI site. Relative abundance was calculated according to <ns0:ref type='bibr' target='#b58'>Scigelova et al. (2011)</ns0:ref>.</ns0:p></ns0:div>
<ns0:div><ns0:head>Results</ns0:head></ns0:div>
<ns0:div><ns0:head>Establishment of callus culture</ns0:head><ns0:p>Dedifferentiation cell was not observed in leaf blade explants. Petiole explants showed tissue dedifferentiation since seventh day of culture and complete process was evident at the day 21 (Fig. <ns0:ref type='figure' target='#fig_7'>1</ns0:ref>). Friable and light green callus was obtained on MS media added with both combinations: NAA (1.5 mg.L -1 ), BAP (1.5 mg.L -1 ), and NAA (3.0 mg.L -1 ) and BAP (3.0 mg.L -1 ).</ns0:p></ns0:div>
<ns0:div><ns0:head>Thin layer chromatography (TLC) analysis</ns0:head><ns0:p>TLC showed differences in compounds accumulation during time culture <ns0:ref type='bibr'>(2, 6, 10, 14, 18, 22, 26, 30, 34 and 38 d)</ns0:ref>. Regard diterpenes, spots with Rf of 0.71 and 0.27 showed higher intensity along this period (Fig. <ns0:ref type='figure' target='#fig_11'>2A</ns0:ref>), unlike flavonoids in which spots with Rf of 0.77 and 0.58, decreased throughout the same culture period (Fig. <ns0:ref type='figure' target='#fig_11'>2B</ns0:ref>). These results suggest that the accumulation of diterpenes and flavonoids was inversely related during callus development. To obtain diterpenes the Soxhlet-methanol extraction was more efficient than sonication-ethanol 80%. TLC analysis of extracts obtained by both methods evidenced differences in the size and intensity of spots in regard to: extraction method, variety (toxic and non-toxic), and plant material (seeds, leaves and callus) (Fig. <ns0:ref type='figure' target='#fig_8'>S1</ns0:ref>).</ns0:p></ns0:div>
<ns0:div><ns0:head>MicrOTOF Q-II and competitive fragmentation modeling for metabolite identification platform (CFM-ID) Phorbol esters (PEs) analysis</ns0:head><ns0:p>Fragmentation profile analysis from seeds extract from both varieties showed several highs signals one of them with m/z of 365 [M+H] + corresponding to 12-deoxy-16-hydroxy-phorbol (DHP), which is the fundamental structural core of the PEs. The MS/MS analysis of this molecular ion showed fragments with m/z of <ns0:ref type='bibr'>295, 276, 234, 203, 185 and 127 [M+H]</ns0:ref> + which is similar to the fragmentation profile of DHP presented in CFM-ID platform (Fig. <ns0:ref type='figure' target='#fig_3'>3</ns0:ref>), this suggests the identification of that molecular structure in all of the extracts obtained from seeds and callus of both toxic and non-toxic varieties. Based on signals intensities from 14 d and 38 d callus extracts from both varieties, it is suggested that the accumulation of DHP is time-dependent. Since, the corresponding signal was higher in callus of 38 d than in those of 14 d (Fig. <ns0:ref type='figure' target='#fig_4'>4</ns0:ref>). Furthermore, two signals with m/z of 547 and 591 [M+H] + were observed, so it is proposed that they are related with the fragmentation profile of the signal with m/z of 711 [M+H] + corresponding to any of the Jatropha factors (C 1 or DHPB to C 6 ) which nucleus structure is DHP <ns0:ref type='bibr' target='#b68'>(Wink et al., 2000;</ns0:ref><ns0:ref type='bibr' target='#b15'>Haas, Sterk & Mittelbach, 2002)</ns0:ref> (Fig. <ns0:ref type='figure' target='#fig_9'>S2</ns0:ref>). Table <ns0:ref type='table' target='#tab_0'>1</ns0:ref> shows the relative abundance of DHP molecular ion (m/z 365 [M+H] + ) on 14 d and 38 d callus extracts from both varieties, evidencing the increment of these compounds through the callus development.</ns0:p></ns0:div>
<ns0:div><ns0:head>Flavonoids analysis</ns0:head><ns0:p>On the other hand, the main group of compounds in Jatropha leaf extracts are flavonoids, among them the apigenin, nevertheless, it is important to refer that the natural condition of flavonoids in the plants is in glycosylated form. On this regard, another of the highest signals observed at the chromatograms was the m/z of 381 [M+H] + ion, the MS-MS experiment of this signal and the proposed structures obtained by CFM-ID platform allowed to relate that molecular ion (m/z 381 [M+H] + ) to the fragmentation profiles of apigenin 6-C-α-L-arabinopyranosyl-8-C-β-Dxylopyranoside, and of apigenin 4'-O-rhamnoside (Fig. <ns0:ref type='figure' target='#fig_5'>5</ns0:ref>). The fragmentation signals and their corresponding predictive structure were also related for vitexin (m/z 433 [M+H] + ), vitexin 4'-Oglucoside-2 ''-O-rhamnoside (m/z 741 [M+H] + ), vicenin-2 (m/z 595 [M+H] + ), and vicenin-2,6''-Oglucoside m/z 757 [M+H] + (Fig. <ns0:ref type='figure' target='#fig_12'>S3</ns0:ref>). Table <ns0:ref type='table' target='#tab_0'>1</ns0:ref> shows the relative abundance of six tentatively identified compounds by relating their molecular ions on 14 d and 38 d callus extracts from both varieties. Inversely to observed on DHP related signal (m/z 365 [M+H] + ), the intensity of the molecular ion related with glycosylated apigenin (m/z 381 [M+H] + ) diminished (Fig. <ns0:ref type='figure' target='#fig_6'>6</ns0:ref>).</ns0:p></ns0:div>
<ns0:div><ns0:head>Discussion</ns0:head><ns0:p>The highest callus induction (95.5%) was observed in petiole explants on MS medium added with NAA (3.0 mg.L -1 ) and BAP (3.0 mg.L -1 ), the second best result (87.7%) was obtained with NAA (1.5 mg.L -1 ) and BAP (1.5 mg.L -1 ), in contrast to reported by <ns0:ref type='bibr' target='#b39'>Nassar et al. (2013)</ns0:ref>, who observed dedifferentiation with NAA and BAP at 0.5 mg.L -1 of each one plant growth regulator. Explants dedifferentiation reported in this work was similar to reported by <ns0:ref type='bibr' target='#b28'>Kumar et al. (2015)</ns0:ref>. The follow up of the explants dedifferentiation process, every 7 d showed callus formation on explants starting on the seventh day. Dedifferentiation began at the cutting sites as expected <ns0:ref type='bibr' target='#b63'>(Sujatha, Makkar & Becker 2005;</ns0:ref><ns0:ref type='bibr' target='#b41'>Nogueira et al., 2011;</ns0:ref><ns0:ref type='bibr' target='#b43'>Ovando-Medina et al., 2016)</ns0:ref>. The callus obtained was light green and friable, similar to reported by <ns0:ref type='bibr' target='#b18'>Hernández et al. (2015)</ns0:ref>. It has been reported that high auxins concentrations could affect production and accumulation of secondary metabolites <ns0:ref type='bibr' target='#b25'>(Kim et al., 2007)</ns0:ref>, hence, according to our results, it is suggested the use of the lowest effective concentration, 1.5 mg.L -1 for both growth regulators. <ns0:ref type='bibr' target='#b37'>Muñoz-Valverde et al. (2003)</ns0:ref> concluded that BAP is determinant to induce callus formation in foliar explants of J. curcas. Likewise, <ns0:ref type='bibr' target='#b62'>Suárez & Salgado (2008)</ns0:ref> reported that the presence of NAA induce callus formation in Stevia rebaudiana, and this effect could be increased when adding BAP. On the other hand, <ns0:ref type='bibr' target='#b60'>Solange et al. (2002)</ns0:ref> determined that the use of NAA and BAP in equal proportion induces callus formation from leaf explants of Tridax procumbens. <ns0:ref type='bibr' target='#b8'>Coutiño-Cortés et al. (2013)</ns0:ref> reported the callus induction in J. curcas leaf explants at 10 d of culture, and total explant-cell dedifferentiation at 20 d using 2, 4-D, BAP and KIN, while in this work, petioles dedifferentiation started at 7 d and total explant-cell dedifferentiation was achieved at 21 d. These results support that synergy between NAA and BAP is essential to achieve a high dedifferentiation degree. Stable callus culture conditions for the two varieties of Jatropha curcas were established.</ns0:p><ns0:p>The PEs are responsible for the toxicity in the plant <ns0:ref type='bibr' target='#b11'>(Devappa, Makkar & Becker, 2011;</ns0:ref><ns0:ref type='bibr' target='#b54'>Sabandar et al., 2013;</ns0:ref><ns0:ref type='bibr' target='#b71'>Zhang et al., 2017)</ns0:ref>. There are varieties of Jatropha curcas denominated as toxic and non-toxic <ns0:ref type='bibr' target='#b33'>(Makkar et al., 1997)</ns0:ref>. The non-toxic varieties have PEs concentration lower than 0.86 mg/g of seed on dry basis <ns0:ref type='bibr' target='#b17'>(He et al., 2011)</ns0:ref>. <ns0:ref type='bibr' target='#b35'>Martínez-Herrera et al. (2006)</ns0:ref> detected high levels of PEs in seed oil from the municipality of Coatzacoalcos, Veracruz, México, but did not detect PEs in seeds from the municipality of Yautepec, Morelos, México. This corroborates the differences between the seeds of the two varieties used in this study.</ns0:p><ns0:p>Regard, to TLC profile analysis, it has been reported that methanolic extraction from seed-oil facilitates separation and availability of methyl ester type compounds, mainly phorbol esters (PEs) <ns0:ref type='bibr' target='#b9'>(Demissie & Lele, 2010;</ns0:ref><ns0:ref type='bibr' target='#b10'>Devappa, Bingham & Khanal, 2013)</ns0:ref>. The detection by TLC of PEs in seed methanolic extracts from toxic and non-toxic J. curcas varieties was reported <ns0:ref type='bibr' target='#b12'>(Devappa, Makkar & Becker (2012)</ns0:ref>, they reported higher spots intensity from toxic variety than from nontoxic, when plates were observed at 366 nm UV light, this result is similar to that observed in this work (Fig. <ns0:ref type='figure' target='#fig_8'>S1</ns0:ref>). <ns0:ref type='bibr' target='#b32'>Makkar & Becker (2009)</ns0:ref> detected higher PEs accumulation in seeds than in leaves extracts. Similar results were obtained in this work, even with different method of extraction. Nevertheless, these results are different of that obtained by Martínez-Herrera, Chel-Guerrero & Martínez-Ayala, 2004, because they reported 96% of PEs extraction through hydroalcoholic extraction, quantified by HPLC; while, in this work the intensity of the spots was higher on Soxhlet-methanol extracts than hydroalcoholic extraction (Fig. <ns0:ref type='figure' target='#fig_8'>S1</ns0:ref>). Using TLC, differences between dedifferentiated cell extracts of both varieties of Jatropha curcas were evidenced.</ns0:p><ns0:p>On the other hand, <ns0:ref type='bibr' target='#b20'>Hirota et al. (2017)</ns0:ref> reported the identification of DHP as the fundamental structural core which is derived from 12-deoxy-16-hydroxy-phorbol-4'-[12',14'-butadienyl]-6'- <ns0:ref type='bibr'>[16',18',20'-nonatrienyl]</ns0:ref> + were also reported in J. curcas seeds <ns0:ref type='bibr' target='#b68'>(Wink et al., 2000)</ns0:ref>. Even more so, <ns0:ref type='bibr' target='#b15'>Haas, Sterk & Mittelbach, (2002)</ns0:ref> reported the identification of diterpenes named Jatropha factors C 2 to C 6 through ESI-MS m/z 711 [M+H] + and of DHP (m/z of 365 [M+H] + ). Furthermore, <ns0:ref type='bibr' target='#b40'>Nishshanka et al. (2016)</ns0:ref> identified six phorbol esters in J. curcas seeds by LC-MS, which have the same core (DHP), at the so named Jatropha factors (C 1 to C 6 ).</ns0:p><ns0:p>Regard to PEs identification by ESI-MS analysis, <ns0:ref type='bibr' target='#b65'>Verardo et al. (2019)</ns0:ref> identified six phorbol esters in J. curcas seeds with m/z of 711 [M+H] + , which have the same fundamental structural core (DHP) m/z of 365 [M+H] + which is coupled to diterpenes of 24 carbon structures named Jatropha factors from C 1 (DHPB) to C 6 . The relative intensity of the molecular ion m/z 365 [M+H] + was higher in seeds extracts from toxic variety, than in seed extracts from non-toxic variety (Data not shown). While in callus, the relative intensity is higher in toxic and non-toxic varieties callus of 38 d of culture (Fig. <ns0:ref type='figure' target='#fig_13'>4B, and 4D</ns0:ref>), than in toxic and non-toxic varieties callus of 14 d of culture (Fig. <ns0:ref type='figure' target='#fig_13'>4A, and 4C</ns0:ref>). These results could suggest the presence of PEs coupled to DHP in the samples analyzed and that their accumulation is differential in regard to the variety-derived cell culture and throughout the time of culture. These results suggest that their accumulation of DHP is time dependent. This ESI-MS analysis allowed to corroborate the results obtained by TLC (Fig. <ns0:ref type='figure' target='#fig_11'>2A</ns0:ref>). Nevertheless, the relative intensity of the signals observed in extracts from callus were lower than that obtained from seeds extracts as reported by <ns0:ref type='bibr' target='#b9'>Demissie & Lele (2010)</ns0:ref>. By ESI-MS, differences in the relative intensity of the signal corresponding to DHP were observed between the callus extracts of both varieties, being higher in the toxic variety in addition to that in calluses at 38 d of culture, it was higher than in the 14 d.</ns0:p><ns0:p>By other hand, phenolic compounds are ubiquitously produced by plants <ns0:ref type='bibr' target='#b27'>(Kumar & Goel, 2019)</ns0:ref>, the main role of phenols in plants is to protect them from biotic or abiotic stress <ns0:ref type='bibr' target='#b46'>(Pereira, 2016)</ns0:ref>. These properties include antimicrobial, insecticidal, antiparasitic, antiviral, anti-ulcerogeBynic, cytotoxic, antioxidant, anti-hepatotoxic, anti-hypertensive and anti-inflammatory activities <ns0:ref type='bibr' target='#b42'>(Oskoueian et al., 2011;</ns0:ref><ns0:ref type='bibr' target='#b45'>Papalia, Barreca & Panuccio, 2017)</ns0:ref>. Flavonoids are recognized as polyphenols. Several of them have been identified in Jatropha genus, such as apigenin glycosides, vitexin, and isovitexin which have been considered as chemiotaxonomic compounds from the genus <ns0:ref type='bibr' target='#b2'>(Abdelgadir &Van Staden, 2013;</ns0:ref><ns0:ref type='bibr' target='#b21'>Huang et al., 2014)</ns0:ref>.</ns0:p><ns0:p>The tentative identification of glycosylates-flavonoids through microQTOF-QII has been already reported <ns0:ref type='bibr' target='#b47'>(Pezzini et al., 2019)</ns0:ref>. In this regard <ns0:ref type='bibr' target='#b69'>Xie et al. (2003)</ns0:ref> reported the apigenin 6-C-α-Larabinopyranosyl-8-C-β-D-xylopyranoside m/z 535 [M+H] + . Likewise, this result may be related to that obtained by <ns0:ref type='bibr' target='#b0'>Abd-Alla et al. (2009)</ns0:ref> who identified apigenin and its aglycone as majoritarian flavonoids in J. curcas leaves, as well as, that obtained by <ns0:ref type='bibr' target='#b52'>Reena, Nand & Sharma, (2008)</ns0:ref> who reported to apigenin as major flavonoid in the same species. Those reports differ from that published by <ns0:ref type='bibr' target='#b45'>Papalia, Barreca & Panuccio, (2017)</ns0:ref> who identified to vitexin and vicenin-2 as the majoritarian flavonoids.</ns0:p><ns0:p>The results obtained by microQTOF-QII of the molecular ion m/z 381 [M+H] + through the MS/MS experiment, and the predictive structures obtained through the CFM-ID platform allowed to suggest the relation of the structures from the molecular ion m/z 381 [M+H] + with the fragmentation profile from apigenin 6-C-α-L-arabinopyranosyl-8-C-β-D-xylopyranoside m/z 535 [M+H] + , which was identified through ESI-MS in Viola yedoensis <ns0:ref type='bibr' target='#b69'>(Xie et al., 2003)</ns0:ref> and apigenin 4'-O-rhamnoside m/z 417 [M+H] + , which was identified in Olea europaea <ns0:ref type='bibr' target='#b48'>(Pieroni et al., 1996)</ns0:ref>.</ns0:p><ns0:p>Based on the molecular ion, MS-MS fragmentation profile and the predictive structures obtained by CFM-ID platform, it is suggested the tentative identification of vitexin m/z of 433 [M+H] + , vicenin-2 m/z of 595 [M+H] + , and vitexin 4'-O-glucoside-2 ''-O-rhamnoside m/z of 741 [M+H] + in leaves and callus from both varieties. These results are similar to obtained by <ns0:ref type='bibr' target='#b21'>Huang et al. (2014)</ns0:ref> who identified vitexin m/z of 433 [M+H] + in J. curcas leaves. This flavonoid was also identified by ESI-MS in Parkinsonea aculeata m/z of 431 [M-H] - <ns0:ref type='bibr' target='#b16'>(Hassan, Abdelaziz & Al Yousef., 2019)</ns0:ref>. In this work it is also suggested the tentative identification of vicenin-2,6''-O-glucoside m/z 757 [M+H] + which has not been reported to Jatropha curcas, but to Stellaria holostea <ns0:ref type='bibr'>(Bouillant et al., 1984)</ns0:ref> (Fig. <ns0:ref type='figure' target='#fig_9'>S2</ns0:ref>). By the MS-MS fragmentation profile, the identification of six glycosylated flavonoids is suggested, it was observed that relative intensities signals related to flavonoid related molecular ion m/z 381 [M+H] + in callus of 14 d was higher than callus of 38 d, differences that were not observed between calluses of the different varieties.</ns0:p></ns0:div>
<ns0:div><ns0:head>Conclusions</ns0:head><ns0:p>Stable dedifferentiated cells culture from petiole explants of Jathopha curcas, were stablished from toxic and non-toxic varieties on MS medium added with NAA and BAP. Thin layer chromatography and mass spectrometry, suggest an inverse relationship between phorbol esters and flavonoids accumulation in callus throughout the time of culture. The tentative identification of diterpene type compounds such as 12-deoxy-16-hydroxy-phorbol and Jatropha factors by ESI-MS in seed and callus (14 and 38 d), as well as, the presence of six flavonoids glycosides in leaf and callus, in extracts from both toxic and non-toxic varieties of J. curcas, is suggested. Both of them, groups of compounds reported with bioactive activity with pharmaceutic/agroindistrial potential. Manuscript to be reviewed Manuscript to be reviewed Manuscript to be reviewed Manuscript to be reviewed Manuscript to be reviewed </ns0:p><ns0:note type='other'>Figure 2</ns0:note><ns0:note type='other'>Figure 3</ns0:note><ns0:note type='other'>Figure 4</ns0:note><ns0:note type='other'>Figure 5</ns0:note></ns0:div><ns0:figure xml:id='fig_0'><ns0:head /><ns0:label /><ns0:figDesc>-bicycle [3.1.0] hexane-(13-O)-2'-[carboxylate]-(16-O)-3'-[8'-butenoic-10']ate (DHPB or Jatropha factor C 1 ), identified as DHPB-Na adduct m/z 733 [M+Na] + . Furthermore, DHPB m/z 711 [M+H] + and DHP m/z of 365 [M+H]</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_1'><ns0:head>Figures description Figure 1 .</ns0:head><ns0:label>description1</ns0:label><ns0:figDesc>Figures description</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_2'><ns0:head>Figure 2 .</ns0:head><ns0:label>2</ns0:label><ns0:figDesc>Figure 2. Identification of both diterpenes-type (A), and flavonoids-type (B) compounds in seeds, leaves, and callus of Jatropha curcas, through thin layer chromatography. Lanes from 2 to 38 correspond to extracts of callus of non-toxic variety throughout 38 d of culture, NTS= Nontoxic variety-seeds, PMA= Phorbol-12-myristate-13-acetate (Sigma) reference standard (Rf 0.42), IV= Isovitexin (Sigma) reference standard (Rf 0.42), and Q= Quercetin (Sigma) reference standard (Rf 0.37). A)The spots intensity increased throughout to culture time (Rfs 0.71, and 0.27), mobile phase chloroform-methanol (94:6). B) The spots intensity decreased throughout to culture time (Rfs 0.77, and 0.58), mobile phase chloroform-methanol (75:25). Plates were revealed with anisaldehyde.</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_3'><ns0:head>Figure 3 .</ns0:head><ns0:label>3</ns0:label><ns0:figDesc>Figure 3. Spectrophotometrical analysis of phorbol esters in extracts of Jatropha curcas seeds. MS/MS fragmentation profile of the molecular ion m/z 365 [M+H] + related to 12-deoxy-16-hydroxy-phorbol, which is the structural core from Jatropha curcas-phorbol esters (referred as Jatropha factors). Predictive structures obtained through CFM-ID platform from each ionized fragment.</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_4'><ns0:head>Figure 4 .</ns0:head><ns0:label>4</ns0:label><ns0:figDesc>Figure 4. Mass spectra from callus extracts of J. curcas showing the relative intensity of the molecular ion m/z 365 [M+H] + related to the structural core of the Jatropha-phorbol esters. Callus extracts from toxic variety: (A) 14 d of culture; (B) 38 d of culture; non-toxic variety: (C) 14 d of culture, (D) 38 d of culture. The relative intensity from molecular ion m/z 365 [M+H] + increased throughout culture time.</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_5'><ns0:head>Figure 5 .</ns0:head><ns0:label>5</ns0:label><ns0:figDesc>Figure 5. Fragmentation profile (MS/MS) of the molecular ion m/z 381 [M+H] + , observed in leaves extracts, and related to fragmentation of two glycosylated apigenin (apigenin (apigenin 6-C-α-L-arabinopyranosyl-8-C-β-D-xylopyranoside m/z 535 [M+H] + , apigenin 4'-O-rhamnoside m/z 417 [M+H] + ). Structures predicted to each molecular ion (381, 355, 335, and 219 m/z), obtained from CFM-ID platform.</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_6'><ns0:head>Figure 6 .</ns0:head><ns0:label>6</ns0:label><ns0:figDesc>Figure 6. Mass spectra of callus extracts from both toxic and non-toxic varieties of J. curcas at 14 and 38 d culture, showing the relative intensity of the molecular ion m/z 381 [M+H] + related to the fragmentation profile from two glycosylated apigenin. A, and C) Extracts of J. curcas callus from J. curcas-toxic variety (14 and 38 d, respectively). B, and D) Extracts of J. curcas callus from non-toxic variety (14 and 38 d, respectively). The relative intensity from molecular ion m/z 381 [M+H] + diminished throughout culture time.</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_7'><ns0:head>Table 1 .</ns0:head><ns0:label>1</ns0:label><ns0:figDesc>Tentative compounds identified by ESI-MS in hydroalcoholic extracts from seeds, leaves, and callus of 14 and 38 d of culture from both toxic and non-toxic Jatropha curcas L.varieties.Supplementary material description</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_8'><ns0:head>Figure S1 .</ns0:head><ns0:label>S1</ns0:label><ns0:figDesc>Figure S1. Thin layer chromatogram of extracts obtained from seeds, leaves, and callus of two J. curcas-varieties with two extraction methods for the identification of phorbol esters (Rf's 0.81, 0.53, and 0.38). The extracts obtained with ethanol 80% -sonication are referred with numbers (1 -8). The extracts obtained with Soxhlet -methanol are referred with letters (A -H). PMA: Phorbol-12-myristate-13-acetate Rf 0.22 (Sigma, PE reference standard). Toxic variety seed (1 and A), Non-toxic variety seed (2 and B), Toxic variety leaves (3 and C), Nontoxic variety leaves (4 and D), Toxic variety-callus 14 d (5 and E), Toxic variety-callus 38 d (6 and F), Non-toxic variety-callus 14 d (7 and G), Non-toxic variety-callus 38 d (8 and H). Mobile phase chloroform-methanol (97:3), cerium sulfate-revealed, observed at 366 nm UV light.</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_9'><ns0:head>Figure S2 .</ns0:head><ns0:label>S2</ns0:label><ns0:figDesc>Figure S2. Predicted structures related with the fragmentation profile from six flavonoids identified through ESI-MS from calluses extracts of non-toxic Jatropha curcas. It is included the predictive structure corresponding to vicenin-2,6''-O-Glucoside m/z 757 [M+H] + which is not reported to Jatropha curcas, but it is to Stellaria holostea.</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_10'><ns0:head>Figure 1 Figure 1</ns0:head><ns0:label>11</ns0:label><ns0:figDesc>Figure 1</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_11'><ns0:head>Figure 2</ns0:head><ns0:label>2</ns0:label><ns0:figDesc>Figure 2</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_12'><ns0:head>Figure 3</ns0:head><ns0:label>3</ns0:label><ns0:figDesc>Figure 3</ns0:figDesc><ns0:graphic coords='20,42.52,204.37,525.00,455.25' type='bitmap' /></ns0:figure>
<ns0:figure xml:id='fig_13'><ns0:head>Figure 4</ns0:head><ns0:label>4</ns0:label><ns0:figDesc>Figure 4</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_14'><ns0:head>Figure 5 Fragmentation</ns0:head><ns0:label>5</ns0:label><ns0:figDesc>Figure 5</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_15'><ns0:head>Figure 6</ns0:head><ns0:label>6</ns0:label><ns0:figDesc>Figure 6</ns0:figDesc></ns0:figure>
<ns0:figure type='table' xml:id='tab_0'><ns0:head>Table 1 (on next page)</ns0:head><ns0:label>1</ns0:label><ns0:figDesc /><ns0:table /></ns0:figure>
<ns0:figure type='table' xml:id='tab_1'><ns0:head>Table 1</ns0:head><ns0:label>1</ns0:label><ns0:figDesc>Tentative compounds identified by ESI-MS in hydroalcoholic extracts from seeds, leaves, and callus of 14 and 38 d of culture from both toxic and non-toxic Jatropha curcas L. varieties</ns0:figDesc><ns0:table /></ns0:figure>
<ns0:figure type='table' xml:id='tab_2'><ns0:head>Table 1 .</ns0:head><ns0:label>1</ns0:label><ns0:figDesc>Tentative compounds identified by ESI-MS in hydroalcoholic extracts from seeds, leaves, and callus of 14 and 38 d of culture from both toxic and non-toxic Jatropha curcas L. varieties.</ns0:figDesc><ns0:table><ns0:row><ns0:cell>Compound type/name</ns0:cell><ns0:cell>Elemental Composition</ns0:cell><ns0:cell>Mass</ns0:cell><ns0:cell>Fragment Ions in Positive Ion Mode (m/z)</ns0:cell><ns0:cell>Plant material</ns0:cell><ns0:cell>Time of culture (d)</ns0:cell><ns0:cell>Variety</ns0:cell><ns0:cell>Relative abundance (%)</ns0:cell></ns0:row><ns0:row><ns0:cell>Phorbol</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell>127, 185,</ns0:cell><ns0:cell>Seeds</ns0:cell><ns0:cell /><ns0:cell>T NT</ns0:cell><ns0:cell>64.70 21.05</ns0:cell></ns0:row><ns0:row><ns0:cell>12-deoxy-16-hydroxy-</ns0:cell><ns0:cell>C20H28O6</ns0:cell><ns0:cell>364.4</ns0:cell><ns0:cell>203, 234,</ns0:cell><ns0:cell /><ns0:cell>14</ns0:cell><ns0:cell>T</ns0:cell><ns0:cell>14.28</ns0:cell></ns0:row><ns0:row><ns0:cell>phorbol</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell>276, 295</ns0:cell><ns0:cell>Callus</ns0:cell><ns0:cell /><ns0:cell>NT</ns0:cell><ns0:cell>8.69</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell>38</ns0:cell><ns0:cell>T</ns0:cell><ns0:cell>30.00</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell>NT</ns0:cell><ns0:cell>25.00</ns0:cell></ns0:row><ns0:row><ns0:cell>Glycosylated Flavonoids</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>Apigenin 6-C-α-L-arabinopyranosyl-8-C-β-D-xylopyranoside (m/z 535 [M+H] + ) and Apigenin 4'-O-rhamnoside (m/z 417 [M+H] + )</ns0:cell><ns0:cell>C25H26O13 and C21H20O9</ns0:cell><ns0:cell>534.47 and 416.4</ns0:cell><ns0:cell>381</ns0:cell><ns0:cell>Leaves Callus</ns0:cell><ns0:cell>14 38</ns0:cell><ns0:cell>T NT T NT T NT</ns0:cell><ns0:cell>45.83 100 100 100 100 62.50</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell>Leaves</ns0:cell><ns0:cell /><ns0:cell>T NT</ns0:cell><ns0:cell>27.27 70.00</ns0:cell></ns0:row><ns0:row><ns0:cell>Vitexin (m/z 433 [M+H] + )</ns0:cell><ns0:cell>C21H20O10</ns0:cell><ns0:cell>432.37</ns0:cell><ns0:cell>415</ns0:cell><ns0:cell>Callus</ns0:cell><ns0:cell>14 38</ns0:cell><ns0:cell>T NT T NT</ns0:cell><ns0:cell>10.34 4.76 11.11 <12.50</ns0:cell></ns0:row><ns0:row><ns0:cell>Vitexin 4'-O-glucoside-2''-</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell>Leaves</ns0:cell><ns0:cell /><ns0:cell>T NT</ns0:cell><ns0:cell>33.33 42.84</ns0:cell></ns0:row><ns0:row><ns0:cell>O-rhamnoside (m/z 741 [M+H] + )</ns0:cell><ns0:cell>C33H40O19</ns0:cell><ns0:cell>740.7</ns0:cell><ns0:cell>577</ns0:cell><ns0:cell>Callus</ns0:cell><ns0:cell>14</ns0:cell><ns0:cell>T NT</ns0:cell><ns0:cell>1.42 6.36</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell>38</ns0:cell><ns0:cell>T NT</ns0:cell><ns0:cell>12.50 8.75</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell>Leaves</ns0:cell><ns0:cell /><ns0:cell>T NT</ns0:cell><ns0:cell>25.00 29.16</ns0:cell></ns0:row><ns0:row><ns0:cell>Vicenin-2 (m/z 595 [M+H] + )</ns0:cell><ns0:cell>C27H30O15</ns0:cell><ns0:cell>594.5</ns0:cell><ns0:cell>503</ns0:cell><ns0:cell>Callus</ns0:cell><ns0:cell>14</ns0:cell><ns0:cell>T NT</ns0:cell><ns0:cell>7.14 4.00</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell>38</ns0:cell><ns0:cell>T NT</ns0:cell><ns0:cell>6.25 3.75</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell>Leaves</ns0:cell><ns0:cell /><ns0:cell>T NT</ns0:cell><ns0:cell><9.09 8.00</ns0:cell></ns0:row><ns0:row><ns0:cell>Vicenin-2,6''-O-glucoside (m/z 757 [M+H] + )</ns0:cell><ns0:cell>C33H40O20</ns0:cell><ns0:cell>756.7</ns0:cell><ns0:cell>757</ns0:cell><ns0:cell>Callus</ns0:cell><ns0:cell>14</ns0:cell><ns0:cell>T NT</ns0:cell><ns0:cell>1.42 1.73</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell>38</ns0:cell><ns0:cell>T NT</ns0:cell><ns0:cell>2.22 1.25</ns0:cell></ns0:row><ns0:row><ns0:cell cols='2'>*T= Toxic, NT= Non-toxic</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row></ns0:table><ns0:note>PeerJ reviewing PDF | (2020:02:45707:1:0:NEW 1 Sep 2020)</ns0:note></ns0:figure>
<ns0:note place='foot'>PeerJ reviewing PDF | (2020:02:45707:1:0:NEW 1 Sep 2020)Manuscript to be reviewed</ns0:note>
</ns0:body>
" | "REBUTAL LETTER
August 30th, 2020
DR. PAWEL URBAN
ACADEMIC EDITOR PEERJ
Dear Dr. Urban:
Since we were allowed a longer period to respond to the comments of the different reviewers of the manuscript 'Chemical analysis of seeds, leaves and callus extracts of toxic and non-toxic varieties of Jatropha curcas L. ID: 45707', we had the opportunity to repeat experimental work and analysis of the results, which is evident in the resulting files. These files have been uploaded to the platform. Hoping they meet PeerJ's requirements.
Thanking you for your attention, I remain at your disposal for any comments.
Dr. Crescencio Bazaldúa
Profesor-Investigador Titular C
Departamento de Biotecnología
Centro de Desarrollo de Productos Bióticos - IPN
Calle Ceprobi No. 8. Col. San Isidro, C.P. 62731
Yautepec, Mor. Tel. +52 735 394 2020 Ext. 82528
Responses to reviewers….
Chemical analysis of seeds, leaves and callus extracts of toxic and non-toxic varieties of Jatropha curcas L.
Gerardo Leyva-Padrón, Pablo Emilio Vanegas-Espinoza, Alma Angélica Del Villar-Martínez, Crescencio Bazaldúa
Reviewer 1 (Anonymous)
Basic Reporting
Q1. Authors need to connect the story as the whole MS is missing the flow.
R. The specific modification was included
Original: The molecular ions related to the extracts were analyzed in positive ion mode (ESI+). 20 μL of sample were injected, capillary potential was -4.5 kV, Lines 123-124
Modified: The molecular ions related to the extracts were analyzed in positive ion mode (ESI+). 20 μL of sample were directly injected into the evaporation chamber, capillary potential was -4.5 kV, Lines 129-131
Q2. The chosen title is not the best one.
R. Title was modified. Lines 1-2
Original: Chemical analysis of seeds, leaves and callus extracts of toxic and non-toxic varieties of Jatropha curcas L.
Modified: Chemical analysis of callus extracts from toxic and non-toxic varieties of Jatropha curcas L.
Experimental design
Q3. Jatropha is a widely studied plant and has important commercial value. I suggest enriching the introduction and discussion part on the basis of current findings. What is the originality of your work?
R. Originality is highlighted
Introduction.
Original: There are a few cases where the use of plant cell culture has allowed the production of active compounds, even more biotechnological production either as pure compounds or as standardized extracts, provides unlimited opportunities for new drug discoveries due to the great chemical diversity (Karuppusamy, 2009). Lines: 63-66
Modified: There are a few cases where the use of plant cell culture of Jatropha curcas has allowed the production of bioactive compounds (Alvero-Bascos & Ungson, 2012; Mahalakshmi, Eganathan & Parida, 2013; Nassar, El-Ahmay & Al-Azizi, 2013; Zaragoza-Martínez et al., 2016), the study of the culture at different stages of toxic and non-toxic varieties, generate the opportunity to design biotechnological models for production of bioactive compounds i.e. terpenoids, alkaloids, flavonoids (Abdelgavir & Van Staden, 2013; Sabandar et al., 2013) providing opportunities for new drugs discovery. Lines: 62-68
Added to Discussion.
Stable callus culture conditions for the two varieties of Jatropha curcas were established. Lines: 206-207
Using TLC, differences between dedifferentiated cell extracts of both varieties of Jatropha curcas were evidenced. Lines 226-227
By ESI-MS, differences in the relative intensity of the signal corresponding to DHP were observed between the callus extracts of both varieties, being higher in the toxic variety in addition to that in calluses at 38 d of culture, it was higher than in the 14 d. Lines 250-253
By the MS-MS fragmentation profile, the identification of six glycosylated flavonoids is suggested, it was observed that relative intensities signals related to flavonoid related molecular ion m/z 381 [M+H]+ in callus of 14 d was higher than callus of 38 d, differences that were not observed between calluses of the different varieties. Lines 284-287
Q4. Why have you chosen toxic and non-toxic varieties?
R. The aim was to know the way in which the phorbol esters accumulate during the development of dedifferentiated cells in toxic and non-toxic varieties and also detect the differential accumulation of other compounds with bioactive potential such as flavonoids.
Original: The aim of this work was to characterize the chemical profile of the extracts from seeds, leaves and callus of both varieties (toxic and non-toxic) of Jatropha curcas, to verify the presence of important compounds in dedifferentiated cells and consider the possibility of using these cultures for the massive production of metabolites. Lines 70-73
Modified: The aim of this work was to characterize the chemical profile of the extracts from callus of both varieties (toxic and non-toxic) of Jatropha curcas, through the cell culture, to verify the presence of important compounds in dedifferentiated cells and consider the possibility of using these cultures for the massive production of bioactive compounds. Lines 72-75
Q5. Are there any similarities or differences in the chemical profiles on your varieties and other varieties reported by other authors?
R. Varieties used in this work have not been analyzed by other authors in the same parameters, there are no studies comparing the phorbol esters content in callus of toxic and non-toxic varieties. Publications in which comparisons between toxic and non-toxic varieties seeds are reported, phorbol esters were not detected in non-toxic varieties (Laviola et al., 2010; Kumar, Anand & Reddy, 2011); in this work phorbol esters were qualitatively detected in non-toxic seeds and callus. Besides, flavonoids were, for first time, identified in leaves and callus from both varieties.
Original: In Mexico, it has been identified a non-toxic variety of this species with very low or non-detectable levels of phorbol esters (PEs) (Martínez-Herrera, Chel-Guerrero & Martínez-Ayala, 2004). The highest accumulation of PEs is at the seeds. PEs are known as Jatropha factors because each one of them has the same nucleus diterpene moiety, namely, 12-deoxy-16-hydroxy-phorbol (DHP) which is coupled to unstables intramolecular diterpenes (named C1–C6 factors) (Hirota et al., 1988; Haas, Sterk & Mittelbach, 2002; Baldini et al., 2014; Nishshanka et al., 2016). Lines 52 -57
Modified: In Mexico, Brazil and India, it have been identified non-toxic varieties of this species with very low or non-detectable levels of phorbol esters (PEs) in seeds (Laviola et al., 2010; Martínez-Herrera, Chel-Guerrero & Martínez-Ayala, 2004; Kumar, Anand & Reddy, 2011). PEs are known as Jatropha factors because each one of them has the same nucleus diterpene moiety, namely, 12-deoxy-16-hydroxy-phorbol (DHP) which is coupled to unstables intramolecular diterpenes (named C1–C6 factors) (Hirota et al., 2017). Lines 51-56
Validity of the findings
Q6. I think the conclusion past can be improved.
R. Conclusions were modified.
Original: During cell dedifferentiation NAA and BAP at the same concentration induced the highest amount of callus in petiole explants from both toxic and non-toxic varieties of Jatropha curcas. The variety of the species did not influenced the cell dedifferentiation. Soxlet-methanol extraction was more efficient than sonication-ethanol 80% to obtain phorbol esters type compounds from seeds and callus. Thin layer chromatography and mass spectrometry, suggest an inverse relationship between phorbol esters and flavonoids accumulation in callus throughout the time of culture. The tentative identification of diterpene type compounds such as 12-deoxy-16-hydroxy-phorbol and Jatropha factors by ESI-MS in seed and callus (14 and 38 d) extracts of J. curcas, from both toxic and non-toxic varieties, as well as, the presence of six flavonoids glycosides in leaf and callus, from both toxic and non-toxic varieties, is suggested. Lines 275-284
Modified: Stable dedifferentiated cells culture from petiole explants of Jathopha curcas, were stablished from toxic and non-toxic varieties on MS medium added with NAA and BAP. Thin layer chromatography and mass spectrometry, suggest an inverse relationship between phorbol esters and flavonoids accumulation in callus throughout the time of culture. The tentative identification of diterpene type compounds such as 12-deoxy-16-hydroxy-phorbol and Jatropha factors by ESI-MS in seed and callus (14 and 38 d), as well as, the presence of six flavonoids glycosides in leaf and callus, in extracts from both toxic and non-toxic varieties of J. curcas, is suggested. Both of them, groups of compounds reported with bioactive activity with pharmaceutic/agroindistrial potential. Lines 289-296
Q7. It would be nice to summarize all findings in the chemical profile of different parts and varieties in a table. Table this will help the readers to understand the findings quickly.
R. Table 1 was modified, and the results were summarized.
Original:
Not mentioned
Inversely to observed on DHPB related signal (m/z 365 [M+H]+), the intensity of the molecular ion related with flavonoids diminished, but there was not difference between type of extracts, leaves or callus from both varieties. Lines 173-175
Modified: Table 1 shows the relative abundance of DHP molecular ion (m/z 365 [M+H]+) on 14 d and 38 d callus extracts from both varieties, evidencing the increment of these compounds through the callus development. Lines: 168-170
Table 1 shows the relative abundance of six tentatively identified compounds by relating their molecular ions on 14 d and 38 d callus extracts from both varieties. Inversely to observed on DHP related signal (m/z 365 [M+H]+), the intensity of the molecular ion related with glycosylated apigenin (m/z 381 [M+H]+) diminished. Lines 181-183
Q8. Some figures have poor quality (i.e., fig 6, fig 4).
R. Figures 1, 2, 4 y 6 were improved
Q9. As a reader, I could not find the importance of this work. Please highlight and specify your findings.
R. Originality is highlighted
Introduction.
Original: There are a few cases where the use of plant cell culture has allowed the production of active compounds, even more biotechnological production either as pure compounds or as standardized extracts, provides unlimited opportunities for new drug discoveries due to the great chemical diversity (Karuppusamy, 2009). Lines: 63-66
Modified: There are a few cases where the use of plant cell culture of Jatropha curcas has allowed the production of bioactive compounds (Alvero-Bascos & Ungson, 2012; Mahalakshmi, Eganathan & Parida, 2013; Nassar, El-Ahmay & Al-Azizi, 2013; Zaragoza-Martínez et al., 2016), the study of the culture at different stages of toxic and non-toxic varieties, generate the opportunity to design biotechnological models for production of bioactive compounds i.e. terpenoids, alkaloids, flavonoids (Abdelgavir & Van Staden, 2013; Sabandar et al., 2013) providing opportunities for new drugs discovery. Lines: 62-68
Added to Discussion.
Stable callus culture conditions for the two varieties of Jatropha curcas were established. Lines: 206-207
Using TLC, differences between dedifferentiated cell extracts of both varieties of Jatropha curcas were evidenced. Lines 226-227
By ESI-MS, differences in the relative intensity of the signal corresponding to DHP were observed between the callus extracts of both varieties, being higher in the toxic variety in addition to that in calluses at 38 d of culture, it was higher than in the 14 d. Lines 250-253
By the MS-MS fragmentation profile, the identification of six glycosylated flavonoids is suggested, it was observed that relative intensities signals related to flavonoid related molecular ion m/z 381 [M+H]+ in callus of 14 d was higher than callus of 38 d, differences that were not observed between calluses of the different varieties. Lines 284-287
Q10. Why did you plant this work? do you have worthy findings?
R. The aim was to know if the accumulation of phorbol esters still varies at level of dedifferentiated cells. Worthy findings are referred in Conclusions
Original: During cell dedifferentiation NAA and BAP at the same concentration induced the highest amount of callus in petiole explants from both toxic and non-toxic varieties of Jatropha curcas. The variety of the species did not influenced the cell dedifferentiation. Soxlet-methanol extraction was more efficient than sonication-ethanol 80% to obtain phorbol esters type compounds from seeds and callus. Thin layer chromatography and mass spectrometry, suggest an inverse relationship between phorbol esters and flavonoids accumulation in callus throughout the time of culture. The tentative identification of diterpene type compounds such as 12-deoxy-16-hydroxy-phorbol and Jatropha factors by ESI-MS in seed and callus (14 and 38 d) extracts of J. curcas, from both toxic and non-toxic varieties, as well as, the presence of six flavonoids glycosides in leaf and callus, from both toxic and non-toxic varieties, is suggested. Lines 275-284
Modified: Stable dedifferentiated cells culture from petiole explants of Jathopha curcas, were stablished from toxic and non-toxic varieties on MS medium added with NAA and BAP. Thin layer chromatography and mass spectrometry, suggest an inverse relationship between phorbol esters and flavonoids accumulation in callus throughout the time of culture. The tentative identification of diterpene type compounds such as 12-deoxy-16-hydroxy-phorbol and Jatropha factors by ESI-MS in seed and callus (14 and 38 d), as well as, the presence of six flavonoids glycosides in leaf and callus, in extracts from both toxic and non-toxic varieties of J. curcas, is suggested. Both of them, groups of compounds reported with bioactive activity with pharmaceutic/agroindistrial potential. Lines 289-296
Reviewer 2 (Anonymous)
Experimental design
Q11. Methods have to be improved for publication: quantification of the different compounds (at least relative) is strongly needed to support the conclusions
R. Methodology was added.
Modificated: Relative abundance was calculated according to Scigelova et al. (2011). Line 136
Table 1 describes relative abundance. It was not possible to carry out the compound’s quantification.
Table 1 was modified, and the results were summarized.
Original:
Non mentioned
Inversely to observed on DHPB related signal (m/z 365 [M+H]+), the intensity of the molecular ion related with flavonoids diminished, but there was not difference between type of extracts, leaves or callus from both varieties. Lines 173-175
Modified:
Table 1 shows the relative abundance of DHP molecular ion (m/z 365 [M+H]+) on 14 d and 38 d callus extracts from both varieties, evidencing the increment of these compounds through the callus development. Lines: 168-170
Table 1 shows the relative abundance of six tentatively identified compounds by relating their molecular ions on 14 d and 38 d callus extracts from both varieties. Inversely to observed on DHP related signal (m/z 365 [M+H]+), the intensity of the molecular ion related with glycosylated apigenin (m/z 381 [M+H]+) diminished. Lines 181-183
Q12. Methods section has to be more clearly described. In particular, US apparatus as well as US extraction parameters are not described in details. The choice of the extraction parameters is not justified. The description of the US apparatus did respect the conventional description. The authors will find these standards in the paper “Ultrasonically assisted extraction (UAE) of natural products some guidelines for good practice and reporting” published in Ultrasonics Sonochemistry, 25 (2015) 94-95.
R. After reviewing the recommended article, the ultrasound frequency, type of flask used during sonication and its position within the ultrasound bath were incorporated.
Original:
Extract obtaining
20 mL of ethanol 80% (v/v) were added to 500 mg of biomass dry weight (dw) and sonicated at 40 ± 5 °C during 30 min (Bransonic Ultrasonic Cleaner, 2510R-MTH, CT, USA) (Bazaldúa et al. 2019), subsequently vortexed. Supernatant was filtered, concentrated to dryness at 25 ± 2 °C, and solubilized in 500 μL of HPLC grade MeOH (Sigma-Aldrich®) for chromatographic analysis (Saeed et al., 2006; Liu et al, 2013). Lines 96-101
Modified:
Ultrasound assisted extraction (UAE)
UAE was performed with an ultrasound bath Branson (2510R-MTH, CT, USA) with automatic control of time and temperature and ultrasound frequency of 40 kHz. 500 mg of grounded biomass dry weight (dw) were placed into a 50 mL borosilicate glass conical Erlenmeyer flask, then 20 mL of ethanol 80% (v/v) were added, and sonicated at 40 ± 5 °C during 30 min (Bazaldúa et al., 2019; Pandey et al., 2018; Dumitraşcu et al., 2019). During sonication flasks were suspended in the water without contact with the bottom of the ultrasonic bath, subsequently they were vortexed. Supernatant was filtered, concentrated to dryness at 25 ± 2 °C, and solubilized in 500 μL of HPLC grade MeOH (Sigma-Aldrich®) for chromatographic analysis (Saeed et al., 2006; Liu et al, 2013). Lines 99-107
Q13. The reference cited by the authors is not related (ie, on podophyllotoxin extraction from another plant source!)
R. References are added regarding the application of the Ultrasound assisted extraction method in the extraction of compounds from different plants
Original: 20 mL of ethanol 80% (v/v) were added to 500 mg of biomass dry weight (dw) and sonicated at 40 ± 5 °C during 30 min (Bransonic Ultrasonic Cleaner, 2510R-MTH, CT, USA) (Bazaldúa et al. 2019), subsequently vortexed. Supernatant was filtered, concentrated to dryness at 25 ± 2 °C, and solubilized in 500 μL of HPLC grade MeOH (Sigma-Aldrich®) for chromatographic analysis (Saeed et al., 2006; Liu et al, 2013). Lines 97-101
Modified: UAE was performed with an ultrasound bath Branson (2510R-MTH, CT, USA) with automatic control of time and temperature and ultrasound frequency of 40 kHz. 500 mg of grounded biomass dry weight (dw) were placed into a 50 mL borosilicate glass conical Erlenmeyer flask, then 20 mL of ethanol 80% (v/v) were added, and sonicated at 40 ± 5 °C during 30 min (Bazaldúa et al., 2019; Pandey et al., 2018; Dumitraşcu et al., 2019). During sonication flasks were suspended in the water without contact with the bottom of the ultrasonic bath, subsequently they were vortexed. Supernatant was filtered, concentrated to dryness at 25 ± 2 °C, and solubilized in 500 μL of HPLC grade MeOH (Sigma-Aldrich®) for chromatographic analysis (Saeed et al., 2006; Liu et al, 2013). Lines 100-107
Validity of the findings
Q14. Impact and novelty are basic
R. Originality is highlighted
Introduction.
Original: There are a few cases where the use of plant cell culture has allowed the production of active compounds, even more biotechnological production either as pure compounds or as standardized extracts, provides unlimited opportunities for new drug discoveries due to the great chemical diversity (Karuppusamy, 2009). Lines: 63-66
Modified: There are a few cases where the use of plant cell culture of Jatropha curcas has allowed the production of bioactive compounds (Alvero-Bascos & Ungson, 2012; Mahalakshmi, Eganathan & Parida, 2013; Nassar, El-Ahmay & Al-Azizi, 2013; Zaragoza-Martínez et al., 2016), the study of the culture at different stages of toxic and non-toxic varieties, generate the opportunity to design biotechnological models for production of bioactive compounds i.e. terpenoids, alkaloids, flavonoids (Abdelgavir & Van Staden, 2013; Sabandar et al., 2013) providing opportunities for new drugs discovery. Lines: 62-68
Added to Discussion.
Stable callus culture conditions for the two varieties of Jatropha curcas were established. Lines: 206-207
Using TLC, differences between dedifferentiated cell extracts of both varieties of Jatropha curcas were evidenced. Lines 226-227
By ESI-MS, differences in the relative intensity of the signal corresponding to DHP were observed between the callus extracts of both varieties, being higher in the toxic variety in addition to that in calluses at 38 d of culture, it was higher than in the 14 d. Lines 250-253
By the MS-MS fragmentation profile, the identification of six glycosylated flavonoids is suggested, it was observed that relative intensities signals related to flavonoid related molecular ion m/z 381 [M+H]+ in callus of 14 d was higher than callus of 38 d, differences that were not observed between calluses of the different varieties. Lines 284-287
Conclusions.
Original: During cell dedifferentiation NAA and BAP at the same concentration induced the highest amount of callus in petiole explants from both toxic and non-toxic varieties of Jatropha curcas. The variety of the species did not influenced the cell dedifferentiation. Soxlet-methanol extraction was more efficient than sonication-ethanol 80% to obtain phorbol esters type compounds from seeds and callus. Thin layer chromatography and mass spectrometry, suggest an inverse relationship between phorbol esters and flavonoids accumulation in callus throughout the time of culture. The tentative identification of diterpene type compounds such as 12-deoxy-16-hydroxy-phorbol and Jatropha factors by ESI-MS in seed and callus (14 and 38 d) extracts of J. curcas, from both toxic and non-toxic varieties, as well as, the presence of six flavonoids glycosides in leaf and callus, from both toxic and non-toxic varieties, is suggested. Lines 275-284
Modified: Stable dedifferentiated cells culture from petiole explants of Jathopha curcas, were stablished from toxic and non-toxic varieties on MS medium added with NAA and BAP. Thin layer chromatography and mass spectrometry, suggest an inverse relationship between phorbol esters and flavonoids accumulation in callus throughout the time of culture. The tentative identification of diterpene type compounds such as 12-deoxy-16-hydroxy-phorbol and Jatropha factors by ESI-MS in seed and callus (14 and 38 d), as well as, the presence of six flavonoids glycosides in leaf and callus, in extracts from both toxic and non-toxic varieties of J. curcas, is suggested. Both of them, groups of compounds reported with bioactive activity with pharmaceutic/agroindistrial potential. Lines 289-296
Q15. no quantification therefore no statistical analysis, this is a drawback for the present work.
R. Relative abundance was calculated according to Scigelova et al. (2011). Line 136
Table 1 describes relative abundance. It was not possible to carry out the compound’s quantification.
Table 1 was modified, and the results were summarized.
Original:
Non mentioned
Inversely to observed on DHPB related signal (m/z 365 [M+H]+), the intensity of the molecular ion related with flavonoids diminished, but there was not difference between type of extracts, leaves or callus from both varieties. Lines 173-175
Modified:
Table 1 shows the relative abundance of DHP molecular ion (m/z 365 [M+H]+) on 14 d and 38 d callus extracts from both varieties, evidencing the increment of these compounds through the callus development. Lines: 168-170
Table 1 shows the relative abundance of six tentatively identified compounds by relating their molecular ions on 14 d and 38 d callus extracts from both varieties. Inversely to observed on DHP related signal (m/z 365 [M+H]+), the intensity of the molecular ion related with glycosylated apigenin (m/z 381 [M+H]+) diminished. Lines 181-183
Q16. Conclusions have to be rewritten to answer the main questions of the present work
R. Conclusions were modified.
Original: During cell dedifferentiation NAA and BAP at the same concentration induced the highest amount of callus in petiole explants from both toxic and non-toxic varieties of Jatropha curcas. The variety of the species did not influenced the cell dedifferentiation. Soxlet-methanol extraction was more efficient than sonication-ethanol 80% to obtain phorbol esters type compounds from seeds and callus. Thin layer chromatography and mass spectrometry, suggest an inverse relationship between phorbol esters and flavonoids accumulation in callus throughout the time of culture. The tentative identification of diterpene type compounds such as 12-deoxy-16-hydroxy-phorbol and Jatropha factors by ESI-MS in seed and callus (14 and 38 d) extracts of J. curcas, from both toxic and non-toxic varieties, as well as, the presence of six flavonoids glycosides in leaf and callus, from both toxic and non-toxic varieties, is suggested. Lines 275-284
Modified: Stable dedifferentiated cells culture from petiole explants of Jathopha curcas, were stablished from toxic and non-toxic varieties on MS medium added with NAA and BAP. Thin layer chromatography and mass spectrometry, suggest an inverse relationship between phorbol esters and flavonoids accumulation in callus throughout the time of culture. The tentative identification of diterpene type compounds such as 12-deoxy-16-hydroxy-phorbol and Jatropha factors by ESI-MS in seed and callus (14 and 38 d), as well as, the presence of six flavonoids glycosides in leaf and callus, in extracts from both toxic and non-toxic varieties of J. curcas, is suggested. Both of them, groups of compounds reported with bioactive activity with pharmaceutic/agroindistrial potential. Lines 289-296
Comments for the Author
Q17. The need of in vitro culture initiation is not described by the authors
R. The aim was to know if the accumulation of phorbol esters still varies at level of dedifferentiated cells. Working with varieties in which callus induction has not been reported, it was necessary to carry out test to obtain dedifferentiated cells.
Original: Nine treatments, as result of the combination of three concentrations (0.0, 1.5 and 3.0 mg.L-1) of both naphthaleneacetic acid (NAA) and 6-benzyl-aminopurine (BAP) were evaluated (Verma, 2013). Lines 85-87
Modified: Since there are no reports of the induction of dedifferentiated cells in the varieties analyzed in this study, the combinations of three concentrations (0.0, 1.5 and 3.0 mg.L-1) of naphthaleneacetic acid (NAA) and 6-benzyl-aminopurine (BAP) were evaluated according to Verma (2013), pH was adjusted to 5.7, media were sterilized at 121 °C for 15 min. Lines 86-90
Q18. Quantification (at least relative) is vital to present firm conclusion
R. Methodology was added.
Modificated: Relative abundance was calculated according to Scigelova et al. (2011). Line 136
It was not possible to carry out the compounds quantification. Table 1 was modified, and the results were summarized.
Q19. Title is not correct. Callus are not stabilized so 'during callus initiation' should be used instead
R. Characteristics of the evaluated callus were specified in Methodology.
Original: Not mentioned
Modified: In order to observe differences in accumulation of compounds during callus development, completely dedifferentiated cells were cultured under the above described conditions for 38 d, samples were taken on days 0, 2 and every 4 d thereafter. Lines 93-95
Q20. Difference between toxic and non-toxic is unclear in the conclusion (again quantification and statistical analysis are required)
R. It was not possible to carry out the compound’s quantification. Table 1 was modified, and the results were summarized.
Conclusions were modified.
Original: During cell dedifferentiation NAA and BAP at the same concentration induced the highest amount of callus in petiole explants from both toxic and non-toxic varieties of Jatropha curcas. The variety of the species did not influenced the cell dedifferentiation. Soxlet-methanol extraction was more efficient than sonication-ethanol 80% to obtain phorbol esters type compounds from seeds and callus. Thin layer chromatography and mass spectrometry, suggest an inverse relationship between phorbol esters and flavonoids accumulation in callus throughout the time of culture. The tentative identification of diterpene type compounds such as 12-deoxy-16-hydroxy-phorbol and Jatropha factors by ESI-MS in seed and callus (14 and 38 d) extracts of J. curcas, from both toxic and non-toxic varieties, as well as, the presence of six flavonoids glycosides in leaf and callus, from both toxic and non-toxic varieties, is suggested. Lines 275-284
Modified: Stable dedifferentiated cells culture from petiole explants of Jathopha curcas, were stablished from toxic and non-toxic varieties on MS medium added with NAA and BAP. Thin layer chromatography and mass spectrometry, suggest an inverse relationship between phorbol esters and flavonoids accumulation in callus throughout the time of culture. The tentative identification of diterpene type compounds such as 12-deoxy-16-hydroxy-phorbol and Jatropha factors by ESI-MS in seed and callus (14 and 38 d), as well as, the presence of six flavonoids glycosides in leaf and callus, in extracts from both toxic and non-toxic varieties of J. curcas, is suggested. Both of them, groups of compounds reported with bioactive activity with pharmaceutic/agroindistrial potential. Lines 289-296
Q21. Objectives of the present work are not clearly exposed
R. Objective was modified.
Original: The aim of this work was to characterize the chemical profile of the extracts from seeds, leaves and callus of both varieties (toxic and non-toxic) of Jatropha curcas, to verify the presence of important compounds in dedifferentiated cells and consider the possibility of using these cultures for the massive production of metabolites. Lines 70-73
Modified: The aim of this work was to characterize the chemical profile of the extracts from callus of both varieties (toxic and non-toxic) of Jatropha curcas, through the cell culture, to verify the presence of important compounds in dedifferentiated cells and consider the possibility of using these cultures for the massive production of bioactive compounds. Lines 72-75
Reviewer 3 (Alberdan Santos)
Basic reporting
Q22. The authors do not refer to the occurrence of J. curcas in other countries to compare their toxicity. In these aspects, the author should introduce some aspects broader out of the local place studying. Its relevance should be mentioned because of the chemotype variety is not discussed to justify the work. The article should include sufficient information if or not chemotypes are one of the important aspects of the presence or absence of the aim metabolite in the introduction, and relate it to J. curcas toxicity.
Original: In Mexico, it has been identified a non-toxic variety of this species with very low or non-detectable levels of phorbol esters (PEs) (Martínez-Herrera, Chel-Guerrero & Martínez-Ayala, 2004). The highest accumulation of PEs is at the seeds. PEs are known as Jatropha factors because each one of them has the same nucleus diterpene moiety, namely, 12-deoxy-16-hydroxy-phorbol (DHP) which is coupled to unstables intramolecular diterpenes (named C1–C6 factors) (Hirota et al., 1988; Haas, Sterk & Mittelbach, 2002; Baldini et al., 2014; Nishshanka et al., 2016). Lines 52 -57
Modified: In Mexico, Brazil and India, it have been identified non-toxic varieties of this species with very low or non-detectable levels of phorbol esters (PEs) in seeds (Laviola et al., 2010; Martínez-Herrera, Chel-Guerrero & Martínez-Ayala, 2004; Kumar, Anand & Reddy, 2011). PEs are known as Jatropha factors because each one of them has the same nucleus diterpene moiety, namely, 12-deoxy-16-hydroxy-phorbol (DHP) which is coupled to unstables intramolecular diterpenes (named C1–C6 factors) (Hirota et al., 2017). Lines 52-56
Q23. Figure 2, it is not so clear, it must be sufficient clearness to show the reference standard. When the extract amount is applied as spot the band appears as spots when applied as a horizontal band (like on electrophoresed gel), it should appear in this paper in the horizontal band form, the results present in this article are not so conventional. It needs an explanation.
R. Figure 2 was modified. A new cell growth kinetics, compounds extraction and chromatographic plates were performed with three reference standards (Phorbol-12-myristate-13-acetate, Quercetin, and Vitexin Sigma).
Q24. Besides, it is important to explain the TLC variation in the volume application to justify the not uniform profiles amounts.
R. In the new chromatographic plate the applied volumes were uniformized. Figure 2
Q25. In line 212, the statement is not appropriate, because TLC does not identify, It does detect the aim metabolite, groups or class. it is important to review.
R. Concept was modified.
Original: The identification by TLC of PEs in seed methanolic extracts from toxic and non-toxic J. curcas varieties was reported Line 208-209
Makkar & Becker (2009) identified higher PEs accumulation in seeds than in leaves extracts. Line 212-213
Modified: The detection by TLC of PEs in seed methanolic extracts from toxic and non-toxic J. curcas varieties was reported Line 217-218
Makkar & Becker (2009) detected higher PEs accumulation in seeds than in leaves extracts. Line 221-222
Reviewer 4 (Usama Aly)
Comments for the Author
Q26. Generally, the paper needs English editing for some spelling and grammar mistakes along with many unclear sentences.
R. Comments indicated in the PDF were addressed.
Original: …evolution… Line 90
Modification: …dedifferentiation… Line 92
Subtitles were modified were it was considered
Original: Extract obtaining Line 96
Modification: Ultrasound assisted extraction (UAE). Line 99
Original: Dedifferentiation began at the cutting sites because of the high response of cells as expected… Line 183-184
Modification: Dedifferentiation began at the cutting sites as expected… Line 192
Original: …reported that it is necessary the presence of NAA in the culture medium to induce... Line 190-191
Modification: …reported that the presence of NAA induce… Line 199
Original: …this effect could be increased when adding a cytokinins like BAP. Line 192
Modification: …this effect could be increased when adding BAP. Line 200
Original: …through hydroalcoholic extraction, while… higher on… Line 215-216
Modification: …quantified by HPLC;… Line 225
Original: …regard to type of organ, variety, and in cell cultures, throughout… Line 236-237
Modification: …regard to the variety-derived cell culture and throughout… Line 246
Q27. What is the main clear idea(s) we get from this paper? Please clarify in conclusion section.
R. Conclusions were modified.
Original: During cell dedifferentiation NAA and BAP at the same concentration induced the highest amount of callus in petiole explants from both toxic and non-toxic varieties of Jatropha curcas. The variety of the species did not influenced the cell dedifferentiation. Soxlet-methanol extraction was more efficient than sonication-ethanol 80% to obtain phorbol esters type compounds from seeds and callus. Thin layer chromatography and mass spectrometry, suggest an inverse relationship between phorbol esters and flavonoids accumulation in callus throughout the time of culture. The tentative identification of diterpene type compounds such as 12-deoxy-16-hydroxy-phorbol and Jatropha factors by ESI-MS in seed and callus (14 and 38 d) extracts of J. curcas, from both toxic and non-toxic varieties, as well as, the presence of six flavonoids glycosides in leaf and callus, from both toxic and non-toxic varieties, is suggested. Lines 275-284
Modified: Stable dedifferentiated cells culture from petiole explants of Jathopha curcas, were stablished from toxic and non-toxic varieties on MS medium added with NAA and BAP. Thin layer chromatography and mass spectrometry, suggest an inverse relationship between phorbol esters and flavonoids accumulation in callus throughout the time of culture. The tentative identification of diterpene type compounds such as 12-deoxy-16-hydroxy-phorbol and Jatropha factors by ESI-MS in seed and callus (14 and 38 d), as well as, the presence of six flavonoids glycosides in leaf and callus, in extracts from both toxic and non-toxic varieties of J. curcas, is suggested. Both of them, groups of compounds reported with bioactive activity with pharmaceutic/agroindistrial potential. Lines 289-296
Q27. Add photo for seeds and leaves of toxic and non-toxic Jatropha varieties at the initial stage of cultivation.
R. The research is based on the study of calluses of both varieties, as there are no substantial differences between the leaves and seeds of the materials, and it is not the subject of the research study, it is not considered necessary to add photographs.
Q28. About 61 references include only five references raised last five years, please update with recent references.
R. Bibliographic references were updated.
Abdelgader MGM, Suleiman EA, Ali SI. 2019. Study of Jatropha curcas as antifungal agent. International Journal of Current Medical and Pharmaceutical Research 5(5), 4202-4210 DOI: 10.24327/23956429.ijcmpr201905657.
Armaly AM, DePorre YC, Groso EJ, Riehl PS, Schindler CS. 2015. Discovery of novel synthetic methodologies and reagents during natural product synthesis in the post-palytoxin era. Chemical Reviews 115(17), 9232-9276 DOI: 10.1021/acs.chemrev.5b00034.
Aye MM, Aung HT, Sein MM, Armijos C. 2019. A review on the phytochemistry, medicinal properties and pharmacological activities of 15 selected Myanmar medicinal plants. Molecules 24(2), 293 DOI: 10.3390/molecules24020293.
Chemat F, Abert-Vian M, Fabiano-Tixier AS, Strube J, Uhlenbrock L, Gunjevic V, Cravotto G. 2019. Green extraction of natural products. Origins, current status, and future challenges. TrAC Trends in Analytical Chemistry 118, 248-263 DOI: 10.1016/j.trac.2019.05.037.
Dumitraşcu L, Enachi E, Stänciuc N, Aprodu J. 2019. Optimization of ultrasound assisted extraction of phenolic compounds from cornelian cherry fruits using response surface methodology. CyTA-Journal of Food 17(1), 814-823 DOI: 10.1080/19476337.2019.1659418.
Ferreira-Rodrigues SC, Rodrigues CM, Dos Santos MG, Gautuz JAA, Silva MG, Cogo JC, Oshima-Franco Y. 2016. Anti-inflammatory and antibothropic properties of Jatropha elliptica, a plant from brazilian cerrado biome. Advanced Pharmaceutical Bulletin 6(4), 573 DOI: 10.15171/apb.2016.071.
Hassan WH, Abdelaziz S, Al Yousef HM. 2019. Chemical composition and biological activities of the aqueous araction of Parkinsonea aculeata L. growing in Saudi Arabia. Arabian Journal of Chemistry 12(3), 377-387 DOI: 10.1016/j.arabjc.2018.08.003.
Hernandez LW, Sarlah D. 2019. Empowering synthesis of complex natural products. Chemistry–A European Journal 25(58), 13248-13270 DOI: 10.1002/chem.201901808.
Hirota F, Maitree S, Anchalee R, Keisuke L, Pornngarm L. Sonthaya U, Masami S. 2017. Phorbol esters in seed oil of Jatropha curcas L. (saboodam in Thai) and their association with cancer prevention: from the initial investigation to the present topics. Journal of Cancer Research and Clinical Oncology 143:1359-1369 DOI: 10.1007/s00432-017-2341-6.
Katagi A, Sui L, Kamitori K, Suzuki T, Katayama T, Hossain A, Tokuda M. 2016. Inhibitory effect of isoamericanol A from Jatropha curcas seeds on the growth of MCF-7 human breast cancer cell line by G2/M cell cycle arrest. Heliyon 2(1), e00055 DOI: 10.1016/j.heliyon.2015.e00055.
Kumar N, Goel N. 2019. Phenolic acids: Natural versatile molecules with promising therapeutic applications. Biotechnology Reports e00370 DOI: 10.1016/j.btre.2019.e00370.
Pandey A, Belwal T, Sekar TC, Bhatt ID, Rawa RS. 2018. Optimization of ultrasonic-assisted (UAE) of phenolics and antioxidant compounds from rhizomes of Rheum moorcroftianum using response surface methodology (RSM). Industrial Crops and Products 119(1), 28-225 DOI: 10.1016/j.indcrop.2018.04.019.
Pereira A. 2016. Plant abiotic stress challenges from the changing environment. Frontiers in Plant Science 7, 1123 DOI: 10.3389/fpls.2016.01123.
Pezzini V, Agostini F, Smiderle F, Touguinha L, Salvador M, Moura S. 2019. Grape juice by-products extracted by ultrasound and microwave-assisted with different solvents: a rich chemical composition. Food Science and Biotechnology 28(3), 691-699 DOI: 10.1007/s10068-018-0531-x.
Rampadarath S, Puchooa D, Jeewon R. 2016. Jatropha curcas L: Phytochemical, antimicrobial and larvicidal properties. Asian Pacific Journal of Tropical Biomedicine 6(10), 858-865 DOI: 10.1016/j.apjtb.2016.01.019.
Salim MN, Masyitha D, Harris A, Balqis U, Iskandar CD, Hambal M. 2018. Anti-inflammatory activity of Jatropha curcas Linn. latex in cream formulation on CD68 expression in mice skin wound. Veterinary world 11(2), 99 DOI: 10.14202/vetworld.2018.99-103.
Verardo G, Baldini M, Ferfuia C, Gorassini A. 2019. Rapid and selective screening for toxic phorbol esters in Jatropha curcas seed oil using high-performance liquid chromatography-electrospray ionization-tandem mass spectrometry. Journal of Chromatography A 1597, 63-75 DOI: 10.1016/j.chroma.2019.03.015.
Zhang Y, Yang Q, Li Ch, Ding M, Lv X, Tao Ch, Yu H, Chen F, Ying Xu Y. 2017. Curcin C, a novel type I ribosome‑inactivating protein from the post‑germinating cotyledons of Jatropha curcas. Amino Acids 49, 1619–1631 DOI 10.1007/s00726-017-2456-8.
Q29. You are kindly requested to follow the corrections made within attachment file.
R. Comments indicated in the PDF were addressed. Q25
" | Here is a paper. Please give your review comments after reading it. |
9,819 | "<ns0:abstract xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:p>In plants, phosphorus (P) uptake occurs via arbuscular mycorrhizal (AM) symbiosis and through plant roots. The phosphate concentration is known to affect colonization by AM fungi, and the effect depends on the plant species. Stevia rebaudiana plants are valuable sources of sweetener compounds called steviol glycosides (SGs), and the principal components of SGs are stevioside and rebaudioside A. However, a detailed analysis describing the effect of the phosphate concentration on the colonization of AM fungi in the roots and the relationship of these factors to the accumulation of SGs and photochemical performance has not been performed; such an analysis was the aim of this study. The results indicated that low phosphate concentrations (20 and 200 µM KH 2 PO 4 ) induced a high percentage of colonization by Rhizophagus irregularis in the roots of S. rebaudiana, while high phosphate concentrations (500 and 1000 µM KH 2 PO 4 ) reduced colonization. The morphology of the colonization structure is a typical Arum-type mycorrhiza, and a mycorrhiza-specific phosphate transporter was identified. Colonization with low phosphate concentrations improved plant growth, chlorophyll and carotenoid concentration, and photochemical performance. The transcription of the genes that encode kaurene oxidase and glucosyltransferase (UGT74G1) was upregulated in colonized plants at 200 µM KH 2 PO 4 , which was consistent with the observed patterns of stevioside accumulation. In contrast, at 200 µM KH 2 PO 4 , the transcription of UGT76G1 and the accumulation of rebaudioside A were higher in noncolonized plants than in colonized plants. The results indicate that a low phosphate concentration improves mycorrhizal colonization and modulates the stevioside and rebaudioside A concentration by regulating the transcription of the genes that encode kaurene oxidase and glucosyltransferases, which are involved in stevioside and rebaudioside A synthesis in S. rebaudiana.</ns0:p></ns0:div>
</ns0:abstract>
<ns0:body xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:p>In plants, phosphorus (P) uptake occurs via arbuscular mycorrhizal (AM) symbiosis and through plant roots. The phosphate concentration is known to affect colonization by AM fungi, and the effect depends on the plant species. Stevia rebaudiana plants are valuable sources of sweetener compounds called steviol glycosides (SGs), and the principal components of SGs are stevioside and rebaudioside A. However, a detailed analysis describing the effect of the phosphate concentration on the colonization of AM fungi in the roots and the relationship of these factors to the accumulation of SGs and photochemical performance has not been performed; such an analysis was the aim of this study. The results indicated that low phosphate concentrations (20 and 200 µM KH 2 PO 4 ) induced a high percentage of colonization by Rhizophagus irregularis in the roots of S. rebaudiana, while high phosphate concentrations (500 and 1000 µM KH 2 PO 4 ) reduced colonization. The morphology of the colonization structure is a typical Arum-type mycorrhiza, and a mycorrhiza-specific phosphate transporter was identified. Colonization with low phosphate concentrations improved plant growth, chlorophyll and carotenoid concentration, and photochemical performance. The transcription of the genes that encode kaurene oxidase and glucosyltransferase (UGT74G1) was upregulated in colonized plants at 200 µM KH 2 PO 4 , which was consistent with the observed patterns of stevioside accumulation. In contrast, at 200 µM KH 2 PO 4 , the transcription of UGT76G1 and the accumulation of rebaudioside A were higher in noncolonized plants than in colonized plants. The results indicate that a low phosphate concentration improves mycorrhizal colonization and modulates the stevioside and rebaudioside A concentration by regulating the transcription of the genes that encode kaurene oxidase and glucosyltransferases, which are involved in stevioside and</ns0:p></ns0:div>
<ns0:div><ns0:head>Introduction</ns0:head><ns0:p>Stevia rebaudiana Bertoni is a plant that belongs to the Asteraceae family and accumulates compounds in its leaves called steviol glycosides (SGs) <ns0:ref type='bibr' target='#b8'>(Brandle, Starratt & Gijzen, 1998)</ns0:ref>. Stevioside and rebaudioside A are the best-known SGs and are important compounds for human health because they are natural low-calorie sweeteners. The sweetening power of stevioside and rebaudioside A is 143 and 320 times higher than that of sucrose, respectively <ns0:ref type='bibr' target='#b26'>(Lemus-Mondaca et al., 2012;</ns0:ref><ns0:ref type='bibr' target='#b9'>Brandle & Telmer, 2007)</ns0:ref>. The biosynthetic pathway for SG synthesis begins in the chloroplasts with the synthesis of geranylgeranyl diphosphate (GGDP) generated from the MEP (methyl-erythritol-4-phosphate) pathway <ns0:ref type='bibr' target='#b46'>(Totté et al., 2000)</ns0:ref>. GGDP is transformed to kaurene by two cyclization steps carried out by a terpene cyclase <ns0:ref type='bibr' target='#b9'>(Brandle & Telmer, 2007)</ns0:ref>. In the endoplasmic reticulum, kaurene is oxidized by kaurene oxidase (KO) to kaurenoic acid; the oxidation of kaurenoic acid produces gibberellins, while the hydroxylation of kaurenoic acid produces steviol. The hydroxyl groups of steviol are glycosylated by the enzymes uridine diphosphate (UDP)-glycosyltransferases (UGTs), and the number of sugars attached by UGTs generates the various SGs <ns0:ref type='bibr' target='#b9'>(Brandle & Telmer, 2007)</ns0:ref>. The UGT74G1 enzyme is involved in the conversion of steviolbioside to stevioside, while the UGT76G1 enzyme converts stevioside into rebaudioside A <ns0:ref type='bibr' target='#b25'>(Kim et al., 2019)</ns0:ref>.</ns0:p><ns0:p>Arbuscular mycorrhizal (AM) symbiosis between the phylum Glomeromycota and plants is a mutualistic association that is useful in the culture of plants and has agricultural and medicinal importance. This symbiosis improves plant growth, photosynthesis and nutrient uptake and increases the production of phytochemicals <ns0:ref type='bibr' target='#b42'>(Smith & Read, 2008;</ns0:ref><ns0:ref type='bibr' target='#b44'>Spatafora et al., 2016;</ns0:ref><ns0:ref type='bibr' target='#b37'>Schoefs et al., 2015)</ns0:ref>.</ns0:p><ns0:p>In plants, there are two modes of P uptake: one mode is by the plant's own Pi transporters, and the other mode occurs via AM symbiosis with mycorrhiza-specific phosphate transporters and takes place in the arbuscules <ns0:ref type='bibr' target='#b20'>(Harrison, Dewbre &Liu, 2002)</ns0:ref>. Phosphate transporters are considered a key feature of this mycorrhizal symbiosis <ns0:ref type='bibr'>(Karandashov & Bucher, 2005)</ns0:ref>. Mycorrhiza-specific phosphate transporters are expressed in arbuscule-containing cortical root cells and are thus considered general markers for AM symbiosis in different model plants <ns0:ref type='bibr' target='#b20'>(Harrison, Dewbre & Liu, 2002;</ns0:ref><ns0:ref type='bibr' target='#b32'>Nagy et al., 2005;</ns0:ref><ns0:ref type='bibr' target='#b33'>Nouri et al., 2014)</ns0:ref>. Likewise, the phosphate concentration affects the colonization of roots by AM. Phosphate application at a high concentration may inhibit the formation of arbuscular mycorrhizae, and the sensitivity to phosphate and the grade of inhibition of arbuscule formation depend on the plant species <ns0:ref type='bibr' target='#b43'>(Smith, Smith & Jakobsen, 2004)</ns0:ref>; for example, these factors differ in Medicago truncatula <ns0:ref type='bibr' target='#b7'>(Bonneau et al., 2013)</ns0:ref> and Petunia hybrida <ns0:ref type='bibr' target='#b10'>(Breuillin et al., 2010;</ns0:ref><ns0:ref type='bibr' target='#b33'>Nouri et al., 2014)</ns0:ref>.</ns0:p><ns0:p>In S. rebaudiana, AM symbiosis enhances the production of stevioside and rebaudioside A and involves nutritional and nonnutritional mechanisms <ns0:ref type='bibr' target='#b28'>(Mandal et al., 2013)</ns0:ref>. AM symbiosis also upregulates the transcription of eleven SG biosynthesis genes as a consequence of the improved nutrition status and the increase in photosynthesis in the plant <ns0:ref type='bibr' target='#b29'>(Mandal et al., 2015)</ns0:ref>. This result suggests the roles of phosphorus nutrition and AM symbiosis in influencing SG content; fertilization with 25 mg P 2 O 5 kg −1 soil in association with AM symbiosis improved SG yield, P uptake and P nutrient use efficiency <ns0:ref type='bibr' target='#b45'>(Tavarini et al., 2018)</ns0:ref>. However, a detailed analysis and systematic description of the morphological type of AM symbiosis and the effect of the different phosphate concentrations on the establishment of AM symbiosis, the identification of mycorrhiza-specific phosphate transporters, the photosynthetic performance, and the relationship with the accumulation of SGs in S. rebaudiana plants have not been addressed. Therefore, in this study, we reported the effects of different phosphate concentrations on the establishment of AM symbiosis between Rhizophagus irregularis and S. rebaudiana, their relationship with photochemical performance and the accumulation of steviol glycosides (SGs) and the expression of two key genes, UGT74G1 and UGT76G1, which encode the (UDP)-glycosyltransferases involved in stevioside and rebaudioside A biosynthesis, respectively. The participation of a mycorrhiza-specific phosphate transporter as a key feature of this mycorrhizal symbiosis was demonstrated.</ns0:p></ns0:div>
<ns0:div><ns0:head>Materials & Methods</ns0:head><ns0:p>Inoculation with Rhizophagus irregularis and plant growth conditions R. irregularis was provided by Dr. Melina López-Meyer from 'Centro Interdisciplinario de Investigación para el Desarrollo Integral Regional, Unidad Sinaloa', Sinaloa state, Mexico. The inoculum was grown on Petri dishes with two compartments containing transformed carrot roots on minimum medium with 2% Gel-rite (Sigma-Aldrich) and incubated in the dark at 23 ± 2°C for six months according to the method reported by <ns0:ref type='bibr' target='#b3'>Bécard & Fortin (1988)</ns0:ref>.</ns0:p><ns0:p>Six-month-old cuttings of S. rebaudiana plants were cultured under greenhouse conditions. Briefly, seven-cm cuttings were disinfected with 70% ethanol for 1 min and 2% sodium hypochlorite for 1 min and washed three times with sterile distilled water for 2 min. To measure root development, the cuttings were transplanted under hydroponic conditions in glass tubes containing Fahraeus medium and cultured in a controlled environment chamber at 25°C with a 16 h light: 8 h dark photoperiod regimen. After ten days of culture, the rooted cuttings were transferred to plastic cones of 125 mm high and 32 mm diameter (M49, Polietilenos del Sur S.A. C.V.), one plant per cone. The substrate used was a 1:1 (v:v) mixture of vermiculite and sand. The substrate was autoclaved twice for 1 h at 121°C and 15 psi. S. rebaudiana plants were inoculated with 150 spores of R. irregularis (M+) that were homogeneously distributed in the substrate; the controls were noncolonized (M-) plants. The plants were watered twice per week with 20 mL of half-strength Hoagland nutrient solution <ns0:ref type='bibr' target='#b21'>(Hoagland & Arnon, 1950)</ns0:ref> with KH 2 PO 4 at the final phosphate concentrations that were evaluated: 20, 200, 500, and 1000 µM. The pH of the nutrient solutions was adjusted to 6.1.</ns0:p><ns0:p>The plants were maintained in a growth chamber at 25°C with a 16 h light:8 h dark photoperiod regimen for 30 days postinoculation (dpi). The experiment was performed utilizing a complete factorial design, and six plants per phosphate concentration and colonization status with R. irregularis were evaluated. The controls were noncolonized plants treated with the different phosphate concentrations. Two independent experiments were performed. Similar trends were obtained in the both experiments, and the results of only one of them is shown.</ns0:p></ns0:div>
<ns0:div><ns0:head>Staining and quantification of mycorrhizal colonization</ns0:head><ns0:p>S. rebaudiana root segments were stained with 0.05% trypan blue in lactoglycerol <ns0:ref type='bibr' target='#b34'>(Phillips & Hayman, 1970)</ns0:ref> and observed by light microscopy (BOECO Germany, BM-180) at 10-40X magnification. Total mycorrhizal colonization by R. irregularis was calculated according to the line-intersection method <ns0:ref type='bibr' target='#b16'>(Giovannetti & Mosse, 1980)</ns0:ref>. For each plant, 90 root segments were assessed, and six plants were evaluated. The arbuscular percentage was calculated with MycoCalc software (http://www.dijon.inra.fr/mychintec/Mycocalc-prg/download.html). To identify the morphological type of the AM symbiosis in S. rebaudiana, mycorrhizal roots were stained with WGA-Alexa Fluor 488 to visualize the arbuscules, and the plant tissue was labeled with propidium iodide according to the methodology reported by <ns0:ref type='bibr' target='#b54'>Xie et al. (2016)</ns0:ref> using confocal laser scanning microscopy (LSM 800, Carl Zeiss).</ns0:p></ns0:div>
<ns0:div><ns0:head>Determination of plant growth</ns0:head><ns0:p>The plants treated with the different phosphate concentrations and colonized (M+) or noncolonized (M-) with R. irregularis were collected at 30 days postinoculation (dpi). Shoots and roots of each plant were separated, total leaves number and the fresh weight of each organ was recorded. For minimize the dependency of the analysis of expression of genes and SG quantification with the leaf position on the plant. The leaves were collected in two equivalent groups, considering their opposite arrangement of the leaves and their position along the stem. Then, leaves positioned in all the nodes of the plant composed each group. The leaf area was determined by image analysis from leaves of the second node, close to the apical meristem. A stereomicroscope (Olympus SZX7, Germany) was used to obtain the micrographs. Image analysis was performed using ImageJ editing software (Version, 1.8.0.112). One half of the leaves were frozen in liquid nitrogen for molecular analysis, and the other half for SG extraction. Root were also collected and separated longitudinally in two sections, one of them was used for determination of mycorrhizal colonization, and the other for mycorrhiza-specific phosphate transporter identification. Six plants per phosphate treatment and R. irregularis colonization status were evaluated.</ns0:p></ns0:div>
<ns0:div><ns0:head>Analysis of phosphorus and magnesium concentration</ns0:head><ns0:p>The leaves of S. rebaudiana plants were prepared according to <ns0:ref type='bibr' target='#b17'>Guerrero-Molina et al. (2014)</ns0:ref>. The leaves were analyzed by a high-resolution scanning electron microscope (SEM) equipped with a field emission cathode and coupled to an energy-dispersive X-ray (EDX, Carl Zeiss, Oberkochen, Germany). The electron energy used was 20 keV. The mapping of P and Mg was determined by EDX to record the two-dimensional elemental composition of the leaf sample surface. For quantitative analyses, EDX spectrograms were recorded and analyzed using QUANTAX ESPRIT, Version 1.9 (BRUKER, Germany). Since the results of the content of each element are given as the percentage of such element with respect to all the components determined in the sample, the concentration of P and Mg in the samples was expressed as the percentage of each element in the leaves.</ns0:p></ns0:div>
<ns0:div><ns0:head>Determination of chlorophyll and carotenoid concentration</ns0:head><ns0:p>The determination of chlorophyll and carotenoids concentration was made from three fresh leaves (approximately 100-150 mg) of each of the six plants of the M-and M + conditions. The leaves were collected from the upper, middle and lower part of each one of the analyzed plants and were ground in a mortar with 80% acetone. The extracts were centrifuged at 3000 g for 15 min; the supernatants were separated, and the absorbance at 646.8, 663.2 and 470 nm was measured in a UV/Vis spectrophotometer (UV-1800, Shimadzu, Japan). The concentration of chlorophylls and carotenoids were calculated following the equations described by <ns0:ref type='bibr' target='#b24'>Khan & Mitchell (1987)</ns0:ref>.</ns0:p></ns0:div>
<ns0:div><ns0:head>Measurement of chlorophyll fluorescence</ns0:head><ns0:p>The chlorophyll fluorescence was measured in the second fully expanded leaf of each plant using a chlorophyll fluorometer (model OS30P, Opti-Sciences Inc., USA). The evaluation was performed at room temperature according to the instructions for the chlorophyll fluorometer. Before the evaluation, the plants were placed in the dark for 30 min, and chlorophyll fluorescence was evaluated after applying a 1 s saturating pulse of actinic light (3500 µmol m -2 s -1 ). The primary fluorescence (Fo), maximal fluorescence (Fm), maximum quantum efficiency of PSII photochemistry (Fv/Fm), and potential photochemical efficiency (Fv/Fo) were calculated. Fv was calculated as Fv = Fm−Fo, and Fv/Fo was calculated as Fv/Fo = Fm/Fo−1 <ns0:ref type='bibr' target='#b38'>(Schreiber, Bilger & Neubauer, 1994)</ns0:ref>.</ns0:p></ns0:div>
<ns0:div><ns0:head>Expression analysis by qRT-PCR</ns0:head><ns0:p>The transcript accumulation levels of the genes for kaurene oxidase and (UDP)glycosyltransferases were evaluated in colonized and noncolonized plants and in plants that were treated with 200 and 1000 µM KH 2 PO 4 . These phosphate concentrations were selected because they were found to be the optimal and suboptimal conditions for inducing root colonization.</ns0:p><ns0:p>Frozen leaves, from one of the groups described in the determination of plant growth section, were ground to a fine powder in liquid nitrogen. Total RNA was isolated from leaves using TRIzol reagent (Invitrogen, Carlsbad, CA) following the manufacturer's protocol. First-strand cDNA synthesis was performed as previously reported by <ns0:ref type='bibr'>Cervantes-Gámez et al. (2016)</ns0:ref>.</ns0:p><ns0:p>The primers used were those designed and reported by <ns0:ref type='bibr' target='#b29'>Mandal et al. (2015)</ns0:ref> for S. rebaudiana plants. The primers correspond to the kaurene oxidase gene (SrKOF 5´-TCTTCACAGTCTCGGTGGTG-3´, and SrKOR 5´-GGTGGTGTCGGTTTATCCTG-3´), the glucosyl transferase UGT74G1 gene (SrUGT74G1F 5´-GGTAGCCTGGTGAAACATGG-3´, and SrUGT74G1R 5´-CTGGGAGCTTTCCCTCTTCT -3´) and the glucosyl transferase UGT76G1 gene (SrUGT76G1F 5´-GACGCGAACTGGAACTGTTG-3´, and SrUGT76G1R 5´-AGCCGTCGGAGGTTAAGACT -3´). qRT-PCR was performed using SYBR ® Green (QIAGEN, USA) and quantified on a Rotor-Gene Q (QIAGEN, USA) real-time PCR thermal cycler. qRT-PCR was programmed for 35 cycles, with denaturing at 95°C for 15 s, annealing at 55°C for 30 s, and extension at 72°C for 30 s. Primer specificity was verified by regular PCR and melting curve analysis. The primers for the S. rebaudiana glyceraldehyde-3-phosphate dehydrogenase (GAPDH) gene (SrGAPDHF 5´-TCAGGGTGGTGCCAAGAAGG-3´, and SrGAPDHR 5´-TTACCTTGGCAAGGGGAGCA -3´) were used as internal controls for normalization, and the quantitative results were evaluated by the 2 −ΔΔCT method described by <ns0:ref type='bibr' target='#b27'>Livak & Schmittgen (2001)</ns0:ref>. To interpret the results, genes with fold change values ≥1.5 were considered 'upregulated', whereas genes with fold change values ≤−0.7 were considered 'downregulated'. Six plants per phosphate treatment and R. irregularis colonization status were evaluated.</ns0:p></ns0:div>
<ns0:div><ns0:head>Cloning the mycorrhiza-associated phosphate transporter gene from S. rebaudiana</ns0:head><ns0:p>A pair of degenerate primers (SrPTF 5´-ATGGGDTTTTTYACYGATGC-3´ and SrPTR 5´-GGNCCAAARTTSGCRAAGAA-3´) were designed by aligning highly conserved regions of AM-specific phosphate transporters from M. truncatula (accession number: AY116210), A. sinicus (accession number: JQ956418), S. lycopersicum (accession number: AF022874), S. tuberosum (accession number: AY793559) and P. hybrida (accession number: EU532763). The PCR product was purified using the QIAquick PCR Purification Kit (QIAGEN, USA) and ligated into the pGEM®-T Easy vector (Promega, USA) in accordance with the manufacturer's protocols. The presence of the correct insert (1350 bp) within the pGEM®-T Easy vector was confirmed by PCR using the universal primers T7 and SP6, and the insert was then sequenced.</ns0:p><ns0:p>Collected roots from one of the groups described in the determination of plant growth section, were ground to a fine powder in liquid nitrogen. Total RNA was obtained from the roots of six colonized (M+) and six noncolonized (M-) plants fertilized with 200 µM KH 2 PO 4 . RNA was isolated using TRIzol reagent (Invitrogen, Carlsbad, CA) following the manufacturer´s protocol. First-strand cDNA synthesis was performed as previously reported by <ns0:ref type='bibr'>Cervantes-Gámez et al. (2016)</ns0:ref>. cDNA synthesis was confirmed by PCR using primers for the S. rebaudiana glyceraldehyde-3-phosphate dehydrogenase (GAPDH) gene (SrGAPDHF 5´-ATGGGDTTTTTYACYGATGC-3´ and SrGAPDHR 5´-GGNCCAAARTTSGCRAAGAA-3´).</ns0:p><ns0:p>For the expression analysis of SrPT in the M-and M+ plants, PCRs were run in a total reaction volume of 10 µL, comprising 0.2 µL of GoTaq® Flexi DNA Polymerase (Promega, USA), 200 nM of each primer, and 50 ng of cDNA. The PCR thermocycler was programmed for 35 cycles, with denaturing at 95°C for 15 s, annealing at 52°C for 1 min, and extension at 72°C for 30 s. The SrGAPDH gene was used as the reference gene.</ns0:p></ns0:div>
<ns0:div><ns0:head>Homology modeling analysis of the AM-specific phosphate transporter from S. rebaudiana</ns0:head><ns0:p>BLAST analysis of the SrPT gene sequence was performed to determine homology predictions using the tools on the NCBI website (https://www.ncbi.nlm.nih.gov). To determine the conserved region of SrPT, multiple sequence alignments of AM-specific phosphate transporter proteins were performed using MULTALIN software. The homology model of SrPT transmembrane domains (TDs) was constructed according to <ns0:ref type='bibr' target='#b55'>Yadav et al. (2010)</ns0:ref>, and the Mtpt4 protein structure from M. truncatula was used as the template for homology modeling for the S. rebaudiana AM-specific phosphate transporters.</ns0:p></ns0:div>
<ns0:div><ns0:head>Steviol glycosides extraction and quantification of concentration</ns0:head><ns0:p>The leaves of colonized (M+) and noncolonized (M-) plants that were treated with 200 and 1000 µM KH 2 PO 4 were used to evaluate the SG concentration. Leaves from one of the groups described in the determination of plant growth section, were dried in in an oven (Thermo Scientific, USA) at 65 ° C for 48 h. The dry leaf tissue (0.1g) was extracted with 1 mL of methanol (J.T. Backer, USA), following the methodology described by <ns0:ref type='bibr' target='#b52'>Woelwer-Rieck et al. (2010)</ns0:ref>. The mixture was stirred for 3 min, allowed to stand for 24 h without stirring, and then centrifuged at 10,000 rpm at 4 °C for 10 min. The supernatant was recovered, placed in Eppendorf tubes, and stored at -4 °C until the analysis by HPTLC (CAMAG, Switzerland). The quantification of SGs was based on the methodology reported by <ns0:ref type='bibr' target='#b4'>Bladt and Zgainski (1996)</ns0:ref> and <ns0:ref type='bibr' target='#b30'>Morlock et al. (2014)</ns0:ref>, and discribed recently by <ns0:ref type='bibr' target='#b49'>Villamarin-Gallegos et al. (2020)</ns0:ref>. Stevioside and rebaudioside A concentration were expressed in mg g DW -1 . Six plants per phosphate treatment and R. irregularis colonization status were evaluated.</ns0:p></ns0:div>
<ns0:div><ns0:head>Statistical analysis</ns0:head><ns0:p>The differences between the total colonization and arbuscular percentages were examined by one-way analysis of variance (ANOVA), and Tukey's post hoc test (P< 0.05) was performed to test the significance of differences between means. Data regarding the effect of the interaction between the KH 2 PO 4 concentration and mycorrhizal colonization on plant growth, P and Mg concentration, pigment concentrations, chlorophyll fluorescence and SG concentration were subjected to factorial two-way analysis of variance (ANOVA). Tukey´s post hoc test was used to analyze the differences. The paired Student's t-test was used to evaluate the significance of differences in the gene expression of kaurene oxidase and (UDP)-glycosyltransferases. All data, used for ANOVA and factorial analysis were checked to normal distributions (Shapiro-Wilk´s test) before statistical analysis. All statistical analyses were performed using the statistical software IBM SPSS for Windows, Version 24.0 (Armonk, NY, IBM Corp.).</ns0:p></ns0:div>
<ns0:div><ns0:head>Results</ns0:head><ns0:p>PeerJ reviewing PDF | (2020:07:50625:1:1:NEW 26 Aug 2020)</ns0:p><ns0:p>Manuscript to be reviewed</ns0:p></ns0:div>
<ns0:div><ns0:head>Phosphate concentration affects the mycorrhizal colonization of S. rebaudiana</ns0:head><ns0:p>The highest percentages of colonization were obtained at 20 and 200 µM KH 2 PO 4 (73.3 and 67.0% colonization, respectively). In contrast, the percentage of total colonization decreased significantly at 500 and 1000 µM KH 2 PO 4 , with 43.3 and 18.4% colonization, respectively (Fig. <ns0:ref type='figure' target='#fig_1'>1</ns0:ref>). The percentage of arbuscules was significantly reduced at 500 and 1000 µM KH 2 PO 4 , with 1.48 and 0.4%, respectively (Fig. <ns0:ref type='figure' target='#fig_1'>1</ns0:ref>).</ns0:p><ns0:p>In the roots of plants treated with 20 µM KH 2 PO 4 , a high number of arbuscules formed (Fig. <ns0:ref type='figure' target='#fig_13'>2A</ns0:ref>, see label *); several intraradical hyphae grew through the cortical cells (Fig. <ns0:ref type='figure' target='#fig_13'>2A</ns0:ref>, see label ih), although vesicle structures were scarce. Similar structures were observed in colonized plants (M+) with 200 µM KH 2 PO 4 ; however, under these experimental conditions, the formation of arbuscules and vesicle structures was more evident (Fig. <ns0:ref type='figure' target='#fig_13'>2B</ns0:ref>, see labels * and v), indicating that 200 µM KH 2 PO 4 created better conditions than 20 µM for the formation of arbuscules. Qualitative differences in mycorrhizal structures were observed when plants were treated with the highest KH 2 PO 4 concentrations (500 and 1000 µM); the formation of arbuscules, for instance, was significantly reduced (Fig. <ns0:ref type='figure' target='#fig_13'>2C</ns0:ref> and 2D, see label *), and shortening of intraradical structures was observed (Fig. <ns0:ref type='figure' target='#fig_13'>2C and D</ns0:ref>, see label ih) in comparison to the structures observed at low KH 2 PO 4 concentrations (Fig. <ns0:ref type='figure' target='#fig_13'>2A and 2B</ns0:ref>).</ns0:p><ns0:p>The activation of specific genes, such as the phosphate transporter specifically induced by the mycorrhizal association, is an important marker for evaluation successful colonization establishment. To our knowledge, there is no information on this transporter type in S. rebaudiana. For this reason, we identified a putative mycorrhiza-specific phosphate transporter in S. rebaudiana (SrPT) by a simple PCR strategy based on degenerate oligonucleotides to clone the corresponding phosphate transporter. A 1175 bp-long genomic fragment containing an open reading frame that encodes a 391-amino acid polypeptide with a molecular mass of 43.39 kDa was cloned (accession number: MN273502). This putative SrPT polypeptide contains 9 of the 12 transmembrane domains from the canonic phosphate transporter (Fig. <ns0:ref type='figure' target='#fig_8'>S1</ns0:ref>). The bioinformatic analysis suggests that SrPT is 72.89% conserved in comparison to the sequences reported for MtPT4 TMDs in Medicago truncatula. Notably, SrPT transcript accumulation increased in the roots of colonized (M+) plants compared with that in the roots of noncolonized (M-) plants. This result suggests that this gene is positively regulated by AM symbiosis in S. rebaudiana, in a similar manner to the genes for other mycorrhiza-specific phosphate transporters reported in other model plants (Fig. <ns0:ref type='figure' target='#fig_9'>S2</ns0:ref>).</ns0:p><ns0:p>Confocal microscopic analysis of the colonized roots and staining with the conjugate WGA-Alexa Fluor® 488 and propidium iodide as fluorescent markers permitted us to differentiate the hyphae and the plant cells, respectively. Intracellular hyphae and the formation of arbuscules from intracellular hyphae growing in the inner cortex were observed (Fig. <ns0:ref type='figure' target='#fig_13'>2E-G</ns0:ref>, see labels ih and *). With this approach, we were able to depict and classify this structure as a typical Arumtype mycorrhiza.</ns0:p></ns0:div>
<ns0:div><ns0:head>Mycorrhizal colonization improves plant growth in S. rebaudiana</ns0:head><ns0:p>In the M-plants, the leaf fresh weight was not different than that in plants treated with 20, 200 and 500 µM KH 2 PO 4 ; the leaf fresh weight only increased significantly at 1000 µM KH 2 PO 4 (Fig. <ns0:ref type='figure' target='#fig_3'>3A</ns0:ref>). In the M+ plants, the leaf fresh weight increased by a factor of 1.74 with 200 µM KH 2 PO 4 in comparison to that in the M-plants (control) at the same phosphate concentration. However, no difference was found between M+ and M-plants at 1000 µM KH 2 PO 4 . The leaves of M+ plants with 200 µM KH 2 PO 4 had a similar fresh weight to those of M+ and M-plants treated with 1000 µM KH 2 PO 4 (Fig. <ns0:ref type='figure' target='#fig_3'>3A</ns0:ref>). In plants treated with 200 µM KH 2 PO 4 , the fresh weight of roots was higher in the roots of M+ plants than in the roots of M-plants. At 1000 µM KH 2 PO 4 , the fresh weight of roots was higher in the M+ plants than in the M-plants (Fig. <ns0:ref type='figure' target='#fig_3'>3B</ns0:ref>). Leaves number and foliar area were determined, and only at 200 µM KH 2 PO 4 , M-plant showed lower number of leaves than M+, as well as foliar area (Fig. <ns0:ref type='figure' target='#fig_10'>S3</ns0:ref>), which is consistent with the pattern of fresh weight of leaves. Foliar area, on the other hand, was only significantly lower in M+ plants at 500 µM KH 2 PO 4 (Fig. <ns0:ref type='figure' target='#fig_10'>S3</ns0:ref>).</ns0:p></ns0:div>
<ns0:div><ns0:head>The phosphorus and magnesium concentration increases in the leaves of colonized plants with a low phosphate concentration.</ns0:head><ns0:p>The P concentration was four times higher in the leaves of M+ plants treated with 200 µM KH 2 PO 4 than that in M-plants. At 500 and 1000 µM KH 2 PO 4 , the P concentration was similar in the leaves of M+ plants and M-plants (Fig. <ns0:ref type='figure' target='#fig_16'>4A</ns0:ref>). In the leaves of M-plants, the Mg content increased significantly at 500 and 1000 µM KH 2 PO 4 , while in the leaves of M+ plants, the Mg concentration increased at 20, 200 and 1000 µM KH 2 PO 4 , but the Mg concentration was lower at 500 µM KH 2 PO 4 (Fig. <ns0:ref type='figure' target='#fig_16'>4B</ns0:ref>). These results suggest that AM symbiosis can stimulate P and Mg accumulation at low KH 2 PO 4 concentrations.</ns0:p></ns0:div>
<ns0:div><ns0:head>Chlorophyll fluorescence and the content of photosynthetic pigments improve in colonized plants at a low phosphate concentration</ns0:head><ns0:p>Chlorophyll fluorescence was used as an indicator of photosynthetic performance in the S. rebaudiana plants. In M-plants and at all KH 2 PO 4 concentrations, the Fv/Fm ratio values were less than 0.8. In M+ plants at 20 and 200 µM KH 2 PO 4 , the Fv/Fm ratio values were greater than 0.8, and at 500 and 1000 µM KH 2 PO 4 , the Fv/Fm ratio values diminished to less than 0.8 (Fig. <ns0:ref type='figure' target='#fig_17'>5A</ns0:ref>). The Fv/Fo ratio is indicative of the photochemical efficiency of photosynthesis. In the Mplants at all phosphate concentrations, the values of the Fv/Fo ratio were less than 4.0. In M+ plants at 20 and 200 µM KH 2 PO 4 , the value of the Fv/Fo ratio was greater than 4.0; at 500 and 1000 µM KH 2 PO 4 , this ratio was less than 4.0 (Fig. <ns0:ref type='figure' target='#fig_17'>5B.</ns0:ref>). The values of Fo, Fm and Fv are presented in Fig. <ns0:ref type='figure' target='#fig_11'>S4</ns0:ref>.</ns0:p><ns0:p>The total concentration of chlorophylls and carotenoids did not change in the M-plants at any phosphate concentration. However, in M+ plants at 20 and 200 µM KH 2 PO 4 , the concentration of chlorophylls and carotenoids was higher as compared to the M-plants at the same phosphate concentrations; at 500 and 1000 µM KH 2 PO 4 , the concentration of the two pigments were similar in M-and M+ plants (Figs. <ns0:ref type='figure' target='#fig_17'>5C and 5D</ns0:ref>).</ns0:p></ns0:div>
<ns0:div><ns0:head>Differential expression of the genes for kaurene oxidase and glucosyltransferases in colonized plants with added phosphate</ns0:head><ns0:p>The transcription of the kaurene oxidase (KO) gene was increased 7.5 times in M+ plants at 200 µM KH 2 PO 4 compared with that in M+ plants without added KH 2 PO 4 . The transcription level of the KO gene did not change in M+ plants at 1000 µM KH 2 PO 4 in comparison to M-plants, since the relative expression (2 -∆∆Ct ) was close to 1 (Fig. <ns0:ref type='figure' target='#fig_18'>6A</ns0:ref>). The UGT74G1 gene encoding the protein involved in stevioside synthesis was upregulated in M+ plants at 200 µM KH 2 PO 4 ; the level of relative expression was over 1.5 (Fig. <ns0:ref type='figure' target='#fig_18'>6B</ns0:ref>). The UGT76G1 gene encoding the protein involved in rebaudioside A synthesis was downregulated in M+ plants at the same KH 2 PO 4 concentration, and its relative expression was less than 0.7 (Fig. <ns0:ref type='figure' target='#fig_18'>6C</ns0:ref>). However, in M+ plants at 1000 µM KH 2 PO 4 , the expression of the UGT74G1 gene was downregulated, and the expression of the UGT76G1 gene was unchanged (Fig. <ns0:ref type='figure' target='#fig_18'>6B and 6C</ns0:ref>).</ns0:p></ns0:div>
<ns0:div><ns0:head>SGs differentially accumulate in the leaves of colonized plants with added phosphate</ns0:head><ns0:p>In M+ plants at 200 µM KH 2 PO 4 , the stevioside concentration was 2.8 times higher and the rebaudioside A concentration was 1.61 times lower than those of M-plants (Fig. <ns0:ref type='figure' target='#fig_19'>7A and 7B</ns0:ref>). This metabolite accumulation is consistent with the transcript levels of the corresponding glucosyl transferases (Fig <ns0:ref type='figure' target='#fig_18'>6B and C</ns0:ref>). At 1000 µM KH 2 PO 4 , the accumulation of the two metabolites in M+ and M-plants was not affected (Fig <ns0:ref type='figure' target='#fig_19'>7A and B</ns0:ref>).</ns0:p></ns0:div>
<ns0:div><ns0:head>Discussion</ns0:head><ns0:p>AM symbiosis enhances P uptake in many plants and plays an important role in agricultural and natural environments <ns0:ref type='bibr' target='#b41'>(Smith et al., 2011)</ns0:ref>. Phosphate availability may change the initial signaling for the establishment, maintenance, and functioning of AM symbiosis <ns0:ref type='bibr' target='#b36'>(Schmitz & Harrison, 2014)</ns0:ref>. In this study, the low phosphate concentrations (20 and 200 µM KH 2 PO 4 ) stimulated a high percentage of total mycorrhizal colonization by R. irregularis in S. rebaudiana plants, while the high KH 2 PO 4 concentrations (500 and 1000 µM) decreased the colonization efficiency of R. irregularis in S. rebaudiana plants by approximately 30 and 70%, respectively. AM symbiosis is inhibited by a high concentration of KH 2 PO 4 <ns0:ref type='bibr' target='#b43'>(Smith, Smith & Jakobsen, 2004)</ns0:ref>, and the sensitivity and inhibition percentage depend on the plant species and AM fungus. In Medicago truncatula, fertilization with 1.3 mM phosphate reduced AM symbiosis by 80% compared to that in plants fertilized with 0.13 mM phosphate <ns0:ref type='bibr' target='#b7'>(Bonneau et al., 2013)</ns0:ref>. In Petunia hybrida, phosphate at 100 µM induces high AM symbiosis, while phosphate at 3 mM and higher concentrations completely suppressed this symbiosis <ns0:ref type='bibr' target='#b33'>(Nouri et al., 2014)</ns0:ref>. Therefore, it was important in our study to define this effect of phosphate on the colonization of R. irregularis in S. rebaudiana plants.</ns0:p><ns0:p>The formation of arbuscules in S. rebaudiana roots was inhibited to a higher extent than the total colonization, indicating that arbuscule formation is more sensitive to high phosphate concentrations than other fungal structures, such as hyphae and vesicles. In addition, shortening of intraradical structures was observed in comparison to the arbuscules of roots at low phosphate concentrations. Similarly, changes in arbuscule structures were observed in P. hybrida plants; fertilization with high phosphate concentrations significantly reduced the development of arbuscules and resulted in malformed arbuscules with fewer branches <ns0:ref type='bibr' target='#b10'>(Breuillin et al., 2010)</ns0:ref>.</ns0:p><ns0:p>Previous studies in S. rebaudiana have reported colonization by R. irregularis, but the AM colonization morphology type has not been documented <ns0:ref type='bibr' target='#b47'>(Vafadar, Amooaghaie & Otroshy, 2014;</ns0:ref><ns0:ref type='bibr' target='#b29'>Mandal et al., 2015;</ns0:ref><ns0:ref type='bibr' target='#b45'>Tavarini et al., 2018)</ns0:ref>. In this study, confocal microscopic analysis with specific fluorescent dyes indicated that the AM colonization is classified as an Arum-type morphology. This AM morphotype is highly sensitive to environmental factors, including soil nutrients <ns0:ref type='bibr' target='#b15'>(Dickson et al., 2003)</ns0:ref>. It is also suggested that Arum-type colonization is more efficient than other morphotypes in the acquisition and transference of phosphate from the soil to the plant, resulting in better plant growth <ns0:ref type='bibr' target='#b48'>(Van Aarle et al., 2005)</ns0:ref> The beneficial effects of AM symbiosis on growth promotion and yield have been reported in plants from the Asteraceae family <ns0:ref type='bibr'>(Rapparini, Llusia & Peñuelas, 2008;</ns0:ref><ns0:ref type='bibr' target='#b0'>Aroca et al., 2013)</ns0:ref>. In this study, AM symbiosis with 200 µM KH 2 PO 4 increased colonization and arbuscule formation but also promoted the leaf and root growth in S. rebaudiana plants; the growth of these plants was comparable to the growth of plants fertilized at higher KH 2 PO 4 concentrations (1000 µM). These results suggest the activation of a high-affinity mycorrhiza-specific phosphate transporter during AM symbiosis, as reported for other plant species. In fact, we found the accumulation of transcripts of a putative phosphate transport gene in response to mycorrhizal colonization. This is the first time that a putative phosphate transporter gene has been identified in S. rebaudiana plants, and we anticipate that this gene encoding the specific phosphate transporter may contribute to growth promotion in leaves and roots. These results indicate that colonized plants are able to compensate for the deficiency in P nutrition using the mycorrhizal P uptake pathway. Similar effects have been reported in M. truncatula and P. hybrida plants at different phosphate concentrations <ns0:ref type='bibr' target='#b2'>(Balzergue et al., 2013;</ns0:ref><ns0:ref type='bibr' target='#b33'>Nouri et al., 2014)</ns0:ref>. Shoots and roots were responsive to mycorrhizal colonization by increasing growth at 200 µM KH 2 PO 4 , whereas at 1000 µM KH 2 PO 4 , roots growth was stimulated only in M+ compared to noncolonized controls (M-). This indicates that the symbiosis differentially affects shoot and root growth, and this depends on the phosphate fertilization regime. In plants at 500 µM KH 2 PO 4 , such differences between fresh weights of shoots and roots in M+ in comparison to M-was not observed. This in contrast to 200 µM and 1000 µM KH 2 PO 4 conditions, in which mycorrhiza-specific and direct plant phosphate uptake pathways dominate the phosphate uptake, respectively. It is possible that 500 µM KH 2 PO 4, the direct and the mycorrhizal-specific phosphate uptake pathways, may be interacting in such a way that no increased in growth is manifested. Although this hypothesis needs to be further studied.</ns0:p><ns0:p>Chlorophyll fluorescence is used as an indicator of light assimilation (electron transport) in the reaction centers of chlorophyll and as an indirect measurement of photosynthetic performance in plants subjected to different conditions of stress <ns0:ref type='bibr' target='#b38'>(Schreiber, Bilger & Neubauer, 1994;</ns0:ref><ns0:ref type='bibr' target='#b19'>Harbinson, 2013)</ns0:ref>. Stress conditions may disrupt components of the photosynthetic apparatus and affect photosynthetic performance. In healthy plants, the values of the Fv/Fm ratio are between 0.79 and 0.82; however, under stress conditions, the photosynthetic performance is affected, and these values decrease to below 0.79. Likewise, values of the Fv/Fo ratio between 4-5 correspond to normal values for healthy plants, and values less than 4 are indicative of a loss of photosynthetic efficiency that may result from a stress condition <ns0:ref type='bibr' target='#b1'>(Baker, 2008)</ns0:ref>. In the noncolonized S. rebaudiana plants at low KH 2 PO 4 concentrations, these values were less than 0.79 and 4.0, which indicates that the plants are subjected to stress conditions. However, these values in colonized plants at the same KH 2 PO 4 concentration were significantly higher than those in noncolonized plants. These results suggest that AM symbiosis may compensate for the nutritional stress induced by low KH 2 PO 4 concentrations by restoring the electron transfer through PSII, regulating the functionality of the reaction center, and improving photosynthetic performance in S. rebaudiana plants. The results are consistent with those of other studies performed in plants under stress conditions such as high salinity, high temperatures, and water stress <ns0:ref type='bibr' target='#b40'>(Sheng et al., 2008;</ns0:ref><ns0:ref type='bibr' target='#b57'>Zhu et al., 2011;</ns0:ref><ns0:ref type='bibr' target='#b22'>Hu et al., 2017)</ns0:ref>.</ns0:p><ns0:p>In colonized S. rebaudiana plants at low phosphate concentrations, the concentration of photosynthetic pigments was higher than that in noncolonized plants at the same phosphate concentration. Similarly, other authors have shown that mycorrhizal colonization stimulates the accumulation of these compounds in plants <ns0:ref type='bibr' target='#b14'>(Colla et al., 2008;</ns0:ref><ns0:ref type='bibr' target='#b31'>Nafady & Elgharably, 2018)</ns0:ref>. These studies support the idea that photosynthesis is improved in colonized plants, which ensures carbon fixation and provides a carbon source to the fungal symbiont under low phosphate conditions <ns0:ref type='bibr' target='#b56'>(Zai et al., 2012)</ns0:ref>. Additionally, increased carotenoid concentration in plants is associated with light harvesting, photoprotection, and antioxidant processes, which may contribute to improving plant growth <ns0:ref type='bibr' target='#b51'>(Walter, 2013)</ns0:ref>.</ns0:p><ns0:p>Mg is bound to the central atom in the porphyrin ring of chlorophyll a and b, and 25-60% of the total Mg in plants exists as chlorophyll-bound Mg <ns0:ref type='bibr' target='#b13'>(Chen et al., 2018)</ns0:ref>. A significant increase in the Mg percentage was found in the leaves of S. rebaudiana plants colonized under the low phosphate concentration. This result is consistent with the increase in chlorophyll concentration and has been reported in other plants <ns0:ref type='bibr' target='#b14'>(Colla et al., 2008;</ns0:ref><ns0:ref type='bibr' target='#b47'>Vafadar, Amooaghaie & Otroshy, 2014)</ns0:ref>. Thus, the increase in the concentration of chlorophylls, Mg, and carotenoids improves plant growth and metabolite accumulation, confirming the positive effect of AM symbiosis.</ns0:p><ns0:p>Kaurene oxidase (KO) plays an important role in SG biosynthesis and represents an important branch point that specifically directs the flow of metabolites towards the biosynthesis of steviol (the central backbone of SGs). In fact, SGs are synthesized from steviol glycosylation, where the conjugation of glucose to steviol is carried out by UDP-glycosyltransferases (UGTs); for example, UGT74G1 is known to convert steviolbioside to stevioside, and UGT76G1 adds the final glucose required to produce rebaudioside A <ns0:ref type='bibr' target='#b9'>(Brandle and Telmer, 2007)</ns0:ref>.</ns0:p><ns0:p>The KO and UGT74G1 gene transcription levels were upregulated in colonized plants compared to those in noncolonized plants at 200 µM KH 2 PO 4 , which was consistent with the observed stevioside accumulation trends. In contrast, in noncolonized plants at 200 µM KH 2 PO 4 , the expression of the UGT76G1 gene and the corresponding accumulation of rebaudioside A were higher than those in colonized plants. These results indicate that plant colonization with AM stimulates stevioside accumulation, while in noncolonized plants, the accumulation of rebaudioside A is stimulated. In addition, the high phosphate concentration (KH 2 PO 4 1000 µM) in colonized plants downregulated the expression of the UGT74G1 gene and did not change the expression of the KO or UGT76G1 genes at the same phosphate concentration. These results support the idea that SG synthesis is sensitive to the phosphate concentration and mycorrhizal interactions and suggest that AM symbiosis causes an increase in stevioside accumulation by modulating the expression of the KO and UGT74G1 genes, which convert steviolbioside to stevioside. The downregulation of the UGT76G1 gene, which encodes a protein involved in transforming stevioside to rebaudioside A, may also contribute to the accumulation of stevioside. In contrast, the accumulation of rebaudioside A, which is the SG with the highest sweetening power, would be favored in noncolonized plants. This information is relevant from a biotechnological perspective. The results demonstrate the effect of the phosphate concentration on the mycorrhizal interaction between S. rebaudiana and R. irregularis as well as the effect on SG concentration after the modulation of the expression of key biosynthetic genes. Manuscript to be reviewed colonization structure was a typical Arum-type mycorrhiza, and a mycorrhiza-specific phosphate transporter was identified. Colonization at low phosphate concentrations improved plant growth, the chlorophyll and carotenoid contents, and photochemical performance. The low phosphate concentration improved mycorrhizal colonization and modulated the stevioside and rebaudioside A concentration by regulating the transcription of genes that encode kaurene oxidase and glucosyltransferases, which are involved in the synthesis of these compounds in S. rebaudiana. This knowledge is important for generating biotechnological strategies that involve manipulating the concentration of stevioside or rebaudioside A by controlling the colonization status and the phosphate concentration of S. rebaudiana plants. The alignment that was used to deduce the amino acid sequence and topology of the 12 SrPT TM domains (TM1-TM12) was predicted using AM-specific phosphate transporters from M. truncatula (MtPT4), A. sinicus (AsPT4), S. tuberosum (StPT4), L. esculentum (LePT4), and P. hybrida (PhPT4). The deduced amino acid sequence was aligned through MULTIALIN, and SrPT was predicted according to <ns0:ref type='bibr' target='#b55'>Yadav et al. (2010)</ns0:ref>. Manuscript to be reviewed Manuscript to be reviewed Manuscript to be reviewed Manuscript to be reviewed </ns0:p></ns0:div>
<ns0:div><ns0:head>Conclusions</ns0:head><ns0:note type='other'>Figure 2</ns0:note><ns0:note type='other'>Figure 5</ns0:note><ns0:note type='other'>Figure 6</ns0:note><ns0:note type='other'>Figure 7</ns0:note></ns0:div><ns0:figure xml:id='fig_0'><ns0:head /><ns0:label /><ns0:figDesc>AM symbiosis between S. rebaudiana and R. irregularis is affected by phosphate concentrations; a low phosphate concentration induces a high percentage of colonization. The morphology of the PeerJ reviewing PDF | (2020:07:50625:1:1:NEW 26 Aug 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_1'><ns0:head>Figure 1 .</ns0:head><ns0:label>1</ns0:label><ns0:figDesc>Figure 1. Quantification of mycorrhizal colonization in S. rebaudiana roots. Percentages of total mycorrhizal colonization (dotted bars) and arbuscule colonization (diagonally hatched bars) by R. irregularis in S. rebaudiana plants fertilized with Hoagland nutrient solution at different KH 2 PO 4 concentrations. Bars represent the mean ± standard deviation (SD) of six replicates. Different letters indicate significant differences according to Tukey´s test (P<0.05). Capital letters were used for total colonization, and lowercase letters were used for arbuscular colonization.</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_2'><ns0:head>Figure 2 .</ns0:head><ns0:label>2</ns0:label><ns0:figDesc>Figure 2. Light and confocal microscopic analysis of R. irregularis colonization structures in S. rebaudiana roots. Mycorrhiza-colonized roots of S. rebaudiana plants fertilized with 20 (A), 200 (B), 500 (C) and 1000 µM KH 2 PO 4 (D) after trypan blue staining depicting the arbuscules, vesicles and intraradical hyphae. Mycorrhiza-colonized roots with 200 µM KH 2 PO 4 were also treated with propidium iodide to label the cell wall (E) and with WGA-Alexa Fluor 488 conjugate to stain the fine details of the intraradical hyphae and arbuscules (F and G). The merged image showing both red and green fluorescence is presented in panel G. Vesicles: V; intraradical hyphae: ih; arbuscules: *.</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_3'><ns0:head>Figure 3 .</ns0:head><ns0:label>3</ns0:label><ns0:figDesc>Figure 3. Effect of AM symbiosis and KH 2 PO 4 concentrations on the fresh weight of S. rebaudiana roots and leaves. Fresh weight of leaves (A) and roots (B) of mycorrhiza-colonized (M+) and noncolonized (M-) S. rebaudiana plants fertilized with Hoagland solution with different KH 2 PO 4 concentrations. Bars represent the mean ± standard deviation (SD) of six replicates. Different letters indicate significant differences according to Tukey's test (P<0.05).</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_4'><ns0:head>Figure 4 .</ns0:head><ns0:label>4</ns0:label><ns0:figDesc>Figure 4. Quantification of P and Mg under AM symbiosis and different KH 2 PO 4 concentrations. Phosphorus (A) and magnesium (B) concentration in the leaves of mycorrhizacolonized (M+) and noncolonized (M-) S. rebaudiana plants fertilized with Hoagland solution at different KH 2 PO 4 concentrations. Bars represent the mean ± standard deviation (SD) of six replicates. Different letters indicate significant differences according to Tukey´s test (P<0.05).</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_5'><ns0:head>Figure 5 .</ns0:head><ns0:label>5</ns0:label><ns0:figDesc>Figure 5. Effects of AM symbiosis and KH 2 PO 4 concentrations on photosynthetic performance in S. rebaudiana. Maximal photochemical efficiency (Fv/Fm) (a), potential photochemical efficiency (Fv/Fo) (b), and total chlorophylls (c), and carotenoids concentration (d) in leaves of mycorrhiza-colonized (M+) and noncolonized (M-) S. rebaudiana plants fertilized with Hoagland solution at different KH 2 PO 4 concentrations. Fv/Fm and Fv/Fo ratios were obtained from chlorophyll fluorescence measurements. Bars represent the mean ± standard deviation (SD) of six replicates. Different letters indicate significant differences according to Tukey´s test (P<0.05).</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_6'><ns0:head>Figure 6 .</ns0:head><ns0:label>6</ns0:label><ns0:figDesc>Figure 6. Differential transcript accumulation of three key genes from the SG biosynthetic pathway under AM symbiosis and different KH 2 PO 4 concentrations. KO (A), UGT74G1 (B) and UGT76G1 (C) relative expression levels in S. rebaudiana plants that were mycorrhiza-</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_7'><ns0:head>Figure 7 .</ns0:head><ns0:label>7</ns0:label><ns0:figDesc>Figure 7. Quantification of SG concentration under AM symbiosis and different KH 2 PO 4 concentrations. Stevioside (A) and rebaudioside A (B) concentration in mycorrhiza-colonized (M+) and noncolonized (M-) S. rebaudiana plants fertilized with Hoagland solution at 200 and 1000 µM KH 2 PO 4 . Bars represent the mean ± standard deviation (SD) of six replicates. Different letters indicate significant differences according to Tukey´s test (P<0.05).</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_8'><ns0:head>Figure S1 .</ns0:head><ns0:label>S1</ns0:label><ns0:figDesc>Figure S1. Protein sequence structure of the mycorrhiza-specific phosphate transporter (SrPT) from S. rebaudiana. The SrPT protein contains 391 amino acids and 9 TM domains. The alignment that was used to deduce the amino acid sequence and topology of the 12 SrPT TM domains (TM1-TM12) was predicted using AM-specific phosphate transporters from M. truncatula (MtPT4), A. sinicus (AsPT4), S. tuberosum (StPT4), L. esculentum (LePT4), and P. hybrida (PhPT4). The deduced amino acid sequence was aligned through MULTIALIN, and SrPT was predicted according to<ns0:ref type='bibr' target='#b55'>Yadav et al. (2010)</ns0:ref>.</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_9'><ns0:head>Figure S2 .</ns0:head><ns0:label>S2</ns0:label><ns0:figDesc>Figure S2. Expression analysis of the SrPT gene in roots of individual M-and M+ S. rebaudiana plants. Lanes 1-6, replicates of noncolonized (M-) and R. irregularis-colonized plants (M+). SrGADPH was used as a reference gene.</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_10'><ns0:head>Figure S3 .</ns0:head><ns0:label>S3</ns0:label><ns0:figDesc>Figure S3. Effects of AM symbiosis and KH 2 PO 4 concentration on aerial part development in S. rebaudiana. Leaves number (A) and foliar area in leaves of mycorrhiza-colonized (M+) and noncolonized (M-) S. rebaudiana plants fertilized with Hoagland solution using different KH 2 PO 4 concentrations Bars represent the mean ± standard deviation (SD) of six replicates. Different letters indicate significant differences according to Tukey´s test (P<0.05).</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_11'><ns0:head>Figure S4 .</ns0:head><ns0:label>S4</ns0:label><ns0:figDesc>Figure S4. Effects of AM symbiosis and KH 2 PO 4 concentration on chlorophyll fluorescence in S. rebaudiana. Primary fluorescence (Fo) (A), maximal fluorescence (Fm) (B), and variable fluorescence (Fv) (C) in leaves of mycorrhiza-colonized (M+) and noncolonized (M-) S. rebaudiana plants fertilized with Hoagland solution using different KH 2 PO 4 concentrations. Bars represent the mean ± standard deviation (SD) of six replicates. Different letters indicate significant differences according to Tukey´s test (P<0.05).</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_12'><ns0:head>Figure 1 Figure 1 .</ns0:head><ns0:label>11</ns0:label><ns0:figDesc>Figure 1</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_13'><ns0:head>Figure 2 .</ns0:head><ns0:label>2</ns0:label><ns0:figDesc>Figure 2. Light and confocal microscopic analysis of R. irregularis colonization structures in S. rebaudiana roots.</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_14'><ns0:head>Figure 3 Figure 3 .</ns0:head><ns0:label>33</ns0:label><ns0:figDesc>Figure 3</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_15'><ns0:head>Figure 4</ns0:head><ns0:label>4</ns0:label><ns0:figDesc>Figure 4</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_16'><ns0:head>Figure 4 .</ns0:head><ns0:label>4</ns0:label><ns0:figDesc>Figure 4. Quantification of P and Mg under AM symbiosis and different KH 2 PO 4 concentrations.</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_17'><ns0:head>Figure 5 .</ns0:head><ns0:label>5</ns0:label><ns0:figDesc>Figure 5. Effects of AM symbiosis and KH 2 PO 4 concentrations on photosynthetic performance in S. rebaudiana.</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_18'><ns0:head>Figure 6 .</ns0:head><ns0:label>6</ns0:label><ns0:figDesc>Figure 6. Differential transcript accumulation of three key genes from the SG biosynthetic pathway under AM symbiosis and different KH 2 PO 4 concentrations.</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_19'><ns0:head>Figure 7 .</ns0:head><ns0:label>7</ns0:label><ns0:figDesc>Figure 7. Quantification of SG concentration under AM symbiosis and different KH 2 PO 4 concentrations.</ns0:figDesc></ns0:figure>
<ns0:note place='foot'>PeerJ reviewing PDF | (2020:07:50625:1:1:NEW 26 Aug 2020)</ns0:note>
<ns0:note place='foot'>PeerJ reviewing PDF | (2020:07:50625:1:1:NEW 26 Aug 2020)Manuscript to be reviewed</ns0:note>
</ns0:body>
" | "INSTITUTO POLITÉCNICO NACIONAL
CENTRO DE DESARROLLO DE PRODUCTOS BIÓTICOS
DEPARTAMENTO DE BIOTECNOLOGÍA
August 21st, 2020
Ana I. Ribeiro-Barros
Editor, PeerJ
Dear Editor
We thank the editor and reviewers for their generous comments on the manuscript “Photosynthetic performance and stevioside concentration are improved by the arbuscular mycorrhizal symbiosis in Stevia rebaudiana under different phosphate concentrations”. In our current manuscript we have carefully addressed all the suggestion raised by the editor regarding the experimental design, missing data and interpretation of the results as well as the reviewers comments, which have substantially improved our manuscript. In the next section we will attend point by point the reviewers questions.
Reviewer 1 indicates that
Basic reporting
See below
Experimental design
See below
Validity of the findings
See below
Comments for the author
Report on ‘Phosphate modulates arbuscular mycorrhizal symbiosis and improves photosynthetic performance and stevioside content in Stevia rebaudiana’
This is a very well written, clear manuscript, that builds on those of the cited publications of Mandal et al. Vafadar et al., Tavarini et al. It deserves publishing because it provides evidence for the type of colonisation structure, for the mode of uptake of PO4 and usefully links the data on photosynthesis [fluorescence and pigments] with growth and SGs [but see my comment below ref Reb-A and stevioside concentrations.
A number of small issues need to be addressed, and then I believe it will be suitable for publication.
I have annotated the text [attached], and herein indicate some of the more important issues that must be addressed.
We appreciate the reviewer comments and all suggestions notes made on the PDF file.
Title: as it read it implies that phosphate improves photosynthetic performance and stevioside… so either add ‘the latter’ or rewrite completely.
We agree with the reviewer comment and we have modified the title as follow:
Photosynthetic performance and stevioside concentration are improved by the arbuscular mycorrhizal symbiosis in Stevia rebaudiana under different phosphate concentration
Also, I disagree with the use throughout of ‘content’ when indeed the authors mean concentration’ For content = concentration * mass, and at no time is mass information given in the text. So, change all uses of ‘content’ to ‘concentration’ [except of course in the references!].
We agree with the reviewer comment and we have changed the words “content” by “concentration”, where was required.
Line 52 Rhizoglomus irregulare or Rhizophagus irregularis I am not an expert on this but ensure the correct one is used.
We appreciate the comment and we have attended all these typo mistakes since the correct name is Rhizophagus irregularis.
See line 97, do not start a sentence with an acronym [and define at first use in the text [separate from the Abstract].
We have addressed this mistake, including the one in line 86.
Likewise, Line 117.
We changed the abbreviated R. irregularis for Rhizophagus irregularis at first mention.
A little more detail in line 138.
As requested by the reviewer more information was added on line 138, the paragraph on lines 139 to 141now reads as follows: After ten days of culture, the rooted cuttings were transferred to plastic cones of 125 mm high and 32 mm diameter (M49, Polietilenos del Sur S.A. C.V.), one plant per cone.
Line 345 any reason the root fwt was higher in M+ than M- at 1000 µM?
We appreciate the reviewer's observation. In response we have added a possible explanation on lines 481-490. Which is as follow:
Shoots and roots were responsive to mycorrhizal colonization by increasing growth at 200 µM KH2PO4, whereas at 1000 µM KH2PO4, roots growth was stimulated only in M+ compared to noncolonized controls (M-). This indicates that the symbiosis differentially affects shoot and root growth, and this depends on the phosphate fertilization regime. In plants at 500 µM KH2PO4, such differences between fresh weights of shoots and roots in M+ in comparison to M- was not observed. This in contrast to 200 µM and 1000 µM KH2PO4 conditions, in which mycorrhiza-specific and direct plant phosphate uptake pathways dominate the phosphate uptake, respectively. It is possible that 500 µM KH2PO4 does not affect the direct and the mycorrhizal-specific phosphate uptake pathways. Although this hypothesis needs to be further studied.
Line 383 compared to what.? Be explicit.
To make the comparison explicit, we changed the sentence “the KO gene did not change in M+ plants at 1000 µM KH2PO4,” to “the KO gene did not change in M+ plants at 1000 µM KH2PO4 in comparison to M- plants, since the relative expression (2-∆∆Ct) was close to 1 (Fig. 6A)”. Lines 415-417.
Line 460 indicate if significant or not.
We revised the observation and added accordingly “significantly” in the corresponding sentence. Line 504
Lines 497-509 need some clarification. For a start, Reb-A is more preferable from a taste perspective than stevioside, and indeed the total SGs is the same with or without M+ at 200 µM. So this needs some extra discussion. And reconcile lines 499 and 505/6.
We have revised the paragraph in lines 551-553 and added new information for clarification, now it reads as follow:
In contrast, the accumulation of rebaudioside A, the SG with the highest sweetening power, would be favored in noncolonized plants. This information is relevant from a biotechnological perspective.
However, concerning to reconcile lines 499 and 505/6, we apologize but we do not understand the concern by the reviewer.
Reviewer 2 indicates that
Basic reporting
The English version is acceptable, although some modifications are required.
The literature references are not always referring to the original literature, e.g., for the Synthesis of steviol which has been published by Totté et al. already in 2000
We completely agree with the reviewer and apologize for the mistake. The suggested reference has been included in the text on line 76 and bibliography.
The structure of the paper is OK
The results are sometimes rather weak, due to the lack of repetitions of the whole experiment.
We appreciate the reviewer comment and we have addressed the raised points. In our current version of the manuscript we make it clear that we performed the experiment by twice. By mistake, we did not provide this information on the previous manuscript, and therefore we really appreciate the comment that allowed attend this point. This information was added in the materials and methods of the new manuscript on lines 153-154. The new paragraph reads as follow: “Two independent experiments were performed. Similar trends were obtained in the both experiments, and the results of only one of them is shown”. Furthermore, we are including in this round of revision the raw data of the second experiment for the reviewer revision.
Experimental design
The intention of the authors is good, but the way that they organized the experiment is not. Only 6 plants per treatment were used, but interesting information on the plants is missing, e.g., the number of leaves, the position of the leaves that were used.
We apologize for omitting this important information about the position, number and condition of the leaves that were used in the different experiments. We added this information in the corresponding Materials and Methods of each experiment.
We also, added the results of leaves number and foliar area as a supplementary data in results section (Fig. S3).
It is not reported whether the leaves were dried or not, the amounts used for the extractions are not always given.
We appreciate the comment and the required information have been added to the manuscript as follows:
On lines 292 to 294 we included
Leaves from one of the groups described in the determination of plant growth section, were dried in in an oven (Thermo Scientific, USA) at 65 ° C for 48 h. The dry leaf tissue (0.1g) was extracted….
Furthermore on lines 298-303.Complementary information about the SGs methodology was added.
The mixture was stirred for 3 min, allowed to stand for 24 h without stirring, and then centrifuged at 10,000 rpm at 4 °C for 10 min. The supernatant was recovered, placed in Eppendorf tubes, and stored at -4 °C until the analysis by HPTLC (CAMAG, Switzerland). The quantification of SGs was based on the methodology reported by Bladt and Zgainski (1996) and Morlock et al. (2014), and discribed recently by Villamarin-Gallegos et al. (2020). Stevioside and rebaudioside A concentration were expressed in mg g DW-1.
It is not reported that validated methods of analysis were used.
Not sufficient information is given to repeat the experiment.
We appreciate the reviewer´s comments. We have double checked the methods used in each section and made sure that the corresponding reference to each one of the procedures used in this work are provided in each one of the sections of the Materials and Methods. We consider that the added information in the revised manuscript will allow reproducing the experiments.
Validity of the findings
It is a good point that the authors used microscopic techniques to see what happens within the plants. However, no information is given on the number and/or condition of the chloroplasts, which are the starting point of SG biosynthesis.
Microscopy techniques were used in this work for the description of the type of mycorrhizal colonization in S. rebaudiana, and for its quantification in the roots system. However, it was not used to perform any characterization of leaf tissue. Instead, in this tissue, we performed SG extractions and gene expression analysis, which revealed valuable information on the biosynthesis and accumulation of these compounds. Even though the starting point of SG biosynthesis resides in chloroplast, the proteins encode by the analyzed genes (SrUGT74G1 and SrUGT76G1) are not chloroplastic, but located in the cytoplasm (Brandle and Telmer, 2007), this is why we decide to determine the SG concentrations and gene expression as relevant information. However, we agree with the reviewer that the description of number and/or condition of chloroplast in leaf tissue would be an interesting issue to investigate; and it will be certainly approached in future work.
In some figures, some outlyers are visible, probably due to sampling methods and/or lack of repetitions of the experiment. (e.g., Fig; 3, 500 µM with colonisation; Fig. 7: A control.
In our current version of the manuscript we make it clear that we performed the experiment by twice. By mistake, we did not provide this information on the previous manuscript, and therefore we really appreciate the comment that allowed attend this point. This information was added in the Materials and Methods of the new manuscript on lines 153-154. The new paragraph reads as follow: “Two independent experiments were performed. Similar trends were obtained in the both experiments, and the results of only one of them is shown”. Furthermore, we are including in this round of revision the raw data of the second experiment for the reviewer’s revision.
We understand the observation by the reviewer regarding Fig 3. Thus, we added in the corresponding part of the discussion on lines 485 to 491 the next phrase:
“In plants at 500 µM KH2PO4, such differences between fresh weights of shoots and roots in M+ in comparison to M- was not observed. This in contrast to 200 µM and 1000 µM KH2PO4 conditions, in which mycorrhiza-specific and direct plant phosphate uptake pathways dominate the phosphate uptake, respectively. It is possible that 500 µM KH2PO4, the direct and the mycorrhizal-specific phosphate uptake pathways, may be interacting in such a way that no increased in growth is manifested. Although this hypothesis needs to be further studied.”
However, concerning Fig 7, we apologize but we do not understand the concern by the reviewer.
Comments for the Author
The authors studied the colonisation of mycorrhiza in Stevia. This part of the work seems well-done.
We thank the reviewer for his comment.
PreIn the introduction, original research papers should be cited, e.g., the first report of the synthesis of steviol by the MEP pathway is by Totté et al. (2000) Tetrahedronn lett. 41, 6407-6410 (not: Tetali 2019). The mention of SGs in Stevia is not from Wölwer-Rieck (2012), etc.
We appreciate the comment and these key works have been included in the bibliography. We also corrected the literature references used in the introduction to describe SGs biosynthesis for the right one.
1) Stevia is not an easy plant to work with.
We agree, however we have been working with stevia plants for some time. During this time, we have developed techniques and procedures to ensure a reproducible management of the plant growth. We consider that stevia is a plant with a wide worldwide demand and interest for its sweetening properties, and therefore the basic knowledge to understand the plant metabolites accumulation in response to certain conditions is highly desirable. Therefore, our work can contribute to a better knowledge about the interaction between arbuscular mycorrhizal fungi with an agronomical interesting plant such as Stevia rebaudiana. Our results show a clear mycorrhizal colonization and activation of the phosphate transport and links the mutualistic interaction and KH2PO4 with photosynthesis and the potential effect on plant growth and SGs biosynthesis.
2) The experiment is done only once, and per treatment 6 plants were used. No repetition of the experiments has been done. Therefore, the results reported should be considered as preliminary as more experiments should be done.
We appreciate the reviewer comment and apologize for not making it clear in the previous manuscript. However, as mentioned before in a related and similar question, this was our mistake and we have included new information in the revised manuscript that make it clear that we repeated our experiments twice with similar results. This information was added in the material and methods section (lines 153-154). Again, we are also including the raw data of the second experiment for the reviewer’s revision.
3) The number and size of the leaves are not given.
We appreciate the observation by both reviewers, in this revised version of the manuscript we are including the required information as supplementary data (Fig. S3). This is also mentioned in the Results section (lines 376 to 379).
Briefly it says as follow: Leaves number and foliar area were determined, and only at 200 µM KH2PO4, M- plant showed lower number of leaves than M+, as well as foliar area (Fig. S3), which is consistent with the pattern of fresh weight of leaves. Foliar area, on the other hand, was only significantly lower in M+ plants at 500 µM KH2PO4 (Fig. S3).
4) Were the leaves used as such or dried before the different extractions? How exactly was the leaf sampling done for the measurements in the different assays? How much material was used?
We appreciate the observation and understand that there was important missing information. Therefore, we added the required specific information in the corresponding part in Materials and Methods of each experiment. All the required information was added as follow:
In lines 174 to185.
For minimize the dependency of the variable evaluated with the leaf position on the plant. The leaves were collected in two equivalent groups, considering their opposite arrangement of the leaves and their position along the stem. Then, leaves positioned in all the nodes of the plant composed each group.
The leaf area was determined by image analysis from leaves of the second node, close to the apical meristem. A stereomicroscope (Olympus SZX7, Germany) was used to obtain the micrographs. Image analysis was performed using ImageJ editing software (Version, 1.8.0.112).
One half of the leaves were frozen in liquid nitrogen for molecular analysis, and the other half for SG extraction. Root were also collected and separated longitudinally in two sections, one of them was used for determination of mycorrhizal colonization, and the other for mycorrhiza-specific phosphate transporter identification.
Lines 203 to 206
The determination of chlorophyll and carotenoids concentration was made from three fresh leaves (approximately 100-150 mg) of each of the six plants of the M- and M + conditions. The leaves were collected from the upper, middle and lower part of each one of the analyzed plants and were ground in a mortar with 80% acetone.
Line 230 to 231
Frozen leaves, from one of the groups described in the determination of plant growth section, were ground to a fine powder in liquid nitrogen.
Line 266 to 267
Collected roots from one of the groups described in the determination of plant growth section, were ground to a fine powder in liquid nitrogen
Line 294 to 295
Leaves from one of the groups described in the determination of plant growth section, were dried in in an oven (Thermo Scientific, USA) at 65 ° C for 48 h. The dry leaf tissue (0.1g) was extracted…
5) line 182: why is the amount of P and Mg given as a percentage, and not just µg/g fresh weight?
The methodology used to determine the amount of P and Mg was using a high resolution scanning electron microscope (Guerrero Molina et al, 2014). They used this technique to know the elemental composition of strawberry plants inoculated with the plant growth-promoting bacterium Azospirillum brasilense. However, the results obtained with this tehnique are given as a percentage. An apology, but with this methodology it is not possible to obtain the results as requested by the reviewer as µg/g fresh weight.
However, in order to clarify this point, the following sentence was added to the Materials and Method section in line 196 to 199.
Since the results of the content of each element are given as the percentage of such element with respect to all the components determined in the sample, the concentration of P and Mg in the samples was expressed as the percentage of each element in the leaves.
6) line 186: how was the leaf tissue sampled for the pigment analysis? Dry or fresh wt?
In order to clarify this point, the next information was included on lines 203 to 206.
The determination of chlorophyll and carotenoids concentration was made from three fresh leaves (approximately 100-150 mg) of each of the six plants of the M- and M + conditions. The leaves were collected from the upper, middle and lower part of each one of the analyzed plants and were ground in a mortar with 80% acetone.
7) Extraction of SGs: What is the quantity of leaves extracted (dry or fresh weigth)? Were 3 extractions done on 3 different (weighed) amounts? As shown by Ceunen & Geuns, Plant Science (2013), 198, 72-82, the photoperiod and leaf position on the plants is very important for the SGs content. To obtain the best results, it is advised to dry a huge number of leaves before and grind them to a fine powder. Thereafter, at least 20 mg dry powder should be extracted to avoid extracting a leaf coming from just 1 position of the plant. This should be done at least 3 times.
Was the method for SG analysis validated by the authors?
We appreciate the comments made by the reviewer regarding the methodology of Ceunen & Geuns, Plant Science (2013), 198, 72-82. Unfortunately, we did not implement that methodology.
We mentioned before in a related and similar question with relation to leaves samples, lines 174 to177.
For minimize the dependency of the variable evaluated with the leaf position on the plant. The leaves were collected in two equivalent groups, considering their opposite arrangement of the leaves and their position along the stem. Then, leaves positioned in all the nodes of the plant composed each group.
In order to address the problem commented by the reviewer, additional information on the methodology used for the extraction and quantification of SGs by HPTLC was incorporated lines 294 to 303.
We also include the citation of the primary source of the methodology used and its corresponding reference in the bibliography section.
“Leaves from one of the groups described in the determination of plant growth section, were dried in in an oven (Thermo Scientific, USA) at 65 ° C for 48 h. The dry leaf tissue (0.1g) was extracted with 1 mL of methanol (J.T. Backer, USA), following the methodology described by Woelwer-Rieck et al. (2010). The mixture was stirred for 3 min, allowed to stand for 24 h without stirring, and then centrifuged at 10,000 rpm at 4 °C for 10 min. The supernatant was recovered, placed in Eppendorf tubes, and stored at -4 °C until the analysis by HPTLC (CAMAG, Switzerland). The quantification of SGs was based on the methodology reported by Bladt and Zgainski (1996) and Morlock et al. (2014), explanin recently by Villamarin-Gallegos et al. (2020). Stevioside and rebaudioside A concentration were expressed in mg g DW-1. Six plants per phosphate treatment and R. irregularis colonization status were evaluated.”
8) Figure 3 (growth) and Fig. 7 (ST and RebA). The authors did not check the number of chloroplasts in the leaves treated with 200 µM phosphate. Probably, the number did not change after 30 d treatment. The first steps of SG biosynthesis are dependent upon the chloroplasts. Therefore, results should be carefully analysed. The growth increase by 200 µM of about 70 %, and the decrease of Reb A content is about 50 %. Is it possible that the decrease just reflects the “dilution effect” of rebA concentration by the larger cell volume? Of course, for ST this is not the case, but the value for the control is extremely low. Is this due to the one-time experiment, the more as the ST content of the 200 µM treatment with colonisation, as well as of both 1000 µM treatment are not different? A similar problem might exist in Fig. 4 with an “outlyer” of Mg at 500 µM with colonisation.
As mentioned before, we did not look at the chloroplast and therefore we cannot comment of chloroplast abundance or differences in the content of chloroplasts in the leaves from the different experimental conditions (M+ and M-). However, the SG profile herein reported were consistent in the two experiments performed, then, we believed that the results reflects the actual concentration of SG, at least in our experimental conditions. Nevertheless, we consider that looking at the morphology and numbers of chloroplast in our experimental conditions will be very important and we are considering to address this issue in further work.
Again, we thank and appreciate all the observations raised by the reviewers that have certainly improved our manuscript. We hope that this improved version of the manuscript will fulfill the reviewer’s request and find it suitable for publication in PeerJ.
We look forward to hearing from you shortly.
Yours sincerely
Dr. Mario Rodríguez Monroy
Departamento de Biotecnología
Investigador del Centro de Desarrollo de Productos Bióticos
Instituto Politécnico Nacional
On behalf of all authors.
" | Here is a paper. Please give your review comments after reading it. |
9,820 | "<ns0:abstract xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:p>In plants, phosphorus (P) uptake occurs via arbuscular mycorrhizal (AM) symbiosis and through plant roots. The phosphate concentration is known to affect colonization by AM fungi, and the effect depends on the plant species. Stevia rebaudiana plants are valuable sources of sweetener compounds called steviol glycosides (SGs), and the principal components of SGs are stevioside and rebaudioside A. However, a detailed analysis describing the effect of the phosphate concentration on the colonization of AM fungi in the roots and the relationship of these factors to the accumulation of SGs and photochemical performance has not been performed; such an analysis was the aim of this study. The results indicated that low phosphate concentrations (20 and 200 µM KH 2 PO 4 ) induced a high percentage of colonization by Rhizophagus irregularis in the roots of S. rebaudiana, while high phosphate concentrations (500 and 1000 µM KH 2 PO 4 ) reduced colonization. The morphology of the colonization structure is a typical Arum-type mycorrhiza, and a mycorrhiza-specific phosphate transporter was identified. Colonization with low phosphate concentrations improved plant growth, chlorophyll and carotenoid concentration, and photochemical performance. The transcription of the genes that encode kaurene oxidase and glucosyltransferase (UGT74G1) was upregulated in colonized plants at 200 µM KH 2 PO 4 , which was consistent with the observed patterns of stevioside accumulation. In contrast, at 200 µM KH 2 PO 4 , the transcription of UGT76G1 and the accumulation of rebaudioside A were higher in noncolonized plants than in colonized plants. The results indicate that a low phosphate concentration improves mycorrhizal colonization and modulates the stevioside and rebaudioside A concentration by regulating the transcription of the genes that encode kaurene oxidase and glucosyltransferases, which are involved in stevioside and rebaudioside A synthesis in S. rebaudiana.</ns0:p></ns0:div>
</ns0:abstract>
<ns0:body xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:p>In plants, phosphorus (P) uptake occurs via arbuscular mycorrhizal (AM) symbiosis and through plant roots. The phosphate concentration is known to affect colonization by AM fungi, and the effect depends on the plant species. Stevia rebaudiana plants are valuable sources of sweetener compounds called steviol glycosides (SGs), and the principal components of SGs are stevioside and rebaudioside A. However, a detailed analysis describing the effect of the phosphate concentration on the colonization of AM fungi in the roots and the relationship of these factors to the accumulation of SGs and photochemical performance has not been performed; such an analysis was the aim of this study. The results indicated that low phosphate concentrations (20 and 200 µM KH 2 PO 4 ) induced a high percentage of colonization by Rhizophagus irregularis in the roots of S. rebaudiana, while high phosphate concentrations (500 and 1000 µM KH 2 PO 4 ) reduced colonization. The morphology of the colonization structure is a typical Arum-type mycorrhiza, and a mycorrhiza-specific phosphate transporter was identified. Colonization with low phosphate concentrations improved plant growth, chlorophyll and carotenoid concentration, and photochemical performance. The transcription of the genes that encode kaurene oxidase and glucosyltransferase (UGT74G1) was upregulated in colonized plants at 200 µM KH 2 PO 4 , which was consistent with the observed patterns of stevioside accumulation. In contrast, at 200 µM KH 2 PO 4 , the transcription of UGT76G1 and the accumulation of rebaudioside A were higher in noncolonized plants than in colonized plants. The results indicate that a low phosphate concentration improves mycorrhizal colonization and modulates the stevioside and rebaudioside A concentration by regulating the transcription of the genes that encode kaurene oxidase and glucosyltransferases, which are involved in stevioside and</ns0:p></ns0:div>
<ns0:div><ns0:head>Introduction</ns0:head><ns0:p>Stevia rebaudiana Bertoni is a plant that belongs to the Asteraceae family and accumulates compounds in its leaves called steviol glycosides (SGs) <ns0:ref type='bibr' target='#b8'>(Brandle, Starratt & Gijzen, 1998)</ns0:ref>. Stevioside and rebaudioside A are the best-known SGs and are important compounds for human health because they are natural low-calorie sweeteners. The sweetening power of stevioside and rebaudioside A is 143 and 320 times higher than that of sucrose, respectively <ns0:ref type='bibr' target='#b26'>(Lemus-Mondaca et al., 2012;</ns0:ref><ns0:ref type='bibr' target='#b9'>Brandle & Telmer, 2007)</ns0:ref>. The biosynthetic pathway for SG synthesis begins in the chloroplasts with the synthesis of geranylgeranyl diphosphate (GGDP) generated from the MEP (methyl-erythritol-4-phosphate) pathway <ns0:ref type='bibr' target='#b47'>(Totté et al., 2000)</ns0:ref>. GGDP is transformed to kaurene by two cyclization steps carried out by a terpene cyclase <ns0:ref type='bibr' target='#b9'>(Brandle & Telmer, 2007)</ns0:ref>. In the endoplasmic reticulum, kaurene is oxidized by kaurene oxidase (KO) to kaurenoic acid; the oxidation of kaurenoic acid produces gibberellins, while the hydroxylation of kaurenoic acid produces steviol. The hydroxyl groups of steviol are glycosylated by the enzymes uridine diphosphate (UDP)-glycosyltransferases (UGTs), and the number of sugars attached by UGTs generates the various SGs <ns0:ref type='bibr' target='#b9'>(Brandle & Telmer, 2007)</ns0:ref>. The UGT74G1 enzyme is involved in the conversion of steviolbioside to stevioside, while the UGT76G1 enzyme converts stevioside into rebaudioside A <ns0:ref type='bibr' target='#b25'>(Kim et al., 2019)</ns0:ref>.</ns0:p><ns0:p>Arbuscular mycorrhizal (AM) symbiosis between the phylum Glomeromycota and plants is a mutualistic association that is useful in the culture of plants and has agricultural and medicinal importance. This symbiosis improves plant growth, photosynthesis and nutrient uptake and increases the production of phytochemicals <ns0:ref type='bibr' target='#b43'>(Smith & Read, 2008;</ns0:ref><ns0:ref type='bibr' target='#b45'>Spatafora et al., 2016;</ns0:ref><ns0:ref type='bibr' target='#b38'>Schoefs et al., 2015)</ns0:ref>.</ns0:p><ns0:p>In plants, there are two modes of P uptake: one mode is by the plant's own Pi transporters, and the other mode occurs via AM symbiosis with mycorrhiza-specific phosphate transporters and takes place in the arbuscules <ns0:ref type='bibr' target='#b20'>(Harrison, Dewbre &Liu, 2002)</ns0:ref>. Phosphate transporters are considered a key feature of this mycorrhizal symbiosis <ns0:ref type='bibr'>(Karandashov & Bucher, 2005)</ns0:ref>. Mycorrhiza-specific phosphate transporters are expressed in arbuscule-containing cortical root cells and are thus considered general markers for AM symbiosis in different model plants <ns0:ref type='bibr' target='#b20'>(Harrison, Dewbre & Liu, 2002;</ns0:ref><ns0:ref type='bibr' target='#b33'>Nagy et al., 2005;</ns0:ref><ns0:ref type='bibr' target='#b34'>Nouri et al., 2014)</ns0:ref>. Likewise, the phosphate concentration affects the colonization of roots by AM. Phosphate application at a high concentration may inhibit the formation of arbuscular mycorrhizae, and the sensitivity to phosphate and the grade of inhibition of arbuscule formation depend on the plant species <ns0:ref type='bibr' target='#b44'>(Smith, Smith & Jakobsen, 2004)</ns0:ref>; for example, these factors differ in Medicago truncatula <ns0:ref type='bibr' target='#b7'>(Bonneau et al., 2013)</ns0:ref> and Petunia hybrida <ns0:ref type='bibr' target='#b10'>(Breuillin et al., 2010;</ns0:ref><ns0:ref type='bibr' target='#b34'>Nouri et al., 2014)</ns0:ref>.</ns0:p><ns0:p>In S. rebaudiana, AM symbiosis enhances the production of stevioside and rebaudioside A and involves nutritional and nonnutritional mechanisms <ns0:ref type='bibr' target='#b28'>(Mandal et al., 2013)</ns0:ref>. AM symbiosis also upregulates the transcription of eleven SG biosynthesis genes as a consequence of the improved nutrition status and the increase in photosynthesis in the plant <ns0:ref type='bibr' target='#b29'>(Mandal et al., 2015)</ns0:ref>. This result suggests the roles of phosphorus nutrition and AM symbiosis in influencing SG concentration; fertilization with 25 mg P 2 O 5 kg −1 soil in association with AM symbiosis improved SG yield, P uptake and P nutrient use efficiency <ns0:ref type='bibr' target='#b46'>(Tavarini et al., 2018)</ns0:ref>. However, a detailed analysis and systematic description of the morphological type of AM symbiosis and the effect of the different phosphate concentrations on the establishment of AM symbiosis, the identification of mycorrhiza-specific phosphate transporters, the photosynthetic performance, and the relationship with the accumulation of SGs in S. rebaudiana plants have not been addressed. Therefore, in this study, we reported the effects of different phosphate concentrations on the establishment of AM symbiosis between Rhizophagus irregularis and S. rebaudiana, their relationship with photochemical performance and the accumulation of steviol glycosides (SGs) and the expression of two key genes, UGT74G1 and UGT76G1, which encode the (UDP)-glycosyltransferases involved in stevioside and rebaudioside A biosynthesis, respectively. The participation of a mycorrhiza-specific phosphate transporter as a key feature of this mycorrhizal symbiosis was demonstrated.</ns0:p></ns0:div>
<ns0:div><ns0:head>Materials & Methods</ns0:head><ns0:p>Inoculation with Rhizophagus irregularis and plant growth conditions R. irregularis was provided by Dr. Melina López-Meyer from 'Centro Interdisciplinario de Investigación para el Desarrollo Integral Regional, Unidad Sinaloa', Sinaloa state, Mexico. The inoculum was grown on Petri dishes with two compartments containing transformed carrot roots on minimum medium with 2% Gel-rite (Sigma-Aldrich) and incubated in the dark at 23 ± 2°C for six months according to the method reported by <ns0:ref type='bibr' target='#b3'>Bécard & Fortin (1988)</ns0:ref>.</ns0:p><ns0:p>Six-month-old cuttings of S. rebaudiana plants were cultured under greenhouse conditions. Briefly, seven-cm cuttings were disinfected with 70% ethanol for 1 min and 2% sodium hypochlorite for 1 min and washed three times with sterile distilled water for 2 min. To measure root development, the cuttings were transplanted under hydroponic conditions in glass tubes containing Fahraeus medium and cultured in a controlled environment chamber at 25°C with a 16 h light: 8 h dark photoperiod regimen. After ten days of culture, the rooted cuttings were transferred to plastic cones of 125 mm high and 32 mm diameter (M49, Polietilenos del Sur S.A. C.V.), one plant per cone. The substrate used was a 1:1 (v:v) mixture of vermiculite and sand. The substrate was autoclaved twice for 1 h at 121°C and 15 psi. S. rebaudiana plants were inoculated with 150 spores of R. irregularis (M+) that were homogeneously distributed in the substrate; the controls were noncolonized (M-) plants. The plants were watered twice per week with 20 mL of half-strength Hoagland nutrient solution <ns0:ref type='bibr' target='#b21'>(Hoagland & Arnon, 1950)</ns0:ref> with KH 2 PO 4 at the final phosphate concentrations that were evaluated: 20, 200, 500, and 1000 µM. The pH of the nutrient solutions was adjusted to 6.1.</ns0:p><ns0:p>The plants were maintained in a growth chamber at 25°C with a 16 h light:8 h dark photoperiod regimen for 30 days postinoculation (dpi). The experiment was performed utilizing a complete factorial design, and six plants per phosphate concentration and colonization status with R. irregularis were evaluated. The controls were noncolonized plants treated with the different phosphate concentrations. Two independent experiments were performed. Similar trends were obtained in the both experiments, and the results of only one of them are shown.</ns0:p></ns0:div>
<ns0:div><ns0:head>Staining and quantification of mycorrhizal colonization</ns0:head><ns0:p>S. rebaudiana root segments were stained with 0.05% trypan blue in lactoglycerol <ns0:ref type='bibr' target='#b35'>(Phillips & Hayman, 1970)</ns0:ref> and observed by light microscopy (BOECO Germany, BM-180) at 10-40X magnification. Total mycorrhizal colonization by R. irregularis was calculated according to the line-intersection method <ns0:ref type='bibr' target='#b16'>(Giovannetti & Mosse, 1980)</ns0:ref>. For each plant, 90 root segments were assessed, and six plants were evaluated. The arbuscular percentage was calculated with MycoCalc software (http://www.dijon.inra.fr/mychintec/Mycocalc-prg/download.html). To identify the morphological type of the AM symbiosis in S. rebaudiana, mycorrhizal roots were stained with WGA-Alexa Fluor 488 to visualize the arbuscules, and the plant tissue was labeled with propidium iodide according to the methodology reported by <ns0:ref type='bibr' target='#b55'>Xie et al. (2016)</ns0:ref> using confocal laser scanning microscopy (LSM 800, Carl Zeiss).</ns0:p></ns0:div>
<ns0:div><ns0:head>Determination of plant growth</ns0:head><ns0:p>The plants treated with the different phosphate concentrations and colonized (M+) or noncolonized (M-) with R. irregularis were collected at 30 days postinoculation (dpi). Shoots and roots of each plant were separated, total leaves number and the fresh weight of each organ was recorded. To minimize the dependency of the analysis of expression of genes and SG quantification with the leaf position on the plant. The leaves were collected in two equivalent groups, considering their opposite arrangement of the leaves and their position along the stem. Then, leaves positioned on each side of all the nodes of the plant composed each group. The leaf area was determined by image analysis from leaves of the second node, close to the apical meristem. A stereomicroscope (Olympus SZX7, Germany) was used to obtain the micrographs. Image analysis was performed using ImageJ editing software (Version, 1.8.0.112). One half of the leaves were frozen in liquid nitrogen for molecular analysis, and the other half for SG extraction. Root were also collected and separated longitudinally in two sections, one of them was used for determination of mycorrhizal colonization, and the other for mycorrhiza-specific phosphate transporter identification. Six plants per phosphate treatment and R. irregularis colonization status were evaluated.</ns0:p></ns0:div>
<ns0:div><ns0:head>Analysis of phosphorus and magnesium concentration</ns0:head><ns0:p>The leaves of S. rebaudiana plants were prepared according to <ns0:ref type='bibr' target='#b17'>Guerrero-Molina et al. (2014)</ns0:ref>. The leaves were analyzed by a high-resolution scanning electron microscope (SEM) equipped with a field emission cathode and coupled to an energy-dispersive X-ray (EDX, Carl Zeiss, Oberkochen, Germany). The electron energy used was 20 keV. The mapping of P and Mg was determined by EDX to record the two-dimensional elemental composition of the leaf sample surface. For quantitative analyses, EDX spectrograms were recorded and analyzed using QUANTAX ESPRIT, Version 1.9 (BRUKER, Germany). Since the results of the concentration of each element is given as the percentage of such element with respect to all the components determined in the sample, the concentration of P and Mg in the samples was expressed as the percentage of each element in the leaves.</ns0:p></ns0:div>
<ns0:div><ns0:head>Determination of chlorophyll and carotenoid concentration</ns0:head><ns0:p>The determination of chlorophyll and carotenoids concentration was made from three fresh leaves (approximately 100-150 mg) of each of the six plants of the M-and M + conditions. The leaves were collected from the upper, middle and lower part of each one of the analyzed plants and were ground in a mortar with 80% acetone. The extracts were centrifuged at 3000 g for 15 min; the supernatants were separated, and the absorbance at 646.8, 663.2 and 470 nm was measured in a UV/Vis spectrophotometer (UV-1800, Shimadzu, Japan). The concentration of chlorophylls and carotenoids were calculated following the equations described by <ns0:ref type='bibr' target='#b24'>Khan & Mitchell (1987)</ns0:ref>.</ns0:p></ns0:div>
<ns0:div><ns0:head>Measurement of chlorophyll fluorescence</ns0:head><ns0:p>The chlorophyll fluorescence was measured in the second fully expanded leaf of each plant using a chlorophyll fluorometer (model OS30P, Opti-Sciences Inc., USA). The evaluation was performed at room temperature according to the instructions for the chlorophyll fluorometer. Before the evaluation, the plants were placed in the dark for 30 min, and chlorophyll fluorescence was evaluated after applying a 1 s saturating pulse of actinic light (3500 µmol m -2 s -1 ). The primary fluorescence (Fo), maximal fluorescence (Fm), maximum quantum efficiency of PSII photochemistry (Fv/Fm), and potential photochemical efficiency (Fv/Fo) were calculated. Fv was calculated as Fv = Fm−Fo, and Fv/Fo was calculated as Fv/Fo = Fm/Fo−1 <ns0:ref type='bibr' target='#b39'>(Schreiber, Bilger & Neubauer, 1994)</ns0:ref>.</ns0:p></ns0:div>
<ns0:div><ns0:head>Expression analysis by qRT-PCR</ns0:head><ns0:p>The transcript accumulation levels of the genes for kaurene oxidase and (UDP)glycosyltransferases were evaluated in colonized and noncolonized plants and in plants that were treated with 200 and 1000 µM KH 2 PO 4 . These phosphate concentrations were selected because they were found to be the optimal and suboptimal conditions for inducing root colonization.</ns0:p><ns0:p>Frozen leaves, from one of the groups described in the determination of plant growth section, were ground to a fine powder in liquid nitrogen. Total RNA was isolated from leaves using TRIzol reagent (Invitrogen, Carlsbad, CA) following the manufacturer's protocol. First-strand cDNA synthesis was performed as previously reported by <ns0:ref type='bibr'>Cervantes-Gámez et al. (2016)</ns0:ref>.</ns0:p><ns0:p>The primers used were those designed and reported by <ns0:ref type='bibr' target='#b29'>Mandal et al. (2015)</ns0:ref> for S. rebaudiana plants. The primers correspond to the kaurene oxidase gene (SrKOF 5´-TCTTCACAGTCTCGGTGGTG-3´, and SrKOR 5´-GGTGGTGTCGGTTTATCCTG-3´), the glucosyl transferase UGT74G1 gene (SrUGT74G1F 5´-GGTAGCCTGGTGAAACATGG-3´, and SrUGT74G1R 5´-CTGGGAGCTTTCCCTCTTCT -3´) and the glucosyl transferase UGT76G1 gene (SrUGT76G1F 5´-GACGCGAACTGGAACTGTTG-3´, and SrUGT76G1R 5´-AGCCGTCGGAGGTTAAGACT -3´). qRT-PCR was performed using SYBR ® Green (QIAGEN, USA) and quantified on a Rotor-Gene Q (QIAGEN, USA) real-time PCR thermal cycler. qRT-PCR was programmed for 35 cycles, with denaturing at 95°C for 15 s, annealing at 55°C for 30 s, and extension at 72°C for 30 s. Primer specificity was verified by regular PCR and melting curve analysis. The primers for the S. rebaudiana glyceraldehyde-3-phosphate dehydrogenase (GAPDH) gene (SrGAPDHF 5´-TCAGGGTGGTGCCAAGAAGG-3´, and SrGAPDHR 5´-TTACCTTGGCAAGGGGAGCA -3´) were used as internal controls for normalization, and the quantitative results were evaluated by the 2 −ΔΔCT method described by <ns0:ref type='bibr' target='#b27'>Livak & Schmittgen (2001)</ns0:ref>. To interpret the results, genes with fold change values ≥1.5 were considered 'upregulated', whereas genes with fold change values ≤−0.7 were considered 'downregulated'. Six plants per phosphate treatment and R. irregularis colonization status were evaluated.</ns0:p></ns0:div>
<ns0:div><ns0:head>Cloning the mycorrhiza-associated phosphate transporter gene from S. rebaudiana</ns0:head><ns0:p>A pair of degenerate primers (SrPTF 5´-ATGGGDTTTTTYACYGATGC-3´ and SrPTR 5´-GGNCCAAARTTSGCRAAGAA-3´) were designed by aligning highly conserved regions of AM-specific phosphate transporters from M. truncatula (accession number: AY116210), A. sinicus (accession number: JQ956418), S. lycopersicum (accession number: AF022874), S. tuberosum (accession number: AY793559) and P. hybrida (accession number: EU532763). The PCR product was purified using the QIAquick PCR Purification Kit (QIAGEN, USA) and ligated into the pGEM®-T Easy vector (Promega, USA) in accordance with the manufacturer's protocols. The presence of the correct insert (1350 bp) within the pGEM®-T Easy vector was confirmed by PCR using the universal primers T7 and SP6, and the insert was then sequenced.</ns0:p><ns0:p>Collected roots from one of the groups described in the determination of plant growth section, were ground to a fine powder in liquid nitrogen. Total RNA was obtained from the roots of six colonized (M+) and six noncolonized (M-) plants fertilized with 200 µM KH 2 PO 4 . RNA was isolated using TRIzol reagent (Invitrogen, Carlsbad, CA) following the manufacturer´s protocol. First-strand cDNA synthesis was performed as previously reported by <ns0:ref type='bibr'>Cervantes-Gámez et al. (2016)</ns0:ref>. cDNA synthesis was confirmed by PCR using primers for the S. rebaudiana glyceraldehyde-3-phosphate dehydrogenase (GAPDH) gene (SrGAPDHF 5´-ATGGGDTTTTTYACYGATGC-3´ and SrGAPDHR 5´-GGNCCAAARTTSGCRAAGAA-3´).</ns0:p><ns0:p>For the expression analysis of SrPT in the M-and M+ plants, PCRs were run in a total reaction volume of 10 µL, comprising 0.2 µL of GoTaq® Flexi DNA Polymerase (Promega, USA), 200 nM of each primer, and 50 ng of cDNA. The PCR thermocycler was programmed for 35 cycles, with denaturing at 95°C for 15 s, annealing at 52°C for 1 min, and extension at 72°C for 30 s. The SrGAPDH gene was used as the reference gene.</ns0:p></ns0:div>
<ns0:div><ns0:head>Homology modeling analysis of the AM-specific phosphate transporter from S. rebaudiana</ns0:head><ns0:p>BLAST analysis of the SrPT gene sequence was performed to determine homology predictions using the tools on the NCBI website (https://www.ncbi.nlm.nih.gov). To determine the conserved region of SrPT, multiple sequence alignments of AM-specific phosphate transporter proteins were performed using MULTALIN software. The homology model of SrPT transmembrane domains (TDs) was constructed according to <ns0:ref type='bibr' target='#b56'>Yadav et al. (2010)</ns0:ref>, and the Mtpt4 protein structure from M. truncatula was used as the template for homology modeling for the S. rebaudiana AM-specific phosphate transporters.</ns0:p></ns0:div>
<ns0:div><ns0:head>Steviol glycosides extraction and quantification of concentration</ns0:head><ns0:p>The leaves of colonized (M+) and noncolonized (M-) plants that were treated with 200 and 1000 µM KH 2 PO 4 were used to evaluate the SG concentration. Leaves from one of the groups described in the determination of plant growth section, were dried in in an oven (Thermo Scientific, USA) at 65 ° C for 48 h. The dry leaf tissue (0.1g) was extracted with 1 mL of methanol (J.T. Backer, USA), following the methodology described by <ns0:ref type='bibr' target='#b54'>Woelwer-Rieck et al. (2010)</ns0:ref>. The mixture was stirred for 3 min, allowed to stand for 24 h without stirring, and then centrifuged at 10,000 rpm at 4 °C for 10 min. The supernatant was recovered, placed in Eppendorf tubes, and stored at -4 °C until the analysis by HPTLC (CAMAG, Switzerland). The quantification of SGs was based on the methodology reported by <ns0:ref type='bibr' target='#b4'>Bladt and Zgainski (1996)</ns0:ref> and <ns0:ref type='bibr' target='#b31'>Morlock et al. (2014)</ns0:ref>, and described recently by <ns0:ref type='bibr' target='#b50'>Villamarin-Gallegos et al. (2020)</ns0:ref>. Stevioside and rebaudioside A concentration were expressed in mg g DW -1 . Six plants per phosphate treatment and R. irregularis colonization status were evaluated.</ns0:p></ns0:div>
<ns0:div><ns0:head>Statistical analysis</ns0:head><ns0:p>The differences between the total colonization and arbuscular percentages were examined by one-way analysis of variance (ANOVA), and Tukey's post hoc test (P< 0.05) was performed to test the significance of differences between means. Data regarding the effect of the interaction between the KH 2 PO 4 concentration and mycorrhizal colonization on plant growth, P and Mg concentration, pigment concentrations, chlorophyll fluorescence and SG concentration were subjected to factorial two-way analysis of variance (ANOVA). Tukey´s post hoc test was used to analyze the differences. The paired Student's t-test was used to evaluate the significance of differences in the gene expression of kaurene oxidase and (UDP)-glycosyltransferases. All data, used for ANOVA and factorial analysis were checked for normal distributions (Shapiro-Wilk´s test) before statistical analysis. All statistical analyses were performed using the statistical software IBM SPSS for Windows, Version 24.0 (Armonk, NY, IBM Corp.).</ns0:p></ns0:div>
<ns0:div><ns0:head>Results</ns0:head><ns0:p>PeerJ reviewing PDF | (2020:07:50625:2:0:NEW 8 Sep 2020)</ns0:p><ns0:p>Manuscript to be reviewed</ns0:p></ns0:div>
<ns0:div><ns0:head>Phosphate concentration affects the mycorrhizal colonization of S. rebaudiana</ns0:head><ns0:p>The highest percentages of colonization were obtained at 20 and 200 µM KH 2 PO 4 (73.3 and 67.0% colonization, respectively). In contrast, the percentage of total colonization decreased significantly at 500 and 1000 µM KH 2 PO 4 , with 43.3 and 18.4% colonization, respectively (Fig. <ns0:ref type='figure' target='#fig_1'>1</ns0:ref>). The percentage of arbuscules was significantly reduced at 500 and 1000 µM KH 2 PO 4 , with 1.48 and 0.4%, respectively (Fig. <ns0:ref type='figure' target='#fig_1'>1</ns0:ref>).</ns0:p><ns0:p>In the roots of plants treated with 20 µM KH 2 PO 4 , a high number of arbuscules formed (Fig. <ns0:ref type='figure' target='#fig_13'>2A</ns0:ref>, see label *); several intraradical hyphae grew through the cortical cells (Fig. <ns0:ref type='figure' target='#fig_13'>2A</ns0:ref>, see label ih), although vesicle structures were scarce. Similar structures were observed in colonized plants (M+) with 200 µM KH 2 PO 4 ; however, under these experimental conditions, the formation of arbuscules and vesicle structures was more evident (Fig. <ns0:ref type='figure' target='#fig_13'>2B</ns0:ref>, see labels * and v), indicating that 200 µM KH 2 PO 4 created better conditions than 20 µM for the formation of arbuscules. Qualitative differences in mycorrhizal structures were observed when plants were treated with the highest KH 2 PO 4 concentrations (500 and 1000 µM); the formation of arbuscules, for instance, was significantly reduced (Fig. <ns0:ref type='figure' target='#fig_13'>2C</ns0:ref> and 2D, see label *), and shortening of intraradical structures was observed (Fig. <ns0:ref type='figure' target='#fig_13'>2C and D</ns0:ref>, see label ih) in comparison to the structures observed at low KH 2 PO 4 concentrations (Fig. <ns0:ref type='figure' target='#fig_13'>2A and 2B</ns0:ref>).</ns0:p><ns0:p>The activation of specific genes, such as the phosphate transporter specifically induced by the mycorrhizal association, is an important marker for evaluation successful colonization establishment. To our knowledge, there is no information on this transporter type in S. rebaudiana. For this reason, we identified a putative mycorrhiza-specific phosphate transporter in S. rebaudiana (SrPT) by a simple PCR strategy based on degenerate oligonucleotides to clone the corresponding phosphate transporter. A 1175 bp-long genomic fragment containing an open reading frame that encodes a 391-amino acid polypeptide with a molecular mass of 43.39 kDa was cloned (accession number: MN273502). This putative SrPT polypeptide contains 9 of the 12 transmembrane domains from the canonic phosphate transporter (Fig. <ns0:ref type='figure' target='#fig_8'>S1</ns0:ref>). The bioinformatic analysis suggests that SrPT is 72.89% conserved in comparison to the sequences reported for MtPT4 TMDs in Medicago truncatula. Notably, SrPT transcript accumulation increased in the roots of colonized (M+) plants compared with that in the roots of noncolonized (M-) plants. This result suggests that this gene is positively regulated by AM symbiosis in S. rebaudiana, in a similar manner to the genes for other mycorrhiza-specific phosphate transporters reported in other model plants (Fig. <ns0:ref type='figure' target='#fig_9'>S2</ns0:ref>).</ns0:p><ns0:p>Confocal microscopic analysis of the colonized roots and staining with the conjugate WGA-Alexa Fluor® 488 and propidium iodide as fluorescent markers permitted us to differentiate the hyphae and the plant cells, respectively. Intracellular hyphae and the formation of arbuscules from intracellular hyphae growing in the inner cortex were observed (Fig. <ns0:ref type='figure' target='#fig_13'>2E-G</ns0:ref>, see labels ih and *). With this approach, we were able to depict and classify this structure as a typical Arumtype mycorrhiza.</ns0:p></ns0:div>
<ns0:div><ns0:head>Mycorrhizal colonization improves plant growth in S. rebaudiana</ns0:head><ns0:p>In the M-plants, the leaf fresh weight was not different than that in plants treated with 20, 200 and 500 µM KH 2 PO 4 ; the leaf fresh weight only increased significantly at 1000 µM KH 2 PO 4 (Fig. <ns0:ref type='figure' target='#fig_3'>3A</ns0:ref>). In the M+ plants, the leaf fresh weight increased by a factor of 1.74 with 200 µM KH 2 PO 4 in comparison to that in the M-plants (control) at the same phosphate concentration. However, no difference was found between M+ and M-plants at 1000 µM KH 2 PO 4 . The leaves of M+ plants with 200 µM KH 2 PO 4 had a similar fresh weight to those of M+ and M-plants treated with 1000 µM KH 2 PO 4 (Fig. <ns0:ref type='figure' target='#fig_3'>3A</ns0:ref>). In plants treated with 200 µM KH 2 PO 4 , the fresh weight of roots was higher in the roots of M+ plants than in the roots of M-plants. At 1000 µM KH 2 PO 4 , the fresh weight of roots was higher in the M+ plants than in the M-plants (Fig. <ns0:ref type='figure' target='#fig_3'>3B</ns0:ref>). Leaf number and foliar area were determined, and only at 200 µM KH 2 PO 4 , M-plant showed fewer leaves than M+, as well as foliar area (Fig. <ns0:ref type='figure' target='#fig_10'>S3</ns0:ref>), which is consistent with the pattern of fresh weight of leaves. Foliar area, on the other hand, was only significantly lower in M+ plants than M-plants at 500 µM KH 2 PO 4 (Fig. <ns0:ref type='figure' target='#fig_10'>S3</ns0:ref>).</ns0:p></ns0:div>
<ns0:div><ns0:head>The phosphorus and magnesium concentration increases in the leaves of colonized plants with a low phosphate concentration.</ns0:head><ns0:p>The P concentration was four times higher in the leaves of M+ plants treated with 200 µM KH 2 PO 4 than that in M-plants. At 500 and 1000 µM KH 2 PO 4 , the P concentration was similar in the leaves of M+ plants and M-plants (Fig. <ns0:ref type='figure' target='#fig_16'>4A</ns0:ref>). In the leaves of M-plants, the Mg concentration increased significantly at 500 and 1000 µM KH 2 PO 4 , while in the leaves of M+ plants, the Mg concentration increased at 20, 200 and 1000 µM KH 2 PO 4 , but the Mg concentration was lower at 500 µM KH 2 PO 4 (Fig. <ns0:ref type='figure' target='#fig_16'>4B</ns0:ref>). These results suggest that AM symbiosis can stimulate P and Mg accumulation at low KH 2 PO 4 concentrations.</ns0:p></ns0:div>
<ns0:div><ns0:head>Chlorophyll fluorescence and the concentration of photosynthetic pigments improve in colonized plants at a low phosphate concentration</ns0:head><ns0:p>Chlorophyll fluorescence was used as an indicator of photosynthetic performance in the S. rebaudiana plants. In M-plants and at all KH 2 PO 4 concentrations, the Fv/Fm ratio values were less than 0.8. In M+ plants at 20 and 200 µM KH 2 PO 4 , the Fv/Fm ratio values were greater than 0.8, and at 500 and 1000 µM KH 2 PO 4 , the Fv/Fm ratio values diminished to less than 0.8 (Fig. <ns0:ref type='figure' target='#fig_17'>5A</ns0:ref>). The Fv/Fo ratio is indicative of the photochemical efficiency of photosynthesis. In the Mplants at all phosphate concentrations, the values of the Fv/Fo ratio were less than 4.0. In M+ plants at 20 and 200 µM KH 2 PO 4 , the value of the Fv/Fo ratio was greater than 4.0; at 500 and 1000 µM KH 2 PO 4 , this ratio was less than 4.0 (Fig. <ns0:ref type='figure' target='#fig_17'>5B.</ns0:ref>). The values of Fo, Fm and Fv are presented in Fig. <ns0:ref type='figure' target='#fig_11'>S4</ns0:ref>.</ns0:p><ns0:p>The total concentration of chlorophylls and carotenoids did not change in the M-plants at any phosphate concentration. However, in M+ plants at 20 and 200 µM KH 2 PO 4 , the concentration of chlorophylls and carotenoids was higher as compared to the M-plants at the same phosphate concentrations; at 500 and 1000 µM KH 2 PO 4 , the concentration of the two pigments were similar in M-and M+ plants (Figs. <ns0:ref type='figure' target='#fig_17'>5C and 5D</ns0:ref>).</ns0:p></ns0:div>
<ns0:div><ns0:head>Differential expression of the genes for kaurene oxidase and glucosyltransferases in colonized plants with added phosphate</ns0:head><ns0:p>The transcription of the kaurene oxidase (KO) gene was increased 7.5 times in M+ plants at 200 µM KH 2 PO 4 compared with that in M+ plants without added KH 2 PO 4 . The transcription level of the KO gene did not change in M+ plants at 1000 µM KH 2 PO 4 in comparison to M-plants, since the relative expression (2 -∆∆Ct ) was close to 1 (Fig. <ns0:ref type='figure' target='#fig_18'>6A</ns0:ref>). The UGT74G1 gene encoding the protein involved in stevioside synthesis was upregulated in M+ plants at 200 µM KH 2 PO 4 ; the level of relative expression was over 1.5 (Fig. <ns0:ref type='figure' target='#fig_18'>6B</ns0:ref>). The UGT76G1 gene encoding the protein involved in rebaudioside A synthesis was downregulated in M+ plants at the same KH 2 PO 4 concentration, and its relative expression was less than 0.7 (Fig. <ns0:ref type='figure' target='#fig_18'>6C</ns0:ref>). However, in M+ plants at 1000 µM KH 2 PO 4 , the expression of the UGT74G1 gene was downregulated, and the expression of the UGT76G1 gene was unchanged (Fig. <ns0:ref type='figure' target='#fig_18'>6B and 6C</ns0:ref>).</ns0:p></ns0:div>
<ns0:div><ns0:head>SGs differentially accumulate in the leaves of colonized plants with added phosphate</ns0:head><ns0:p>In M+ plants at 200 µM KH 2 PO 4 , the stevioside concentration was 2.8 times higher and the rebaudioside A concentration was 1.61 times lower than those of M-plants (Fig. <ns0:ref type='figure' target='#fig_19'>7A and 7B</ns0:ref>). This metabolite accumulation is consistent with the transcript levels of the corresponding glucosyl transferases (Fig <ns0:ref type='figure' target='#fig_18'>6B and C</ns0:ref>). At 1000 µM KH 2 PO 4 , the accumulation of the two metabolites in M+ and M-plants was not affected (Fig <ns0:ref type='figure' target='#fig_19'>7A and B</ns0:ref>).</ns0:p></ns0:div>
<ns0:div><ns0:head>Discussion</ns0:head><ns0:p>AM symbiosis enhances P uptake in many plants and plays an important role in agricultural and natural environments <ns0:ref type='bibr' target='#b42'>(Smith et al., 2011)</ns0:ref>. Phosphate availability may change the initial signaling for the establishment, maintenance, and functioning of AM symbiosis <ns0:ref type='bibr' target='#b37'>(Schmitz & Harrison, 2014)</ns0:ref>. In this study, the low phosphate concentrations (20 and 200 µM KH 2 PO 4 ) stimulated a high percentage of total mycorrhizal colonization by R. irregularis in S. rebaudiana plants, while the high KH 2 PO 4 concentrations (500 and 1000 µM) decreased the colonization efficiency of R. irregularis in S. rebaudiana plants by approximately 30 and 70%, respectively. AM symbiosis is inhibited by a high concentration of KH 2 PO 4 <ns0:ref type='bibr' target='#b44'>(Smith, Smith & Jakobsen, 2004)</ns0:ref>, and the sensitivity and inhibition percentage depend on the plant species and AM fungus. In Medicago truncatula, fertilization with 1.3 mM phosphate reduced AM symbiosis by 80% compared to that in plants fertilized with 0.13 mM phosphate <ns0:ref type='bibr' target='#b7'>(Bonneau et al., 2013)</ns0:ref>. In Petunia hybrida, phosphate at 100 µM induces high AM symbiosis, while phosphate at 3 mM and higher concentrations completely suppressed this symbiosis <ns0:ref type='bibr' target='#b34'>(Nouri et al., 2014)</ns0:ref>. Therefore, it was important in our study to define this effect of phosphate on the colonization of R. irregularis in S. rebaudiana plants.</ns0:p><ns0:p>The formation of arbuscules in S. rebaudiana roots was inhibited to a higher extent than the total colonization, indicating that arbuscule formation is more sensitive to high phosphate concentrations than other fungal structures, such as hyphae and vesicles. In addition, shortening of intraradical structures was observed in comparison to the arbuscules of roots at low phosphate concentrations. Similarly, changes in arbuscule structures were observed in P. hybrida plants; fertilization with high phosphate concentrations significantly reduced the development of arbuscules and resulted in malformed arbuscules with fewer branches <ns0:ref type='bibr' target='#b10'>(Breuillin et al., 2010)</ns0:ref>.</ns0:p><ns0:p>Previous studies in S. rebaudiana have reported colonization by R. irregularis, but the AM colonization morphology type has not been documented <ns0:ref type='bibr' target='#b48'>(Vafadar, Amooaghaie & Otroshy, 2014;</ns0:ref><ns0:ref type='bibr' target='#b29'>Mandal et al., 2015;</ns0:ref><ns0:ref type='bibr' target='#b46'>Tavarini et al., 2018)</ns0:ref>. In this study, confocal microscopic analysis with specific fluorescent dyes indicated that the AM colonization is classified as an Arum-type morphology. This AM morphotype is highly sensitive to environmental factors, including soil nutrients <ns0:ref type='bibr' target='#b15'>(Dickson et al., 2003)</ns0:ref>. It is also suggested that Arum-type colonization is more efficient than other morphotypes in the acquisition and transference of phosphate from the soil to the plant, resulting in better plant growth <ns0:ref type='bibr' target='#b49'>(Van Aarle et al., 2005)</ns0:ref> The beneficial effects of AM symbiosis on growth promotion and yield have been reported in plants from the Asteraceae family <ns0:ref type='bibr'>(Rapparini, Llusia & Peñuelas, 2008;</ns0:ref><ns0:ref type='bibr' target='#b0'>Aroca et al., 2013)</ns0:ref>. In this study, AM symbiosis with 200 µM KH 2 PO 4 increased colonization and arbuscule formation but also promoted the leaf and root growth in S. rebaudiana plants; the growth of these plants was comparable to the growth of plants fertilized at higher KH 2 PO 4 concentrations (1000 µM). These results suggest the activation of a high-affinity mycorrhiza-specific phosphate transporter during AM symbiosis, as reported for other plant species. In fact, we found the accumulation of transcripts of a putative phosphate transport gene in response to mycorrhizal colonization. This is the first time that a putative phosphate transporter gene has been identified in S. rebaudiana plants, and we anticipate that this gene encoding the specific phosphate transporter may contribute to growth promotion in leaves and roots. These results indicate that colonized plants are able to compensate for the deficiency in P nutrition using the mycorrhizal P uptake pathway. Similar effects have been reported in M. truncatula and P. hybrida plants at different phosphate concentrations <ns0:ref type='bibr' target='#b2'>(Balzergue et al., 2013;</ns0:ref><ns0:ref type='bibr' target='#b34'>Nouri et al., 2014)</ns0:ref>. Shoots and roots were responsive to mycorrhizal colonization by increasing growth at 200 µM KH 2 PO 4 , whereas at 1000 µM KH 2 PO 4 , roots growth was stimulated only in M+ compared to noncolonized controls (M-). This indicates that the symbiosis differentially affects shoot and root growth, and this depends on the phosphate fertilization regime. In plants at 500 µM KH 2 PO 4 , such differences between fresh weights of shoots and roots in M+ in comparison to M-was not observed. This in contrast to 200 µM and 1000 µM KH 2 PO 4 conditions, in which mycorrhiza-specific and direct plant phosphate uptake pathways dominate the phosphate uptake, respectively. It is possible that 500 µM KH 2 PO 4, the direct and the mycorrhizal-specific phosphate uptake pathways, may be interacting in such a way that no increased in growth is manifested. Although this hypothesis needs to be further studied.</ns0:p><ns0:p>Chlorophyll fluorescence is used as an indicator of light assimilation (electron transport) in the reaction centers of chlorophyll and as an indirect measurement of photosynthetic performance in plants subjected to different conditions of stress <ns0:ref type='bibr' target='#b39'>(Schreiber, Bilger & Neubauer, 1994;</ns0:ref><ns0:ref type='bibr' target='#b19'>Harbinson, 2013)</ns0:ref>. Stress conditions may disrupt components of the photosynthetic apparatus and affect photosynthetic performance. In healthy plants, the values of the Fv/Fm ratio are between 0.79 and 0.82; however, under stress conditions, the photosynthetic performance is affected, and these values decrease to below 0.79. Likewise, values of the Fv/Fo ratio between 4-5 correspond to normal values for healthy plants, and values less than 4 are indicative of a loss of photosynthetic efficiency that may result from a stress condition <ns0:ref type='bibr' target='#b1'>(Baker, 2008)</ns0:ref>. In the noncolonized S. rebaudiana plants at low KH 2 PO 4 concentrations, these values were less than 0.79 and 4.0, which indicates that the plants are subjected to stress conditions. However, these values in colonized plants at the same KH 2 PO 4 concentration were significantly higher than those in noncolonized plants. These results suggest that AM symbiosis may compensate for the nutritional stress induced by low KH 2 PO 4 concentrations by restoring the electron transfer through PSII, regulating the functionality of the reaction center, and improving photosynthetic performance in S. rebaudiana plants. The results are consistent with those of other studies performed in plants under stress conditions such as high salinity, high temperatures, and water stress <ns0:ref type='bibr' target='#b41'>(Sheng et al., 2008;</ns0:ref><ns0:ref type='bibr' target='#b58'>Zhu et al., 2011;</ns0:ref><ns0:ref type='bibr' target='#b22'>Hu et al., 2017)</ns0:ref>.</ns0:p><ns0:p>In colonized S. rebaudiana plants at low phosphate concentrations, the concentration of photosynthetic pigments was higher than that in noncolonized plants at the same phosphate concentration. Similarly, other authors have shown that mycorrhizal colonization stimulates the accumulation of these compounds in plants <ns0:ref type='bibr' target='#b14'>(Colla et al., 2008;</ns0:ref><ns0:ref type='bibr' target='#b32'>Nafady & Elgharably, 2018)</ns0:ref>. These studies support the idea that photosynthesis is improved in colonized plants, which ensures carbon fixation and provides a carbon source to the fungal symbiont under low phosphate conditions <ns0:ref type='bibr' target='#b57'>(Zai et al., 2012)</ns0:ref>. Additionally, increased carotenoid concentration in plants is associated with light harvesting, photoprotection, and antioxidant processes, which may contribute to improving plant growth <ns0:ref type='bibr' target='#b52'>(Walter, 2013)</ns0:ref>.</ns0:p><ns0:p>Mg is bound to the central atom in the porphyrin ring of chlorophyll a and b, and 25-60% of the total Mg in plants exists as chlorophyll-bound Mg <ns0:ref type='bibr' target='#b13'>(Chen et al., 2018)</ns0:ref>. A significant increase in the Mg percentage was found in the leaves of S. rebaudiana plants colonized under the low phosphate concentration. This result is consistent with the increase in chlorophyll concentration and has been reported in other plants <ns0:ref type='bibr' target='#b14'>(Colla et al., 2008;</ns0:ref><ns0:ref type='bibr' target='#b48'>Vafadar, Amooaghaie & Otroshy, 2014)</ns0:ref>. Thus, the increase in the concentration of chlorophylls, Mg, and carotenoids improves plant growth and metabolite accumulation, confirming the positive effect of AM symbiosis.</ns0:p><ns0:p>Kaurene oxidase (KO) plays an important role in SG biosynthesis and represents an important branch point that specifically directs the flow of metabolites towards the biosynthesis of steviol (the central backbone of SGs). In fact, SGs are synthesized from steviol glycosylation, where the conjugation of glucose to steviol is carried out by UDP-glycosyltransferases (UGTs); for example, UGT74G1 is known to convert steviolbioside to stevioside, and UGT76G1 adds the final glucose required to produce rebaudioside A <ns0:ref type='bibr' target='#b9'>(Brandle and Telmer, 2007)</ns0:ref>.</ns0:p><ns0:p>The KO and UGT74G1 gene transcription levels were upregulated in colonized plants compared to those in noncolonized plants at 200 µM KH 2 PO 4 , which was consistent with the observed stevioside accumulation trends. In contrast, in noncolonized plants at 200 µM KH 2 PO 4 , the expression of the UGT76G1 gene and the corresponding accumulation of rebaudioside A were higher than those in colonized plants. These results indicate that plant colonization with AM stimulates stevioside accumulation, while in noncolonized plants, the accumulation of rebaudioside A is stimulated. In addition, the high phosphate concentration (KH 2 PO 4 1000 µM) in colonized plants downregulated the expression of the UGT74G1 gene and did not change the expression of the KO or UGT76G1 genes at the same phosphate concentration. These results support the idea that SG synthesis is sensitive to the phosphate concentration and mycorrhizal interactions and suggest that AM symbiosis causes an increase in stevioside accumulation by modulating the expression of the KO and UGT74G1 genes, which convert steviolbioside to stevioside. The downregulation of the UGT76G1 gene, which encodes a protein involved in transforming stevioside to rebaudioside A, may also contribute to the accumulation of stevioside. In contrast, the accumulation of rebaudioside A, which is the SG with the highest sweetening power, would be favored in noncolonized plants. This information is relevant from a biotechnological perspective. The results demonstrate the effect of the phosphate concentration on the mycorrhizal interaction between S. rebaudiana and R. irregularis as well as the effect on SG concentration after the modulation of the expression of key biosynthetic genes. Manuscript to be reviewed colonization structure was a typical Arum-type mycorrhiza, and a mycorrhiza-specific phosphate transporter was identified. Colonization at low phosphate concentrations improved plant growth, the chlorophyll and carotenoid concentrations, and photochemical performance. The low phosphate concentration improved mycorrhizal colonization and modulated the stevioside and rebaudioside A concentration by regulating the transcription of genes that encode kaurene oxidase and glucosyltransferases, which are involved in the synthesis of these compounds in S. rebaudiana. This knowledge is important for generating biotechnological strategies that involve manipulating the concentration of stevioside or rebaudioside A by controlling the colonization status and the phosphate concentration of S. rebaudiana plants. The alignment that was used to deduce the amino acid sequence and topology of the 12 SrPT TM domains (TM1-TM12) was predicted using AM-specific phosphate transporters from M. truncatula (MtPT4), A. sinicus (AsPT4), S. tuberosum (StPT4), L. esculentum (LePT4), and P. hybrida (PhPT4). The deduced amino acid sequence was aligned through MULTIALIN, and SrPT was predicted according to <ns0:ref type='bibr' target='#b56'>Yadav et al. (2010)</ns0:ref>. Manuscript to be reviewed Manuscript to be reviewed Manuscript to be reviewed Manuscript to be reviewed </ns0:p></ns0:div>
<ns0:div><ns0:head>Conclusions</ns0:head><ns0:note type='other'>Figure 2</ns0:note><ns0:note type='other'>Figure 5</ns0:note><ns0:note type='other'>Figure 6</ns0:note><ns0:note type='other'>Figure 7</ns0:note></ns0:div><ns0:figure xml:id='fig_0'><ns0:head /><ns0:label /><ns0:figDesc>AM symbiosis between S. rebaudiana and R. irregularis is affected by phosphate concentrations; a low phosphate concentration induces a high percentage of colonization. The morphology of the PeerJ reviewing PDF | (2020:07:50625:2:0:NEW 8 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_1'><ns0:head>Figure 1 .</ns0:head><ns0:label>1</ns0:label><ns0:figDesc>Figure 1. Quantification of mycorrhizal colonization in S. rebaudiana roots. Percentages of total mycorrhizal colonization (dotted bars) and arbuscule colonization (diagonally hatched bars) by R. irregularis in S. rebaudiana plants fertilized with Hoagland nutrient solution at different KH 2 PO 4 concentrations. Bars represent the mean ± standard deviation (SD) of six replicates. Different letters indicate significant differences according to Tukey´s test (P<0.05). Capital letters were used for total colonization, and lowercase letters were used for arbuscular colonization.</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_2'><ns0:head>Figure 2 .</ns0:head><ns0:label>2</ns0:label><ns0:figDesc>Figure 2. Light and confocal microscopic analysis of R. irregularis colonization structures in S. rebaudiana roots. Mycorrhiza-colonized roots of S. rebaudiana plants fertilized with 20 (A), 200 (B), 500 (C) and 1000 µM KH 2 PO 4 (D) after trypan blue staining depicting the arbuscules, vesicles and intraradical hyphae. Mycorrhiza-colonized roots with 200 µM KH 2 PO 4 were also treated with propidium iodide to label the cell wall (E) and with WGA-Alexa Fluor 488 conjugate to stain the fine details of the intraradical hyphae and arbuscules (F and G). The merged image showing both red and green fluorescence is presented in panel G. Vesicles: V; intraradical hyphae: ih; arbuscules: *.</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_3'><ns0:head>Figure 3 .</ns0:head><ns0:label>3</ns0:label><ns0:figDesc>Figure 3. Effect of AM symbiosis and KH 2 PO 4 concentrations on the fresh weight of S. rebaudiana roots and leaves. Fresh weight of leaves (A) and roots (B) of mycorrhiza-colonized (M+) and noncolonized (M-) S. rebaudiana plants fertilized with Hoagland solution with different KH 2 PO 4 concentrations. Bars represent the mean ± standard deviation (SD) of six replicates. Different letters indicate significant differences according to Tukey's test (P<0.05).</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_4'><ns0:head>Figure 4 .</ns0:head><ns0:label>4</ns0:label><ns0:figDesc>Figure 4. Quantification of P and Mg under AM symbiosis and different KH 2 PO 4 concentrations. Phosphorus (A) and magnesium (B) concentration in the leaves of mycorrhizacolonized (M+) and noncolonized (M-) S. rebaudiana plants fertilized with Hoagland solution at different KH 2 PO 4 concentrations. Bars represent the mean ± standard deviation (SD) of six replicates. Different letters indicate significant differences according to Tukey´s test (P<0.05).</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_5'><ns0:head>Figure 5 .</ns0:head><ns0:label>5</ns0:label><ns0:figDesc>Figure 5. Effects of AM symbiosis and KH 2 PO 4 concentrations on photosynthetic performance in S. rebaudiana. Maximal photochemical efficiency (Fv/Fm) (a), potential photochemical efficiency (Fv/Fo) (b), and total chlorophylls (c), and carotenoids concentration (d) in leaves of mycorrhiza-colonized (M+) and noncolonized (M-) S. rebaudiana plants fertilized with Hoagland solution at different KH 2 PO 4 concentrations. Fv/Fm and Fv/Fo ratios were obtained from chlorophyll fluorescence measurements. Bars represent the mean ± standard deviation (SD) of six replicates. Different letters indicate significant differences according to Tukey´s test (P<0.05).</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_6'><ns0:head>Figure 6 .</ns0:head><ns0:label>6</ns0:label><ns0:figDesc>Figure 6. Differential transcript accumulation of three key genes from the SG biosynthetic pathway under AM symbiosis and different KH 2 PO 4 concentrations. KO (A), UGT74G1 (B) and UGT76G1 (C) relative expression levels in S. rebaudiana plants that were mycorrhizacolonized and fertilized with Hoagland solution at 200 and 1000 µM KH 2 PO 4 . For each condition, the transcript levels of the KO, UGT74G1 and UGT76G1 genes were first normalized against SrGAPDH and then normalized against the gene expression of S. rebaudiana without inoculation. The analysis of the relative gene expression data used the 2 -ΔΔCT method. Bars represent the mean ± standard deviation (SD) of three biological and three technical replicates. Different letters indicate significant differences according to Student´s t test (P<0.05).</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_7'><ns0:head>Figure 7 .</ns0:head><ns0:label>7</ns0:label><ns0:figDesc>Figure 7. Quantification of SG concentration under AM symbiosis and different KH 2 PO 4 concentrations. Stevioside (A) and rebaudioside A (B) concentration in mycorrhiza-colonized (M+) and noncolonized (M-) S. rebaudiana plants fertilized with Hoagland solution at 200 and 1000 µM KH 2 PO 4 . Bars represent the mean ± standard deviation (SD) of six replicates. Different letters indicate significant differences according to Tukey´s test (P<0.05).</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_8'><ns0:head>Figure S1 .</ns0:head><ns0:label>S1</ns0:label><ns0:figDesc>Figure S1. Protein sequence structure of the mycorrhiza-specific phosphate transporter (SrPT) from S. rebaudiana. The SrPT protein contains 391 amino acids and 9 TM domains. The alignment that was used to deduce the amino acid sequence and topology of the 12 SrPT TM domains (TM1-TM12) was predicted using AM-specific phosphate transporters from M. truncatula (MtPT4), A. sinicus (AsPT4), S. tuberosum (StPT4), L. esculentum (LePT4), and P. hybrida (PhPT4). The deduced amino acid sequence was aligned through MULTIALIN, and SrPT was predicted according to<ns0:ref type='bibr' target='#b56'>Yadav et al. (2010)</ns0:ref>.</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_9'><ns0:head>Figure S2 .</ns0:head><ns0:label>S2</ns0:label><ns0:figDesc>Figure S2. Expression analysis of the SrPT gene in roots of individual M-and M+ S. rebaudiana plants. Lanes 1-6, replicates of noncolonized (M-) and R. irregularis-colonized plants (M+). SrGADPH was used as a reference gene.</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_10'><ns0:head>Figure S3 .</ns0:head><ns0:label>S3</ns0:label><ns0:figDesc>Figure S3. Effects of AM symbiosis and KH 2 PO 4 concentration on aerial part development in S. rebaudiana. Leaves number (A) and foliar area in leaves of mycorrhiza-colonized (M+) and noncolonized (M-) S. rebaudiana plants fertilized with Hoagland solution using different KH 2 PO 4 concentrations Bars represent the mean ± standard deviation (SD) of six replicates. Different letters indicate significant differences according to Tukey´s test (P<0.05).</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_11'><ns0:head>Figure S4 .</ns0:head><ns0:label>S4</ns0:label><ns0:figDesc>Figure S4. Effects of AM symbiosis and KH 2 PO 4 concentration on chlorophyll fluorescence in S. rebaudiana. Primary fluorescence (Fo) (A), maximal fluorescence (Fm) (B), and variable fluorescence (Fv) (C) in leaves of mycorrhiza-colonized (M+) and noncolonized (M-) S. rebaudiana plants fertilized with Hoagland solution using different KH 2 PO 4 concentrations. Bars represent the mean ± standard deviation (SD) of six replicates. Different letters indicate significant differences according to Tukey´s test (P<0.05).</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_12'><ns0:head>Figure 1 Figure 1 .</ns0:head><ns0:label>11</ns0:label><ns0:figDesc>Figure 1</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_13'><ns0:head>Figure 2 .</ns0:head><ns0:label>2</ns0:label><ns0:figDesc>Figure 2. Light and confocal microscopic analysis of R. irregularis colonization structures in S. rebaudiana roots.</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_14'><ns0:head>Figure 3 Figure 3 .</ns0:head><ns0:label>33</ns0:label><ns0:figDesc>Figure 3</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_15'><ns0:head>Figure 4</ns0:head><ns0:label>4</ns0:label><ns0:figDesc>Figure 4</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_16'><ns0:head>Figure 4 .</ns0:head><ns0:label>4</ns0:label><ns0:figDesc>Figure 4. Quantification of P and Mg under AM symbiosis and different KH 2 PO 4 concentrations.</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_17'><ns0:head>Figure 5 .</ns0:head><ns0:label>5</ns0:label><ns0:figDesc>Figure 5. Effects of AM symbiosis and KH 2 PO 4 concentrations on photosynthetic performance in S. rebaudiana.</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_18'><ns0:head>Figure 6 .</ns0:head><ns0:label>6</ns0:label><ns0:figDesc>Figure 6. Differential transcript accumulation of three key genes from the SG biosynthetic pathway under AM symbiosis and different KH 2 PO 4 concentrations.</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_19'><ns0:head>Figure 7 .</ns0:head><ns0:label>7</ns0:label><ns0:figDesc>Figure 7. Quantification of SG concentration under AM symbiosis and different KH 2 PO 4 concentrations.</ns0:figDesc></ns0:figure>
<ns0:note place='foot' n='910'>PeerJ reviewing PDF | (2020:07:50625:2:0:NEW 8 Sep 2020)Manuscript to be reviewed</ns0:note>
</ns0:body>
" | "INSTITUTO POLITÉCNICO NACIONAL
CENTRO DE DESARROLLO DE PRODUCTOS BIÓTICOS
DEPARTAMENTO DE BIOTECNOLOGÍA
September 08th, 2020
Ana I. Ribeiro-Barros
Editor, PeerJ
Ref. 50625
Dear Editor
We thank the editor and reviewer for their generous suggestions on the manuscript “Photosynthetic performance and stevioside concentration are improved by the arbuscular mycorrhizal symbiosis in Stevia rebaudiana under different phosphate concentrations”.
We incorporated the suggestions of the reviewer in the new version of MS.
Reviewer 1 indicates that
Basic reporting
See below
Experimental design
See below
Validity of the findings
See below
Comments for the Author
New report on “Photosynthetic performance and stevioside concentration are improved by the arbuscular mycorrhizal symbiosis in Stevia rebaudiana under different phosphate concentrations”
The authors have taken on board most of my suggestions, and I commend them for improving the manuscript.
A few minor issues to be attended to during editing:
There are still instances of content where concentration should be used, on line 109, 196, 386, 392 and 563.
We have changed the words “content” by “concentration”, in lines 109, 196, 386, 392 and 563
Line 154 to read ..only one of them are shown.
We agree with the reviewer comment and we have taken his suggestion, line 154
Line 174 to read ..recorded. To minimize the..
We agree with the reviewer comment and we have taken his suggestion, line 174
Line 177 do you mean ..positioned on each side of all the nodes of the..
We revised the observation and we have changed the sentence “leaves positioned in all the nodes of the plant” by “leaves positioned on each side of all the nodes of the plant”, line 177
Line 302 mispelled described
We have addressed this mistake, including the one in line 302.
Line 315 no comma after data
We have addressed this mistake, 86
Line 316 to read .. were checked for normal distributions.
We agree with the reviewer comment and we have taken his suggestion, line 316
Line 376 to read Leaf number and foliar area were..
We have taken his suggestion, line 376
Line 377 to read fewer leaves than M+,
We agree and we changed the sentence “M- plant showed lower number of leaves than M+ ” by M- plant showed fewer leaves lower number of leaves than M+, line 377
Line 378 significantly lower than what..??
We appreciate the reviewer's observation. In response we have added a possible explanation on lines 378. Which is as follow:
Foliar area, on the other hand, was only significantly lower in M+ plants than M- plants at 500 µM KH2PO4
Reviewer 2 indicates that
Basic reporting
The revised manuscript is much better than the original one.
Experimental design
More information is given which makes it better understandable.
Validity of the findings
OK, no problems.
Comments for the Author
No further comments
Again, we thank and appreciate all the observations raised by the reviewer that have certainly improved our manuscript. We hope that this improved version of the manuscript will be suitable for publication in PeerJ.
We look forward to hearing from you shortly.
Yours sincerely
Dr. Mario Rodríguez Monroy
Departamento de Biotecnología
Investigador del Centro de Desarrollo de Productos Bióticos
Instituto Politécnico Nacional
On behalf of all authors.
" | Here is a paper. Please give your review comments after reading it. |
9,821 | "<ns0:abstract xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:p>Maternal thyroid hormones (THs) are known to be crucial in embryonic development in humans, but their influence on other, especially wild, animals remains poorly understood.</ns0:p><ns0:p>So far, the studies that experimentally investigated the consequences of maternal THs focused on short-term effects, while early organisational effects with long-term consequences, as shown for other prenatal hormones, could also be expected. In this study, we aimed at investigating both the short-and long-term effects of prenatal THs in a bird species, the Japanese quail Coturnix japonica. We experimentally elevated yolk TH content (the prohormone T 4 , and its active metabolite T 3 , as well as a combination of both hormones). We analysed hatching success, embryonic development, offspring growth and oxidative stress as well as their potential organisational effects on reproduction, moult, and oxidative stress in adulthood. We found that eggs injected with T 4 had a higher hatching success compared with control eggs, suggesting conversion of T 4 into T 3 by the embryo. We detected no evidence for other short-term or long-term effects of yolk THs.</ns0:p><ns0:p>These results suggest that yolk thyroid hormones are important in the embryonic stage of precocial birds, but other short-and long-term consequences remain unclear. Research on maternal thyroid hormones will greatly benefit from studies investigating how embryos use and respond to this maternal signalling. Long-term studies on prenatal THs in other taxa in the wild are needed for a better understanding of this hormone-mediated maternal pathway.</ns0:p></ns0:div>
</ns0:abstract>
<ns0:body xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:head>Introduction</ns0:head><ns0:p>Maternal effects represent all the non-genetic influences of a mother on her offspring and have received increasing attention in evolutionary and behavioural ecology. Through maternal effects, mothers can influence the fitness of their progeny by adapting their phenotype to expected environmental conditions ('adaptive maternal effects' in <ns0:ref type='bibr' target='#b8'>Marshall and Uller, 2007;</ns0:ref><ns0:ref type='bibr'>Mousseau and Fox, 1998)</ns0:ref>, and this view is now also incorporated in the human disease literature <ns0:ref type='bibr'>(Gluckman, Hanson & Spencer, 2005)</ns0:ref>. Maternal hormones transferred to the offspring can mediate important maternal effects. Historically, research on maternal hormones has mostly focused on steroid hormones <ns0:ref type='bibr'>(Groothuis et al., 2005;</ns0:ref><ns0:ref type='bibr'>von Engelhardt & Groothuis, 2011)</ns0:ref>. While research on maternal thyroid hormones has emerged between the 80s and the 90s in several taxa <ns0:ref type='bibr'>(mammals, Morreale De Escobar et al., 1985;</ns0:ref><ns0:ref type='bibr'>fish, Brown et al., 1988;</ns0:ref><ns0:ref type='bibr'>birds, Wilson & McNabb, 1997)</ns0:ref>, these hormones are still underrepresented in the literature on hormone-mediated maternal effects (reviewed in <ns0:ref type='bibr'>Ruuskanen & Hsu, 2018)</ns0:ref>.</ns0:p><ns0:p>Thyroid hormones (THs) are metabolic hormones produced by the thyroid gland and are present in two main forms: the prohormone thyroxine (T 4 ) and the biologically active form triiodothyronine (T 3 ). THs play a crucial role in various aspects of an individual's life, e.g.</ns0:p><ns0:p>development, metabolism and reproduction, across vertebrates, including humans (Morreale de Escobar, Obregon & Escobar del <ns0:ref type='bibr' target='#b18'>Rey, 2004;</ns0:ref><ns0:ref type='bibr'>Krassas, Poppe & Glinoer, 2010)</ns0:ref>. In humans, physiological variation of maternal THs (i.e. no clinical symptoms in both mothers and foetuses) is found to be associated with infant birth weight and IQ in older children <ns0:ref type='bibr' target='#b14'>(Medici et al., 2013;</ns0:ref><ns0:ref type='bibr'>Korevaar et al., 2016)</ns0:ref>. In other vertebrates, THs in general play a role in brain development and neuronal turnover (mammals, Morreale de Escobar, Obregon & Escobar del <ns0:ref type='bibr' target='#b18'>Rey, 2004;</ns0:ref><ns0:ref type='bibr'>birds, McNabb, 2007)</ns0:ref>. THs control the endothermic heat production, and are therefore important in</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:03:47322:1:1:NEW 22 Jun 2020)</ns0:p><ns0:p>Manuscript to be reviewed thermoregulation in homeothermic species <ns0:ref type='bibr'>(mammals, Danforth & Burger, 1984;</ns0:ref><ns0:ref type='bibr'>birds, McNabb & Darras, 2015)</ns0:ref>. THs can act, in concert with other hormonal axes, as mediators of life stage transitions across vertebrates (reviewed in <ns0:ref type='bibr'>Watanabe et al., 2016)</ns0:ref>. The interaction between thyroid hormones and corticosteroids on amphibian metamorphosis is a well-known example of such effect on life stage transition <ns0:ref type='bibr'>(Kikuyama et al., 1993;</ns0:ref><ns0:ref type='bibr'>Wada, 2008)</ns0:ref>. THs are involved in gonadal development, and hyperthyroidism tends to accelerate maturation <ns0:ref type='bibr'>(Holsberger & Cooke, 2005)</ns0:ref>, and coordinate the transition between reproduction and moult <ns0:ref type='bibr' target='#b11'>(McNabb and Darras, 2015)</ns0:ref>.</ns0:p><ns0:p>Administration of exogenous THs is known to stop egg laying and induce moult in birds <ns0:ref type='bibr'>(Sekimoto et al., 1987;</ns0:ref><ns0:ref type='bibr'>Keshavarz & Quimby, 2002)</ns0:ref>. THs are also involved in photoperiodic control in seasonal breeding <ns0:ref type='bibr'>(Dardente, Hazlerigg & Ebling, 2014)</ns0:ref>. For example, thyroidectomised starlings transferred to long photoperiods became insensitive to future changes in photoperiod, and short photoperiod did not induce gonadal regression <ns0:ref type='bibr'>(Dawson, 1993)</ns0:ref>. While there has been recent research effort on the influence of maternal THs on offspring traits across vertebrate taxa, there are still substantial gaps in our knowledge. Manipulating yolk hormones within the natural range of a species is necessary to better understand the role of maternal THs in an eco-evolutionary context. In humans, studies have essentially looked at the consequences of clinical hyper-or hypothyroidism (but see <ns0:ref type='bibr' target='#b14'>Medici et al., 2013)</ns0:ref>. Research in fish has applied supra-physiological doses for aquaculture purposes <ns0:ref type='bibr'>(Brown et al., 2014)</ns0:ref>. However, these studies do not give information on how variations within the natural range of the species would shape offspring phenotype and affect its fitness, in turn influencing evolution. While recent literature on birds has shown that even physiological variations of prenatal THs can have phenotypic consequences <ns0:ref type='bibr'>(Ruuskanen et al., 2016;</ns0:ref><ns0:ref type='bibr'>Hsu et al., 2017</ns0:ref><ns0:ref type='bibr'>Hsu et al., , 2019;;</ns0:ref><ns0:ref type='bibr'>Sarraude et al., 2020)</ns0:ref>,</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:03:47322:1:1:NEW 22 Jun 2020)</ns0:p><ns0:p>Manuscript to be reviewed this view is still underrepresented in maternal THs research. Besides, research on maternal thyroid hormones up to date has mainly investigated the short-term effects of prenatal THs on developing fish <ns0:ref type='bibr'>(Brown et al., 1988;</ns0:ref><ns0:ref type='bibr'>Raine et al., 2004)</ns0:ref> and amphibians <ns0:ref type='bibr'>(Duarte-Guterman et al., 2010;</ns0:ref><ns0:ref type='bibr'>Fini et al., 2012)</ns0:ref> and pre-fledging birds <ns0:ref type='bibr'>(Ruuskanen et al., 2016;</ns0:ref><ns0:ref type='bibr'>Hsu et al., 2017</ns0:ref><ns0:ref type='bibr'>Hsu et al., , 2019;;</ns0:ref><ns0:ref type='bibr'>Sarraude et al., 2020)</ns0:ref>. So far, only a study on rock pigeons has looked at the influence of yolk THs on post-fledging survival and found no effect <ns0:ref type='bibr'>(Hsu et al., 2017)</ns0:ref>. None of these studies in any taxa investigated the potential organisational effects of prenatal THs on life-history stage transitions in adult life. Early exposure to elevated THs may affect the hypothalamic-pituitary-thyroid (HPT) axis (humans and mice: <ns0:ref type='bibr'>Alonso et al., 2007;</ns0:ref><ns0:ref type='bibr'>Srichomkwun et al., 2017;</ns0:ref><ns0:ref type='bibr' target='#b3'>Anselmo et al., 2019)</ns0:ref>, via epigenetic modifications for example, such as those induced by adverse early life conditions <ns0:ref type='bibr'>(Jimeno et al., 2019)</ns0:ref> or yolk testosterone <ns0:ref type='bibr' target='#b6'>(Bentz, Becker & Navara, 2016)</ns0:ref>.</ns0:p><ns0:p>Oviparous species, such as birds, are suitable models for studying the role of maternal hormones on the progeny because embryos develop in eggs outside mothers' body. The content of an egg cannot be adjusted by the mother after laying, which facilitates the quantification of hormones transmitted by the mothers. In addition, the measurement and experimental manipulation of maternal hormones in the egg after it has been laid is not confounded by maternal physiology. These advantages combined with their well-known ecology and evolution, birds have become the most extensively studied taxa in research on the function of maternal hormones <ns0:ref type='bibr'>(Groothuis et al. 2019)</ns0:ref>.</ns0:p><ns0:p>Previous studies on prenatal THs in birds focused only on altricial species (great tits, <ns0:ref type='bibr'>Ruuskanen et al., 2016;</ns0:ref><ns0:ref type='bibr'>rock pigeons, Hsu et al., 2017;</ns0:ref><ns0:ref type='bibr'>collared flycatchers, Hsu et al., 2019</ns0:ref><ns0:ref type='bibr'>, pied flycatchers, Sarraude et al., 2020)</ns0:ref>. Embryonic development differs substantially between</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:03:47322:1:1:NEW 22 Jun 2020)</ns0:p><ns0:p>Manuscript to be reviewed altricial and precocial species. In the latter, embryonic development is more advanced than in the former. In addition, precocial embryos start their endogenous production of TH around midincubation, considerably earlier than their altricial counterparts, in which endogenous TH production begins only after hatching <ns0:ref type='bibr' target='#b13'>(McNabb, Scanes & Zeman, 1998)</ns0:ref>. While embryonic hormone production may limit the influence of maternal hormones, prenatal hormones have been shown to affect chick endogenous production and sensitivity <ns0:ref type='bibr'>(Pfannkuche et al., 2011)</ns0:ref>. Overall, exposure to maternal hormones may be of different importance in these two developmental modes.</ns0:p><ns0:p>Previous research has studied the effects of T 3 only <ns0:ref type='bibr'>(Raine et al., 2004;</ns0:ref><ns0:ref type='bibr'>Walpita et al., 2007;</ns0:ref><ns0:ref type='bibr'>Fini et al., 2012)</ns0:ref> or a combination of T 3 and T 4 <ns0:ref type='bibr'>(Ruuskanen et al., 2016;</ns0:ref><ns0:ref type='bibr'>Hsu et al., 2017</ns0:ref><ns0:ref type='bibr'>Hsu et al., , 2019;;</ns0:ref><ns0:ref type='bibr'>Sarraude et al., 2020)</ns0:ref>, where the effects of the two forms cannot be separated. Although T 3 is the biologically active form that binds to the receptors, both T 3 and T 4 are deposited in eggs <ns0:ref type='bibr'>(Prati et al., 1992)</ns0:ref> and T 4 may be converted to T 3 via deiodinases from the mother or the developing embryo <ns0:ref type='bibr'>(Van Herck et al., 2015)</ns0:ref> or may still exert non-genomic actions (reviewed in <ns0:ref type='bibr'>Davis et al., 2016)</ns0:ref>. Manipulating yolk T 4 and T 3 independently would help understanding the relative contribution of these two hormones.</ns0:p><ns0:p>In this study, we aimed at assessing the effects of maternal THs on development and lifehistory traits in a precocial bird species, the Japanese quail (Coturnix japonica). Japanese quails are easy to maintain in captivity, and their short generation time makes it a good model to investigate the long-term effects of maternal hormones. Rearing birds in captivity allowed us to apply a powerful within-female experimental design (i.e. knowing which chick hatched from which egg which is not feasible in field studies), thus reducing the effect of random variation among females. Moreover, studying the role of natural variation of prenatal THs in precocial</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:03:47322:1:1:NEW 22 Jun 2020)</ns0:p><ns0:p>Manuscript to be reviewed species may give additional information to previous studies in altricial species. Finally, Japanese quail is a commonly used model in maternal hormone research with substantial literature available (e.g. <ns0:ref type='bibr' target='#b10'>McNabb, Blackman & Cherry, 1985;</ns0:ref><ns0:ref type='bibr' target='#b12'>McNabb, Dicken & Cherry, 1985;</ns0:ref><ns0:ref type='bibr'>Wilson & McNabb, 1997;</ns0:ref><ns0:ref type='bibr'>Okuliarova et al., 2011)</ns0:ref>. We manipulated eggs received either an injection of T 4 or T 3 separately, a combination of both hormones, or a control injection of the vehicle saline solution. This design allowed us to explore the effects of T 4 and T 3 separately, which has not been done in previous studies. The elevation in yolk THs remained within the natural range of this species, a crucial condition to obtain relevant results for an eco-evolutionary context. We measured traits known to be influenced by circulating and yolk THs: hatching success, age at embryonic mortality, growth, transition between life-history stages (i.e., reproductive state and moult) and oxidative stress. First, we hypothesise that elevation of yolk THs in Japanese quails positively affects hatching success, as found in two studies on collared flycatchers and rock pigeons <ns0:ref type='bibr'>(Hsu et al., 2017;</ns0:ref><ns0:ref type='bibr'>Hsu et al., 2019</ns0:ref><ns0:ref type='bibr'>, but see Ruuskanen et al., 2016</ns0:ref><ns0:ref type='bibr'>and Sarraude et al., 2020)</ns0:ref>. Second, elevation of yolk THs is predicted to increase the proportion of well-developed embryos before hatching, as found in rock pigeons <ns0:ref type='bibr'>(Hsu et al., 2017)</ns0:ref>. We therefore looked at the age at mortality in unhatched eggs. Third, we expect elevated yolk THs to affect chick growth (in body mass, tarsus and wing length) either positively <ns0:ref type='bibr'>(Wilson & McNabb, 1997;</ns0:ref><ns0:ref type='bibr'>Hsu et al., 2019;</ns0:ref><ns0:ref type='bibr'>weak effect in Sarraude et al., 2020</ns0:ref><ns0:ref type='bibr'>), negatively (Hsu et al., 2017)</ns0:ref>, or in a sex-specific manner <ns0:ref type='bibr'>(Ruuskanen et al., 2016)</ns0:ref>. Prenatal THs may exert most of their effects in the offspring early life;</ns0:p><ns0:p>this is why we separately tested both posthatch morphological traits and the growth curve.</ns0:p><ns0:p>Similarly, we also independently analysed morphological traits at adulthood, as these traits may affect the fitness of an individual. For example, small adult females may lay smaller eggs and larger males may be more dominant. Fourth, we predict that yolk THs will have organisational Manuscript to be reviewed effects on life-history stage transitions; that is, age at sexual maturity and male gonadal regression (using cloacal gland size as a proxy), and moult when birds are exposed to short photoperiod. Based on the literature mentioned above we expect elevated yolk THs to advance the timing of puberty, gonadal regression and moult. The rate of moult should also be influenced, with birds receiving experimental TH elevation moulting faster. Previous studies have reported that gravid female three-spined sticklebacks (Gasterosteus aculeatus) exposed to predatory cues produced eggs with higher corticosterone <ns0:ref type='bibr'>(Giesing et al., 2011)</ns0:ref>, disturbed embryonic transcriptome <ns0:ref type='bibr' target='#b16'>(Mommer & Bell, 2014)</ns0:ref>, offspring with altered anti-predator behaviour <ns0:ref type='bibr'>(Giesing et al., 2011)</ns0:ref> and modified cortisol response in adulthood <ns0:ref type='bibr' target='#b15'>(Mommer & Bell, 2013)</ns0:ref>. We may therefore expect elevated yolk THs to similarly induce long-term behavioural changes in response to environmental cues (i.e. photoperiod), via organising effects during the embryonic development. We also explored the effects of yolk THs on reproductive investment in females, another important fitness aspect. Finally, yolk THs may increase oxidative stress due to their stimulating effects on metabolism.</ns0:p></ns0:div>
<ns0:div><ns0:head>Material and Methods</ns0:head></ns0:div>
<ns0:div><ns0:head>Parental generation and egg collection</ns0:head><ns0:p>The parental generation was composed of adult Japanese quails provided by Finnish private local breeders that were kept in two acclimated rooms. Twenty-four breeding pairs were formed by pairing birds from different breeders. Individuals were identified using metal leg bands. The floor was covered with 3-5cm sawdust bedding. A hiding place, sand and calcium grit were provided. Each pair was housed in indoor aviary dividing into pens of 1 m² floor area. The temperature was set to 20°C with a 16L:8D photoperiod (light from 06.00 to 22.00). Food</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:03:47322:1:1:NEW 22 Jun 2020)</ns0:p><ns0:p>Manuscript to be reviewed (Poultry complete feed, 'Kanan Paras Täysrehu', Hankkija, Finland) was provided ad libitum and water was changed every day.</ns0:p><ns0:p>Pairs were monitored every morning to collect eggs for 7 days. Eggs were individually marked (non-toxic marker), weighed and stored in a climate-controlled chamber at 15°C and 50% relative humidity. On the last day of collection, a total of 4 to 8 eggs per pair were injected with a solution (see next section).</ns0:p></ns0:div>
<ns0:div><ns0:head>Preparation of the solution, injection procedure and incubation</ns0:head><ns0:p>The preparation of hormone solution and the procedure of injection were based on previous studies <ns0:ref type='bibr'>(Ruuskanen et al., 2016;</ns0:ref><ns0:ref type='bibr'>Hsu et al., 2017)</ns0:ref>. In brief, crystal T 4 (L-thyroxine, ≥ 98% HPCL, CAS number 51-48-9, Sigma-Aldrich) and T 3 (3,3',5-triiodo-L-thyronine, > 95% HPCL, CAS number 6893-02-3, Sigma-Aldrich) were first dissolved in 0.1M NaOH and then diluted in 0.9% NaCl. The injection of thyroid hormones resulted in an increase of two standard deviations (T 4 = 8.9 ng/egg, equivalent to 1.79 pg/mg yolk; T 3 = 4.7 ng/egg, equivalent to 1.24 pg/mg yolk), a recommended procedure for hormone manipulation within the natural range <ns0:ref type='bibr'>(Ruuskanen et al., 2016;</ns0:ref><ns0:ref type='bibr'>Hsu et al., 2017;</ns0:ref><ns0:ref type='bibr'>Podmokła, Drobniak & Rutkowska, 2018)</ns0:ref>. The control solution (CO) was a saline solution (0.9% NaCl). The concentrations of the hormone solutions were based on previous measurements of 15 eggs from the same flock (content per egg (SD) T 4 = 15.3 (4.4) ng, T 3 = 7.6 (2.3) ng; concentrations (SD), T 4 = 4.20 (0.89) pg/mg yolk, T 3 = 2.10 (0.62) pg/mg yolk). Hormone injections were performed at room temperature in a laminar hood. Eggs were put sideways, allowing yolks to float up to the middle position. Before injection, the shell was disinfected with a cotton pad dipped in 70% EtOH. We used a 27G needle (BD Microlance ™)</ns0:p><ns0:p>to pierce the eggshell and then used a 0.3 ml syringe to deliver 50 µl of the respective hormone</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:03:47322:1:1:NEW 22 Jun 2020)</ns0:p><ns0:p>Manuscript to be reviewed solution or control. After injection, the hole was sealed with a sterile plaster (OPSITE Flexigrid, Smith&Nephew). In total, 158 eggs were injected and divided as follows over the treatments: T 3 treatment (N = 39); T 4 treatment (N = 39); T 3 +T 4 treatment (N = 40); and control, CO (N = 40). To balance the genetic background of the parents and the effect of storage, each egg laid by the same female was sequentially assigned to a different treatment and the order of treatments was rotated among females. After injection, eggs were placed in an incubator at 37.8°C and 55% relative humidity.</ns0:p><ns0:p>Until day 14 after starting incubation, eggs were automatically tilted every hour by 90°. On day 14, tilting was halted and each egg was transferred to an individual container to monitor which chick hatched from which egg. On day 16 after injection, (normal incubation time = 17 days), the temperature was set to 37.5°C and the relative humidity to 70%. Eggs were checked for hatching every 4 hours from day 16 onwards. Four days after the first egg hatched, all unhatched eggs were stored in a freezer and dissected to determine the presence of an embryo. The age of developed embryos was assessed according to <ns0:ref type='bibr' target='#b0'>Ainsworth et al. (2010)</ns0:ref>.</ns0:p></ns0:div>
<ns0:div><ns0:head>Rearing conditions of the experimental birds</ns0:head><ns0:p>In total, 66 chicks hatched (N = 10 CO, 15 T 3 , 20 T 4 and 21 T 3 T 4 ), yielding a rather low overall hatching success (ca. 40%). Among the unhatched eggs, 33.7% (31 out of 92) had no developed embryos, and these were evenly distributed between the treatments (CO = 9/40, T 3 = 8/39, T 3 T 4 = 8/40, and T 4 = 6/39 eggs). Discarding the unfertilised eggs gives an overall hatching success of ca. 51%. Previous studies on Japanese quails have reported comparable hatching success, even in unmanipulated eggs (e.g. 40% in Okuliarová, <ns0:ref type='bibr'>Škrobánek & Zeman, 2007;</ns0:ref><ns0:ref type='bibr'>ca. 60% in Pick et al., 2016 and</ns0:ref><ns0:ref type='bibr'>in Stier, Metcalfe &</ns0:ref><ns0:ref type='bibr'>Monaghan, 2019)</ns0:ref>. In addition, the injection procedure itself is also known to reduce hatching success to some extent (Groothuis & von</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:03:47322:1:1:NEW 22 Jun 2020)</ns0:p><ns0:p>Manuscript to be reviewed <ns0:ref type='bibr'>Engelhardt, 2005)</ns0:ref>. Twelve hours after hatching, the chicks were marked by a unique combination of coloured rings and nail coding and transferred to two cages of 1 m² floor area and ca. 30 cm height (ca. 30 chicks/cage, sex and treatments mixed together). The chicks were provided with heating mats and lamps as extra heat sources for the first two weeks. The chicks were fed with sieved commercial poultry feed ('Punaheltta paras poikanen', Hankkija, Finland), and provided with Calcium and bathing sand. Two weeks after hatching, the chicks were separated in four 1 m² cages (ca. 30 cm high) of about 16 individuals. Around 3 weeks after hatching, coloured rings were replaced by unique metal rings. On week 4 after hatching, birds were transferred to eight pens of 1 m² floor area (average of 7.1 birds/pen, range = 4-9), under the same conditions as the parents. Around the age of sexual maturity (ca. 6-8 weeks after hatching), the birds were separated by sex in twelve 1 m² pens (average of 4.8 birds/pen, range = 4-5). The chicks were under the same photoperiod as the adults (i.e. 16L:8D).</ns0:p></ns0:div>
<ns0:div><ns0:head>Monitoring of growth and reproductive maturation</ns0:head><ns0:p>Body mass and wing length were measured twelve hours after hatching. Tarsus was not measured because it bends easily, resulting in inaccurate measures and potential harm for the young. From day 3 to day 15, these three traits were monitored every 3 days. From day 15 to day 78 (ca. 12 weeks), chicks were measured once a week. Body mass was recorded using a digital balance to the nearest 0.1 g. Wing and tarsus lengths were respectively measured with a ruler and a calliper to the nearest 0.5 mm and 0.1 mm. The sample size for the growth analysis was 7 CO, 11 T 3 , 18 T 4 and 21 T 3 T 4 . From week 6 to week 10, we monitored cloacal gland development and foam production in 28 males. Cloacal glands were measured every other day with a calliper to the nearest 0.1 mm as a proxy for testes development and sexual maturation <ns0:ref type='bibr' target='#b7'>(Biswas et al., 2007)</ns0:ref>. Foam production (by gently squeezing the cloacal gland) was assessed at the same time Manuscript to be reviewed and coded from 0 (no foam) to 3 (high production of foam), as a proxy of cloacal gland function <ns0:ref type='bibr'>(Cheng et al., 1989a;</ns0:ref><ns0:ref type='bibr'>Cheng et al., 1989b)</ns0:ref>. The same observer performed all measurements. We collected eggs produced by 10-week-old females over a 6-day period and recorded their mass to the nearest 0.1 g. We collected on average 5.7 eggs (range = 4-7) per female from 28 females.</ns0:p></ns0:div>
<ns0:div><ns0:head>Monitoring of cloacal gland regression and moult</ns0:head><ns0:p>In Japanese quails, exposure to short photoperiod and cold temperature triggers reproductive inhibition and postnuptial moulting <ns0:ref type='bibr'>(Tsuyoshi & Wada, 1992)</ns0:ref>. Thyroid hormones are known to coordinate these two responses (see introduction). When the birds reached the age of ca. 7 months, we exposed birds to short photoperiod (8L:16D, i.e., light from 08.00 to 16.00) with a 12:12-h cycle of normal (20°C) and low (9°C) temperature (low temperature was effective from 18.00 to 06.00). Cloacal gland regression (as a proxy for testes regression) was monitored every other day for 2 weeks with a calliper by measuring the width and length to obtain the area of the gland to the nearest 0.1 mm² (N = 26 males; 4 CO, 4 T3, 8 T4 and 12 T3T4). Primary moult was recorded from a single wing by giving a score to each primary from 0 (old feather) to 5 (new fully-grown feather) following Ginn and Melville (1983) (N = 54 males and females; 7 CO, 11 T3, 16 T4 and 20 T3T4). The total score of moulting was obtained by adding the score of all feathers. The sample size for the moult analysis was 7 CO, 11 T 3 , 16 T 4 and 20 T 3 T 4 .</ns0:p></ns0:div>
<ns0:div><ns0:head>Oxidative status biomarker analyses</ns0:head><ns0:p>Two blood samples were drawn, when birds were 2 weeks (N = 51 chicks) and 4 months old (N = 49 adults), respectively. The sample size per treatment was 7 CO, 11 T 3 , 17 T 4 and 20 T 3 T 4 .</ns0:p><ns0:p>200 µl of blood was collected from the brachial vein in heparinized capillaries and directly frozen in liquid nitrogen. Then, the samples were stored at -80°C until analyses. We measured </ns0:p></ns0:div>
<ns0:div><ns0:head>Ethics</ns0:head><ns0:p>The study complied with Finnish regulation and was approved by the Finnish Animal Experiment Board (ESAVI/1018/04.10.07/2016). In case of signs of harassment or disease, birds were placed in quarantine and monitored daily until they had recovered. Criteria for humane endpoints were defined as follow: passive behaviour, loss of appetite, loss of 30% of body weight, moving abnormally, trouble breathing. If we observed no clear improvement after two days, we would consult the veterinarian. A bird would be euthanised if it does not show signs of improvement in the next two days, though some judgement can be applied based on the alleged cause. One male was euthanised before the end of the experiment due to severe head injury. At the end of the experiment, all birds were euthanised by decapitation for collection of tissue samples (not used in this study).</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:03:47322:1:1:NEW 22 Jun 2020)</ns0:p><ns0:p>Manuscript to be reviewed</ns0:p></ns0:div>
<ns0:div><ns0:head>Statistical analysis</ns0:head><ns0:p>Data were analysed with the software R version 3.5.3 (R core team, 2019). In this study, two different statistical approaches were used: null-hypothesis testing with Generalised Linear Mixed Models (GLMMs) and Linear Mixed Models (LMMs), and multimodel inference with Generalised Additive Mixed Models (GAMMs). GAMMs were used to analyse the data on body and cloacal gland growth to account for its non-linear pattern (see Growth). In this analysis, we preferred multimodel inference as GAMMs generate many candidate models that cannot be directly compared (e.g., by the Kenward-Roger approach). Instead, candidate models were ranked based on their Akaike Information Criterion (AIC) values. Models with a ΔAIC ≤ 2 from the top-ranked model were retained in the set of best models. Akaike weights of all models were calculated following <ns0:ref type='bibr'>(Burnham & Anderson, 2002)</ns0:ref>, and evidence ratios of the top-ranked models were calculated as the weight of a model divided by the weight of the null model <ns0:ref type='bibr'>(Burnham, Anderson & Huyvaert, 2011)</ns0:ref>. To estimate the effect of the predictors, we computed the 95% confidence intervals from the best models using the nlme package <ns0:ref type='bibr'>(Pinheiro et al., 2018)</ns0:ref>.</ns0:p><ns0:p>GLMMs and LMMs were fitted using the R package lme4 <ns0:ref type='bibr' target='#b5'>(Bates et al., 2015)</ns0:ref>, and GAMMs were fitted using the package mgcv (Wood, 2017 Manuscript to be reviewed for independent means with GPower <ns0:ref type='bibr'>(Faul et al., 2009)</ns0:ref> with the effect size values calculated.</ns0:p><ns0:p>When presenting and discussing our results, we use the language of statistical 'clarity' rather than statistical 'significance' as suggested by <ns0:ref type='bibr'>Dushoff et al. (2019)</ns0:ref>.</ns0:p></ns0:div>
<ns0:div><ns0:head>Hatching success</ns0:head><ns0:p>To analyse hatching success, each egg was given a binary score: 0 for unhatched egg and 1 for hatched egg. A GLMM was fitted with a binomial error distribution (logit link) and mother identity as a random intercept and the 4-level treatment as the predictor. Egg mass might affect hatchability and was therefore added as a covariate in both models. The potential effect of storage duration on hatchability (Reis, Gama & Soares, 1997) was accounted for by including laying order as a covariate in both models. This covariate allowed us to control for the age of the egg as well.</ns0:p></ns0:div>
<ns0:div><ns0:head>Duration of embryonic period, age at embryonic mortality and early morphological traits</ns0:head><ns0:p>Duration of embryonic period and early morphological traits (mass and wing length at hatching, and tarsus length at day 3) were modelled with separate LMMs. Treatment, sex of the individuals and egg mass were included as fixed factors. Laying order was added as a covariate to account for potential effects of storage duration on hatching time and on chick weight (Reis, Gama & Soares, 1997). Mother identity was included as a random intercept.</ns0:p><ns0:p>The data for embryonic age had a skewed distribution and residuals were not normally distributed and heterogenous, which violated LMM assumptions on residual distribution. We therefore performed a simple Kruskal-Wallis test.</ns0:p></ns0:div>
<ns0:div><ns0:head>Growth</ns0:head><ns0:p>As growth curves typically reach an asymptote, we fitted non-linear GAMMs to these curves.</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:03:47322:1:1:NEW 22 Jun 2020)</ns0:p><ns0:p>Manuscript to be reviewed</ns0:p><ns0:p>Growth in body mass, tarsus and wing length were analysed in separate GAMMs. Growth was analysed until week 10 after hatching as all birds appeared to have reached their maximum body mass and tarsus and wing length. The data are composed of repeated measurements of the same individuals over time; therefore, we first corrected for temporal autocorrelation between the measurements using an ARMA(1,1) model for the residuals <ns0:ref type='bibr'>(Zuur et al., 2009)</ns0:ref>. Second, as mothers produced several eggs, the models included nested random effects, with measured individuals nested into mother identity, allowing for random intercepts. GAMMs allow modelling the vertical shift of the curves (i.e., changes in intercepts) and their shape. Treatment and sex were included as predictors. A smoothing function for the age of the birds was included to model the changes in the growth curves, and was allowed to vary by sex or treatment only, or none of these predictors. The interaction between sex and treatment was not analysed due to low statistical power. Additive effect of treatment and sex was tested for the intercept but could not be computed for curve shape. All combinations of the relevant predictors were tested for both shape parameters (i.e., intercept and curve shape). Prenatal THs may exert most of their effects in the offspring early life; this is why we additionally tested hatchlings morphological traits apart from the growth curve. Likewise, we also analysed separately morphological traits at adulthood (ca. 9 weeks old), as these traits may condition the fitness of an individual. Because of sex differences and low sex-specific sample sizes, we standardised the measures within sex and regressed the standardised responses against treatment in a linear regression.</ns0:p></ns0:div>
<ns0:div><ns0:head>Reproductive maturation, regression and investment</ns0:head><ns0:p>Due to low sample sizes in these sex-specific responses, we could not perform robust statistical analyses. We therefore present these analyses and results in the supplementary material and only</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:03:47322:1:1:NEW 22 Jun 2020)</ns0:p><ns0:p>Manuscript to be reviewed briefly discuss them (Figs. S6 to S9; Table <ns0:ref type='table'>S3</ns0:ref>).</ns0:p></ns0:div>
<ns0:div><ns0:head>Oxidative stress</ns0:head><ns0:p>A principal component analysis (PCA) was first performed on measured antioxidant markers (SOD, CAT, GPx, tGSH and GST), to reduce the number of metrics for subsequent analyses.</ns0:p><ns0:p>The first and the second principal components (PCs) explained together 60.2% of the variance (Table <ns0:ref type='table'>1</ns0:ref>). PC1 and PC2 were then used as dependent variables in separate LMMs. LMMs included the treatment, sex and age of individuals (2 weeks and 4 months old) as fixed factors and the 2-way interactions between treatment and sex, and treatment and age. Mother and individual identities, to account for repeated measures, were added as random intercepts.</ns0:p><ns0:p>Malondialdehyde (MDA) is a marker of oxidative damage, which is a different measure from antioxidant activity, and was therefore analysed in a separate LMM using the same parameters as for PC1 and PC2, adding the batch of the assay as an additional random intercept. The marker of cell oxidative status (GSH:GSSG ratio) was analysed with the same model used for PC1 and PC2.</ns0:p></ns0:div>
<ns0:div><ns0:head>Moult</ns0:head><ns0:p>Two parameters of moult were analysed in separate LMMs: the timing of moult (i.e., the moult score after one week of short photoperiod), and the rate of moult (i.e., how fast birds moulted).</ns0:p><ns0:p>Both models included treatment and sex as fixed factors, and mother identity as a random intercept. The rate of moult was tested by fitting an interaction between treatment and age. This model also included the main effect of age and individual identity, nested within mother identity, as a random intercept to account for repeated measures. Estimated marginal means and standard errors (EMMs ± SE) were derived from the model using the package emmeans (Lenth, 2019).</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:03:47322:1:1:NEW 22 Jun 2020)</ns0:p><ns0:p>Manuscript to be reviewed</ns0:p></ns0:div>
<ns0:div><ns0:head>Results</ns0:head></ns0:div>
<ns0:div><ns0:head>Effects of prenatal THs on hatching success and age of embryo mortality</ns0:head><ns0:p>There was a clear effect of elevated prenatal THs on hatching success (GLMM, p = 0.03, Fig. <ns0:ref type='figure' target='#fig_6'>1</ns0:ref>).</ns0:p><ns0:p>Tukey post-hoc analysis revealed that hatching success in the T 3 T 4 (66%) group was statistically higher than in the CO group (32%) (Tukey z = 2.77, p = 0.03) . There was a non-significant trend between the T4 (61%) and the CO groups (z = 2.37, p = 0.08). There were no clear differences in hatching success between the T 3 (48%) and the CO group (z = 1.25, p = 0.45), or between the hormone treatments (all z < 1.61, all p > 0.37). Dissection of the unhatched eggs showed that age of embryo mortality did not differ between the treatments (Kruskal-Wallis χ² = 7.22, df = 3, p = 0.07; Fig. <ns0:ref type='figure' target='#fig_6'>S1</ns0:ref>). Finally, the manipulation of yolk THs did not affect the duration of embryonic period (LMM, F 3,42.0 = 0.57, p = 0.64, Fig. <ns0:ref type='figure' target='#fig_7'>S2</ns0:ref>). Sex of the embryo or egg mass (LMM sex, F 1,49.7 = 2.63, p = 0.11; LMM egg mass, F 1,19.3 = 0.01, p = 0.92) were also not associated with the duration of the embryonic period.</ns0:p><ns0:p>Laying order (i.e. the effect of storage duration) was not correlated with any of the responses (all p ≥ 0.25).</ns0:p></ns0:div>
<ns0:div><ns0:head>Effects of prenatal THs on growth</ns0:head><ns0:p>Mass at hatching was not influenced by the elevation of prenatal THs (LMM, F 3,35.0 = 0.81, p = 0.50, Fig. <ns0:ref type='figure' target='#fig_8'>S3</ns0:ref>). Mass at hatching was positively correlated with egg mass (LMM, Estimate±SE = 0.72±0.10 g, F 1,24.1 = 46.9, p < 0.001). Although we detected no clear differences on hatchling morphological traits (body mass, wing and tarsus length) due to prenatal THs (all p > 0.12), the calculated effect sizes (Cohen's d[95%CI]) and achieved statistical power yielded additional information regarding the potential effects of prenatal THs (Table <ns0:ref type='table' target='#tab_1'>2</ns0:ref>). For body mass, the effect PeerJ reviewing PDF | (2020:03:47322:1:1:NEW 22 Jun 2020)</ns0:p><ns0:p>Manuscript to be reviewed sizes were low and the achieved statistical power was very low. For wing length, the effect sizes were moderate and the achieved statistical power was low. For tarsus length, the effect sizes were moderate to large and the achieved statistical power was low to moderate. Similarly, adult morphology was not affected by the treatment (all p > 0.13), but effect sizes indicate small to large effects of prenatal THs (Table <ns0:ref type='table' target='#tab_1'>2</ns0:ref>). For body mass, the effect sizes were small and the achieved power was low. For wing length, the effect sizes were large and the achieved power was moderate. For tarsus length, the effect sizes were small to large and the achieved power was moderate to high.</ns0:p><ns0:p>Regarding body mass growth, the top-ranked model showed that the curve shape and the intercept differ according to sex (Table <ns0:ref type='table'>3</ns0:ref>). After 10 weeks, females had a larger body mass than males (mean±SE females = 214.4±5.7 g, males = 172.4±4.5 g, Fig. <ns0:ref type='figure' target='#fig_7'>2</ns0:ref>), which was supported by the 95% CIs (Table <ns0:ref type='table'>4</ns0:ref>). Based on model selection we conclude that the treatment had no effect on body mass growth (Table <ns0:ref type='table'>3</ns0:ref>).</ns0:p><ns0:p>For wing growth, the top-ranked model (ΔAIC ≤ 2) included sex in the intercept, while treatment was not included in the best supported model (Table <ns0:ref type='table'>S1</ns0:ref>). The 95% CIs (Table <ns0:ref type='table'>3</ns0:ref>) confirmed that males had a lower wing length than females (Fig. <ns0:ref type='figure' target='#fig_9'>S4</ns0:ref>). Concerning tarsus growth, the models within ΔAIC ≤ 2 included no predictors for the curve shape but included treatment for the intercept (Table <ns0:ref type='table' target='#tab_1'>S2</ns0:ref>). The 95% CIs of the parameter estimates from these models suggested that there was a slight negative effect of T 3 T 4 treatment on tarsus growth (Table <ns0:ref type='table'>4</ns0:ref>, Fig. <ns0:ref type='figure'>S5</ns0:ref>). However, as the estimates were close to 0 (Table <ns0:ref type='table'>4</ns0:ref>) and evidence ratios showed that the model with treatment as a predictor was only 3.5 times more supported than the null model (Table <ns0:ref type='table' target='#tab_1'>S2</ns0:ref>), we conclude that the effect of THs on tarsus length is likely to be very small. Likewise, the second model for tarsus length included sex as a predictor for the intercept, but its 95% CIs overlapped with 0 (Table <ns0:ref type='table'>4</ns0:ref>). We therefore conclude that sex had no effect on tarsus growth.</ns0:p></ns0:div>
<ns0:div><ns0:head>Effects of prenatal THs on postnuptial moult</ns0:head><ns0:p>As expected, birds started to moult soon after being exposed to short photoperiod, with an average increase of moult score by 6 per week (SE = 0.2, F 1,254.0 = 827.4, p < 0.001, Fig. <ns0:ref type='figure' target='#fig_8'>3</ns0:ref>). The first moult score (assessed one week after switching to short photoperiod) was not affected by the treatment (LMM, F 3,42.7 = 0.36, p = 0.78), but was influenced by sex, with females having a higher score than male (EMMs ± SE: female = 21.4 ± 1.6, male = 7.2 ± 1.7; LMM F 1,45.3 = 41.9, p < 0.001). Yolk TH elevation did not affect the rate of moult (LMM interaction treatment × time, F 3,251.0 = 0.59, p = 0.62, Fig. <ns0:ref type='figure' target='#fig_8'>3</ns0:ref>).</ns0:p></ns0:div>
<ns0:div><ns0:head>Effects of prenatal THs on oxidative stress</ns0:head><ns0:p>The elevation of yolk THs had no effect on PC1 or PC2 of antioxidants at either 2 weeks ('chicks') or 4 months ('adults') old (LMM on PC1, F 3,40.3 = 2.40 , p = 0.08; LMM on PC2, F 3,42.2 = 0.92, p = 0.44, treatment × age, F < 0.91, p > 0.44). The age of the birds had a highly significant effect on PC1, with chicks generally having higher antioxidant capacities (CAT, GST and tGSH) than adults (LMM, Estimate±SE = -1.34±0.19, F 1,49.2 = 52.1, p < 0.001). All the other predictors had no effect on either PC1 or PC2 (all F < 2.93 and all p > 0.09).</ns0:p><ns0:p>The marker of oxidative damage, MDA, was affected by the elevation of yolk THs (LMM, F 3,43.6 = 3.08, p = 0.04, Fig. <ns0:ref type='figure' target='#fig_9'>4</ns0:ref>). Tukey post-hoc analysis showed that the T4 group had higher MDA values than the T3 group (Estimate±SE = 0.01±0.004, Tukey contrast p = 0.01), but none of the groups differed from the control (Tukey p-values > 0.19). However, this result became non-significant when removing the outlier in the T 4 group (LMM, F 3,43.1 = 2.68, p = PeerJ reviewing PDF | (2020:03:47322:1:1:NEW 22 Jun 2020)</ns0:p><ns0:p>Manuscript to be reviewed 0.06). MDA levels were not affected by the age or the sex of individuals (LMM age, F 1,54.4 = 0.30, p = 0.59; LMM sex, F 1,42.0 = 1.47, p = 0.23).</ns0:p><ns0:p>The marker of cell oxidative balance, GSH:GSSG, was not influenced by the yolk THs nor by the sex of the birds (LMM treatment, F 3,33.0 = 0.85, p = 0.48; LMM sex, F 1,40.6 = 0.57, p = 0.45). However, chicks had a higher GSH:GSSG ratio than adults (LMM, Estimate±SE = 0.17±0.04, F 1,50.0 = 18.3, p < 0.001).</ns0:p></ns0:div>
<ns0:div><ns0:head>Discussion</ns0:head><ns0:p>The aim of this experimental study was to investigate the potential short-term and organisational effects (with long-term consequences) of maternal thyroid hormones (THs) in a precocial species, the Japanese quail, by experimental elevation of THs in eggs. Our study is the first to investigate the effects of yolk T 3 and T 4 separately, within the natural range of the study model.</ns0:p><ns0:p>In addition we studied both short-and long-term effects on embryonic development, growth, life stage transitions and oxidative stress. We detected a positive effect of yolk THs on hatching success. All other response variables studied were not affected by elevated prenatal THs.</ns0:p></ns0:div>
<ns0:div><ns0:head>Effects of prenatal THs on hatching success and embryonic development</ns0:head><ns0:p>The overall low hatching success, and especially in the control group, forces us to interpret these results with caution. In addition, we cannot exclude that our results may be partly due to selective disappearance of lower quality embryos in the control group and with injected THs helping lower quality chicks to hatch. This might have biased the results after hatching, but is still a relevant effect of the hormone treatment. We found that hatching success almost doubled when the eggs received an injection of both T 4 and T 3 , or an injection of T 4 only. Previous similar studies reported comparable effects of yolk THs in rock pigeons <ns0:ref type='bibr'>(Hsu et al., 2017)</ns0:ref> </ns0:p></ns0:div>
<ns0:div><ns0:head>and in</ns0:head><ns0:p>PeerJ reviewing PDF | (2020:03:47322:1:1:NEW 22 Jun 2020)</ns0:p><ns0:p>Manuscript to be reviewed collared flycatchers <ns0:ref type='bibr'>(Hsu et al., 2019)</ns0:ref>. In these studies, injections consisted of a mixture of both T 3 and T 4 . Given that mostly T 3 binds to receptors, these results suggest that embryos likely express deiodinase enzymes to convert T 4 to T 3 , and/or yolk may contain maternally derived deiodinase mRNA, as injection with T 3 only did not differ from control. Indeed, deiodinase expression has previously been characterised in chicken embryos already 24h after the onset of incubation <ns0:ref type='bibr'>(Darras et al., 2009)</ns0:ref>. An old study found that injecting T 4 close to hatching can advance hatching time, which suggests that yolk THs may help embryos overcoming hurdles close to hatching <ns0:ref type='bibr' target='#b4'>(Balaban & Hill, 1971)</ns0:ref>. In contrast with our study, two similar studies in altricial species detected no increased hatching success due to the injection of THs <ns0:ref type='bibr'>(Ruuskanen et al., 2016;</ns0:ref><ns0:ref type='bibr'>Sarraude et al., 2020)</ns0:ref>. The dissimilarities between the studies may come from interspecific differences in terms of utilisation of yolk THs by the embryos or from contextdependent effects (e.g. due to other egg components). Further comparative and mechanistic studies could help understanding the dynamic of yolk THs during incubation. Increased yolk THs did not influence age of embryo mortality. Similar to our study, <ns0:ref type='bibr'>Ruuskanen et al. (2016)</ns0:ref> did not find any difference in the timing of mortality in great tit embryos. Conversely, the study on rock pigeons found that yolk THs increased the proportion of well-developed embryos <ns0:ref type='bibr'>(Hsu et al., 2017)</ns0:ref>. Similarly to our result on hatching success, yolk TH effects on embryonic development may differ in a species-specific manner.</ns0:p><ns0:p>Our results on hatching success may partly be attributed to yolk THs balancing the negative effects of injections on embryonic survivability. Further studies may aim at understanding the contribution of THs to counteract the effect of injection. To do so, such studies may use an non-invasive method to manipulate yolk THs (e.g., egg-dipping method as in <ns0:ref type='bibr'>Perrin et al. 1995)</ns0:ref>, in addition to injected controls, like in our study.</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:03:47322:1:1:NEW 22 Jun 2020)</ns0:p><ns0:p>Manuscript to be reviewed</ns0:p></ns0:div>
<ns0:div><ns0:head>Effects of prenatal THs on growth</ns0:head><ns0:p>We found no apparent influence of yolk THs on growth, contrary to our expectations based on the recent literature. Other comparable studies found either a positive <ns0:ref type='bibr'>(Hsu et al., 2019;</ns0:ref><ns0:ref type='bibr'>weak effect in Sarraude et al., 2020</ns0:ref><ns0:ref type='bibr'>), a negative (Hsu et al., 2017)</ns0:ref> or a sex-specific effect <ns0:ref type='bibr'>(Ruuskanen et al., 2016)</ns0:ref> of yolk THs on growth. This notable difference may be due to the captive conditions experienced by the Japanese quails in our study, with unrestricted access to food and water.</ns0:p><ns0:p>Although the pigeon study also provided ad libitum food, parents still needed to process food before feeding their nestlings in the form of crop milk, whereas precocial quails have no such limitation. In addition, the Japanese quail has been domesticated for many generations, and probably selected for rapid growth for economic reasons. Whole-genome sequencing in chickens showed that domestication induced a strong positive selection on genes associated with growth <ns0:ref type='bibr'>(Rubin et al., 2010)</ns0:ref>. Interestingly, that study also found a strong selection for a locus associated with thyroid stimulating hormone (TSH) receptor. TSH controls most of the TH production by the thyroid gland <ns0:ref type='bibr' target='#b11'>(McNabb & Darras, 2015)</ns0:ref>, and this artificial selection may overshadow the effects of natural variations of prenatal THs on growth. Besides, the low number of individuals in the control and T 3 groups (7 and 11, respectively) limited the statistical power to detect differences between the treatments. Indeed, we were able to detect small to moderate negative effects of yolk THs on morphological traits at hatching and in adulthood. Such negative effects, although small, may still be biologically relevant. Repeating the study with a larger sample size may allow us to ascertain the effects of yolk THs on growth in precocial study models. Research on the influences of prenatal THs on growth will also benefit from experimental studies on wild precocial species.</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:03:47322:1:1:NEW 22 Jun 2020)</ns0:p><ns0:p>Manuscript to be reviewed</ns0:p></ns0:div>
<ns0:div><ns0:head>Effects of prenatal THs on postnuptial moult</ns0:head><ns0:p>Short photoperiod in combination with cold temperature triggered primary moult, as expected.</ns0:p><ns0:p>However, we detected no effect of yolk THs on the timing or speed of moult. Thyroid hormones are important in moult and feather growth (reviewed in <ns0:ref type='bibr'>Dawson, 2015)</ns0:ref>. For example, thyroidectomised birds fail to moult after being exposed to long photoperiods <ns0:ref type='bibr'>(Dawson, 2015)</ns0:ref>.</ns0:p><ns0:p>In addition, thyroidectomised nestling starlings failed to grow normal adult plumage and grown feathers presented an abnormal structure <ns0:ref type='bibr'>(Dawson et al., 1994)</ns0:ref>. By removing the thyroid gland, these two studies implemented extreme pharmacological protocols that differ drastically from our injection of physiological doses. In addition, our experimental design, increasing TH exposure (vs decreased TH exposure in the above-mentioned studies), may have different consequences. For example, there may be a threshold above which any, additional hormones may not affect moult. Overall, our results show no support for the hypothesis of organising effect of prenatal THs on life stage transitions. Yet, due to small sample sizes in sex-specific analyses (i.e., male gonadal maturation and regression, and female reproductive investment), there remains a</ns0:p><ns0:p>relatively high uncertainty about the potential organising effects of prenatal THs. Replicate studies with larger samples sizes and different study models will reduce this uncertainty.</ns0:p></ns0:div>
<ns0:div><ns0:head>Effects of prenatal THs on oxidative stress</ns0:head><ns0:p>In contrast to our predictions, elevated yolk THs did not affect oxidative status during chick or adult phase. We found no changes in antioxidant activities in relation to yolk THs and no imbalance in the oxidative cell status. Nevertheless, T 4 birds had a higher level of oxidative damage on lipids than T 3 birds, but this was a weak effect driven by one outlier. The lack of effects on chick oxidative status among the treatment groups could be explained by the absence Manuscript to be reviewed of treatment effects on growth, given that high growth rates usually result in higher oxidative stress and damage (e.g. <ns0:ref type='bibr' target='#b2'>Alonso-Alvarez et al., 2007)</ns0:ref>. In turn, the lack of treatment effects on adult oxidative status may suggest no organisational effects of prenatal THs on adult metabolism.</ns0:p><ns0:p>Two recent studies in altricial species also found no influence of yolk THs on nestling oxidative stress <ns0:ref type='bibr'>(Hsu et al., 2019;</ns0:ref><ns0:ref type='bibr'>Sarraude et al., 2020</ns0:ref>), yet telomere length, a biomarker of aging was affected <ns0:ref type='bibr'>(Stier et al., 2020)</ns0:ref>. Our study shows for the first time that prenatal THs have no influence on adult oxidative stress either. The previous study focused on a limited set of biomarkers: one antioxidant enzyme, oxidative damage on lipids and oxidative balance. In the present study, we measured 7 biomarkers, thus providing broader support to the absence of effects of prenatal THs on post-natal/hatching oxidative stress.</ns0:p></ns0:div>
<ns0:div><ns0:head>Conclusion</ns0:head><ns0:p>To our knowledge, this study is the first one to experimentally investigate the consequences of natural variations of maternal THs not only early but also in adult physiology and postnuptial moult in any vertebrate. Furthermore, this study explored for the first time the effects of maternal T 3 and T 4 separately. We found no evidence for differential effects of maternal T 4 and T 3, while an effect of T 4 , alone or in combination with T 3 , on hatching success suggests that T 4 is converted into T 3 , the biologically active form during embryonic development. Contrary to similar studies on wild altricial species, we found no influence of maternal THs on growth.</ns0:p><ns0:p>Further research on embryos utilisation of maternal THs may help understand the differences observed between precocial and altricial species. Studies in other vertebrates are urgently needed to understand the potential organising effects of maternal THs with long-term consequences.</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:03:47322:1:1:NEW 22 Jun 2020)</ns0:p><ns0:p>Manuscript to be reviewed Manuscript to be reviewed Cohen's d, 95% CIs and achieved statistical power for post-hatching and adult morphological measures (body mass, wing and tarsus length). Manuscript to be reviewed Manuscript to be reviewed Manuscript to be reviewed Manuscript to be reviewed </ns0:p></ns0:div><ns0:figure xml:id='fig_0'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:03:47322:1:1:NEW 22 Jun 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_1'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:03:47322:1:1:NEW 22 Jun 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_2'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:03:47322:1:1:NEW 22 Jun 2020)Manuscript to be reviewed various biomarkers of antioxidant status; the antioxidant glutathione (tGSH), the ratio of reduced and oxidised glutathione (GSH:GSSG) and activity of the antioxidant enzymes glutathione peroxidase (GPx), catalase (CAT) and superoxide dismutase (SOD) from the blood. Measuring multiple biomarkers of oxidative and antioxidant status allows a broader understanding of the mechanism, and the interpretation of the results is more reliable if multiple markers show similar patterns. The GSH:GSSG ratio represents the overall oxidative state of cells and a low ratio reveals oxidative stress(Hoffman, 2002; Isaksson et al., 2005; Lilley et al., 2013; Rainio et al., 2013; Halliwell & Gutteridge, 2015). GPx enzymes catalyse the glutathione cycle, whereas CAT and SOD directly regulate the level of reactive oxygen species (ROS)(Ercal, Gurer-Orhan & Aykin-Burns, 2001; Halliwell & Gutteridge, 2015). The methodology for measuring each biomarker is described in detail inRainio et al. (2015). All analyses were conducted blindly of the treatment following Ruuskanen et al (2017).</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_3'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:03:47322:1:1:NEW 22 Jun 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_4'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:03:47322:1:1:NEW 22 Jun 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_5'><ns0:head>95%</ns0:head><ns0:label /><ns0:figDesc>CIs were calculated by bootstrap resampling with 5,000 resamples. CO = control, T 4 (thyroxine) = injection of T 4 , T 3 (triiodothyronine) = injection of T 3 , T 3 T 4 = injection of T 3 and T 4. PeerJ reviewing PDF | (2020:03:47322:1:1:NEW 22 Jun 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_6'><ns0:head>Figure 1</ns0:head><ns0:label>1</ns0:label><ns0:figDesc>Figure 1</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_7'><ns0:head>Figure 2</ns0:head><ns0:label>2</ns0:label><ns0:figDesc>Figure 2</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_8'><ns0:head>Figure 3</ns0:head><ns0:label>3</ns0:label><ns0:figDesc>Figure 3</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_9'><ns0:head>Figure 4 MDA</ns0:head><ns0:label>4</ns0:label><ns0:figDesc>Figure 4</ns0:figDesc></ns0:figure>
<ns0:figure type='table' xml:id='tab_0'><ns0:head /><ns0:label /><ns0:figDesc>). P-values for GLMMs were obtained by parametric bootstrapping with 1,000 simulations and p-values for LMMs were calculated by model comparison using Kenward-Roger approximation, using the package pbkrtest in both cases(Halekoh & Højsgaard, 2014). Post-hoc Tukey analyses were conducted with the package multcomp(Hothorn et al., 2008). Model residuals were checked visually for normality and homoscedasticity. Covariates and interactions were removed when non-significant (α = 0.05).</ns0:figDesc><ns0:table /><ns0:note>Effect size calculations (Cohen's d and 95%CI) were performed with the website estimationstats.com(Ho et al., 2019) and statistical power analyses were performed using t-tests PeerJ reviewing PDF | (2020:03:47322:1:1:NEW 22 Jun 2020)</ns0:note></ns0:figure>
<ns0:figure type='table' xml:id='tab_1'><ns0:head>Table 2 (on next page)</ns0:head><ns0:label>2</ns0:label><ns0:figDesc /><ns0:table /></ns0:figure>
</ns0:body>
" | "Dear Editors and Reviewers,
Thank you very much for your interest in the manuscript and for your constructive and
thorough comments. We have followed the reviewers’ comments carefully and revised the manuscript accordingly, particularly
1) we rearranged the introduction to have a more general introduction before we move to birds, and included additional references to other species (mammals, fish).
2) we explained in more details why birds are a good study model for research on maternal hormones, and more specifically why Japanese quails are a good model for our study.
3) we added information on embryonic development of unhatched eggs, and added a figure on this in the supplements. We also recalculated the Cohen’s d index and 95%CI.
4) we tempered our interpretation of the results on hatching success, due to potential selection of injected THs on embryos
and 5) we have corrected all minor issues suggested.
The changes have been made with tracked changes. Our detailed responses to the questions from the reviewers appear below.
We hope that our revision has improved the quality of the manuscript and that you can now consider our manuscript for publication.
Sincerely,
On behalf of my co-authors,
Tom Sarraude
Editor's Decision Major Revisions
As you will see both reviewers provided thorough and constructive reviews of your manuscript. They thought the study addresses an important area of avian development. However, they had some concerned that I would encourage you to address fully. Specifically, issues regarding hatching success and unequal sample sizes between groups.
Comments from the reviewers
Reviewer 1
Basic reporting
The manuscript is well written and has the correct structure. It is a well written manuscript with what I find a very interesting topic. They provide good context for the study. All figures and tables are well done, although they should note what the box plots are showing (median and quartiles?). I appreciate the inclusion of all data points along with the mean (median) values on the figures.
We have added in the caption of figure 4 “Boxplots show median and quartiles.”
Experimental design
The experiments are within the scope of the journal. The research question is well defined and is meaningful. Our understanding of the influence of maternally derived thyroid hormones on development in birds is incomplete and this study attempts to address this by supplementing the yolk with additional thyroid hormones. The methods are all well-defined. I wonder how effective the dosing of the yolk is for actually delivering the hormone to the developing embryo - see comments below.
Validity of the findings
I hate to say this, as I really like the idea of the experiment, but given the low hatching success of the control group I tend to agree with the previous reviewers of an earlier version of this manuscript. It would be nice for them to repeat the hatching success portion of the experiment from initial exposure through hatching to confirm their findings on hatching success, especially given this is the only parameter where they found a significant response. Your control group has a hatching success of only 25% and you are comparing it to 60% in other “control” studies. This means your control eggs had 2 to 3 times worse hatching success than other control studies. How does taking into account those eggs with no development change the hatching success of the control?
Unfortunately, repeating the hatching success experiment is not possible for logistical constraints, including the fact that we do not have the bird lines anymore. Since the low hatching success seems to be a major concern for both current reviewers (in addition to previous reviewers), we tempered the importance of our finding with regard to the effect of treatment on hatching success by acknowledging the uncertainty induced by the low hatching success (lines 479-483). Nevertheless, this result of higher hatching success in the T3T4 and T4 groups compared with the controls is consistent with previous studies in rock pigeons (Hsu et al. 2017) and collared flycatchers (Hsu et al. 2019), which suggests that this enhancement in hatching success could be a real effect and not just an artefact because of the low hatching success in our study. The low hatching success seems to be an, albeit unexpected and unfortunate, feature of our bird line: in another experiment with the same parental birds and incubator, the hatching success of unmanipulated eggs (i.e. non-injected) was only around 50% (Ruuskanen et al. unpubl data). If we take into account that in the current study 9 eggs in the control group did not show any embryonic development and are likely not fertilised, the hatching success was 32%. The difference between 50 and 32% therefore may be attributed to the detrimental effects of egg injection, piercing several membranes in the egg. This magnitude of reduction in hatching success is comparable to those in other species, as discussed in Hsu et al. 2019. This suggests that the low hatching success in our study is due to a combination of the quail line we used and the injection of the eggs.
I would like to see the data on mortality. When did the embryos die during development?
Please see our detailed response below.
Comments for the author
A major question I have is in regard to the dosing of the yolk with TH. I know that these authors have used this method in the past with other species. While I appreciate that the authors are attempting to elevate TH within a somewhat natural range, I wonder how well injecting in the yolk is as a delivery method. Given this method and the lack of a response, we really don’t know when the excess TH are actually taken up by the embryo. Are the embryos actually experiencing “higher” levels of TH during development? Do they become diluted in the yolk, so that it is a very minimal dosing or does it stay concentrated and they take it up all at once? Is there a study that has injected an inert dye into the yolk and watched how it diffuses and moves through development (not suggesting you do this, just wondering if you know what happens once it gets in the yolk)? Can you provide the concentration in terms of ng T3 per ml (or mg) yolk?
The large body of the literature that applied such a method is predominantly focusing on yolk androgens (see Podmokła et al. 2018 Biol. Rev. for a detailed review and meta-analysis). For androgens, a previous experiment using chicken eggs has validated the gradual uptake of radioisotope-labelled androgens from the injection side into the embryo (von Engelhardt et al. GCE 2009). In terms of THs, previous studies in chicken have also shown that injection of THs in the yolk increased TH concentrations in embryonic extract (Darras, Front. Endocrinol. 2019). Although the concentrations used in these studies were supraphysiological, we expect our physiological manipulation to increase the exposure of embryos to THs. Moreover, as chicks have consumed all yolk and albumin 2-3 days after hatching (after absorption of the yolk sac), we expect that all injected hormones (or their metabolites) have reached the embryo. To what extent the injected hormone would be metabolised by the embryo is very interesting and open for further study. In other studies applying a similar injection protocol, TH-elevation led to significant effects on growth and metabolism (Hsu et al. 2019, 2017, Ruuskanen et al 2016), suggesting that embryos should be able to utilise yolk THs.
The concentration of THs in the eggs after injection will vary according to the egg and yolk mass. This is why we provide information on the actual amounts injected. We added concentration values as additional information (lines 184 and 189).
Line 28: “These results suggest that yolk thyroid hormones are important in the embryonic stage of precocial birds…”
Line 60 – change “fasten” to accelerate
This has been changed.
Line 179 - I am a little worried about the low hatching success in this study, as noted above. While they show provide references of studies where others have a 50% hatching success, there are other studies with hatching successes between 80-95% for fertilized eggs (for one see Martin and Arnold, 1991, The Condor). Additionally, their control eggs had a 25% hatching success, which is almost half of the worst hatching success they reference. It would be nice to see the distribution of ages at which the embyros died. What was the distribution between the treatments of the 31 eggs with no developed embryo? If a larger majority of these eggs were in the control, this might skew your hatching success rates away from the controls. Is the hatch rate you present in Figure 1 including the total number of eggs set or just those that showed developed embryos (discarded the infertile and those whose yolks may have been damaged resulting in no development) as the denominator? I would argue that it should be not include the infertile eggs that did not have a developed embryo. How far along were the additional 61 animals that did not hatch? I would recommend you repeat this aspect of the study to confirm the findings.
The number of unfertilised eggs was evenly distributed between the treatments: CO = 9/40, T3 = 8/39, T3T4 = 8/40, T4 = 6/39 (X-squared = 0.68, df = 3, p-value = 0.88). These numbers have been added at line 212.
We excluded unfertilised eggs in Figure 1 to present new hatching success values (CO = 32%, T3 = 48%, T3T4 = 66% and T4 = 61%). We also modified these values in the result section (lines 394-399) and re-ran the statistical models.
Below is the age distribution of the embryos that did not hatch (median and quartiles). There was no statistical difference between the groups (Kruskal-Wallis χ² = 7.22, df = 3, p = 0.07).
In Japanese quails, hatching occurs around 17 days after incubation. We have added the graph as a supplement (Suppl Fig 1).
Although the overall p-value of the Kruskall-Wallis test indicates that none of the groups differs from the other 3 groups, the trend seems to suggest that the embryos from the control group died at a later stage compared to the TH groups (median = 11 in CO, larger than all other groups). We therefore still conducted a post-hoc pairwise analysis (Conover test) with Bonferroni correction and did not find differences between the groups.
However, the relatively slow sample sizes force us to interpret this non-significant trend with caution. If the reviewers wish, we could add a short discussion along these lines on embryonic mortality in the main text.
Line 291 – I applaud your use of a measure of effect sizes as a means to determine how large a response in the data is. I suggest you include the 95% CI for your Cohen’s d index. There is a nice website: estimationstats.com which provides the R script to calculate Cohen’s d along with bootstrap 95% CI. Here is the citation they provide: Moving beyond P values: Everyday data analysis with estimation plots. Joses Ho, Tayfun Tumkaya, Sameer Aryal, Hyungwon Choi, Adam Claridge-Chang. Nature Methods 2019, 1548-7105.
Thank you for the recommendation, we did not know that such a website existed. We have now used the website to compute the 95%CI of the Cohen’s d, and we added them in a new table with the power analyses (Table 4). We also modified the method (line 319-327) and results (line 413-426) sections accordingly.
Line 326 – “… additionally tested early morphological traits apart from the growth curve…” Can you define early? Is this during the embryonic period or neonatal period?
We meant hatchlings’ morphological traits. This has now been made explicit.
Line 358 – 362 – Did you remove the undeveloped (infertile) eggs from the denominator?
No, we did not. Thank you for the suggestion. We have modified the analysis and the new values do not include the infertile eggs anymore.
Line 363 – It might be useful for you to provide some data as to when the embryos died. Was it early on in development or just prior to hatching?
Please see the detailed answer above
Line 368 – add period at the end of the sentence.
It has been added
Line 373 – I would change early to neonate or hatchling. When I read early, I was thinking of during the embryonic stages.
This has been made more explicit (line 362).
Line 374-379 – might be nice to include the 95% CI for Cohen’s d. Will give us an idea of the range of d and how “strong” the moderate effect size is.
Please see the response above
Line 380-381 – were there differences in the sex ratios for the three treatments? Just curious.
The number of females and males per treatment was as follow: CO: 3 F, 4 M and 3 unsexed; T3: 7 F, 4 M and 4 unsexed; T3T4: 9 F, 12 M; T4: 10 F, 8 M, and 2 unsexed. We did not sex those individuals that died before reaching sexual maturity.
Figure 4 figure legend – what do the boxes represent? Please provide description in the legend.
Boxplots represent median values and quartiles. The figure legend has been updated.
Line 438 – Do they know at what point in the embryonic development that the embryo starts to produce T3 and deiodinase? How far along into embryonic development? Can you add the timing to this sentence?
McNichols and McNabb (GCE, 1988) have shown that T3 in the thyroid gland of Japanese quail embryos started to have clear increase from incubation day 13 onwards. Darras et al. (GCE, 2009) have detected the presence of the three DIOs already 24h after incubation (mRNA).
The information on deiodinases has been added, but we removed the part on endogenous T3 as it now seems a bit out of place (line 490).
Line 446-448 – I am not sure what you mean by improving the age of embryo mortality. If an embryo dies, does changing the timing of this improve it? Because you do not provide the data for timing of embryo mortality, it is hard to judge this statement.
Indeed, “improving” was a wrong term. It has been replaced with “influence”
Figure S1 – what do the line and boxes represent? Provide in the figure legend.
Boxplots show median and quartiles. This has been added in Figs S1, S2, S3 and S8.
Figures S5-S8 are not references in the text.
Since we do not present these results in the main text, we mention the Figs S6 to S9 and Table S3 in the method section (line 370).
Reviewer 2
Basic reporting
This study sets out to investigate the effects of maternal thyroid hormones on hatching success and longer-term metrics in Japanese quails. While I disagree that this avenue of research is particularly novel, it would nonetheless be a welcome addition to the literature. Perhaps an enhanced focus on T3 vs. T4 would help set it apart — to this end, however, there is very little in terms of their endocrinological differences discussed at the outset. To this end, I think the introduction could use a restructure to showcase i) why we care about maternal thyroid hormones in Japanese quails, or why Japanese quails (or birds in general) represent a good model to investigate these questions. While the methods were selected for ecological relevance, the overall manuscript really lacked the endocrinological and ecological context to make it matter.
Thank you for the suggestions.
The ecological value of this study lies in the use of physiological manipulation of yolk THs, in contrast to most other studies on THs (reviewed in Ruuskanen and Hsu 2018). Please see our detailed response to a later comment.
We added a short paragraph at line 93-100 to describe why birds are very good models to study the effects of maternal hormones on offspring:
Oviparous species, such as birds, are suitable models for studying the role of maternal hormones on the progeny because embryos develop in eggs outside mothers’ body. The content of an egg cannot be adjusted by the mother after laying, which facilitates the quantification of hormones transmitted by the mothers. In addition, the measurement and experimental manipulation of maternal hormones in the egg after it has been laid is not confounded by maternal physiology. These advantages combined with their well-known ecology and evolution, birds have become the most extensively studied taxa in research on the function of maternal hormones (Groothuis et al. 2019).
For choosing Japanese quails as our model, we provided our reasons detailed below and in the introduction (lines 121-130):
Japanese quails are easy to maintain in captivity, and their short generation time makes it a good model to investigate the long-term effects of maternal hormones. Rearing birds in captivity allowed us to apply a powerful within-female experimental design (i.e. knowing which chick hatched from which egg which is not feasible in field studies), thus reducing the effect of random variation among females Moreover, studying the role of natural variation of prenatal THs in precocial species may give additional information to previous studies in altricial species. Finally, Japanese quail is a commonly used model in maternal hormone research with substantial literature available (e.g. McNabb, Blackman & Cherry, 1985; McNabb, Dicken & Cherry, 1985; Wilson & McNabb, 1997; Okuliarova et al., 2011).
Introduction. We veer into avian territory petty quickly with not setup into why looking at maternal thyroid hormones in this system would be particularly interesting. In other words, there is very little context for how an avian system can help us understand this phenomenon. I assumed its largely because embryonic THs can be more easily manipulated? Other reasons? I would recommend strengthening this link, where you go from humans to birds.
Thank you for suggestions, we restructured the introduction accordingly. Now the first entire two pages are dealing with what we know about thyroid hormones in vertebrates in general, where we have added additional literature (lines 43, 56, 58, 75-76). Then we added a paragraph on why birds are a good model, next we discuss the literature on birds from line 101 onwards and end with why we chose quail as a study species.
43-44. Maternal effects of thyroid hormones is not a novel concept. Of the many examples in the literatures, Wilson and McNabb (1997) show that maternal thyroid hormones influence embryonic development in the same species of quail used here — perhaps this would be relevant to in the ms.
We agree with the comment. We wanted to argue that maternal THs have received less attention than steroids in the context of maternal hormones. We modified the sentence accordingly (lines 42-46):
While research on maternal thyroid hormones has emerged between the 80s and the 90s in several taxa (mammals, Morreale De Escobar et al., 1985; fish, Brown et al., 1988; birds, Wilson & McNabb, 1997), these hormones are still underrepresented in the literature of hormone-mediated maternal effects (reviewed in Ruuskanen & Hsu, 2018).
67-79. Now we’re back to all vertebrates. I would recommend restructuring the introduction so that there is a logical flow.
See our detailed reply above
104-133. There are a lot of references to studies showing answers to these questions in other species. In this context, I think the introduction would benefit from justifying why these same questions are important in the Japanese quail — is the novel angle specifically looking at differences between T3 and T4? If this is the case, I recommend tailoring the introduction to this question and contextualising differences in T3/T4 action.
We do not fully agree, because except the recent studies on birds that did find effects of physiological variations of yolk THs, very few other studies (in other models) have, to our knowledge, looked at the effects of physiological variation in maternal THs and none have discussed why such effects from physiological variation in maternal THs may have evolutionary consequence. In mammals (humans and rats), research has essentially focused on consequences of clinical hypo- and hyperthyroidism with a few exception showing correlations between healthy mothers and children. In fish, research has mostly been applied to aquaculture by using supra-physiological doses. We have no intention to depreciate the values of those research. However, neither clinical TH disorders nor supra-physiological TH manipulation can provide relevant information from an eco-evolutionary view, because the former will be unlikely to survive in the wild and the latter simply does not happen in nature. Therefore, for eco-evolutionary relevant studies, manipulations need to reflect variations occurring in nature. We added a short paragraph along these lines (lines 73-81).
The purpose of using Japanese quails in this study is severalfold and now detailed from line 121-130. See our detailed response above.
Concerning the potential difference between T3 and T4, we agree that it is an important point of our study. This is why we have written a paragraph about it from line 112-119. We emphasised the novelty of the design at 133, and the new title also highlights this aspect of our study. Finally, we provide effects sizes for the contrasts between all treatments in the new Table 4.
Experimental design
I also echo concerns over statistical power of previous reviewers. I do not necessarily have an issue with the smaller sample sizes themselves, but rather the unequal distribution in sample size across treatment. This means the chances of finding significant differences between treatment X and control for a given metric are substantially higher for some treatments than others — i.e. small samples more prone to type II error. Further, it is difficult to appreciate the findings on hatching success without the actual numbers of unfertilized eggs for each treatment. Could this be added to the results? Did I miss it? As mentioned by a previous reviewer, I also think further replication with a strict control treatment (in addition to the sham control used here) would be useful.
We agree that these low sample sizes decrease statistical power. Unfortunately, it is impossible for us to replicate this experiment. The unequal sample size across the four treatment groups (hence different probability of type II error) is also a consequence of the different hatching success in each group. However, it is not the unbalanced sample size between controls on the one hand and treatment groups on the other that in itself would strongly affect the statistical power to detect differences with the control groups, as the referee seem to suggest, but rather the somewhat low sample size (especially in the control group). Given these low sample sizes, we think the best way to deal with this is, as we already did, is to report effect sizes and statistical power, so readers will be able to evaluate whether the “non-significance” could be possibly true or due to insufficient power.
The number of unfertilised eggs per treatment has been added at line 212 and has been discussed above.
Methods + Results
146-147. Just use the range — 6.6 eggs is not particularly meaningful, but the range gives an idea of how sample sizes may be spread out.
The average value has been removed.
I would expect methodology this invasive to include both a control and a sham control — in other words, injection into an egg alone does not in itself represent an ecologically relevant control. A strict control would allow you to test for the “normal” hatching success rate in these birds and investigate potential interactions between injection and thyroid hormone actions.
Having a strict control would give information on the effect of the injection itself and on the interaction between injection and yolk THs. Although this is of interest, it was not the purpose of our study. As a matter of fact, the majority of the research on maternal hormones typically do not include a group of unmanipulated control.
The only way to know exactly how THs interact with the negative effect of the injection, would be to include a method of yolk TH manipulation that does not involve injection (e.g., egg-dipping method as in Perrin et al., J. Reprod. Fertil. 1995). Thus, with the following treatments: TH injection, sham injection, TH-dipping solution, and sham-dipping solution, one would be able to tease apart the effects of injection and yolk THs on hatching success.
Nevertheless, the enhanced hatching success in the T3T4 and T4 groups in our study may partly be explained by the fact that injected THs helped the embryos to overcome the negative effects induced by the injection procedure. Since it is not possible for us to repeat this experiment, we briefly discuss this possibility and suggest that, for future studies, it might be worthwhile including a group of non-injected eggs (lines 505-509).
181-183. Good. This is an important point.
183-184. This is not a problem so much as the idea that manipulated hormone levels may allow some embryos to deal with injection trauma better than others. Again, this is where a strict control treatment would have been useful. Were all eggs injected by the same individual?
Yes, all the eggs were injected by the same individual and over one day. Therefore, we expect the negative effects induced by the injection procedure (i.e. the injection trauma termed by the reviewer) should be similar across all treatment groups. Nevertheless, as we explained above, the higher hatching success in T3T4 group (and T4 group) may be partly attributed to injected THs helping overcome the “injection trauma”. Therefore, we discuss this in the discussion (line 505-509).
209. Was this observer blind?
Yes, the observer did not know the treatment of the birds.
220. Is this the sample size of males per treatment? Probably not, right? Unless I’m mistaken, this seems misleading. It would be more appropriate to denote the sample size of males per treatment (i.e. N=6-7)?
The sample size per treatment was 4 CO, 4 T3, 8 T4 and 12 T3T4 and has been added.
223-224. Use the sample size notation for the above.
The sample size per treatment was 7 CO, 11 T3, 16 T4 and 20 T3T4 and has been added.
281. Why only 1,000 simulations? 10,000 is commonly the “unofficial” minimum and would be much more believable, especially when p hovers around 0.05.
We ran again the model on data without the unfertilised eggs and tested different number of simulations (500, 1000, 5000 and 10000). It turned out that the p value was always hovering between 0.02 and 0.03, suggesting that in this model, the estimation on p values is robust against the number of simulations. Because parametric bootstrapping for high numbers of simulations can be time consuming, we therefore decide to use 1 000 simulations.
292. I assume this takes unfertilized eggs into account?
Previous percentages and numbers included unfertilised eggs. This has been changed in this version.
296-299. Good.
337. I don’t understand this sentence — was the point to avoid over-parameterizing the model? Could you not have looked at each one separately?
These are the oxidative stress responses we measured. As these responses are often correlated, we performed first a PCA to reduce the number of models in this paper. This is a common way to analyse such traits (e.g. Rainio et al. 2015).
360-362. I’m trying to understand how 21 out of 40 eggs for T3/T4 was significant, but 20 out of 40 for T4 was p>0.09? This is where more information in lines 179-181 would be more useful.
The p values have changed after removing the unfertilised eggs (hatched eggs: T3T4 = 21/31 and T4 = 20/32), yet the difference remains (0.03 vs 0.08, lines 393-398). It seems that one more hatched egg in this case is just sufficient to cross the 0.05 threshold of significance level. This is also due to correction for multiple comparisons between the treatments (p value without correction = 0.02 ; p value with correction = 0.08). This difference also demonstrates how adherence to strict p-values only can be misleading in biological terms.
Validity of the findings
I would consider changing the title further to reflect that small sample sizes may have promoted type II errors — i.e. you did not find evidence that there are clear effects of thyroid hormones into adulthood, not that there “are no clear effects”.
Our use of the “clarity” language followed the suggestions by Dushoff et al. (2019) (added line 326), and served the same purpose as this comment. However, to avoid any misunderstanding by the readers, the title has been changed to a more general form.
" | Here is a paper. Please give your review comments after reading it. |
9,822 | "<ns0:abstract xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:p>Maternal thyroid hormones (THs) are known to be crucial in embryonic development in humans, but their influence on other, especially wild, animals remains poorly understood.</ns0:p><ns0:p>So far, the studies that experimentally investigated the consequences of maternal THs focused on short-term effects, while early organisational effects with long-term consequences, as shown for other prenatal hormones, could also be expected. In this study, we aimed at investigating both the short-and long-term effects of prenatal THs in a bird species, the Japanese quail Coturnix japonica. We experimentally elevated yolk TH content (the prohormone T 4 , and its active metabolite T 3 , as well as a combination of both hormones). We analysed hatching success, embryonic development, offspring growth and oxidative stress as well as their potential organisational effects on reproduction, moult, and oxidative stress in adulthood. We found that eggs injected with T 4 had a higher hatching success compared with control eggs, suggesting conversion of T 4 into T 3 by the embryo. We detected no evidence for other short-term or long-term effects of yolk THs.</ns0:p><ns0:p>These results suggest that yolk thyroid hormones are important in the embryonic stage of precocial birds, but other short-and long-term consequences remain unclear. Research on maternal thyroid hormones will greatly benefit from studies investigating how embryos use and respond to this maternal signalling. Long-term studies on prenatal THs in other taxa in the wild are needed for a better understanding of this hormone-mediated maternal pathway.</ns0:p></ns0:div>
</ns0:abstract>
<ns0:body xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:head>Introduction</ns0:head><ns0:p>Maternal effects represent all the non-genetic influences of a mother on her offspring and have received increasing attention in evolutionary and behavioural ecology. Through maternal effects, mothers can influence the fitness of their progeny by adapting their phenotype to expected environmental conditions ('adaptive maternal effects' in <ns0:ref type='bibr' target='#b8'>Marshall and Uller, 2007;</ns0:ref><ns0:ref type='bibr'>Mousseau and Fox, 1998)</ns0:ref>, and this view is now also incorporated in the human disease literature <ns0:ref type='bibr'>(Gluckman, Hanson & Spencer, 2005)</ns0:ref>. Maternal hormones transferred to the offspring can mediate important maternal effects. Historically, research on maternal hormones has mostly focused on steroid hormones <ns0:ref type='bibr'>(Groothuis et al., 2005;</ns0:ref><ns0:ref type='bibr'>von Engelhardt & Groothuis, 2011)</ns0:ref>. While research on maternal thyroid hormones has emerged between the 80s and the 90s in several taxa <ns0:ref type='bibr'>(mammals, Morreale De Escobar et al., 1985;</ns0:ref><ns0:ref type='bibr'>fish, Brown et al., 1988;</ns0:ref><ns0:ref type='bibr'>birds, Wilson & McNabb, 1997)</ns0:ref>, these hormones are still underrepresented in the literature on hormone-mediated maternal effects (reviewed in <ns0:ref type='bibr'>Ruuskanen & Hsu, 2018)</ns0:ref>.</ns0:p><ns0:p>Thyroid hormones (THs) are metabolic hormones produced by the thyroid gland and are present in two main forms: the prohormone thyroxine (T 4 ) and the biologically active form triiodothyronine (T 3 ). THs play a crucial role in various aspects of an individual's life, e.g.</ns0:p><ns0:p>development, metabolism and reproduction, across vertebrates, including humans (Morreale de Escobar, Obregon & Escobar del <ns0:ref type='bibr' target='#b18'>Rey, 2004;</ns0:ref><ns0:ref type='bibr'>Krassas, Poppe & Glinoer, 2010)</ns0:ref>. In humans, physiological variation of maternal THs (i.e. no clinical symptoms in both mothers and foetuses) is found to be associated with infant birth weight and IQ in older children <ns0:ref type='bibr' target='#b14'>(Medici et al., 2013;</ns0:ref><ns0:ref type='bibr'>Korevaar et al., 2016)</ns0:ref>. In other vertebrates, THs in general play a role in brain development and neuronal turnover (mammals, Morreale de Escobar, Obregon & Escobar del <ns0:ref type='bibr' target='#b18'>Rey, 2004;</ns0:ref><ns0:ref type='bibr'>birds, McNabb, 2007)</ns0:ref>. THs control the endothermic heat production, and are therefore important in Manuscript to be reviewed thermoregulation in homeothermic species <ns0:ref type='bibr'>(mammals, Danforth & Burger, 1984;</ns0:ref><ns0:ref type='bibr'>birds, McNabb & Darras, 2015)</ns0:ref>. THs can act, in concert with other hormonal axes, as mediators of life stage transitions across vertebrates (reviewed in <ns0:ref type='bibr'>Watanabe et al., 2016)</ns0:ref>. The interaction between thyroid hormones and corticosteroids on amphibian metamorphosis is a well-known example of such effect on life stage transition <ns0:ref type='bibr'>(Kikuyama et al., 1993;</ns0:ref><ns0:ref type='bibr'>Wada, 2008)</ns0:ref>. THs are involved in gonadal development, and hyperthyroidism tends to accelerate maturation <ns0:ref type='bibr'>(Holsberger & Cooke, 2005)</ns0:ref>, and coordinate the transition between reproduction and moult <ns0:ref type='bibr' target='#b11'>(McNabb and Darras, 2015)</ns0:ref>.</ns0:p><ns0:p>Administration of exogenous THs is known to stop egg laying and induce moult in birds <ns0:ref type='bibr'>(Sekimoto et al., 1987;</ns0:ref><ns0:ref type='bibr'>Keshavarz & Quimby, 2002)</ns0:ref>. THs are also involved in photoperiodic control in seasonal breeding <ns0:ref type='bibr'>(Dardente, Hazlerigg & Ebling, 2014)</ns0:ref>. For example, thyroidectomised starlings transferred to long photoperiods became insensitive to future changes in photoperiod, and short photoperiod did not induce gonadal regression <ns0:ref type='bibr'>(Dawson, 1993)</ns0:ref>. While there has been recent research effort on the influence of maternal THs on offspring traits across vertebrate taxa, there are still substantial gaps in our knowledge. Manipulating yolk hormones within the natural range of a species is necessary to better understand the role of maternal THs in an eco-evolutionary context. In humans, studies have essentially looked at the consequences of clinical hyper-or hypothyroidism (but see <ns0:ref type='bibr' target='#b14'>Medici et al., 2013)</ns0:ref>. Research in fish has applied supra-physiological doses for aquaculture purposes <ns0:ref type='bibr'>(Brown et al., 2014)</ns0:ref>. However, these studies do not give information on how variations within the natural range of the species would shape offspring phenotype and affect its fitness, in turn influencing evolution. While recent literature on birds has shown that even physiological variations of prenatal THs can have phenotypic consequences <ns0:ref type='bibr'>(Ruuskanen et al., 2016;</ns0:ref><ns0:ref type='bibr'>Hsu et al., 2017</ns0:ref><ns0:ref type='bibr'>Hsu et al., , 2019;;</ns0:ref><ns0:ref type='bibr'>Sarraude et al., 2020)</ns0:ref>, Manuscript to be reviewed this view is still underrepresented in maternal THs research. Besides, research on maternal thyroid hormones up to date has mainly investigated the short-term effects of prenatal THs on developing fish <ns0:ref type='bibr'>(Brown et al., 1988;</ns0:ref><ns0:ref type='bibr'>Raine et al., 2004)</ns0:ref> and amphibians <ns0:ref type='bibr'>(Duarte-Guterman et al., 2010;</ns0:ref><ns0:ref type='bibr'>Fini et al., 2012)</ns0:ref> and pre-fledging birds <ns0:ref type='bibr'>(Ruuskanen et al., 2016;</ns0:ref><ns0:ref type='bibr'>Hsu et al., 2017</ns0:ref><ns0:ref type='bibr'>Hsu et al., , 2019;;</ns0:ref><ns0:ref type='bibr'>Sarraude et al., 2020)</ns0:ref>. So far, only a study on rock pigeons has looked at the influence of yolk THs on post-fledging survival and found no effect <ns0:ref type='bibr'>(Hsu et al., 2017)</ns0:ref>. None of these studies in any taxa investigated the potential organisational effects of prenatal THs on life-history stage transitions in adult life. Early exposure to elevated THs may affect the hypothalamic-pituitary-thyroid (HPT) axis (humans and mice: <ns0:ref type='bibr'>Alonso et al., 2007;</ns0:ref><ns0:ref type='bibr'>Srichomkwun et al., 2017;</ns0:ref><ns0:ref type='bibr' target='#b3'>Anselmo et al., 2019)</ns0:ref>, via epigenetic modifications for example, such as those induced by adverse early life conditions <ns0:ref type='bibr'>(Jimeno et al., 2019)</ns0:ref> or yolk testosterone <ns0:ref type='bibr' target='#b6'>(Bentz, Becker & Navara, 2016)</ns0:ref>.</ns0:p><ns0:formula xml:id='formula_0'>PeerJ</ns0:formula><ns0:p>Oviparous species, such as birds, are suitable models for studying the role of maternal hormones on the progeny because embryos develop in eggs outside mothers' body. The content of an egg cannot be adjusted by the mother after laying, which facilitates the quantification of hormones transmitted by the mothers. In addition, the measurement and experimental manipulation of maternal hormones in the egg after it has been laid is not confounded by maternal physiology. These advantages combined with their well-known ecology and evolution, birds have become the most extensively studied taxa in research on the function of maternal hormones <ns0:ref type='bibr'>(Groothuis et al. 2019)</ns0:ref>.</ns0:p><ns0:p>Previous studies on prenatal THs in birds focused only on altricial species (great tits, <ns0:ref type='bibr'>Ruuskanen et al., 2016;</ns0:ref><ns0:ref type='bibr'>rock pigeons, Hsu et al., 2017;</ns0:ref><ns0:ref type='bibr'>collared flycatchers, Hsu et al., 2019</ns0:ref><ns0:ref type='bibr'>, pied flycatchers, Sarraude et al., 2020)</ns0:ref>. Embryonic development differs substantially between of well-developed embryos before hatching, as found in rock pigeons <ns0:ref type='bibr'>(Hsu et al., 2017)</ns0:ref>. We therefore looked at the age at mortality in unhatched eggs. Third, we expect elevated yolk THs to affect chick growth (in body mass, tarsus and wing length) either positively <ns0:ref type='bibr'>(Wilson & McNabb, 1997;</ns0:ref><ns0:ref type='bibr'>Hsu et al., 2019;</ns0:ref><ns0:ref type='bibr'>weak effect in Sarraude et al., 2020</ns0:ref><ns0:ref type='bibr'>), negatively (Hsu et al., 2017)</ns0:ref>, or in a sex-specific manner <ns0:ref type='bibr'>(Ruuskanen et al., 2016)</ns0:ref>. Prenatal THs may exert most of their effects in the offspring early life; this is why we separately tested both posthatch morphological traits and the growth curve. Similarly, we also independently analysed morphological traits at adulthood, as these traits may affect the fitness of an individual. For example, small adult females may lay smaller eggs and larger males may be more dominant. Fourth, we predict that yolk THs will have organisational effects on life-history stage transitions; that is, age at sexual maturity and male gonadal regression (using cloacal gland size as a proxy), and moult when birds are exposed to short photoperiod. Based on the literature mentioned above we expect elevated yolk THs to advance the timing of puberty, gonadal regression and moult. The rate of moult should also be influenced, with birds receiving experimental TH elevation moulting faster. Previous studies have reported that gravid female three-spined sticklebacks (Gasterosteus aculeatus) exposed to predatory cues produced eggs with higher corticosterone <ns0:ref type='bibr'>(Giesing et al., 2011)</ns0:ref>, disturbed embryonic transcriptome <ns0:ref type='bibr' target='#b16'>(Mommer & Bell, 2014)</ns0:ref>, offspring with altered anti-predator behaviour <ns0:ref type='bibr'>(Giesing et al., 2011)</ns0:ref> and modified cortisol response in adulthood <ns0:ref type='bibr' target='#b15'>(Mommer & Bell, 2013)</ns0:ref>. We may therefore expect elevated yolk THs to similarly induce long-term behavioural changes in response to environmental cues (i.e. photoperiod), via organising effects during the embryonic development. We also explored the effects of yolk THs on reproductive investment in females, another important fitness aspect. Finally, yolk THs may increase oxidative stress due to their stimulating effects on metabolism.</ns0:p><ns0:p>bands. The floor was covered with 3-5cm sawdust bedding. A hiding place, sand and calcium grit were provided. Each pair was housed in indoor aviary dividing into pens of 1 m² floor area.</ns0:p><ns0:p>The temperature was set to 20°C with a 16L:8D photoperiod (light from 06.00 to 22.00). Food (Poultry complete feed, 'Kanan Paras Täysrehu', Hankkija, Finland) was provided ad libitum and water was changed every day.</ns0:p><ns0:p>Pairs were monitored every morning to collect eggs for 7 days. Eggs were individually marked (non-toxic marker), weighed and stored in a climate-controlled chamber at 15°C and 50% relative humidity. On the last day of collection, a total of 4 to 8 eggs per pair were injected with a solution (see next section).</ns0:p></ns0:div>
<ns0:div><ns0:head>Preparation of the solution, injection procedure and incubation</ns0:head><ns0:p>The preparation of hormone solution and the procedure of injection were based on previous studies <ns0:ref type='bibr'>(Ruuskanen et al., 2016;</ns0:ref><ns0:ref type='bibr'>Hsu et al., 2017)</ns0:ref>. In brief, crystal T 4 (L-thyroxine, ≥ 98% HPCL, CAS number 51-48-9, Sigma-Aldrich) and T 3 (3,3',5-triiodo-L-thyronine, > 95% HPCL, CAS number 6893-02-3, Sigma-Aldrich) were first dissolved in 0.1M NaOH and then diluted in 0.9% NaCl. The injection of thyroid hormones resulted in an increase of two standard deviations (T 4 = 8.9 ng/egg, equivalent to 1.79 pg/mg yolk; T 3 = 4.7 ng/egg, equivalent to 1.24 pg/mg yolk), a recommended procedure for hormone manipulation within the natural range <ns0:ref type='bibr'>(Ruuskanen et al., 2016;</ns0:ref><ns0:ref type='bibr'>Hsu et al., 2017;</ns0:ref><ns0:ref type='bibr'>Podmokła, Drobniak & Rutkowska, 2018)</ns0:ref>. The control solution (CO) was a saline solution (0.9% NaCl). Manuscript to be reviewed put sideways, allowing yolks to float up to the middle position. Before injection, the shell was disinfected with a cotton pad dipped in 70% EtOH. We used a 27G needle (BD Microlance ™)</ns0:p><ns0:p>to pierce the eggshell and then used a 0.3 ml syringe to deliver 50 µl of the respective hormone solution or control. After injection, the hole was sealed with a sterile plaster (OPSITE Flexigrid, Smith&Nephew). In total, 158 eggs were injected and divided as follows over the treatments: T 3 treatment (N = 39); T 4 treatment (N = 39); T 3 +T 4 treatment (N = 40); and control, CO (N = 40). To balance the genetic background of the parents and the effect of storage, each egg laid by the same female was sequentially assigned to a different treatment and the order of treatments was rotated among females. After injection, eggs were placed in an incubator at 37.8°C and 55% relative humidity.</ns0:p><ns0:p>Until day 14 after starting incubation, eggs were automatically tilted every hour by 90°. On day 14, tilting was halted and each egg was transferred to an individual container to monitor which chick hatched from which egg. On day 16 after injection, (normal incubation time = 17 days), the temperature was set to 37.5°C and the relative humidity to 70%. Eggs were checked for hatching every 4 hours from day 16 onwards. Four days after the first egg hatched, all unhatched eggs were stored in a freezer and dissected to determine the presence of an embryo. The age of developed embryos was assessed according to <ns0:ref type='bibr' target='#b0'>Ainsworth et al. (2010)</ns0:ref>.</ns0:p></ns0:div>
<ns0:div><ns0:head>Rearing conditions of the experimental birds</ns0:head><ns0:p>In total, 66 chicks hatched (N = 10 CO, 15 T 3 , 20 T 4 and 21 T 3 T 4 ), yielding a rather low overall hatching success (ca. 40%). Among the unhatched eggs, 33.7% (31 out of 92) had no developed embryos, and these were evenly distributed between the treatments (CO = 9/40, T 3 = 8/39, T 3 T 4 = 8/40, and T 4 = 6/39 eggs). Discarding the unfertilised eggs gives an overall hatching success of ca. 51%. Previous studies on Japanese quails have reported comparable hatching</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:03:47322:2:0:NEW 10 Sep 2020)</ns0:p><ns0:p>Manuscript to be reviewed success, even in unmanipulated eggs (e.g. 40% in Okuliarová, <ns0:ref type='bibr'>Škrobánek & Zeman, 2007;</ns0:ref><ns0:ref type='bibr'>ca. 60% in Pick et al., 2016 and</ns0:ref><ns0:ref type='bibr'>in Stier, Metcalfe &</ns0:ref><ns0:ref type='bibr'>Monaghan, 2019)</ns0:ref>. In addition, the injection procedure itself is also known to reduce hatching success to some extent <ns0:ref type='bibr'>(Groothuis & von Engelhardt, 2005)</ns0:ref>. Twelve hours after hatching, the chicks were marked by a unique combination of coloured rings and nail coding and transferred to two cages of 1 m² floor area and ca. 30 cm height (ca. 30 chicks/cage, sex and treatments mixed together). The chicks were provided with heating mats and lamps as extra heat sources for the first two weeks. The chicks were fed with sieved commercial poultry feed ('Punaheltta paras poikanen', Hankkija, Finland), and provided with Calcium and bathing sand. Two weeks after hatching, the chicks were separated in four 1 m² cages (ca. 30 cm high) of about 16 individuals. Around 3 weeks after hatching, coloured rings were replaced by unique metal rings. On week 4 after hatching, birds were transferred to eight pens of 1 m² floor area (average of 7.1 birds/pen, range = 4-9), under the same conditions as the parents. Around the age of sexual maturity (ca. 6-8 weeks after hatching), the birds were separated by sex in twelve 1 m² pens (average of 4.8 birds/pen, range = 4-5). The chicks were under the same photoperiod as the adults (i.e. 16L:8D).</ns0:p></ns0:div>
<ns0:div><ns0:head>Monitoring of growth and reproductive maturation</ns0:head><ns0:p>Body mass and wing length were measured twelve hours after hatching. Tarsus was not measured because it bends easily, resulting in inaccurate measures and potential harm for the young. From day 3 to day 15, these three traits were monitored every 3 days. From day 15 to day 78 (ca. 12 weeks), chicks were measured once a week. Body mass was recorded using a digital balance to the nearest 0.1 g. Wing and tarsus lengths were respectively measured with a ruler and a calliper to the nearest 0.5 mm and 0.1 mm. The sample size for the growth analysis was 7 CO, 11 T 3 , 18 T 4 and 21 T 3 T 4 . From week 6 to week 10, we monitored cloacal gland development Manuscript to be reviewed and foam production in 28 males. Cloacal glands were measured every other day with a calliper to the nearest 0.1 mm as a proxy for testes development and sexual maturation <ns0:ref type='bibr' target='#b7'>(Biswas et al., 2007)</ns0:ref>. Foam production (by gently squeezing the cloacal gland) was assessed at the same time and coded from 0 (no foam) to 3 (high production of foam), as a proxy of cloacal gland function <ns0:ref type='bibr'>(Cheng et al., 1989a;</ns0:ref><ns0:ref type='bibr'>Cheng et al., 1989b)</ns0:ref>. The same observer performed all measurements. We collected eggs produced by 10-week-old females over a 6-day period and recorded their mass to the nearest 0.1 g. We collected on average 5.7 eggs (range = 4-7) per female from 28 females.</ns0:p></ns0:div>
<ns0:div><ns0:head>Monitoring of cloacal gland regression and moult</ns0:head><ns0:p>In Japanese quails, exposure to short photoperiod and cold temperature triggers reproductive inhibition and postnuptial moulting <ns0:ref type='bibr'>(Tsuyoshi & Wada, 1992)</ns0:ref>. Thyroid hormones are known to coordinate these two responses (see introduction). When the birds reached the age of ca. 7 months, we exposed birds to short photoperiod (8L:16D, i.e., light from 08.00 to 16.00) with a 12:12-h cycle of normal (20°C) and low (9°C) temperature (low temperature was effective from 18.00 to 06.00). Cloacal gland regression (as a proxy for testes regression) was monitored every other day for 2 weeks with a calliper by measuring the width and length to obtain the area of the gland to the nearest 0.1 mm² (N = 26 males; 4 CO, 4 T 3 , 8 T 4 and 12 T 3 T 4 ). Primary moult was recorded from a single wing by giving a score to each primary from 0 (old feather) to 5 (new fully-grown feather) following Ginn and Melville (1983) (N = 54 males and females; 7 CO, 11 T 3 , 16 T 4 and 20 T 3 T 4 ). The total score of moulting was obtained by adding the score of all feathers. The sample size for the moult analysis was 7 CO, 11 T 3 , 16 T 4 and 20 T 3 T 4 .</ns0:p></ns0:div>
<ns0:div><ns0:head>Oxidative status biomarker analyses</ns0:head><ns0:p>Two blood samples were drawn, when birds were 2 weeks (N = 51 chicks) and 4 months old (N </ns0:p></ns0:div>
<ns0:div><ns0:head>Ethics</ns0:head><ns0:p>The study complied with Finnish regulation and was approved by the Finnish Animal Manuscript to be reviewed cause. One male was euthanised before the end of the experiment due to severe head injury. At the end of the experiment, all birds were euthanised by decapitation for collection of tissue samples (not used in this study).</ns0:p></ns0:div>
<ns0:div><ns0:head>Statistical analysis</ns0:head><ns0:p>Data were analysed with the software R version 3.5.3 (R core team, 2019). In this study, two different statistical approaches were used: null-hypothesis testing with Generalised Linear Mixed Models (GLMMs) and Linear Mixed Models (LMMs), and multimodel inference with Generalised Additive Mixed Models (GAMMs). GAMMs were used to analyse the data on body and cloacal gland growth to account for its non-linear pattern (see Growth). In this analysis, we preferred multimodel inference as GAMMs generate many candidate models that cannot be directly compared (e.g., by the Kenward-Roger approach). Instead, candidate models were ranked based on their Akaike Information Criterion (AIC) values. Models with a ΔAIC ≤ 2 from the top-ranked model were retained in the set of best models. Akaike weights of all models were calculated following <ns0:ref type='bibr'>(Burnham & Anderson, 2002)</ns0:ref>, and evidence ratios of the top-ranked models were calculated as the weight of a model divided by the weight of the null model <ns0:ref type='bibr'>(Burnham, Anderson & Huyvaert, 2011)</ns0:ref>. To estimate the effect of the predictors, we computed the 95% confidence intervals from the best models using the nlme package <ns0:ref type='bibr'>(Pinheiro et al., 2018)</ns0:ref>.</ns0:p><ns0:p>GLMMs and LMMs were fitted using the R package lme4 <ns0:ref type='bibr' target='#b5'>(Bates et al., 2015)</ns0:ref>, and GAMMs When presenting and discussing our results, we use the language of statistical 'clarity' rather than statistical 'significance' as suggested by <ns0:ref type='bibr'>Dushoff et al. (2019)</ns0:ref>.</ns0:p></ns0:div>
<ns0:div><ns0:head>Hatching success</ns0:head><ns0:p>To analyse hatching success, each egg was given a binary score: 0 for unhatched egg and 1 for hatched egg. A GLMM was fitted with a binomial error distribution (logit link) and mother identity as a random intercept and the 4-level treatment as the predictor. Egg mass might affect hatchability and was therefore added as a covariate in both models. The potential effect of storage duration on hatchability (Reis, Gama & Soares, 1997) was accounted for by including laying order as a covariate in both models. This covariate allowed us to control for the age of the egg as well.</ns0:p></ns0:div>
<ns0:div><ns0:head>Duration of embryonic period, age at embryonic mortality and early morphological traits</ns0:head><ns0:p>Duration of embryonic period and early morphological traits (mass and wing length at hatching, and tarsus length at day 3) were modelled with separate LMMs. Treatment, sex of the individuals and egg mass were included as fixed factors. Laying order was added as a covariate to account for potential effects of storage duration on hatching time and on chick weight (Reis, Gama & Soares, 1997). Mother identity was included as a random intercept.</ns0:p><ns0:p>The data for embryonic age had a skewed distribution and residuals were not normally distributed and heterogenous, which violated LMM assumptions on residual distribution. We Manuscript to be reviewed therefore performed a simple Kruskal-Wallis test.</ns0:p></ns0:div>
<ns0:div><ns0:head>Growth</ns0:head><ns0:p>As growth curves typically reach an asymptote, we fitted non-linear GAMMs to these curves.</ns0:p><ns0:p>Growth in body mass, tarsus and wing length were analysed in separate GAMMs. Growth was analysed until week 10 after hatching as all birds appeared to have reached their maximum body mass and tarsus and wing length. The data are composed of repeated measurements of the same individuals over time; therefore, we first corrected for temporal autocorrelation between the measurements using an ARMA(1,1) model for the residuals <ns0:ref type='bibr'>(Zuur et al., 2009)</ns0:ref>. Second, as mothers produced several eggs, the models included nested random effects, with measured individuals nested into mother identity, allowing for random intercepts. GAMMs allow modelling the vertical shift of the curves (i.e., changes in intercepts) and their shape. Treatment and sex were included as predictors. A smoothing function for the age of the birds was included to model the changes in the growth curves, and was allowed to vary by sex or treatment only, or none of these predictors. The interaction between sex and treatment was not analysed due to low statistical power. Additive effect of treatment and sex was tested for the intercept but could not be computed for curve shape. All combinations of the relevant predictors were tested for both shape parameters (i.e., intercept and curve shape). Prenatal THs may exert most of their effects in the offspring early life; this is why we additionally tested hatchlings morphological traits apart from the growth curve. Likewise, we also analysed separately morphological traits at adulthood (ca. 9 weeks old), as these traits may condition the fitness of an individual. Because of sex differences and low sex-specific sample sizes, we standardised the measures within sex and regressed the standardised responses against treatment in a linear regression.</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:03:47322:2:0:NEW 10 Sep 2020)</ns0:p><ns0:p>Manuscript to be reviewed</ns0:p></ns0:div>
<ns0:div><ns0:head>Reproductive maturation, regression and investment</ns0:head><ns0:p>Due to low sample sizes in these sex-specific responses, we could not perform robust statistical analyses. We therefore present these analyses and results in the supplementary material and only briefly discuss them (Figs. S6 to S9; Table <ns0:ref type='table'>S3</ns0:ref>).</ns0:p></ns0:div>
<ns0:div><ns0:head>Moult</ns0:head><ns0:p>Two parameters of moult were analysed in separate LMMs: the timing of moult (i.e., the moult score after one week of short photoperiod), and the rate of moult (i.e., how fast birds moulted).</ns0:p><ns0:p>Both models included treatment and sex as fixed factors, and mother identity as a random intercept. The rate of moult was tested by fitting an interaction between treatment and age. This model also included the main effect of age and individual identity, nested within mother identity, as a random intercept to account for repeated measures. Estimated marginal means and standard errors (EMMs ± SE) were derived from the model using the package emmeans (Lenth, 2019).</ns0:p></ns0:div>
<ns0:div><ns0:head>Oxidative stress</ns0:head><ns0:p>A principal component analysis (PCA) was first performed on measured antioxidant markers (SOD, CAT, GPx, tGSH and GST), to reduce the number of metrics for subsequent analyses.</ns0:p><ns0:p>The first and the second principal components (PCs) explained together 60.2% of the variance (Table <ns0:ref type='table'>1</ns0:ref>). PC1 and PC2 were then used as dependent variables in separate LMMs. LMMs included the treatment, sex and age of individuals (2 weeks and 4 months old) as fixed factors and the 2-way interactions between treatment and sex, and treatment and age. Mother and individual identities, to account for repeated measures, were added as random intercepts.</ns0:p><ns0:p>Malondialdehyde (MDA) is a marker of oxidative damage, which is a different measure from antioxidant activity, and was therefore analysed in a separate LMM using the same parameters as Manuscript to be reviewed for PC1 and PC2, adding the batch of the assay as an additional random intercept. The marker of cell oxidative status (GSH:GSSG ratio) was analysed with the same model used for PC1 and PC2.</ns0:p></ns0:div>
<ns0:div><ns0:head>Results</ns0:head></ns0:div>
<ns0:div><ns0:head>Effects of prenatal THs on hatching success and age of embryo mortality</ns0:head><ns0:p>There was a clear effect of elevated prenatal THs on hatching success (GLMM, p = 0.03, Fig. <ns0:ref type='figure' target='#fig_14'>1</ns0:ref>).</ns0:p><ns0:p>Tukey post-hoc analysis revealed that hatching success in the T 3 T 4 (66%) group was statistically higher than in the CO group (32%) (Tukey z = 2.77, p = 0.03) . There was a non-significant trend between the T 4 (61%) and the CO groups (z = 2.37, p = 0.08). There were no clear differences in hatching success between the T 3 (48%) and the CO group (z = 1.25, p = 0.45), or between the hormone treatments (all z < 1.61, all p > 0.37). Dissection of the unhatched eggs showed that age of embryo mortality did not differ between the treatments (Kruskal-Wallis χ² = 7.22, df = 3, p = 0.07; Fig. <ns0:ref type='figure' target='#fig_14'>S1</ns0:ref>). Finally, the manipulation of yolk THs did not affect the duration of embryonic period (LMM, F 3,42.0 = 0.57, p = 0.64, Fig. <ns0:ref type='figure' target='#fig_15'>S2</ns0:ref>). Sex of the embryo or egg mass (LMM sex, F 1,49.7 = 2.63, p = 0.11; LMM egg mass, F 1,19.3 = 0.01, p = 0.92) were also not associated with the duration of the embryonic period.</ns0:p><ns0:p>Laying order (i.e. the effect of storage duration) was not correlated with any of the responses (all p ≥ 0.25).</ns0:p></ns0:div>
<ns0:div><ns0:head>Effects of prenatal THs on growth</ns0:head><ns0:p>Mass at hatching was not influenced by the elevation of prenatal THs (LMM, F 3,35.0 = 0.81, p = 0.50, Fig. <ns0:ref type='figure' target='#fig_16'>S3</ns0:ref>). Mass at hatching was positively correlated with egg mass (LMM, Estimate±SE = 0.72±0.10 g, F 1,24.1 = 46.9, p < 0.001). Although we detected no clear differences on hatchling Manuscript to be reviewed morphological traits (body mass, wing and tarsus length) due to prenatal THs (all p > 0.12), the calculated effect sizes (Cohen's d[95%CI]) and achieved statistical power yielded additional information regarding the potential effects of prenatal THs (Table <ns0:ref type='table' target='#tab_2'>2</ns0:ref>). For body mass, the effect sizes were low and the achieved statistical power was very low. For wing length, the effect sizes were moderate and the achieved statistical power was low. For tarsus length, the effect sizes were moderate to large and the achieved statistical power was low to moderate. Similarly, adult morphology was not affected by the treatment (all p > 0.13), but effect sizes indicate small to large effects of prenatal THs (Table <ns0:ref type='table' target='#tab_2'>2</ns0:ref>). For body mass, the effect sizes were small and the achieved power was low. For wing length, the effect sizes were large and the achieved power was moderate. For tarsus length, the effect sizes were small to large and the achieved power was moderate to high.</ns0:p><ns0:p>Regarding body mass growth, the top-ranked model showed that the curve shape and the intercept differ according to sex (Table <ns0:ref type='table'>3</ns0:ref>). After 10 weeks, females had a larger body mass than males (mean±SE females = 214.4±5.7 g, males = 172.4±4.5 g, Fig. <ns0:ref type='figure' target='#fig_15'>2</ns0:ref>), which was supported by the 95% CIs (Table <ns0:ref type='table'>4</ns0:ref>). Based on model selection we conclude that the treatment had no effect on body mass growth (Table <ns0:ref type='table'>3</ns0:ref>). For wing growth, the top-ranked model (ΔAIC ≤ 2) included sex in the intercept, while treatment was not included in the best supported model (Table <ns0:ref type='table'>S1</ns0:ref>). The 95% CIs (Table <ns0:ref type='table'>3</ns0:ref>) confirmed that males had a lower wing length than females (Fig. <ns0:ref type='figure' target='#fig_17'>S4</ns0:ref>). Concerning tarsus growth, the models within ΔAIC ≤ 2 included no predictors for the curve shape but included treatment for the intercept (Table <ns0:ref type='table' target='#tab_2'>S2</ns0:ref>). The 95% CIs of the parameter estimates from these models suggested that there was a slight negative effect of T 3 T 4 treatment on tarsus growth (Table <ns0:ref type='table'>4</ns0:ref>, Fig. <ns0:ref type='figure'>S5</ns0:ref>). However, as the estimates were close to 0 (Table <ns0:ref type='table'>4</ns0:ref>) and Manuscript to be reviewed evidence ratios showed that the model with treatment as a predictor was only 3.5 times more supported than the null model (Table <ns0:ref type='table' target='#tab_2'>S2</ns0:ref>), we conclude that the effect of THs on tarsus length is likely to be very small. Likewise, the second model for tarsus length included sex as a predictor for the intercept, but its 95% CIs overlapped with 0 (Table <ns0:ref type='table'>4</ns0:ref>). We therefore conclude that sex had no effect on tarsus growth.</ns0:p></ns0:div>
<ns0:div><ns0:head>Effects of prenatal THs on postnuptial moult</ns0:head><ns0:p>As expected, birds started to moult soon after being exposed to short photoperiod, with an average increase of moult score by 6 per week (SE = 0.2, F 1,254.0 = 827.4, p < 0.001, Fig. <ns0:ref type='figure' target='#fig_16'>3</ns0:ref>). The first moult score (assessed one week after switching to short photoperiod) was not affected by the treatment (LMM, F 3,42.7 = 0.36, p = 0.78), but was influenced by sex, with females having a higher score than male (EMMs ± SE: female = 21.4 ± 1.6, male = 7.2 ± 1.7; LMM F 1,45.3 = 41.9, p < 0.001). Yolk TH elevation did not affect the rate of moult (LMM interaction treatment × time, F 3,251.0 = 0.59, p = 0.62, Fig. <ns0:ref type='figure' target='#fig_16'>3</ns0:ref>).</ns0:p></ns0:div>
<ns0:div><ns0:head>Effects of prenatal THs on oxidative stress</ns0:head><ns0:p>The elevation of yolk THs had no effect on PC1 or PC2 of antioxidants at either 2 weeks ('chicks') or 4 months ('adults') old (LMM on PC1, F 3,40.3 = 2.40 , p = 0.08; LMM on PC2, F 3,42.2 = 0.92, p = 0.44, treatment × age, F < 0.91, p > 0.44). The age of the birds had a highly significant effect on PC1, with chicks generally having higher antioxidant capacities (CAT, GST and tGSH) than adults (LMM, Estimate±SE = -1.34±0.19, F 1,49.2 = 52.1, p < 0.001). All the other predictors had no effect on either PC1 or PC2 (all F < 2.93 and all p > 0.09).</ns0:p><ns0:p>The marker of oxidative damage, MDA, was affected by the elevation of yolk THs (LMM, F 3,43.6 = 3.08, p = 0.04, Fig. <ns0:ref type='figure' target='#fig_17'>4</ns0:ref>). Tukey post-hoc analysis showed that the T 4 group had</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:03:47322:2:0:NEW 10 Sep 2020)</ns0:p><ns0:p>Manuscript to be reviewed higher MDA values than the T 3 group (Estimate±SE = 0.01±0.004, Tukey contrast p = 0.01), but none of the groups differed from the control (Tukey p-values > 0.19). However, this result became non-significant when removing the outlier in the T 4 group (LMM, F 3,43.1 = 2.68, p = 0.06). MDA levels were not affected by the age or the sex of individuals (LMM age, F 1,54.4 = 0.30, p = 0.59; LMM sex, F 1,42.0 = 1.47, p = 0.23).</ns0:p><ns0:p>The marker of cell oxidative balance, GSH:GSSG, was not influenced by the yolk THs nor by the sex of the birds (LMM treatment, F 3,33.0 = 0.85, p = 0.48; LMM sex, F 1,40.6 = 0.57, p = 0.45). However, chicks had a higher GSH:GSSG ratio than adults (LMM, Estimate±SE = 0.17±0.04, F 1,50.0 = 18.3, p < 0.001).</ns0:p></ns0:div>
<ns0:div><ns0:head>Discussion</ns0:head><ns0:p>The aim of this experimental study was to investigate the potential short-term and organisational effects (with long-term consequences) of maternal thyroid hormones (THs) in a precocial species, the Japanese quail, by experimental elevation of THs in eggs. Our study is the first to investigate the effects of yolk T 3 and T 4 separately, within the natural range of the study model.</ns0:p><ns0:p>In addition we studied both short-and long-term effects on embryonic development, growth, life stage transitions and oxidative stress. We detected a positive effect of yolk THs on hatching success. All other response variables studied were not clearly affected by elevated prenatal THs.</ns0:p></ns0:div>
<ns0:div><ns0:head>Effects of prenatal THs on hatching success and embryonic development</ns0:head><ns0:p>The overall low hatching success, and especially in the control group, forces us to interpret these results with caution. In addition, we cannot exclude that our results may be partly due to selective disappearance of lower quality embryos in the control group and with injected THs helping lower quality chicks to hatch. This might have biased the results after hatching, but is Manuscript to be reviewed still a relevant effect of the hormone treatment. We found that hatching success almost doubled when the eggs received an injection of both T 4 and T 3 , or an injection of T 4 only. Previous similar studies reported comparable effects of yolk THs in rock pigeons <ns0:ref type='bibr'>(Hsu et al., 2017)</ns0:ref> and in collared flycatchers <ns0:ref type='bibr'>(Hsu et al., 2019)</ns0:ref>. In these studies, injections consisted of a mixture of both T 3 and T 4 . Given that mostly T 3 binds to receptors, these results suggest that embryos likely express deiodinase enzymes to convert T 4 to T 3 , and/or yolk may contain maternally derived deiodinase mRNA, as injection with T 3 only did not differ from control. Indeed, deiodinase expression has previously been characterised in chicken embryos already 24h after the onset of incubation <ns0:ref type='bibr'>(Darras et al., 2009)</ns0:ref>. An old study found that injecting T 4 close to hatching can advance hatching time, which suggests that yolk THs may help embryos overcoming hurdles close to hatching <ns0:ref type='bibr' target='#b4'>(Balaban & Hill, 1971)</ns0:ref>. In contrast with our study, two similar studies in altricial species detected no increased hatching success due to the injection of THs <ns0:ref type='bibr'>(Ruuskanen et al., 2016;</ns0:ref><ns0:ref type='bibr'>Sarraude et al., 2020)</ns0:ref>. The dissimilarities between the studies may come from interspecific differences in terms of utilisation of yolk THs by the embryos or from contextdependent effects (e.g. due to other egg components). Further comparative and mechanistic studies could help understanding the dynamic of yolk THs during incubation. Increased yolk THs did not influence age of embryo mortality. Similar to our study, <ns0:ref type='bibr'>Ruuskanen et al. (2016)</ns0:ref> did not find any difference in the timing of mortality in great tit embryos. Conversely, the study on rock pigeons found that yolk THs increased the proportion of well-developed embryos <ns0:ref type='bibr'>(Hsu et al., 2017)</ns0:ref>. Similar to our result on hatching success, yolk TH effects on embryonic development may differ in a species-specific manner.</ns0:p><ns0:p>Our results on hatching success may partly be attributed to yolk THs balancing the negative effects of injections on embryonic survivability. Further studies may aim at Manuscript to be reviewed understanding the contribution of THs to counteract the effect of injection. To do so, such studies may use an non-invasive method to manipulate yolk THs (e.g., egg-dipping method as in <ns0:ref type='bibr'>Perrin et al. 1995)</ns0:ref>, in addition to injected controls, like in our study.</ns0:p></ns0:div>
<ns0:div><ns0:head>Effects of prenatal THs on growth</ns0:head><ns0:p>We found no apparent influence of yolk THs on growth, contrary to our expectations based on the recent literature. Other comparable studies found either a positive <ns0:ref type='bibr'>(Hsu et al., 2019;</ns0:ref><ns0:ref type='bibr'>weak effect in Sarraude et al., 2020)</ns0:ref>, a negative <ns0:ref type='bibr'>(Hsu et al., 2017)</ns0:ref> or a sex-specific effect <ns0:ref type='bibr'>(Ruuskanen et al., 2016)</ns0:ref> of yolk THs on growth. This notable difference may be due to the captive conditions experienced by the Japanese quails in our study, with unrestricted access to food and water.</ns0:p><ns0:p>Although the pigeon study also provided ad libitum food, parents still needed to process food before feeding their nestlings in the form of crop milk, whereas precocial quails have no such limitation. In addition, the Japanese quail has been domesticated for many generations, and probably selected for rapid growth for economic reasons. Whole-genome sequencing in chickens showed that domestication induced a strong positive selection on genes associated with growth <ns0:ref type='bibr'>(Rubin et al., 2010)</ns0:ref>. Interestingly, that study also found a strong selection for a locus associated with thyroid stimulating hormone (TSH) receptor. TSH controls most of the TH production by the thyroid gland <ns0:ref type='bibr' target='#b11'>(McNabb & Darras, 2015)</ns0:ref>, and this artificial selection may overshadow the effects of natural variations of prenatal THs on growth. Besides, the low number of individuals in the control and T 3 groups (7 and 11, respectively) limited the statistical power to detect differences between all the treatments. Indeed, we were able to detect small to moderate negative effects of yolk THs on morphological traits at hatching and in adulthood. Such negative effects, although small, may still be biologically relevant. Repeating the study with a larger sample size may allow us to ascertain the effects of yolk THs on growth in precocial study models. Research Manuscript to be reviewed on the influences of prenatal THs on growth will also benefit from experimental studies on wild precocial species.</ns0:p></ns0:div>
<ns0:div><ns0:head>Effects of prenatal THs on postnuptial moult</ns0:head><ns0:p>Short photoperiod in combination with cold temperature triggered primary moult, as expected.</ns0:p><ns0:p>However, we detected no effect of yolk THs on the timing or speed of moult. Thyroid hormones are important in moult and feather growth (reviewed in <ns0:ref type='bibr'>Dawson, 2015)</ns0:ref>. For example, thyroidectomised birds fail to moult after being exposed to long photoperiods <ns0:ref type='bibr'>(Dawson, 2015)</ns0:ref>.</ns0:p><ns0:p>In addition, thyroidectomised nestling starlings failed to grow normal adult plumage and grown feathers presented an abnormal structure <ns0:ref type='bibr'>(Dawson et al., 1994)</ns0:ref>. By removing the thyroid gland, these two studies implemented extreme pharmacological protocols that differ drastically from our injection of physiological doses. In addition, our experimental design, increasing TH exposure (vs decreased TH exposure in the above-mentioned studies), may have different consequences. For example, there may be a threshold above which any additional hormones may not affect moult. Overall, our results show no support for the hypothesis of organising effect of prenatal THs on life stage transitions. Yet, due to small sample sizes in sex-specific analyses (i.e., male gonadal maturation and regression, and female reproductive investment), there remains a</ns0:p><ns0:p>relatively high uncertainty about the potential organising effects of prenatal THs. Replicate studies with larger samples sizes and different study models will reduce this uncertainty.</ns0:p></ns0:div>
<ns0:div><ns0:head>Effects of prenatal THs on oxidative stress</ns0:head><ns0:p>In contrast to our predictions, elevated yolk THs did not affect oxidative status during chick or adult phase. We found no changes in antioxidant activities in relation to yolk THs and no</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:03:47322:2:0:NEW 10 Sep 2020)</ns0:p><ns0:p>Manuscript to be reviewed imbalance in the oxidative cell status. Nevertheless, T 4 birds had a higher level of oxidative damage on lipids than T 3 birds, but this was a weak effect driven by one outlier. The lack of effects on chick oxidative status among the treatment groups could be explained by the absence of treatment effects on growth, given that high growth rates usually result in higher oxidative stress and damage (e.g. <ns0:ref type='bibr' target='#b2'>Alonso-Alvarez et al., 2007)</ns0:ref>. In turn, the lack of treatment effects on adult oxidative status may suggest no organisational effects of prenatal THs on adult metabolism.</ns0:p><ns0:p>Two recent studies in altricial species also found no influence of yolk THs on nestling oxidative stress <ns0:ref type='bibr'>(Hsu et al., 2019;</ns0:ref><ns0:ref type='bibr'>Sarraude et al., 2020</ns0:ref>), yet telomere length, a biomarker of aging was affected <ns0:ref type='bibr'>(Stier et al., 2020)</ns0:ref>. Our study shows for the first time that prenatal THs have no influence on adult oxidative stress either. The previous study focused on a limited set of biomarkers: one antioxidant enzyme, oxidative damage on lipids and oxidative balance. In the present study, we measured 7 biomarkers, thus providing broader support to the absence of effects of prenatal THs on post-natal/hatching oxidative stress.</ns0:p></ns0:div>
<ns0:div><ns0:head>Conclusion</ns0:head><ns0:p>To our knowledge, this study is the first one to experimentally investigate the consequences of natural variations of maternal THs not only early but also in adult physiology and postnuptial moult in any vertebrate. Furthermore, this study explored for the first time the effects of maternal T 3 and T 4 separately. We found no evidence for differential effects of maternal T 4 and T 3, while an effect of T 4 , alone or in combination with T 3 , on hatching success suggests that T 4 is converted into T 3 , the biologically active form during embryonic development. Contrary to similar studies on wild altricial species, we found no influence of maternal THs on growth.</ns0:p><ns0:p>Further research on embryos utilisation of maternal THs may help understand the differences observed between precocial and altricial species. Studies in other vertebrates are urgently needed Manuscript to be reviewed Cohen's d, 95% CIs and achieved statistical power for post-hatching and adult morphological measures (body mass, wing and tarsus length). Manuscript to be reviewed Manuscript to be reviewed Manuscript to be reviewed Manuscript to be reviewed </ns0:p></ns0:div><ns0:figure xml:id='fig_0'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:03:47322:2:0:NEW 10 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_1'><ns0:head /><ns0:label /><ns0:figDesc>reviewing PDF | (2020:03:47322:2:0:NEW 10 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_2'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:03:47322:2:0:NEW 10 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_3'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:03:47322:2:0:NEW 10 Sep 2020)Manuscript to be reviewed = 49 adults), respectively. The sample size per treatment was 7 CO, 11 T 3 , 17 T 4 and 20 T 3 T 4 . 200 µl of blood was collected from the brachial vein in heparinized capillaries and directly frozen in liquid nitrogen. Then, the samples were stored at -80°C until analyses. We measured various biomarkers of antioxidant status; the antioxidant glutathione (tGSH), the ratio of reduced and oxidised glutathione (GSH:GSSG) and activity of the antioxidant enzymes glutathione peroxidase (GPx), catalase (CAT) and superoxide dismutase (SOD) from the blood. Measuring multiple biomarkers of oxidative and antioxidant status allows a broader understanding of the mechanism, and the interpretation of the results is more reliable if multiple markers show similar patterns. The GSH:GSSG ratio represents the overall oxidative state of cells and a low ratio reveals oxidative stress(Hoffman, 2002; Isaksson et al., 2005; Lilley et al., 2013; Rainio et al., 2013; Halliwell & Gutteridge, 2015). GPx enzymes catalyse the glutathione cycle, whereas CAT and SOD directly regulate the level of reactive oxygen species (ROS)(Ercal, Gurer-Orhan & Aykin-Burns, 2001; Halliwell & Gutteridge, 2015). The methodology for measuring each biomarker is described in detail inRainio et al. (2015). All analyses were conducted blindly of the treatment following Ruuskanen et al (2017).</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_4'><ns0:head /><ns0:label /><ns0:figDesc>were fitted using the package mgcv(Wood, 2017). P-values for GLMMs were obtained by parametric bootstrapping with 1,000 simulations and p-values for LMMs were calculated by model comparison using Kenward-Roger approximation, using the package pbkrtest in both cases(Halekoh & Højsgaard, 2014). Post-hoc Tukey analyses were conducted with the package multcomp(Hothorn et al., 2008). Model residuals were checked visually for normality andPeerJ reviewing PDF | (2020:03:47322:2:0:NEW 10 Sep 2020)Manuscript to be reviewed homoscedasticity. Covariates and interactions were removed when non-significant (α = 0.05).Effect size calculations (Cohen's d and 95%CI) were performed with the website estimationstats.com(Ho et al., 2019) and statistical power analyses were performed using t-tests for independent means withGPower (Faul et al., 2009) with the effect size values calculated.</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_5'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:03:47322:2:0:NEW 10 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_6'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:03:47322:2:0:NEW 10 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_7'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:03:47322:2:0:NEW 10 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_8'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:03:47322:2:0:NEW 10 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_9'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:03:47322:2:0:NEW 10 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_10'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:03:47322:2:0:NEW 10 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_11'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:03:47322:2:0:NEW 10 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_12'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:03:47322:2:0:NEW 10 Sep 2020) Manuscript to be reviewed 553 to understand the potential organising effects of maternal THs with long-term consequences. PeerJ reviewing PDF | (2020:03:47322:2:0:NEW 10 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_13'><ns0:head>95%</ns0:head><ns0:label /><ns0:figDesc>CIs were calculated by bootstrap resampling with 5,000 resamples. CO = control, T 4 (thyroxine) = injection of T 4 , T 3 (triiodothyronine) = injection of T 3 , T 3 T 4 = injection of T 3 and T 4. PeerJ reviewing PDF | (2020:03:47322:2:0:NEW 10 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_14'><ns0:head>Figure 1</ns0:head><ns0:label>1</ns0:label><ns0:figDesc>Figure 1</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_15'><ns0:head>Figure 2</ns0:head><ns0:label>2</ns0:label><ns0:figDesc>Figure 2</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_16'><ns0:head>Figure 3</ns0:head><ns0:label>3</ns0:label><ns0:figDesc>Figure 3</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_17'><ns0:head>Figure 4 MDA</ns0:head><ns0:label>4</ns0:label><ns0:figDesc>Figure 4</ns0:figDesc></ns0:figure>
<ns0:figure type='table' xml:id='tab_2'><ns0:head>Table 2 (on next page)</ns0:head><ns0:label>2</ns0:label><ns0:figDesc /><ns0:table /></ns0:figure>
</ns0:body>
" | " Dear Editors and Reviewers,
We are glad to see that overall reviewers are satisfied with our revisions and happy to address the further issues. Concerning the low and unbalanced sample sizes, raised again by reviewer 2, we agree that this is a limitation of our manuscript. As discussed in our previous revision, we have tried to account for this uncertainty by providing effect sizes, confidence intervals and power analyses. We also acknowledge throughout the manuscript the limitations induced by our low sample sizes, and we also avoid making strong claims.
In the statistical analysis section, we moved the oxidative stress paragraph to the end of the section, to remain consistent with the method, result and discussion sections.
We hope that these further clarifications and discussions will be satisfying to the reviewers and that you can now consider our manuscript for publication.
Sincerely,
On behalf of my co-authors,
Tom Sarraude
Reviewer 1
Basic reporting
The authors have addressed all my major concerns.
Experimental design
The authors have addressed all my major concerns.
Validity of the findings
The authors have addressed all my major concerns.
Comments for the author
When Figures 2, 3, S4, and S5 are printed on a black and white printer, one can not tell which treatment each line is for.
Indeed as the reviewer said, Fig2 in black and white is much less readable than the coloured version. This is also true when fitting different types of lines for each group. We expect the same for the other figures. Since to our knowledge, PeerJ allows coloured figures, we prefer to keep the figures in colour. Besides, we a chose colour-blind friendly theme to make sure the figures are accessible to everyone.
Figure legend for S1: it references itself (Fig. S1) for the description of treatments. I think, since you added a new S1, the material to be referenced needs to move from S2 up to S1.
Indeed, it was a mistake that has been corrected.
Reviewer 2
Basic reporting
The intro is much stronger and provides great justification for the study. It now runs a little long — now that you have the general bird justification, I recommend moving the Japanese quail justification (118-127) to the methods. I also recommend condensing the text in the intro thereafter so that it focuses on hypotheses but leaves much of the more specific details for the methods.
Thank you for the positive comment. We moved the paragraph on Japanese quails a new method subsection (overview of the method). We also moved a few sentences on the experimental design to this subsection. However, we believe that our hypotheses and predictions need some background information so they can be understood by the readers.
Experimental design
I have no major issues with the design itself though I think a true control would have been very informative.
Validity of the findings
In terms of the statistical problems, I maintain there is an issue with the unequal sample sizes. The imbalance is not just between the control and treatment groups as the authors assert but also between treatment groups themselves. In the cloacal gland recession analyses, for instance, there are 2-fold and 3-fold differences in sample size between T3 vs T4 and T3 vs T3/T4 treatments, respectively. This trend is similarly true for the other analyses. These differences, coupled with the especially low sample size of the control group, means that you have more power to find statistical differences between the T3/T4 treatment and control (for instance) than for the T3 treatment and control. The same is true for T3 vs T4 to a lesser but nonetheless worrisome extent. The added issue arises from the authors reporting largely negative results (as indicated by their previous title). In short, I simply have trouble accepting that the negative results may reflect anything but a lack of statistical power (or unequal power in the experimental design).
We agree with the fact that our low and unbalanced sample sizes limited the statistical power of our analyses. This is why, as mentioned in our first letter, we also provided the readers with effect sizes and power analyses for morphological traits in hatchlings and adults (Table 2). Therefore, readers can judge themselves whether the absence of effects we report are true negative results or simply due to a lack of power.
Lack of power in studies with unequal sample sizes is often aggravated by unequal variances across the treatment groups (Rusticus and Lovato 2014, PARE 11:1–10). However, based on model residuals and tests on the homogeneity of variance (Bartlet and Levene tests), variance across the groups was uniform in this study, thus this aspect may not contribute strongly to the power. We do not provide these details in the manuscript since we mention the visual checking of the residuals (L. 304-305). However, if the reviewer wants, we can add a sentence about the tests on the homogeneity of variance.
In the discussion (L. 500-505, 521-524), we acknowledged that we may have lacked power to detect differences between groups, and we encourage repeating this study with larger sample sizes. As suggested by the reviewer, we have now also indicated that this also concerns comparisons among the different treatment groups (not only control vs THs, L. 502). In addition, we remained cautious in the abstract by saying that the short- and long-term effects of prenatal THs remain unclear (L. 27).
Besides, for our sex-specific analyses (e.g. cloacal gland development and regression) we explicitly wrote (L. 350-353) that we could not perform reliable analyses. This is why we only present these results in the supplements. In the supplements, we also explicitly admit that the low sample sizes do not allow us to make robust comparisons.
All in all, throughout the manuscript we analysed and presented our results in a conservative fashion, avoiding strong claims such as prenatal THs having no effects on early development and adults.
I also now read the statistical methods more skeptically. This is a lot of effort and work-around analyses for an experimental design that should be statistically straightforward to analyse via GLM ANOVA. It does not seem to be a particularly parsimonious way of analyzing the results. I also hesitate to even use descriptive statistics when N<5.
We disagree with the suggestion that the data presented here can be analysed with simple GLMs.
The variety of the analyses presented reflects the nature and the structure of the data collected: non-linear models (GAMMs) were necessary to fit S-shaped growth curves; mixed models (GLMMs and LMMs) are used to account for repeated measures within individuals as well as shared genetic background between individuals from the same mother; the PCA analysis (for the oxidative stress response) is also a commonly used method to analyse multiple traits that may correlate with each other.
Furthermore, we have now checked all our mixed models again to see how much of variation was actually explained by the random effects, to justify the need of the random term (although it is a critical part of the experimental design). In cases when the variance estimate of a random factor collapses to 0, it may be because the model is too complex for the data, and in which case one might want to have the random factor removed from the model (singularity issue). We found one singularity issue with the LMM on tarsus length after hatching (motherID variance = 0), suggesting low variation across maternal genetic background. Such a model is therefore equivalent to an LM with the same predictors (treatment and egg mass). Since this is the only model where we detected singularity, we therefore decide to keep our LMM for consistency. All other models found that individual and/or mother identities explained substantial variance (variance estimate > 0) to be retained in the models.
Besides, although the sample size of the control group is low, it is always equal to 7 individuals (for the results present in the main text). That being said, it is true that our sex-specific data include groups with < 5 individuals. Nevertheless, we already admit (in the main text L. 350-353, and in the supplements) that our sample sizes are too low to properly interpret our results. This is why we did not consider using only descriptive statistics.
" | Here is a paper. Please give your review comments after reading it. |
9,823 | "<ns0:abstract xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:p>Background. Antimicrobial peptides have a broad spectrum of antimicrobial activities and are attracting attention as promising next-generation antibiotics against multidrug-resistant (MDR) bacteria. The all-denantiomer [ D (KLAKLAK) 2 ] has been reported to have antimicrobial activity against Escherichia coli and Pseudomonas aeruginosa, and to be resistant to protein degradation in bacteria because it is composed of D-enantiomer compounds. In this study, we demonstrated that modification of [ D (KLAKLAK) 2 ] by the addition of an L-cysteine residue to its N-or C-terminus markedly enhanced its antimicrobial activities against Gram-negative bacteria such as MDR Acinetobacter baumannii, E. coli, and P. aeruginosa.</ns0:p></ns0:div>
<ns0:div><ns0:head>Methods.</ns0:head><ns0:p>The peptides [ D (KLAKLAK) 2 ] (DP), DP to which L-cysteine was added at the N-terminus C-DP, and DP to which L-cysteine was added at the C-terminus DP-C, were synthesized at >95% purity. The minimum inhibitory concentrations of peptides and antibiotics were determined by the broth microdilution method. The synergistic effects of the peptides and the antibiotics against MDR P. aeruginosa were evaluated using the checkerboard dilution method. In order to assess how these peptides affect the survival of human cells, cell viability was determined using a Cell Counting Kit-8.</ns0:p><ns0:p>Results. C-DP and DP-C enhanced the antimicrobial activities of the peptide against MDR Gram-negative bacteria, including A. baumannii, E. coli, and P. aeruginosa. The antimicrobial activity of DP-C was greater than that of C-DP, with these peptides also having antimicrobial activity against drug-susceptible P. aeruginosa and drug-resistant P. aeruginosa overexpressing the efflux pump components. C-DP and DP-C also showed antimicrobial activity against colistin-resistant E. coli harboring mcr-1, which encodes a lipid A modifying enzyme. DP-C showed synergistic antimicrobial activity against MDR P. aeruginosa when combined with colistin. The LD 50 of DP-C against a human cell line HepG2 was six times higher than the MIC of DP-C against MDR P. aeruginosa. The LD 50 of DP-C was not altered by incubation with low-dose colistin.</ns0:p></ns0:div>
<ns0:div><ns0:head>Conclusion.</ns0:head><ns0:p>Attachment of an L-cysteine residue to the N-or C-terminus of [ D (KLAKLAK) 2 ] enhanced its antimicrobial activity against A. baumannii, E. coli, and P. aeruginosa. The combination of C-DP or DP-C and colistin had synergistic effects against multi-drug resistant P. aeruginosa. In addition, DP-C and C-DP showed much stronger antimicrobial activity against MDR A. baumannii and E. coli than against P. aeruginosa.</ns0:p></ns0:div>
</ns0:abstract>
<ns0:body xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:head>Introduction</ns0:head><ns0:p>The emergence and spread of multidrug-resistant (MDR) Gram-negative pathogens has become a serious public health problem worldwide. A global priority list of antibiotic-resistant bacteria published by the World Health Organization (WHO) to guide research, discovery, and development of new antibiotics listed carbapenem-resistant Acinetobacter baumannii, Pseudomonas aeruginosa and carbapenem-resistant and third-generation cephalosporin-resistant Enterobacteriaceae as first priority pathogens <ns0:ref type='bibr'>(Shai, 2002)</ns0:ref>.</ns0:p><ns0:p>Antimicrobial peptides (AMPs) are produced by various host organisms and partially contribute to the host's innate immunity <ns0:ref type='bibr'>(Plesniak et al., 2004;</ns0:ref><ns0:ref type='bibr'>Onuchic, Jennings & Ben-Jacob, 2013)</ns0:ref>. These peptides exhibit potent antimicrobial activities against a wide range of microorganisms, including viruses, bacteria, protozoa, and fungi <ns0:ref type='bibr'>(Shai, 2002)</ns0:ref>. AMPs are chemically amphiphilic polycationic peptides, generally comprising 6-50 amino acid residues, and they constitute a unique and diverse group of molecules <ns0:ref type='bibr'>(Peters, Shirtliff & Jabra-Rizk, 2010)</ns0:ref>. AMPs are classified according to their secondary structures including mixtures of α-helices, β-sheets, loops, and extended peptides. Most of these peptides are thought to bind to the cytoplasmic membrane, forming micelle-like aggregates that destroy the membrane <ns0:ref type='bibr'>(Peters, Shirtliff & Jabra-Rizk, 2010)</ns0:ref>. These peptides orient parallel to the interface, and associate with the membrane surface. After reaching a threshold concentration on the bilayer surface, they aggregate promoting channel formation through the bilayer and disrupt the membrane <ns0:ref type='bibr'>(Matsuzaki et al., 1995;</ns0:ref><ns0:ref type='bibr'>Plesniak et al., 2004;</ns0:ref><ns0:ref type='bibr'>Onuchic, Jennings & Ben-Jacob, 2013)</ns0:ref>. Because of their mechanisms of action, AMPs show antimicrobial activities against MDR as well as drug-susceptible Gramnegative bacteria. Thus, AMPs are attracting attention as promising next-generation antibiotics for the treatment of MDR bacterial infections <ns0:ref type='bibr'>(Hancock & Lehrer, 1998;</ns0:ref><ns0:ref type='bibr'>Marr, Gooderham & Hancock, 2006)</ns0:ref>. The all-D-enantiomer, [ D (KLAKLAK) 2 ] is an amphipathic lysine-leucine-rich α-helical peptide with high antimicrobial activity against Escherichia coli and P. aeruginosa, but with low toxicity against mouse <ns0:ref type='bibr'>3T3 cells (Javadpour et al., 1996;</ns0:ref><ns0:ref type='bibr'>McGrath et al., 2013)</ns0:ref>. [ D (KLAKLAK) 2 ] orients parallel to the interface and associates with the outer membrane surface. After reaching a threshold concentration on the outer membrane surface, the peptides aggregate to promote channel formation through the bilayer <ns0:ref type='bibr'>(Matsuzaki et al., 1995;</ns0:ref><ns0:ref type='bibr'>Plesniak et al., 2004;</ns0:ref><ns0:ref type='bibr'>Onuchic, Jennings & Ben-Jacob, 2013)</ns0:ref>. [ D (KLAKLAK) 2 ] was reported to selectively interfere with the bilayer of the outer membranes of E. coli and P. aeruginosa, leading to cell death by membrane disruption and loss of membrane potential. [ D (KLAKLAK) 2 ] has shown antimicrobial activity against Gram-negative bacteria but not Gram-positive bacteria, as this peptide was unlikely to disrupt the thick peptidoglycan layer of the latter <ns0:ref type='bibr'>(McGrath et al., 2013)</ns0:ref>. One of the strategies used to protect AMPs from protease degradation was sequence modification of D-amino acids to replace L-amino acids <ns0:ref type='bibr'>(Choi et al., 1993;</ns0:ref><ns0:ref type='bibr'>Braunstein, Papo & Shai, 2004;</ns0:ref><ns0:ref type='bibr'>Lee & Lee, 2008)</ns0:ref>. Because this peptide is an all-D-enantiomer, it is highly resistant to proteolysis in bacteria and has low immunogenicity. This stability and low immunogenicity may prolong its half-life and enhance its efficacy at low doses in vivo. Initially we added L-cysteine to the N-or C-terminus of [ D (KLAKLAK) 2 ] in order to conjugate with protein such as a single chain antibody. The side chain of cysteine contains sulfhydryl group, which can make a covalent coupling with an amino group of the protein via a cross-linker molecule such as sulfo-SMCC (sulfosuccinimidyl 4-(Nmaleimidomethyl) cyclohexane-1-carboxylate). The cysteine-rich AMPs were isolated from leguminous plants and the granular hemocytes of mangrove crabs <ns0:ref type='bibr'>(Sivakamavalli, Nirosha & Vaseeharan, 2015;</ns0:ref><ns0:ref type='bibr'>Maróti, Downie & Kondorosi, 2015)</ns0:ref>. The functionalized textiles and nasal prongs modified with L-cysteine exhibited antimicrobial activity against Staphylococcus aureus and Klebsiella pneumoniae <ns0:ref type='bibr'>(Gouveia, Sá & Henriques, 2012;</ns0:ref><ns0:ref type='bibr'>Caldeira et al., 2013;</ns0:ref><ns0:ref type='bibr'>Xu et al., 2017;</ns0:ref><ns0:ref type='bibr'>Odeberg et al., 2018)</ns0:ref>.</ns0:p><ns0:p>Modification of [ D (KLAKLAK) 2 ] by the addition of an L-cysteine residue to its N-or Cterminus markedly enhanced its antimicrobial activities against Gram-negative bacteria such as P. aeruginosa, E. coli, and A. baumannii. The present study describes the antimicrobial activities of the modified peptides against clinical isolates of MDR P. aeruginosa, A. baumannii and E. coli, and its synergistic effects with low dose colistin.</ns0:p></ns0:div>
<ns0:div><ns0:head>Materials & Methods</ns0:head></ns0:div>
<ns0:div><ns0:head>Peptide design and synthesis</ns0:head><ns0:p>The peptides [(D-Lys-D-Leu-D-Ala-D-Lys-D-Leu-D-Ala-D-Lys) 2 ] (DP) (9), DP to which Lcysteine was added at the N-terminus, [L-Cys-(D-Lys-D-Leu-D-Ala-D-Lys-D-Leu-D-Ala-D-Lys) 2 ] (C-DP) and DP to which L-cysteine was added at the C-terminus [(D-Lys-D-Leu-D-Ala-D-Lys-D-Leu-D-Ala-D-Lys) 2 -L-Cys] (DP-C), were synthesized at >95% purity (Scrum Inc., Tokyo, Japan). DP-C was dimerized by heating at 60 ° C for 30 min (DP-C dimer) to convert cysteine to cystine. Purity of synthetic peptides was checked on a SunFire C18 column (100Å, 5 µm, 4.6 mm inner diameter × 250 mm, Waters). Each peptides were gradiently eluted with solution A (water containing 0.1% trifluoroacetic acid) and solution B (acetonitrile containing 0.1% trifluoroacetic acid) at a flow rate of 1.0 mL/min. The elution program for DP and C-DP was as follows: at 0 min, 10% of B; at 20 min, 60% of B. The elution program for DP-C was as follows: at 0 min, 0% of B; at 20 min, 100% of B. The separated components were detected at 220 nm. The DP-C dimer was separated on a COSMOSIL 5C18-AR-300 reversed phase column (4.6 mm inner diameter × 250 mm, Nacalai Tesque, Kyoto, Japan), using an automated HPLC system (LC-2010AHT; Shimadzu, Kyoto, Japan). The reaction products were gradiently eluted with solution A (water containing 0.086% trifluoroacetic acid) and solution B (acetonitrile containing 0.086% trifluoroacetic acid) at a flow rate of 1.0 mL/min. The elution program for DP-C dimer was as follows: at 0 min, 20% of B; at 20 min, 50% of B. The peptide masses were determined by MALDI-TOF MS on a microflex (Bruker, Billerica MA). </ns0:p></ns0:div>
<ns0:div><ns0:head>Drug susceptibility testing</ns0:head><ns0:p>The minimum inhibitory concentrations (MICs) of peptides and antibiotics, including meropenem, amikacin, ofloxacin, and colistin, were determined by the broth microdilution method according to Clinical Laboratory Standards Institute (CLSI) guidelines <ns0:ref type='bibr'>(Weinstein, 2018)</ns0:ref>. Bacterial strains were inoculated at 5 x 10 5 CFU/ml per well into 96-well round-bottom microtiter plates (Watson Bio Lab, Kobe, Japan) containing an equal volume of serially diluted peptide or antibiotic. Three independent experiments were performed to confirm reproducibility.</ns0:p><ns0:p>The synergistic effects of the peptides and the antibiotics amikacin, colistin, meropenem, ofloxacin, and rifampicin against P. aeruginosa NCGM2.S1 were evaluated using the checkerboard dilution method. The peptides were two-fold serially diluted to final concentrations ranging from 0.125-to 2-times the MIC longitudinally in 96-well round-bottom microtiter plates (Watson Bio Lab). Subsequently, antibiotic was two-fold serially diluted to final concentrations ranging from 0.125-to 2-times the MIC transversely into the plates. NCGM2.S1 was inoculated at 5 x 10 5 CFU/ml per well at a volume equal to that of the diluted peptide and antibiotic. Three independent experiments were performed to confirm reproducibility. The synergistic effect of the peptides and antibiotics was assessed by determining the fractional inhibitory concentration (FIC) index <ns0:ref type='bibr'>(Berenbaum, 1978)</ns0:ref>, using the formula: MIC of peptide in combination MIC of antimicrobial agent in combination FIC = MIC of peptide alone MIC of antimicrobial agent alone  An FIC index ≤0.5 was defined as synergistic, an FIC index >0.5 to 4.0 was defined as additive or unrelated, and an FIC index > 4.0 was defined as antagonistic.</ns0:p></ns0:div>
<ns0:div><ns0:head>Cytotoxicity tests</ns0:head><ns0:p>The human hepatoblastoma cell line, HepG2 (ATCC HB-8065), was obtained from American Type Culture Collection and cultured in DMEM supplemented with 10% fetal bovine serum (FBS). HepG2 cells were seeded at 3000 cells/well in 96-well cell culture-treated flat bottom microtiter plates (Falcon, Corning NY). The cells were incubated at 37 °C for 48 h in an atmosphere containing 5% CO 2 , followed by the addition of peptides, at a final concentration of 0-256 µg/ml, or colistin, at a final concentration of 0-3000 µg/ml, by serial dilution. The plates were incubated with 0.2% FBS in 5% CO 2 at 37 °C for 48 h; under these conditions, HepG2 cells were alive but did not grow. Cell viability was determined using a Cell Counting Kit-8 (Dojin, Tokyo, Japan), and colorimetric changes were determined at OD 450 nm with a microplate reader (Corona Electric Co Ltd., Ibaraki, Japan). The LD 50 was defined as the concentration of peptides or colistin that resulted in 50% cell viability. Three independent experiments were performed to confirm reproducibility.</ns0:p></ns0:div>
<ns0:div><ns0:head>Statistical analysis</ns0:head><ns0:p>The Mann-Whitney U test was used to compare the MIC values of C-DP, DP-C, and DP-C dimer with DP in P. aeruginosa and A. baumannii. P-values less than 0.05 were considered statistically significant. Cell survival was expressed as a percentage of the control were obtained as mean ± Standard Deviation (SD) of three independent experiments done in three replicates for each treatment. Significant differences of cell survival rate between each concentration and the control were statistically evaluated by Student's t-test.</ns0:p></ns0:div>
<ns0:div><ns0:head>Results</ns0:head></ns0:div>
<ns0:div><ns0:head>Addition of L-cysteine to the N-or C-terminus enhanced the antimicrobial activity of the original peptide</ns0:head><ns0:p>The HPLC analysis indicated that the purities of synthetic DP, C-DP, and DP-C were 100, 95.52 and 98.68%, respectively (Fig. <ns0:ref type='figure'>S1-3</ns0:ref>). Additionally the masses of DP, C-DP, and DP-C by MALDI-TOF MS analysis were <ns0:ref type='bibr'>1525.444, 1626.595, and 1627.151</ns0:ref>, respectively that matched well with the theoretical molecular weights (1524.0, 1627.1, and 1627.1) (Fig. <ns0:ref type='figure'>S1-3</ns0:ref>). Similarly the formation of DP-C dimer was assessed by HPLC and MALDI-TOF MS. The HPLC analysis of heat treated DP-C showed one large peak estimated as DP-C dimer and one small peak estimated as DP-C monomer, the peak area ratio was 6.3：1 (Fig. <ns0:ref type='figure'>S4</ns0:ref>). The MALDI-TOF MS result of heat treated DP-C was 3252.7, which was in consistent with the estimated molecular weight of DP-C dimer (Fig. <ns0:ref type='figure'>S4</ns0:ref>). The grand average hydropathy (GRAVY) values were -0.07 for DP and 0.1 each for C-DP and DP-C, indicating that the addition of L-cysteine affected the hydrophobicity of DP.</ns0:p><ns0:p>Assessment of antibiotic susceptibility showed that drug-susceptible P. aeruginosa PAO-1 (Weinstein, 2018) and PAO4290 expressing a normal level of <ns0:ref type='bibr'>MexAB-OprM (Yoneyama et al., 1997)</ns0:ref>, were susceptible to all antibiotics tested; MDR P. aeruginosa NCGM2.S1 <ns0:ref type='bibr'>(Miyoshi-Akiyama et al., 2011)</ns0:ref> was susceptible to colistin, but resistant to amikacin, meropenem and ofloxacin; and OCR1, a nalB multidrug-resistant mutant that overproduces the outer membrane protein OprM <ns0:ref type='bibr'>(Poole et al., 1996)</ns0:ref> was susceptible to amikacin and colistin, intermediately susceptible to ofloxacin, but resistant to meropenem (Table <ns0:ref type='table'>1</ns0:ref>). DP showed antimicrobial activity against PAO-1, with an MIC of 300 µg/mL, consistent with previous findings <ns0:ref type='bibr'>(McGrath et al., 2013)</ns0:ref>. DP also had antimicrobial activity against NCGM2.S1 and PAO4290, with MICs of 300 µg/mL, but DP showed no antimicrobial activity against OCR1 within the tested concentration range. C-DP had greater antimicrobial activities than DP against all of these strains (The Mann-Whitney U test, p < 0.05), with MICs of 64-128 µg/mL, and DP-C had greater antimicrobial activities than C-DP and DP (The Mann-Whitney U test, p < 0.05), with MICs 16-32 µg/mL (Table <ns0:ref type='table'>1</ns0:ref>). The DP-C dimer also had antimicrobial activities and showed MICs identical to DP-C against all the strains tested. These results indicate that the addition of L-cysteine to the N-or Cterminus of [ D (KLAKLAK) 2 ] increased its antimicrobial activity. DP showed antimicrobial activity against four A. baumannii strains, with MICs of 64 to 300 µg/mL, and against a clinical isolate of colistin-and carbapenem-resistant E. coli NCCHD1261-5 co-harboring mcr-1 and bla NDM-5 genes <ns0:ref type='bibr'>(Uchida et al., 2018)</ns0:ref>, with an MIC of 64 µg/mL. In contrast, DP was inactive against drug-susceptible E. coli ATCC 25922, two K. pneumoniae strains, S marcescens NBRC102204 and S. aureus ATCC 25923 (Table <ns0:ref type='table' target='#tab_1'>2</ns0:ref>). C-DP and DP-C showed higher antimicrobial activities than DP (The Mann-Whitney U test, p < 0.05), with MICs of 4 to 8 µg/mL against the four A. baumannii strains and MICs of 4 to 16 µg/mL against the two E. coli strains. C-DP and DP-C showed antimicrobial activity against carbapenem-resistant K. pneumoniae ATCC BAA-2146 harboring bla NDM-1 , with both having MICs of 16 µg/mL, but not against penicillin-resistant, β-lactamase-producing K. pneumoniae ATCC 15380, S. marcescens NBRC102204 and S. aureus ATCC 25923 (Table <ns0:ref type='table' target='#tab_1'>2</ns0:ref>).</ns0:p></ns0:div>
<ns0:div><ns0:head>Synergistic effects of peptides and antibiotics</ns0:head><ns0:p>The combinations of DP, C-DP and DP-C with colistin had synergistic effects on antimicrobial activity (Table <ns0:ref type='table'>3</ns0:ref>). For example, the growth of P. aeruginosa NCGM2.S1 was inhibited by a combination of one-sixteenth the MIC of DP (19 µg/mL) and one-fourth the MIC of colistin (0.25 µg/mL), by a combination of one-fourth the MIC of C-DP (32 µg/mL) and oneeighth the MIC of colistin (0.125 µg/mL) and by a combination of one-eighth the MIC of DP-C (4.0 µg/mL) and one-fourth the MIC of colistin (0.25 µg/mL). Although AMPs that induced susceptibility to rifampicin were reported in clinical MDR isolates of P. aeruginosa <ns0:ref type='bibr' target='#b0'>(Baker et al., 2019)</ns0:ref>, DP-C only slightly enhanced the susceptibility to rifampicin. The combinations of C-DP with meropenem or ofloxacin had additive effect on antimicrobial activity. The synergistic effects were not observed when DP-C was combined on amikacin.</ns0:p></ns0:div>
<ns0:div><ns0:head>Cytotoxicity of antimicrobial peptides to HepG2 cells</ns0:head><ns0:p>The LD 50 values of each peptide in HepG2 cells were >256, >256, 192, and 2100 µg/ml for DP, C-DP, DP-C, and colistin, respectively (Fig. <ns0:ref type='figure'>S5</ns0:ref>, Fig. <ns0:ref type='figure'>S6</ns0:ref>). The raw data of Fig. <ns0:ref type='figure'>S5</ns0:ref> and Fig. <ns0:ref type='figure'>S6</ns0:ref> are available in Supplementary file 2 and Supplementary file 3 respectively. Since the antimicrobial activity of these peptides against P. aeruginosa were synergistic with colistin, we examined whether a combination of these peptides and colistin was more toxic than the peptide alone. The combination of DP-C and colistin dose-dependently induced the death of HepG2 cells (Figure <ns0:ref type='figure'>1</ns0:ref>), with more than 50% of the cells dying at 256 µg/ml DP-C and 25.6 µg/ml colistin. The cytotoxicity of the peptide was not enhanced by combination of low doses of colistin. There was no cytotoxicity to HepG2 at 4 µg/ml DP-C and 0.4 µg/ml colistin, which concentrations showed synergistic effects of colistin and DP-C. The raw data of Fig. <ns0:ref type='figure'>1</ns0:ref> is available in Supplementary file 2.</ns0:p></ns0:div>
<ns0:div><ns0:head>Discussion</ns0:head><ns0:p>We added L-cysteine to the N-or C-terminus of [ D (KLAKLAK) 2 ] in order to bind it with another protein like a single-chain antibody in the beginning of the experiment. This attachment yielded two compounds, DP-C and C-DP, with significantly greater antimicrobial activities against A. baumannii, E. coli and P. aeruginosa than the original DP. Similarly, the addition of L-cysteine to AMPs, Andersonin-Y1, HBcARD, buforinII or lysin, enhanced its antimicrobial activity, with higher membrane disruption activity than the original peptide <ns0:ref type='bibr'>(Chen et al., 2018;</ns0:ref><ns0:ref type='bibr'>Pal et al., 2019)</ns0:ref>. The cysteine-derived cationic dipeptides lysine-cysteine, arginine-cysteine and histidine-cysteine presented antimicrobial activity, SEM analysis suggests that these dipeptides interact with cell walls to disrupt membrane integrity <ns0:ref type='bibr'>(Tsai et al., 2020)</ns0:ref>. Whereas addition of an L-cysteine to the C-terminus of indolicidin, magainine or epinecidin-1 did not change their antimicrobial activity <ns0:ref type='bibr'>(Chen et al., 2018)</ns0:ref>. It remains unclear what could be the mechanism by which the addition of cysteine to the N-or C-terminus to AMPs enhances antimicrobial activity. It is unlikely that this effect is simply due to peptide dimerization via cysteine disulfide bond formation because DP-C dimer showed the same MIC value as DP-C monomer against P. aeruginosa strains tested. The cysteine-rich region in Factor C receptors in the horseshoe crab specifically binds to bacterial lipopolysaccharides on Gram-negative bacteria <ns0:ref type='bibr'>(Koshiba, Hashii & Kawabata, 2006)</ns0:ref>. The addition of L-cysteine to the N-or C-terminus of the peptide may have facilitated its binding to the bacterial membrane surface and form structures that disrupt their cell wall. The potential mechanism of efficacy enhancement by the attachment of cysteine residue to AMPs requires further investigation. Systematic hybridization of two lead peptides from unrelated classes of AMPs showed no associations of net charge, charge density, and antipneumococcal activity among the hybrid peptides, although AMPs with higher hydrophobicity values have been reported to have greater antimicrobial activity against Streptococcus pneumoniae <ns0:ref type='bibr'>(Le et al., 2015)</ns0:ref>. The peptides we tested showed a similar trend, when GRAVY was calculated, DPC showed higher hydrophobicity than DP. L-cysteine exhibited preferred antimicrobial activity against S. aureus compared with Dcysteine, whereas D-cysteine showed stronger antimicrobial activity against E. coli, Listeria monocytogenes and Salmonella enteritis <ns0:ref type='bibr'>(Wang et al., 2019)</ns0:ref>. D-amino acid is highly resistant to proteolysis in bacteria, addition of D-cystein instead of L-cystein may be effective. DP-C and C-DP had potent activity against multidrug-resistant Gram-negative pathogens. The emergence and spread of these drug-resistant pathogens has become a serious worldwide public health problem <ns0:ref type='bibr'>(Boucher et al., 2009;</ns0:ref><ns0:ref type='bibr'>Tacconelli et al., 2018)</ns0:ref>. Carbapenem is a last resort βlactam antibiotic administered to treat infections with drug-resistant Gram-negative pathogens. The development of new antibiotics against carbapenem-resistance pathogens is of top priority <ns0:ref type='bibr'>(Tacconelli et al., 2018)</ns0:ref>. DP-C and C-DP also had antimicrobial activity against E. coli strains harboring the plasmid-mediated colistin resistance mcr-1 gene. Colistin is a last line polycationic peptide antibiotic which is used to treat infections with carbapenem-resistant Gram-negative pathogens <ns0:ref type='bibr'>(Paterson & Harris, 2016)</ns0:ref>. However, colistin-resistant mcr-1 producers have emerged in humans and animals in China <ns0:ref type='bibr'>(Liu et al., 2016)</ns0:ref> and have spread worldwide. DP-C and C-DP, like DP, possess a positive charge and hydrophobic regions, suggesting that they target the lipid bilayer of the membrane and destroy it, causing loss of membrane potential and ultimately cell death <ns0:ref type='bibr'>(McGrath et al., 2013)</ns0:ref>. These peptides had antimicrobial activity against Gram-negative but not Gram-positive pathogens. These peptides were also inactive against intrinsically colistinresistant S. marcescens, indicating that the addition of positively charged 4-amino-4-deoxy-Larabinopyranose 1 to lipopolysaccharide changes the membrane charge and prevents peptide binding. The combination of DP-C and colistin reduced effective doses of both and may reduce peptide toxicity and colistin clinical nephrotoxicity. Although D-amino acid-based AMPs have been used clinically in the topical treatment of acne <ns0:ref type='bibr'>(Gordon, Romanowski & McDermott, 2005)</ns0:ref> but not yet for systemic infectious diseases. It should be rewarding to explore in the systemic treatment whether attachment of cysteine to the N-or C-terminus of AMPs could help broaden the spectrum and enhance the activity of AMPs against various drug-resistant microorganisms.</ns0:p><ns0:p>An algorithm predicting the effectiveness of silico designed stapled AMPs that are stable, active and selective toward bacterial membranes in vivo, has enabled the modification of magainin II (Mag2) and other known AMPs <ns0:ref type='bibr'>(Mourtada et al., 2019)</ns0:ref>. Modified Mag(i+4)1,15(A9K) was found to have MICs <4 µg/ml for MDR P. aeruginosa, A. baumannii and E. coli, with concentrations as high as ~100 µg/ml having almost no red blood cell hemolytic activity. Combinations of colistin with DP-C and C-DP may achieve the same level of antimicrobial activity against these MDR bacteria, as well as widening the safety windows of both drugs. Furthermore, linking of chimeric DP-C and C-DP to macrocycles derived from polymyxin and colistin could have synergistic antimicrobial activity. Chimeric peptidomimetic antibiotics, b Generated by heating DP-C at 60 °C for 30 minutes to convert cysteine to cystine.</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:06:50539:1:1:NEW 2 Sep 2020)</ns0:p><ns0:p>Manuscript to be reviewed MICs of antimicrobial peptides against strains of bacteria.</ns0:p><ns0:p>The Mann-Whitney U test was used to compare the MIC values of C-DP and DP-C with DP in species. P-values less than 0.05 were considered statistically significant (*p < 0.05).</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:06:50539:1:1:NEW 2 Sep 2020)</ns0:p><ns0:p>Manuscript to be reviewed </ns0:p></ns0:div><ns0:figure xml:id='fig_0'><ns0:head /><ns0:label /><ns0:figDesc>Bacterial strainsMDR P. aeruginosa NCGM2.S1(Miyoshi-Akiyama et al., 2011); drug-susceptible P. aeruginosa PAO1 (Weinstein, 2018); P. aeruginosaOCR1 (Poole et al., 1996); and P. aeruginosa PAO4290 (Yoneyama et al., 1997), were grown in Luria Bertani broth (LB Broth; BD Japan, Tokyo, Japan) or on LB plates containing 15 g/L agar, at 37° C. Drug-susceptible E. coli ATCC 25922, a clinical isolate of E.coli NCCHD1261-5 (Uchida et al., 2018), and S. aureus ATCC 25923 were grown at 37℃ in tryptic soy broth (TSB; BD Japan). Drug-susceptible A. baumannii ATCC 15308; a clinical isolate of MDR A. baumannii IOMTU433 (Tada et al., 2015) (GenBank accession no. AP014649); a clinical isolate of MDR A. baumannii NCGM237 (Tada et al., 2015) (GenBank accession no. AP013357); a clinical isolate of MDR A. baumannii NCGM253 (Tada et al., 2015) (GenBank accession no. AB823544); K. pneumoniae ATCC-BAA-2146; K. pneumoniae ATCC15380 (Reading & Cole, 1977) and Serratia marcescens NBRC102204 T , were grown at 37° C in Difco TM Nutrient broth (BD).</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_1'><ns0:head>a</ns0:head><ns0:label /><ns0:figDesc>A. baumannii strains were wild-type strain ATCC 15308 and multi-drug resistant strains IOMTU433(Tada et al., 2015) (GenBank accession no. AP014649), NCGM237(Tada et al., 2015) (GenBank accession no. AP013357) and NCGM253(Tada et al., 2015) (GenBank accession no. AB823544). E. coli strains were wild-type strain ATCC25922 and multi-drug resistant strain NCCHD1261-5 (Uchida et al., 2018). K. pneumoniae strains were multidrug-resistant strain ATCC15380(Reading & Cole, 1977) and the penicillin resistant strain ATCC-BAA-2146, a resistance caused by the production of β-lactamase. The S. marcescens strain NBRC102204 and the S. aureus strain ATCC 25923 were wild-type strain.</ns0:figDesc></ns0:figure>
<ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='23,42.52,283.42,525.00,371.25' type='bitmap' /></ns0:figure>
<ns0:figure type='table' xml:id='tab_0'><ns0:head>MIC (µg/mL)</ns0:head><ns0:label /><ns0:figDesc /><ns0:table><ns0:row><ns0:cell>Strains of</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell>Antibiotics</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell cols='3'>Antimicrobial peptides</ns0:cell></ns0:row><ns0:row><ns0:cell>P. aeruginosa a</ns0:cell><ns0:cell cols='7'>Amikacin Colistin Meropenem Ofloxacin DP C-DP* DP-C*</ns0:cell><ns0:cell>DP-C Dimer* b</ns0:cell></ns0:row><ns0:row><ns0:cell>PAO-1</ns0:cell><ns0:cell>4</ns0:cell><ns0:cell>1</ns0:cell><ns0:cell>2</ns0:cell><ns0:cell>0.5</ns0:cell><ns0:cell>300</ns0:cell><ns0:cell>64</ns0:cell><ns0:cell>16</ns0:cell><ns0:cell>16</ns0:cell></ns0:row><ns0:row><ns0:cell>NCGM2.S1</ns0:cell><ns0:cell>128</ns0:cell><ns0:cell>1</ns0:cell><ns0:cell>>512</ns0:cell><ns0:cell>64</ns0:cell><ns0:cell>300</ns0:cell><ns0:cell>128</ns0:cell><ns0:cell>32</ns0:cell><ns0:cell>32</ns0:cell></ns0:row><ns0:row><ns0:cell>OCR1</ns0:cell><ns0:cell>8</ns0:cell><ns0:cell>1</ns0:cell><ns0:cell>8</ns0:cell><ns0:cell>2</ns0:cell><ns0:cell>>300</ns0:cell><ns0:cell>128</ns0:cell><ns0:cell>16</ns0:cell><ns0:cell>16</ns0:cell></ns0:row><ns0:row><ns0:cell>PAO4290</ns0:cell><ns0:cell>4</ns0:cell><ns0:cell>1</ns0:cell><ns0:cell>1</ns0:cell><ns0:cell>1</ns0:cell><ns0:cell>300</ns0:cell><ns0:cell>128</ns0:cell><ns0:cell>16</ns0:cell><ns0:cell>16</ns0:cell></ns0:row></ns0:table><ns0:note>a P. aeruginosa strains used in this study were wild typePAO-1 (Weinstein, 2018), the MDR clinical strain NCGM2.S1(Miyoshi-Akiyama et al., 2011), the OprM overexpressing mutant OCR1(Poole et al., 1996) and PAO4290 (Yoneyama et al., 1997) which expressed a wild-type level of MexAB-OprM.</ns0:note></ns0:figure>
<ns0:figure type='table' xml:id='tab_1'><ns0:head>Table 2 (on next page)</ns0:head><ns0:label>2</ns0:label><ns0:figDesc /><ns0:table /></ns0:figure>
<ns0:figure type='table' xml:id='tab_2'><ns0:head>MIC (μg/ml) Genes or mutations associated with drug resistance Strains a</ns0:head><ns0:label /><ns0:figDesc /><ns0:table><ns0:row><ns0:cell /><ns0:cell>DP</ns0:cell><ns0:cell>C-DP</ns0:cell><ns0:cell>DP-C</ns0:cell><ns0:cell>β-lactamase(s)</ns0:cell><ns0:cell>16S rRNA methylase</ns0:cell><ns0:cell>colistin-resistance gene</ns0:cell></ns0:row><ns0:row><ns0:cell cols='2'>Acinetobacter baumannii*</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>ATCC15308</ns0:cell><ns0:cell>300</ns0:cell><ns0:cell>8</ns0:cell><ns0:cell>8</ns0:cell><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>IOMTU433</ns0:cell><ns0:cell>64</ns0:cell><ns0:cell>4</ns0:cell><ns0:cell>8</ns0:cell><ns0:cell>bla NDM -1 , bla OXA -23 , bla PER -7</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>NCGM237</ns0:cell><ns0:cell>128</ns0:cell><ns0:cell>4</ns0:cell><ns0:cell>8</ns0:cell><ns0:cell>bla OXA-23</ns0:cell><ns0:cell>armA</ns0:cell></ns0:row><ns0:row><ns0:cell>NCGM253</ns0:cell><ns0:cell>128</ns0:cell><ns0:cell>4</ns0:cell><ns0:cell>8</ns0:cell><ns0:cell>bla OXA-72</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>Escherichia coli</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>ATCC 25922</ns0:cell><ns0:cell>>300</ns0:cell><ns0:cell>16</ns0:cell><ns0:cell>8</ns0:cell><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>NCCHD1261-5</ns0:cell><ns0:cell>64</ns0:cell><ns0:cell>4</ns0:cell><ns0:cell>8</ns0:cell><ns0:cell>bla NDM-5</ns0:cell><ns0:cell /><ns0:cell>mcr-1</ns0:cell></ns0:row><ns0:row><ns0:cell cols='2'>Klebsiella pneumoniae</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>ATCC 15380</ns0:cell><ns0:cell cols='2'>>128 >128</ns0:cell><ns0:cell>128</ns0:cell><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell cols='2'>ATCC BAA-2146 >128</ns0:cell><ns0:cell>16</ns0:cell><ns0:cell>16</ns0:cell><ns0:cell>bla NDM-1</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>Serratia marcescens</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>NBRC102204</ns0:cell><ns0:cell cols='2'>>256 >256</ns0:cell><ns0:cell>>256</ns0:cell><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell cols='2'>Staphylococcus aureus</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>ATCC 25923</ns0:cell><ns0:cell cols='2'>>128 >128</ns0:cell><ns0:cell>>128</ns0:cell><ns0:cell /><ns0:cell /></ns0:row></ns0:table></ns0:figure>
<ns0:note place='foot'>PeerJ reviewing PDF | (2020:06:50539:1:1:NEW 2 Sep 2020)</ns0:note>
</ns0:body>
" | "Dr. Jack Leo
Academic Editor
PeerJ
August 30, 2020
Dear Dr. Jack Leo
Thank you very much for the thoughtful and constructive feedback you provided regarding our manuscript, article number 50539 (Addition of L-cysteine to the N- or C-terminus of the all-d-enantiomer [D(KLAKLAK)2] increases antimicrobial activities against multidrug-resistant Pseudomonas aeruginosa, Acinetobacter baumannii and Escherichia coli). We thank two reviewers for providing constructive comments regarding the improvement of the original manuscript. It is with great pleasure that we resubmit our article for further consideration. We have incorporated changes that reflect the detailed suggestions you have graciously provided. We also hope that our edits and the responses we
provide below satisfactorily address all the issues and concerns you and the reviewers have noted.
To facilitate your review of our revisions, the following is a point-by-point response to the questions and comments delivered in your letter dated July 24, 2020.
We believe that the manuscript is now suitable for publication in PeerJ.
Sincerely,
Maki Kamiya Ohno, Ph.D
Faculty of pharmaceutical Sciences, Teikyo Heisei University
4-21-2, Nakano, Nakano-ku, Tokyo, 164-8530, Japan
Tel: +81 80 5418 7433
Email: m.ohno@thu.ac.jp
On behalf of all authors.
RESPONSE TO EDITOR (Dr. Jack Leo):
Editor comments
Two experts have reviewed the paper, and both found the study to be technically sound and had only minor comments on it. These relate mainly to statistical analyses, inclusion of HPLC profiles and mass spectra, and some small textual issues. Please address the points raised by the reviewers and submit a revised version of the manuscript.
Reply:
Thank you for providing these insights. We have carried out the HPLC and the MALDI-TOF MS analysis for all synthetic peptides. The HPLC analysis indicated that the purities of all the peptides were >95%. The mass of all synthetic peptides by MALDI-TOF MS analysis matched well with the theoretical molecular weights, it was confirmed disulfide formation of the DP-C dimer. We have added new Figures (Fig S1-4). The statistical analysis has been performed (Table 1-2, Figure 1).
RESPONSE TO REVIEWER #1:
Thank you very much for your review on our paper. Your comments helped us enhance the quality of the paper.
Basic reporting
Reviewer #1 comments
The report is easy to read, relevant references are included.
The manuscript investigates how the addition of L-cysteine to a known antimicrobial peptide (in complete D-configuration) improves the antibacterial effect against certain Gram-negative pathogens and retains synergistic activity in combination with colistin. The scope of this work is rather limited by focusing on one peptide sequence. It would be interesting to explore whether incorporation of thiol residues into antimicrobial peptides is a more general strategy to improve the antibacterial effect of antimicrobial peptides.
1) Why was L-cysteine incorporated into the peptide sequence? The D-cysteine likely would enhance metabolic stability. This is not clear.
Reply:
Thank you for providing these insights. Initially we added L-cysteine to the N- or C-terminus of [D(KLAKLAK)2] in order to conjugate with protein such as a single chain antibody. L-cysteine-rich antimicrobial peptides were isolated from leguminous plants and the granular hemocytes of mangrove crabs (Sivakamavalli, Nirosha & Vaseeharan, 2015; Maróti, Downie & Kondorosi, 2015). The functionalized textiles and nasal prongs modified with L-cysteine exhibited antimicrobial activity against Staphylococcus aureus and Klebsiella pneumoniae (Gouveia, Sá & Henriques, 2012; Caldeira et al., 2013; Xu et al., 2017; Odeberg et al., 2018). We have added these observations to the introduction section (lines 91-100).
L-cysteine exhibited preferred antimicrobial activity against S. aureus compared with D-cysteine (Wang et al., 2019). However, D-cysteine showed stronger antimicrobial activity against E. coli, Listeria monocytogenes and Salmonella enteritis (Wang et al., 2019). D-amino acid is highly resistant to proteolysis in bacteria, addition of D-cystein instead of L-cystein may be effective. We have added these observations to the discussion section (lines 280-283).
2) Very little information was provided why the incorporation of L-cysteine enhances the antibacterial effect. ?
Reply:
It remains unclear what could be the mechanism by which the addition of cysteine to the N- or C-terminus to antimicrobial peptides enhances antimicrobial activity. It is unlikely that this effect is simply due to peptide dimerization via cysteine disulfide bond formation because DP-C dimer showed the same MIC value as DP-C monomer against all P. aeruginosa strains tested. The cysteine-rich region in Factor C receptors in the horseshoe crab specifically binds to bacterial lipopolysaccharides on Gram-negative bacteria (Koshiba, Hashii & Kawabata, 2006). The addition of L-cysteine to the N- or C-terminus of Andersonin-Y1, a potent antimicrobial peptide, enhanced its antimicrobial activity, with higher membrane disruption activity than the original peptide (Pal et al., 2019). The addition of L-cysteine to the N- or C-terminus of the peptide may have facilitated its binding to the bacterial membrane surface and form structures that disrupt their cell wall. The potential mechanism of efficacy enhancement by the attachment of cysteine residue to the antimicrobial peptides requires further investigation. We have added these observations to the discussion section (lines 265-274).
Experimental design
The experiments performed follow previously established standard protocols. Peptides were purchased and MIC, checkerboard studies and toxicity studies were performed using standard protocols.
Validity of the findings
3) I was unable to locate data that confirm complete disulfide formation of the dimeric peptides (MS spectra should be provided). It is usually standard to show the HPLC profiles of the purified peptides to demonstrate purity even when the peptides were purchased.
Reply:
Thank you for providing these insights. We have carried out the HPLC and the MALDI-TOF MS analysis for synthetic DP, C-DP DP-C, and DP-C dimer. We have added new Figures (Fig S1-4). The HPLC analysis indicated that the purities of all synthetic peptides were >95% (Fig S1-S3). The HPLC analysis of heat treated DP-C showed one large peak estimated as DP-C dimer and one small peak estimated as DP-C monomer, the peak area ratio was 6.3：1 (Fig. S4). The MALDI-TOF MS result of heat treated DP-C was 3252.7, which was in consistent with the estimated molecular weight of DP-C dimer (Fig. S4).
Comments for the Author
4) The scope is rather limited. It would be nice to see that this approach can be transferred to other antimicrobial peptides.
Reply:
We agree with you and have incorporated this suggestion throughout our paper. The addition of L-cysteine to the antimicrobial peptides, Andersonin-Y1, HBcARD, buforinII or lysin, enhanced its antimicrobial activity, with higher membrane disruption activity than the original peptide (Chen et al., 2018; Pal et al., 2019). The cysteine-derived cationic dipeptides lysine–cysteine, arginine–cysteine and histidine–cysteine presented antimicrobial activity, SEM analysis suggests that these dipeptides interact with cell walls to disrupt membrane integrity (Tsai et al., 2020). Whereas addition of an L-cysteine to the C-terminus of indolicidin, magainine or epinecidin-1 did not change their antimicrobial activity (Chen et al., 2018). It should be rewarding to explore in the systemic treatment whether attachment of cysteine to the N- or C-terminus of antimicrobial peptides could help broaden the spectrum and enhance the activity of AMPs against various drug-resistant microorganisms. We have added these observations to the discussion section (lines 258-265, 303-305).
5) There is little understanding for the observed effect. In absence of mechanistic studies the results remain unexplained and do not enhance understanding.
Reply:
We agree with you and have incorporated this suggestion throughout our paper. Please see point #2 above. We have added these observations to the discussion section (lines 265-274).
6) Incorporation of D-amino acids into peptides can increase metabolic stability and proteolysis but why selecting L-cysteine instead of D-cysteine needs to be explained.
Reply:
We agree with you and have incorporated this suggestion throughout our paper. Please see point #1 above. We have added these observations to the introduction section (lines 91-100) and the discussion section (lines 280-283).
7) The synergistic effect of the thiol-modified peptides with other antibiotics could be investigated. For instance rifampicin and novobiocin are usually potentiated by antimicrobial peptides and it would be interesting to explore whether the effect is retained.
Reply:
Thank you for providing these insights. We have carried out the checkerboard dilution method to evaluate the synergistic effect of DP-C and rifampicin against MDR P.aeruginosa (Table 3). The combinations of C-DP with rifanpicin had additive effect on antimicrobial activity. We have added these data to Table 3 and the results section (lines 234-236).
RESPONSE TO REVIEWER #2:
Thank you very much for your review on our paper. Your comments helped us enhance the quality of the paper.
Basic reporting
The manuscript is written in clear, unambiguous English with only minor typographical points observed as listed below. The abstract is clear and concise, and the introduction provides a good background to all areas that the manuscript investigates and is supported by the relevant literature. The manuscript is self-contained with all discussion points relevant to the results and the results are clearly relevant to the aims of the manuscript. Tables and figures are generally well presented but with some minor omissions that should be addressed (see elsewhere in the review). The manuscript is self-contained with all relevant information provided.
Minor typographical points:
1) Abstract: line 36 “C-DP or DP-C” should be changed to “C-DP and DP-C”
Reply:
The correction has been made (line 37).
2) Abstract: lines 39-40 “overexpressing the efflux” should be changed to “overexpressing efflux”
Reply:
The correction has been made (lines 40-41).
3) Abstract: line 49 “showed much stronger effects against” this should be altered to be more specific in describing what the stronger effect was e.g. stronger antimicrobial activity
Reply:
The correction has been made (line 50).
4) Introduction: line 66 “thought to bind to cytoplasmic membrane” consider “thought to bind to the cytoplasmic membrane”
Reply:
The correction has been made (line 67).
5) Results line 171: “but DP not against OCR1” suggested change to “but showed no antimicrobial activity against OCR1 within the tested concentration range” or similar to improve clarity.
Reply:
The correction has been made (lines 209-210).
Experimental design
This manuscript provides original primary research within the aims and scope of PeerJ. The human cell line HepG2 is described correctly with ATCC number provided. Methods are described with sufficient detail and information to replicate experiments. The research question well defined, relevant & meaningful. It is stated how research fills an identified knowledge gap to further improve the efficacy of antimicrobial peptides.
6) Statistical analyses should be performed when comparing the difference between MICs of the different AMPs to show that the increase in sensitivity is statistically significant.
Reply:
The statistical analysis has been performed. We have added these data to Table 1-2 and described in the Materials and Methods section (lines 176-182). The Mann-Whitney U test was used to compare the MIC values of C-DP and DP-C with DP in P. aeruginosa and A. baumannii (Table 1-2). C-DP, DP-C had significantly greater antimicrobial activities than DP against all P. aeruginosa and A.baumannii strains tested.
7) Statistical analyses should also be used to determine at what concentration of DP-C and colistin cell line viability is significantly decreased.
Reply:
The statistical analysis has been performed. We have added these data to Figure 1 and described in the Materials and Methods section (lines 176-182). Significant differences of cell survival rate between each concentration of the combination of DP-C and colistin and the control were statistically evaluated by Student's t-test.
Minor points
8) The authors state that peptides were synthesised to >95% purity. Can the authors state how the >95% purity of peptides was measured or if synthesised externally explicitly state that the synthesis was conducted by the named external company.
Reply:
Thank you for providing these insights. We have carried out the HPLC and the MALDI-TOF MS analysis for synthetic DP, C-DP DP-C, and DP-C dimer. We have added new Figures (Fig S1-4). The HPLC analysis indicated that the purities of all synthetic peptides were >95% (Fig S1-S3). The HPLC analysis of heat treated DP-C showed one large peak estimated as DP-C dimer and one small peak estimated as DP-C monomer, the peak area ratio was 6.3：1 (Fig. S4). The mass of all synthetic peptides by MALDI-TOF MS analysis matched well with the theoretical molecular weights (Fig S1-S3). The MALDI-TOF MS result of heat treated DP-C was 3252.7, which was in consistent with the estimated molecular weight of DP-C dimer (Fig. S4).
9) Lines 111-113 and 115-117: description of what the strains do and are for should not be in the methods section, they should be in the results section.
Reply:
Description of what the strains do and are for has been moved in the methods section to the results section.
10) Table 1: It should be stated clearly within the table that the antibiotic concentrations shown are the MIC concentrations.
Reply:
The correction has been made (Table 1).
Validity of the findings
11) Raw data has been provided for all human cell line assays in the form of percentage viability, an appropriate number of replicates is provided with mean, SD and SEM are calculated. This information is however unclear in figure 1, number of replicates, and statement of whether SD or SEM is used to create the graph are needed in the legend.
Reply:
The correction has been made (Figure 1).
12) The conclusions reflect the results presented with little speculation.
Reply:
Thank you for providing these insights. We have rewritten the conclusion section.
" | Here is a paper. Please give your review comments after reading it. |
9,824 | "<ns0:abstract xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:p>Background. Antimicrobial peptides have a broad spectrum of antimicrobial activities and are attracting attention as promising next-generation antibiotics against multidrug-resistant (MDR) bacteria. The all-denantiomer [ D (KLAKLAK) 2 ] has been reported to have antimicrobial activity against Escherichia coli and Pseudomonas aeruginosa, and to be resistant to protein degradation in bacteria because it is composed of D-enantiomer compounds. In this study, we demonstrated that modification of [ D (KLAKLAK) 2 ] by the addition of an L-cysteine residue to its N-or C-terminus markedly enhanced its antimicrobial activities against Gram-negative bacteria such as MDR Acinetobacter baumannii, E. coli, and P. aeruginosa.</ns0:p></ns0:div>
<ns0:div><ns0:head>Methods.</ns0:head><ns0:p>The peptides [ D (KLAKLAK) 2 ] (DP), DP to which L-cysteine was added at the N-terminus C-DP, and DP to which L-cysteine was added at the C-terminus DP-C, were synthesized at >95% purity. The minimum inhibitory concentrations of peptides and antibiotics were determined by the broth microdilution method. The synergistic effects of the peptides and the antibiotics against MDR P. aeruginosa were evaluated using the checkerboard dilution method. In order to assess how these peptides affect the survival of human cells, cell viability was determined using a Cell Counting Kit-8.</ns0:p><ns0:p>Results. C-DP and DP-C enhanced the antimicrobial activities of the peptide against MDR Gram-negative bacteria, including A. baumannii, E. coli, and P. aeruginosa. The antimicrobial activity of DP-C was greater than that of C-DP, with these peptides also having antimicrobial activity against drug-susceptible P. aeruginosa and drug-resistant P. aeruginosa overexpressing the efflux pump components. C-DP and DP-C also showed antimicrobial activity against colistin-resistant E. coli harboring mcr-1, which encodes a lipid A modifying enzyme. DP-C showed synergistic antimicrobial activity against MDR P. aeruginosa when combined with colistin. The LD 50 of DP-C against a human cell line HepG2 was six times higher than the MIC of DP-C against MDR P. aeruginosa. The LD 50 of DP-C was not altered by incubation with low-dose colistin.</ns0:p></ns0:div>
<ns0:div><ns0:head>Conclusion.</ns0:head><ns0:p>Attachment of an L-cysteine residue to the N-or C-terminus of [ D (KLAKLAK) 2 ] enhanced its antimicrobial activity against A. baumannii, E. coli, and P. aeruginosa. The combination of C-DP or DP-C and colistin had synergistic effects against multi-drug resistant P. aeruginosa. In addition, DP-C and C-DP showed much stronger antimicrobial activity against MDR A. baumannii and E. coli than against P. aeruginosa.</ns0:p></ns0:div>
</ns0:abstract>
<ns0:body xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:head>Introduction</ns0:head><ns0:p>The emergence and spread of multidrug-resistant (MDR) Gram-negative pathogens has become a serious public health problem worldwide. A global priority list of antibiotic-resistant bacteria published by the World Health Organization (WHO) to guide research, discovery, and development of new antibiotics listed carbapenem-resistant Acinetobacter baumannii, Pseudomonas aeruginosa and carbapenem-resistant and third-generation cephalosporin-resistant Enterobacteriaceae as first priority pathogens <ns0:ref type='bibr'>(Shai, 2002)</ns0:ref>.</ns0:p><ns0:p>Antimicrobial peptides (AMPs) are produced by various host organisms and partially contribute to the host's innate immunity <ns0:ref type='bibr'>(Plesniak et al., 2004;</ns0:ref><ns0:ref type='bibr'>Onuchic, Jennings & Ben-Jacob, 2013)</ns0:ref>. These peptides exhibit potent antimicrobial activities against a wide range of microorganisms, including viruses, bacteria, protozoa, and fungi <ns0:ref type='bibr'>(Shai, 2002)</ns0:ref>. AMPs are chemically amphiphilic polycationic peptides, generally comprising 6-50 amino acid residues, and they constitute a unique and diverse group of molecules <ns0:ref type='bibr'>(Peters, Shirtliff & Jabra-Rizk, 2010)</ns0:ref>. AMPs are classified according to their secondary structures including mixtures of α-helices, β-sheets, loops, and extended peptides. Most of these peptides are thought to bind to the cytoplasmic membrane, forming micelle-like aggregates that destroy the membrane <ns0:ref type='bibr'>(Peters, Shirtliff & Jabra-Rizk, 2010)</ns0:ref>. These peptides orient parallel to the interface, and associate with the membrane surface. After reaching a threshold concentration on the bilayer surface, they aggregate promoting channel formation through the bilayer and disrupt the membrane <ns0:ref type='bibr'>(Matsuzaki et al., 1995;</ns0:ref><ns0:ref type='bibr'>Plesniak et al., 2004;</ns0:ref><ns0:ref type='bibr'>Onuchic, Jennings & Ben-Jacob, 2013)</ns0:ref>. Because of their mechanisms of action, AMPs show antimicrobial activities against MDR as well as drug-susceptible Gramnegative bacteria. Thus, AMPs are attracting attention as promising next-generation antibiotics for the treatment of MDR bacterial infections <ns0:ref type='bibr'>(Hancock & Lehrer, 1998;</ns0:ref><ns0:ref type='bibr'>Marr, Gooderham & Hancock, 2006)</ns0:ref>. The all-D-enantiomer, [ D (KLAKLAK) 2 ] is an amphipathic lysine-leucine-rich α-helical peptide with high antimicrobial activity against Escherichia coli and P. aeruginosa, but with low toxicity against mouse <ns0:ref type='bibr'>3T3 cells (Javadpour et al., 1996;</ns0:ref><ns0:ref type='bibr'>McGrath et al., 2013)</ns0:ref>. [ D (KLAKLAK) 2 ] orients parallel to the interface and associates with the outer membrane surface. After reaching a threshold concentration on the outer membrane surface, the peptides aggregate to promote channel formation through the bilayer <ns0:ref type='bibr'>(Matsuzaki et al., 1995;</ns0:ref><ns0:ref type='bibr'>Plesniak et al., 2004;</ns0:ref><ns0:ref type='bibr'>Onuchic, Jennings & Ben-Jacob, 2013)</ns0:ref>. [ D (KLAKLAK) 2 ] was reported to selectively interfere with the bilayer of the outer membranes of E. coli and P. aeruginosa, leading to cell death by membrane disruption and loss of membrane potential. [ D (KLAKLAK) 2 ] has shown antimicrobial activity against Gram-negative bacteria but not Gram-positive bacteria, as this peptide was unlikely to disrupt the thick peptidoglycan layer of the latter <ns0:ref type='bibr'>(McGrath et al., 2013)</ns0:ref>. One of the strategies used to protect AMPs from protease degradation was sequence modification of D-amino acids to replace L-amino acids <ns0:ref type='bibr'>(Choi et al., 1993;</ns0:ref><ns0:ref type='bibr'>Braunstein, Papo & Shai, 2004;</ns0:ref><ns0:ref type='bibr'>Lee & Lee, 2008)</ns0:ref>. Because this peptide is an all-D-enantiomer, it is highly resistant to proteolysis in bacteria and has low immunogenicity. This stability and low immunogenicity may prolong its half-life and enhance its efficacy at low doses in vivo. Initially we added L-cysteine to the N-or C-terminus of [ D (KLAKLAK) 2 ] in order to conjugate with protein such as a single chain antibody. The side chain of cysteine contains sulfhydryl group, which can make a covalent coupling with an amino group of the protein via a cross-linker molecule such as sulfo-SMCC (sulfosuccinimidyl 4-(Nmaleimidomethyl) cyclohexane-1-carboxylate). The cysteine-rich AMPs were isolated from leguminous plants and the granular hemocytes of mangrove crabs <ns0:ref type='bibr'>(Sivakamavalli, Nirosha & Vaseeharan, 2015;</ns0:ref><ns0:ref type='bibr'>Maróti, Downie & Kondorosi, 2015)</ns0:ref>. The functionalized textiles and nasal prongs modified with L-cysteine exhibited antimicrobial activity against Staphylococcus aureus and Klebsiella pneumoniae <ns0:ref type='bibr'>(Gouveia, Sá & Henriques, 2012;</ns0:ref><ns0:ref type='bibr'>Caldeira et al., 2013;</ns0:ref><ns0:ref type='bibr'>Xu et al., 2017;</ns0:ref><ns0:ref type='bibr'>Odeberg et al., 2018)</ns0:ref>.</ns0:p><ns0:p>Modification of [ D (KLAKLAK) 2 ] by the addition of an L-cysteine residue to its N-or Cterminus markedly enhanced its antimicrobial activities against Gram-negative bacteria such as P. aeruginosa, E. coli, and A. baumannii. The present study describes the antimicrobial activities of the modified peptides against clinical isolates of MDR P. aeruginosa, A. baumannii and E. coli, and its synergistic effects with low dose colistin.</ns0:p></ns0:div>
<ns0:div><ns0:head>Materials & Methods</ns0:head></ns0:div>
<ns0:div><ns0:head>Peptide design and synthesis</ns0:head><ns0:p>The peptides [(D-Lys-D-Leu-D-Ala-D-Lys-D-Leu-D-Ala-D-Lys) 2 ] (DP) (9), DP to which Lcysteine was added at the N-terminus, [L-Cys-(D-Lys-D-Leu-D-Ala-D-Lys-D-Leu-D-Ala-D-Lys) 2 ] (C-DP) and DP to which L-cysteine was added at the C-terminus [(D-Lys-D-Leu-D-Ala-D-Lys-D-Leu-D-Ala-D-Lys) 2 -L-Cys] (DP-C), were synthesized at >95% purity (Scrum Inc., Tokyo, Japan). DP-C was dimerized by heating at 60 ° C for 30 min (DP-C dimer) to convert cysteine to cystine. Purity of synthetic peptides was checked on a SunFire C18 column (100Å, 5 µm, 4.6 mm inner diameter × 250 mm, Waters). Each peptides were gradiently eluted with solution A (water containing 0.1% trifluoroacetic acid) and solution B (acetonitrile containing 0.1% trifluoroacetic acid) at a flow rate of 1.0 mL/min. The elution program for DP and C-DP was as follows: at 0 min, 10% of B; at 20 min, 60% of B. The elution program for DP-C was as follows: at 0 min, 0% of B; at 20 min, 100% of B. The separated components were detected at 220 nm. The DP-C dimer was separated on a COSMOSIL 5C18-AR-300 reversed phase column (4.6 mm inner diameter × 250 mm, Nacalai Tesque, Kyoto, Japan), using an automated HPLC system (LC-2010AHT; Shimadzu, Kyoto, Japan). The reaction products were gradiently eluted with solution A (water containing 0.086% trifluoroacetic acid) and solution B (acetonitrile containing 0.086% trifluoroacetic acid) at a flow rate of 1.0 mL/min. The elution program for DP-C dimer was as follows: at 0 min, 20% of B; at 20 min, 50% of B. The peptide masses were determined by MALDI-TOF MS on a microflex (Bruker, Billerica MA). </ns0:p></ns0:div>
<ns0:div><ns0:head>Drug susceptibility testing</ns0:head><ns0:p>The minimum inhibitory concentrations (MICs) of peptides and antibiotics, including meropenem, amikacin, ofloxacin, and colistin, were determined by the broth microdilution method according to Clinical Laboratory Standards Institute (CLSI) guidelines <ns0:ref type='bibr'>(Weinstein, 2018)</ns0:ref>. Bacterial strains were inoculated at 5 x 10 5 CFU/ml per well into 96-well round-bottom microtiter plates (Watson Bio Lab, Kobe, Japan) containing an equal volume of serially diluted peptide or antibiotic. Three independent experiments were performed to confirm reproducibility.</ns0:p><ns0:p>The synergistic effects of the peptides and the antibiotics amikacin, colistin, meropenem, ofloxacin, and rifampicin against P. aeruginosa NCGM2.S1 were evaluated using the checkerboard dilution method. The peptides were two-fold serially diluted to final concentrations ranging from 0.125-to 2-times the MIC longitudinally in 96-well round-bottom microtiter plates (Watson Bio Lab). Subsequently, antibiotic was two-fold serially diluted to final concentrations ranging from 0.125-to 2-times the MIC transversely into the plates. NCGM2.S1 was inoculated at 5 x 10 5 CFU/ml per well at a volume equal to that of the diluted peptide and antibiotic. Three independent experiments were performed to confirm reproducibility. The synergistic effect of the peptides and antibiotics was assessed by determining the fractional inhibitory concentration (FIC) index <ns0:ref type='bibr'>(Berenbaum, 1978)</ns0:ref>, using the formula: MIC of peptide in combination MIC of antimicrobial agent in combination FIC = MIC of peptide alone MIC of antimicrobial agent alone  An FIC index ≤0.5 was defined as synergistic, an FIC index >0.5 to 4.0 was defined as additive or unrelated, and an FIC index > 4.0 was defined as antagonistic.</ns0:p></ns0:div>
<ns0:div><ns0:head>Cytotoxicity tests</ns0:head><ns0:p>The human hepatoblastoma cell line, HepG2 (ATCC HB-8065), was obtained from American Type Culture Collection and cultured in DMEM supplemented with 10% fetal bovine serum (FBS). HepG2 cells were seeded at 3000 cells/well in 96-well cell culture-treated flat bottom microtiter plates (Falcon, Corning NY). The cells were incubated at 37 °C for 48 h in an atmosphere containing 5% CO 2 , followed by the addition of peptides, at a final concentration of 0-256 µg/ml, or colistin, at a final concentration of 0-3000 µg/ml, by serial dilution. The plates were incubated with 0.2% FBS in 5% CO 2 at 37 °C for 48 h; under these conditions, HepG2 cells were alive but did not grow. Cell viability was determined using a Cell Counting Kit-8 (Dojin, Tokyo, Japan), and colorimetric changes were determined at OD 450 nm with a microplate reader (Corona Electric Co Ltd., Ibaraki, Japan). The LD 50 was defined as the concentration of peptides or colistin that resulted in 50% cell viability. Three independent experiments were performed to confirm reproducibility.</ns0:p></ns0:div>
<ns0:div><ns0:head>Statistical analysis</ns0:head><ns0:p>The Mann-Whitney U test was used to compare the MIC values of C-DP, DP-C, and DP-C dimer with DP in P. aeruginosa and A. baumannii. P-values less than 0.05 were considered statistically significant. Cell survival was expressed as a percentage of the control were obtained as mean ± Standard Deviation (SD) of three independent experiments done in three replicates for each treatment. Significant differences of cell survival rate between each concentration and the control were statistically evaluated by Student's t-test.</ns0:p></ns0:div>
<ns0:div><ns0:head>Results</ns0:head></ns0:div>
<ns0:div><ns0:head>Addition of L-cysteine to the N-or C-terminus enhanced the antimicrobial activity of the original peptide</ns0:head><ns0:p>The HPLC analysis indicated that the purities of synthetic DP, C-DP, and DP-C were 100, 95.52 and 98.68%, respectively (Fig. <ns0:ref type='figure'>S1-3</ns0:ref>). Additionally the masses of DP, C-DP, and DP-C by MALDI-TOF MS analysis were <ns0:ref type='bibr'>1525.444, 1626.595, and 1627.151</ns0:ref>, respectively that matched well with the theoretical molecular weights (1524.0, 1627.1, and 1627.1) (Fig. <ns0:ref type='figure'>S1-3</ns0:ref>). Similarly the formation of DP-C dimer was assessed by HPLC and MALDI-TOF MS. The HPLC analysis of heat treated DP-C showed one large peak estimated as DP-C dimer and one small peak estimated as DP-C monomer, the peak area ratio was 6.3：1 (Fig. <ns0:ref type='figure'>S4</ns0:ref>). The MALDI-TOF MS result of heat treated DP-C was 3252.7, which was in consistent with the estimated molecular weight of DP-C dimer (Fig. <ns0:ref type='figure'>S4</ns0:ref>). The grand average hydropathy (GRAVY) values were -0.07 for DP and 0.1 each for C-DP and DP-C, indicating that the addition of L-cysteine affected the hydrophobicity of DP.</ns0:p><ns0:p>Assessment of antibiotic susceptibility showed that drug-susceptible P. aeruginosa PAO-1 (Weinstein, 2018) and PAO4290 expressing a normal level of <ns0:ref type='bibr'>MexAB-OprM (Yoneyama et al., 1997)</ns0:ref>, were susceptible to all antibiotics tested; MDR P. aeruginosa NCGM2.S1 <ns0:ref type='bibr'>(Miyoshi-Akiyama et al., 2011)</ns0:ref> was susceptible to colistin, but resistant to amikacin, meropenem and ofloxacin; and OCR1, a nalB multidrug-resistant mutant that overproduces the outer membrane protein OprM <ns0:ref type='bibr'>(Poole et al., 1996)</ns0:ref> was susceptible to amikacin and colistin, intermediately susceptible to ofloxacin, but resistant to meropenem (Table <ns0:ref type='table'>1</ns0:ref>). DP showed antimicrobial activity against PAO-1, with an MIC of 300 µg/mL, consistent with previous findings <ns0:ref type='bibr'>(McGrath et al., 2013)</ns0:ref>. DP also had antimicrobial activity against NCGM2.S1 and PAO4290, with MICs of 300 µg/mL, but DP showed no antimicrobial activity against OCR1 within the tested concentration range. C-DP had greater antimicrobial activities than DP against all of these strains (The Mann-Whitney U test, p < 0.05), with MICs of 64-128 µg/mL, and DP-C had greater antimicrobial activities than C-DP and DP (The Mann-Whitney U test, p < 0.05), with MICs 16-32 µg/mL (Table <ns0:ref type='table'>1</ns0:ref>). The DP-C dimer also had antimicrobial activities and showed MICs identical to DP-C against all the strains tested. These results indicate that the addition of L-cysteine to the N-or Cterminus of [ D (KLAKLAK) 2 ] increased its antimicrobial activity. DP showed antimicrobial activity against four A. baumannii strains, with MICs of 64 to 300 µg/mL, and against a clinical isolate of colistin-and carbapenem-resistant E. coli NCCHD1261-5 co-harboring mcr-1 and bla NDM-5 genes <ns0:ref type='bibr'>(Uchida et al., 2018)</ns0:ref>, with an MIC of 64 µg/mL. In contrast, DP was inactive against drug-susceptible E. coli ATCC 25922, two K. pneumoniae strains, S marcescens NBRC102204 and S. aureus ATCC 25923 (Table <ns0:ref type='table' target='#tab_1'>2</ns0:ref>). C-DP and DP-C showed higher antimicrobial activities than DP (The Mann-Whitney U test, p < 0.05), with MICs of 4 to 8 µg/mL against the four A. baumannii strains and MICs of 4 to 16 µg/mL against the two E. coli strains. C-DP and DP-C showed antimicrobial activity against carbapenem-resistant K. pneumoniae ATCC BAA-2146 harboring bla NDM-1 , with both having MICs of 16 µg/mL, but not against penicillin-resistant, β-lactamase-producing K. pneumoniae ATCC 15380, S. marcescens NBRC102204 and S. aureus ATCC 25923 (Table <ns0:ref type='table' target='#tab_1'>2</ns0:ref>).</ns0:p></ns0:div>
<ns0:div><ns0:head>Synergistic effects of peptides and antibiotics</ns0:head><ns0:p>The combinations of DP, C-DP and DP-C with colistin had synergistic effects on antimicrobial activity (Table <ns0:ref type='table'>3</ns0:ref>). For example, the growth of P. aeruginosa NCGM2.S1 was inhibited by a combination of one-sixteenth the MIC of DP (19 µg/mL) and one-fourth the MIC of colistin (0.25 µg/mL), by a combination of one-fourth the MIC of C-DP (32 µg/mL) and oneeighth the MIC of colistin (0.125 µg/mL) and by a combination of one-eighth the MIC of DP-C (4.0 µg/mL) and one-fourth the MIC of colistin (0.25 µg/mL). Although AMPs that induced susceptibility to rifampicin were reported in clinical MDR isolates of P. aeruginosa <ns0:ref type='bibr' target='#b0'>(Baker et al., 2019)</ns0:ref>, DP-C only slightly enhanced the susceptibility to rifampicin. The combinations of C-DP with meropenem or ofloxacin had additive effect on antimicrobial activity. The synergistic effects were not observed when DP-C was combined on amikacin.</ns0:p></ns0:div>
<ns0:div><ns0:head>Cytotoxicity of antimicrobial peptides to HepG2 cells</ns0:head><ns0:p>The LD 50 values of each peptide in HepG2 cells were >256, >256, 192, and 2100 µg/ml for DP, C-DP, DP-C, and colistin, respectively (Fig. <ns0:ref type='figure'>S5</ns0:ref>, Fig. <ns0:ref type='figure'>S6</ns0:ref>). The raw data of Fig. <ns0:ref type='figure'>S5</ns0:ref> and Fig. <ns0:ref type='figure'>S6</ns0:ref> are available in Supplementary file 2 and Supplementary file 3 respectively. Since the antimicrobial activity of these peptides against P. aeruginosa were synergistic with colistin, we examined whether a combination of these peptides and colistin was more toxic than the peptide alone. The combination of DP-C and colistin dose-dependently induced the death of HepG2 cells (Figure <ns0:ref type='figure'>1</ns0:ref>), with more than 50% of the cells dying at 256 µg/ml DP-C and 25.6 µg/ml colistin. The cytotoxicity of the peptide was not enhanced by combination of low doses of colistin. There was no cytotoxicity to HepG2 at 4 µg/ml DP-C and 0.4 µg/ml colistin, which concentrations showed synergistic effects of colistin and DP-C. The raw data of Fig. <ns0:ref type='figure'>1</ns0:ref> is available in Supplementary file 2.</ns0:p></ns0:div>
<ns0:div><ns0:head>Discussion</ns0:head><ns0:p>We added L-cysteine to the N-or C-terminus of [ D (KLAKLAK) 2 ] in order to bind it with another protein like a single-chain antibody in the beginning of the experiment. This attachment yielded two compounds, DP-C and C-DP, with significantly greater antimicrobial activities against A. baumannii, E. coli and P. aeruginosa than the original DP. Similarly, the addition of L-cysteine to AMPs, Andersonin-Y1, HBcARD, buforinII or lysin, enhanced its antimicrobial activity, with higher membrane disruption activity than the original peptide <ns0:ref type='bibr'>(Chen et al., 2018;</ns0:ref><ns0:ref type='bibr'>Pal et al., 2019)</ns0:ref>. The cysteine-derived cationic dipeptides lysine-cysteine, arginine-cysteine and histidine-cysteine presented antimicrobial activity, SEM analysis suggests that these dipeptides interact with cell walls to disrupt membrane integrity <ns0:ref type='bibr'>(Tsai et al., 2020)</ns0:ref>. Whereas addition of an L-cysteine to the C-terminus of indolicidin, magainine or epinecidin-1 did not change their antimicrobial activity <ns0:ref type='bibr'>(Chen et al., 2018)</ns0:ref>. It remains unclear what could be the mechanism by which the addition of cysteine to the N-or C-terminus to AMPs enhances antimicrobial activity. It is unlikely that this effect is simply due to peptide dimerization via cysteine disulfide bond formation because DP-C dimer showed the same MIC value as DP-C monomer against P. aeruginosa strains tested. The cysteine-rich region in Factor C receptors in the horseshoe crab specifically binds to bacterial lipopolysaccharides on Gram-negative bacteria <ns0:ref type='bibr'>(Koshiba, Hashii & Kawabata, 2006)</ns0:ref>. The addition of L-cysteine to the N-or C-terminus of the peptide may have facilitated its binding to the bacterial membrane surface and form structures that disrupt their cell wall. The potential mechanism of efficacy enhancement by the attachment of cysteine residue to AMPs requires further investigation. Systematic hybridization of two lead peptides from unrelated classes of AMPs showed no associations of net charge, charge density, and antipneumococcal activity among the hybrid peptides, although AMPs with higher hydrophobicity values have been reported to have greater antimicrobial activity against Streptococcus pneumoniae <ns0:ref type='bibr'>(Le et al., 2015)</ns0:ref>. The peptides we tested showed a similar trend, when GRAVY was calculated, DPC showed higher hydrophobicity than DP. L-cysteine exhibited preferred antimicrobial activity against S. aureus compared with Dcysteine, whereas D-cysteine showed stronger antimicrobial activity against E. coli, Listeria monocytogenes and Salmonella enteritis <ns0:ref type='bibr'>(Wang et al., 2019)</ns0:ref>. D-amino acid is highly resistant to proteolysis in bacteria, addition of D-cystein instead of L-cystein may be effective. DP-C and C-DP had potent activity against multidrug-resistant Gram-negative pathogens. The emergence and spread of these drug-resistant pathogens has become a serious worldwide public health problem <ns0:ref type='bibr'>(Boucher et al., 2009;</ns0:ref><ns0:ref type='bibr'>Tacconelli et al., 2018)</ns0:ref>. Carbapenem is a last resort βlactam antibiotic administered to treat infections with drug-resistant Gram-negative pathogens. The development of new antibiotics against carbapenem-resistance pathogens is of top priority <ns0:ref type='bibr'>(Tacconelli et al., 2018)</ns0:ref>. DP-C and C-DP also had antimicrobial activity against E. coli strains harboring the plasmid-mediated colistin resistance mcr-1 gene. Colistin is a last line polycationic peptide antibiotic which is used to treat infections with carbapenem-resistant Gram-negative pathogens <ns0:ref type='bibr'>(Paterson & Harris, 2016)</ns0:ref>. However, colistin-resistant mcr-1 producers have emerged in humans and animals in China <ns0:ref type='bibr'>(Liu et al., 2016)</ns0:ref> and have spread worldwide. DP-C and C-DP, like DP, possess a positive charge and hydrophobic regions, suggesting that they target the lipid bilayer of the membrane and destroy it, causing loss of membrane potential and ultimately cell death <ns0:ref type='bibr'>(McGrath et al., 2013)</ns0:ref>. These peptides had antimicrobial activity against Gram-negative but not Gram-positive pathogens. These peptides were also inactive against intrinsically colistinresistant S. marcescens, indicating that the addition of positively charged 4-amino-4-deoxy-Larabinopyranose 1 to lipopolysaccharide changes the membrane charge and prevents peptide binding. The combination of DP-C and colistin reduced effective doses of both and may reduce peptide toxicity and colistin clinical nephrotoxicity. Although D-amino acid-based AMPs have been used clinically in the topical treatment of acne <ns0:ref type='bibr'>(Gordon, Romanowski & McDermott, 2005)</ns0:ref> but not yet for systemic infectious diseases. It should be rewarding to explore in the systemic treatment whether attachment of cysteine to the N-or C-terminus of AMPs could help broaden the spectrum and enhance the activity of AMPs against various drug-resistant microorganisms.</ns0:p><ns0:p>An algorithm predicting the effectiveness of silico designed stapled AMPs that are stable, active and selective toward bacterial membranes in vivo, has enabled the modification of magainin II (Mag2) and other known AMPs <ns0:ref type='bibr'>(Mourtada et al., 2019)</ns0:ref>. Modified Mag(i+4)1,15(A9K) was found to have MICs < 4 µg/ml for MDR P. aeruginosa, A. baumannii and E. coli, with concentrations as high as ~100 µg/ml having almost no red blood cell hemolytic activity. Combinations of colistin with DP-C and C-DP may achieve the same level of antimicrobial activity against these MDR bacteria, as well as widening the safety windows of both drugs. Furthermore, linking of chimeric DP-C and C-DP to macrocycles derived from polymyxin and colistin could have synergistic antimicrobial activity. Chimeric peptidomimetic antibiotics, Manuscript to be reviewed MICs of antimicrobial peptides against strains of bacteria.</ns0:p><ns0:p>The Mann-Whitney U test was used to compare the MIC values of C-DP and DP-C with DP in species. P-values less than 0.05 were considered statistically significant (*p < 0.05).</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:06:50539:2:0:NEW 16 Sep 2020)</ns0:p><ns0:p>Manuscript to be reviewed </ns0:p></ns0:div><ns0:figure xml:id='fig_0'><ns0:head /><ns0:label /><ns0:figDesc>Bacterial strainsMDR P. aeruginosa NCGM2.S1(Miyoshi-Akiyama et al., 2011); drug-susceptible P. aeruginosa PAO1 (Weinstein, 2018); P. aeruginosaOCR1 (Poole et al., 1996); and P. aeruginosa PAO4290 (Yoneyama et al., 1997), were grown in Luria Bertani broth (LB Broth; BD Japan, Tokyo, Japan) or on LB plates containing 15 g/L agar, at 37° C. Drug-susceptible E. coli ATCC 25922, a clinical isolate of E.coli NCCHD1261-5 (Uchida et al., 2018), and S. aureus ATCC 25923 were grown at 37℃ in tryptic soy broth (TSB; BD Japan). Drug-susceptible A. baumannii ATCC 15308; a clinical isolate of MDR A. baumannii IOMTU433 (Tada et al., 2015) (GenBank accession no. AP014649); a clinical isolate of MDR A. baumannii NCGM237 (Tada et al., 2015) (GenBank accession no. AP013357); a clinical isolate of MDR A. baumannii NCGM253 (Tada et al., 2015) (GenBank accession no. AB823544); K. pneumoniae ATCC-BAA-2146; K. pneumoniae ATCC15380 (Reading & Cole, 1977) and Serratia marcescens NBRC102204 T , were grown at 37° C in Difco TM Nutrient broth (BD).</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_1'><ns0:head>a</ns0:head><ns0:label /><ns0:figDesc>A. baumannii strains were wild-type strain ATCC 15308 and multi-drug resistant strains IOMTU433(Tada et al., 2015) (GenBank accession no. AP014649), NCGM237(Tada et al., 2015) (GenBank accession no. AP013357) and NCGM253(Tada et al., 2015) (GenBank accession no. AB823544). E. coli strains were wild-type strain ATCC25922 and multi-drug resistant strain NCCHD1261-5 (Uchida et al., 2018). K. pneumoniae strains were multidrug-resistant strain ATCC15380(Reading & Cole, 1977) and the penicillin resistant strain ATCC-BAA-2146, a resistance caused by the production of β-lactamase. The S. marcescens strain NBRC102204 and the S. aureus strain ATCC 25923 were wild-type strain.</ns0:figDesc></ns0:figure>
<ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='23,42.52,359.92,525.00,371.25' type='bitmap' /></ns0:figure>
<ns0:figure type='table' xml:id='tab_0'><ns0:head>MIC (µg/mL)</ns0:head><ns0:label /><ns0:figDesc /><ns0:table><ns0:row><ns0:cell>Strains of</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell>Antibiotics</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell cols='3'>Antimicrobial peptides</ns0:cell></ns0:row><ns0:row><ns0:cell>P. aeruginosa a</ns0:cell><ns0:cell cols='7'>Amikacin Colistin Meropenem Ofloxacin DP C-DP* DP-C*</ns0:cell><ns0:cell>DP-C Dimer* b</ns0:cell></ns0:row><ns0:row><ns0:cell>PAO-1</ns0:cell><ns0:cell>4</ns0:cell><ns0:cell>1</ns0:cell><ns0:cell>2</ns0:cell><ns0:cell>0.5</ns0:cell><ns0:cell>300</ns0:cell><ns0:cell>64</ns0:cell><ns0:cell>16</ns0:cell><ns0:cell>16</ns0:cell></ns0:row><ns0:row><ns0:cell>NCGM2.S1</ns0:cell><ns0:cell>128</ns0:cell><ns0:cell>1</ns0:cell><ns0:cell>>512</ns0:cell><ns0:cell>64</ns0:cell><ns0:cell>300</ns0:cell><ns0:cell>128</ns0:cell><ns0:cell>32</ns0:cell><ns0:cell>32</ns0:cell></ns0:row><ns0:row><ns0:cell>OCR1</ns0:cell><ns0:cell>8</ns0:cell><ns0:cell>1</ns0:cell><ns0:cell>8</ns0:cell><ns0:cell>2</ns0:cell><ns0:cell>>300</ns0:cell><ns0:cell>128</ns0:cell><ns0:cell>16</ns0:cell><ns0:cell>16</ns0:cell></ns0:row><ns0:row><ns0:cell>PAO4290</ns0:cell><ns0:cell>4</ns0:cell><ns0:cell>1</ns0:cell><ns0:cell>1</ns0:cell><ns0:cell>1</ns0:cell><ns0:cell>300</ns0:cell><ns0:cell>128</ns0:cell><ns0:cell>16</ns0:cell><ns0:cell>16</ns0:cell></ns0:row></ns0:table><ns0:note>a P. aeruginosa strains used in this study were wild type PAO-1 (Weinstein, 2018), the MDR clinical strain NCGM2.S1 (Miyoshi-Akiyama et al., 2011), the OprM overexpressing mutant OCR1 (Poole et al., 1996) and PAO4290 (Yoneyama et al., 1997) which expressed a wild-type level of MexAB-OprM. b Generated by heating DP-C at 60 °C for 30 minutes to convert cysteine to cystine. PeerJ reviewing PDF | (2020:06:50539:2:0:NEW 16 Sep 2020)</ns0:note></ns0:figure>
<ns0:figure type='table' xml:id='tab_1'><ns0:head>Table 2 (on next page)</ns0:head><ns0:label>2</ns0:label><ns0:figDesc /><ns0:table /></ns0:figure>
<ns0:figure type='table' xml:id='tab_2'><ns0:head>MIC (μg/ml) Genes or mutations associated with drug resistance Strains a</ns0:head><ns0:label /><ns0:figDesc /><ns0:table><ns0:row><ns0:cell /><ns0:cell>DP</ns0:cell><ns0:cell>C-DP</ns0:cell><ns0:cell>DP-C</ns0:cell><ns0:cell>β-lactamase(s)</ns0:cell><ns0:cell>16S rRNA methylase</ns0:cell><ns0:cell>colistin-resistance gene</ns0:cell></ns0:row><ns0:row><ns0:cell cols='2'>Acinetobacter baumannii*</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>ATCC15308</ns0:cell><ns0:cell>300</ns0:cell><ns0:cell>8</ns0:cell><ns0:cell>8</ns0:cell><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>IOMTU433</ns0:cell><ns0:cell>64</ns0:cell><ns0:cell>4</ns0:cell><ns0:cell>8</ns0:cell><ns0:cell>bla NDM -1 , bla OXA -23 , bla PER -7</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>NCGM237</ns0:cell><ns0:cell>128</ns0:cell><ns0:cell>4</ns0:cell><ns0:cell>8</ns0:cell><ns0:cell>bla OXA-23</ns0:cell><ns0:cell>armA</ns0:cell></ns0:row><ns0:row><ns0:cell>NCGM253</ns0:cell><ns0:cell>128</ns0:cell><ns0:cell>4</ns0:cell><ns0:cell>8</ns0:cell><ns0:cell>bla OXA-72</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>Escherichia coli</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>ATCC 25922</ns0:cell><ns0:cell>>300</ns0:cell><ns0:cell>16</ns0:cell><ns0:cell>8</ns0:cell><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>NCCHD1261-5</ns0:cell><ns0:cell>64</ns0:cell><ns0:cell>4</ns0:cell><ns0:cell>8</ns0:cell><ns0:cell>bla NDM-5</ns0:cell><ns0:cell /><ns0:cell>mcr-1</ns0:cell></ns0:row><ns0:row><ns0:cell cols='2'>Klebsiella pneumoniae</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>ATCC 15380</ns0:cell><ns0:cell cols='2'>>128 >128</ns0:cell><ns0:cell>128</ns0:cell><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell cols='2'>ATCC BAA-2146 >128</ns0:cell><ns0:cell>16</ns0:cell><ns0:cell>16</ns0:cell><ns0:cell>bla NDM-1</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>Serratia marcescens</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>NBRC102204</ns0:cell><ns0:cell cols='2'>>256 >256</ns0:cell><ns0:cell>>256</ns0:cell><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell cols='2'>Staphylococcus aureus</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>ATCC 25923</ns0:cell><ns0:cell cols='2'>>128 >128</ns0:cell><ns0:cell>>128</ns0:cell><ns0:cell /><ns0:cell /></ns0:row></ns0:table></ns0:figure>
</ns0:body>
" | "Dr. Jack Leo
Academic Editor
PeerJ
September 16, 2020
Dear Dr. Jack Leo
Thank you for inviting us to submit a revised draft of our manuscript entitled, article number 50539 (Addition of L-cysteine to the N- or C-terminus of the all-d-enantiomer [D(KLAKLAK)2] increases antimicrobial activities against multidrug-resistant Pseudomonas aeruginosa, Acinetobacter baumannii and Escherichia coli) to PeerJ. We also appreciated the time and effort you and each of the reviewers have dedicated to providing insightful feedback on ways to strengthen our paper. We have incorporated changes that reflect the detailed suggestions you have graciously provided. We also hope that our edits and the responses we provide below satisfactorily address all the issues and concerns you and the reviewers have noted.
To facilitate your review of our revisions, the following is a point-by-point response to the questions and comments delivered in your letter dated September 16, 2020.
We look forward to hearing from you regarding our submission. We would be glad to respond to any further questions and comments that you may have.
Sincerely,
Maki Kamiya Ohno, Ph.D
Faculty of pharmaceutical Sciences, Teikyo Heisei University
4-21-2, Nakano, Nakano-ku, Tokyo, 164-8530, Japan
Tel: +81 80 5418 7433
Email: m.ohno@thu.ac.jp
On behalf of all authors.
RESPONSE TO EDITOR (Dr. Jack Leo):
Editor comments
The reviewer would still like to see amendments to the legend of Figure 1. Please make these and resubmit a revised version.
Reply:
Thank you for this comment. We have rewritten the legend of Figure 1.
RESPONSE TO REVIEWER #2:
Thank you very much for your review on our paper. Your comments helped us enhance the quality of the paper.
Basic reporting
Requested improvements were made to the text in the revised manuscript.
Experimental design
No comment.
Validity of the findings
No comment.
Comments for the Author
I cannot see the changes that the authors said they made to the legend of Figure 1 in their rebuttal letter - Raw data has been provided for all human cell line assays in the form of percentage viability, an appropriate number of replicates is provided with mean, SD and SEM are calculated. This information is however unclear in figure 1, number of replicates, and statement of whether SD or SEM is used to create the graph are needed in the legend.
This information is now found in the methods section, however can the authors include it in the figure legend for completeness and ease of interpreting the data (e.g. N=3 +/- SD). Also the meaning of the * in the figure should be clearly described in the figure legend (e.g. Students t-test *p<0.05).
Reply:
We are sorry. The modifications were not made. We have revised the legend of Figure 1 and Supplementary file 1.
" | Here is a paper. Please give your review comments after reading it. |
9,825 | "<ns0:abstract xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:p>Koalas (Phascolarctos cinereus) are highly specialized herbivorous marsupials that feed almost exclusively on Eucalyptus leaves, which are known to contain varying concentrations of many different toxic chemical compounds. The literature suggests that Lonepinella koalarum, a bacterium in the Pasteurellaceae family, can break down some of these toxic chemical compounds. Furthermore, in a previous study, we identified L. koalarum as the most predictive taxon of koala survival during antibiotic treatment. Therefore, we believe that this bacterium may be important for koala health. Here, we isolated a strain of L. koalarum from a healthy koala female and sequenced its genome using a combination of short-read and long-read sequencing. We placed the genome assembly into a phylogenetic tree based on 120 genome markers using the Genome Taxonomy Database (GTDB), which currently does not include any L. koalarum assemblies.</ns0:p><ns0:p>Our genome assembly fell in the middle of a group of Haemophilus, Pasteurella and Basfia species. According to average nucleotide identity and a 16S rRNA gene tree, the closest relative of our isolate is L. koalarum strain Y17189. Then, we annotated the gene sequences and compared them to 55 closely related, publicly available genomes. Several genes that are known to be involved in carbohydrate metabolism could exclusively be found in L. koalarum relative to the other taxa in the pangenome, including glycoside hydrolase families GH2, GH31, GH32, GH43 and GH77. Among the predicted genes of L. koalarum were 79 candidates putatively involved in the degradation of plant secondary metabolites. Additionally, several genes coding for amino acid variants were found that had been shown to confer antibiotic resistance in other bacterial species against pulvomycin, beta-lactam antibiotics and the antibiotic efflux pump KpnH. In summary, this genetic characterization allows us to build hypotheses to explore the potentially beneficial role that L. koalarum might play in the koala intestinal microbiome. Characterizing and understanding beneficial symbionts at the whole genome level is important for the</ns0:p></ns0:div>
</ns0:abstract>
<ns0:body xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:head>Introduction</ns0:head><ns0:p>Koalas (Phascolarctos cinereus) are arboreal marsupials that are highly specialized herbivores in that they feed almost exclusively on the foliage of select Eucalyptus species <ns0:ref type='bibr'>(Moore & Foley, 2005;</ns0:ref><ns0:ref type='bibr' target='#b17'>Callaghan et al., 2011)</ns0:ref>. All Eucalyptus species contain chemical defenses against isolates associated with the intestinal microbial communities of koalas have been characterized in the context of degradation of PCDs found in Eucalyptus leaves <ns0:ref type='bibr' target='#b79'>(Osawa, 1990</ns0:ref><ns0:ref type='bibr' target='#b80'>(Osawa, , 1992;;</ns0:ref><ns0:ref type='bibr' target='#b81'>Osawa et al., 1993</ns0:ref><ns0:ref type='bibr' target='#b82'>Osawa et al., , 1995;;</ns0:ref><ns0:ref type='bibr' target='#b64'>Looft, Levine & Stanton, 2013)</ns0:ref>. One of these cultured isolates is a bacterium known as Lonepinella koalarum. It had been first isolated from the mucus around the caecum in koalas and was shown to degrade tannin-protein complexes that can be found in Eucalyptus leaves <ns0:ref type='bibr' target='#b82'>(Osawa et al., 1995;</ns0:ref><ns0:ref type='bibr' target='#b37'>Goel et al., 2005)</ns0:ref>. Briefly, tannin-protein complexes are extremely diverse and result from the reaction between plant defense secondary metabolites; i.e., tannins, and proteins. Tannins bind proteins followed by the formation of a precipitate, which leads to the inhibition of digestive enzymes in herbivores <ns0:ref type='bibr' target='#b0'>(Adamczyk et al., 2017)</ns0:ref>. In our previous work, L. koalarum was identified as the most predictive taxon of koala survival during antibiotic treatment <ns0:ref type='bibr' target='#b24'>(Dahlhausen et al., 2018)</ns0:ref>. Briefly, a co-occurrence network analysis identified four bacterial taxa, including one of the genus Lonepinella, that could be found in feces of koalas that lived at both the beginning and end of their antibiotic treatment after Chlamydia infection.</ns0:p><ns0:p>However, these four taxa were absent from feces of koalas that died. Furthermore, in the same study a random forest analysis revealed that the most predictive taxon of whether a koala would live or die during their antibiotic treatment was identified as L. koalarum. This finding suggests that L. koalarum could be important for koala health, but the study did not present any evidence relating to PCD degradation in the highly specialized diet of koalas.</ns0:p><ns0:p>It is well understood that animals with highly specialized diets also are likely to have highly specialized intestinal microbial communities <ns0:ref type='bibr' target='#b43'>(Higgins et al., 2011;</ns0:ref><ns0:ref type='bibr' target='#b54'>Kohl et al., 2014;</ns0:ref><ns0:ref type='bibr' target='#b2'>Alfano et al., 2015;</ns0:ref><ns0:ref type='bibr' target='#b56'>Kohl, Stengel & Denise Dearing, 2016)</ns0:ref>. Disturbances of a specialized microbial community, such as the introduction of antibiotics, can have profound effects on the host's health <ns0:ref type='bibr'>(Kohl & Denise Dearing, 2016;</ns0:ref><ns0:ref type='bibr' target='#b11'>Brice et al., 2019)</ns0:ref>. Yet, koalas are regularly treated with antibiotics due to the high prevalence of Chlamydia infections in many populations <ns0:ref type='bibr' target='#b87'>(Polkinghorne, Hanger & Timms, 2013)</ns0:ref>. While recent advances in Chlamydia pecorum vaccines for koalas are a promising alternative for managing koala populations, antibiotics are still the current treatment method for bacterial infections in koalas <ns0:ref type='bibr' target='#b106'>(Waugh et al., 2016;</ns0:ref><ns0:ref type='bibr' target='#b27'>Desclozeaux et al., 2017;</ns0:ref><ns0:ref type='bibr' target='#b78'>Nyari et al., 2018)</ns0:ref>. The antibiotics used in practice might not only target Chlamydia pecorum but also beneficial koala gut symbionts as a side effect. Therefore, it is important to learn about bacteria associated with koala health, such as L. koalarum, in order to further the development of alternative treatments for bacterial infections in koalas and to recommend antibiotic compounds that are potentially less disruptive to members of the koala gut microbiome.</ns0:p><ns0:p>Here we isolated a strain of L. koalarum (hereafter called strain UCD-LQP1) from the feces of a healthy koala (P. cinereus) female at the San Francisco Zoo. We sequenced the genome of L. koalarum UCD-LQP1 using a combination of long-and short-read sequencing, and then assembled and annotated the genome. We compared the genome assembly to the most closely related genomes that are currently publicly available. The genome assembly of L. koalarum UCD-LQP1 was placed in a phylogenetic tree and screened for genes putatively involved in the degradation of plant secondary metabolites, carbohydrate metabolism, and antibiotic resistance.</ns0:p><ns0:p>Additionally, we identified and characterized putative genes that were unique to this strain and two recently sequenced genomes of L. koalarum from Australia.</ns0:p></ns0:div>
<ns0:div><ns0:head>Materials and methods</ns0:head></ns0:div>
<ns0:div><ns0:head>Sampling of koala feces and preparation of culturing media</ns0:head><ns0:p>A koala fecal pellet was collected, with permission from the San Francisco Zoo, from a healthy, adult, captive, female koala (Phascolarctos cinereus). We do not have any information on the geographical origin of this koala. Koalas at the SF Zoo are fed blue gum leaves (Eucalyptus globulus), which grow quite abundantly in California. Jim Nappi and Graham Crawford of the San Francisco Zoo organized and permitted koala fecal sample collection. The fresh fecal pellet was collected from the floor with sterilized tweezers and stored in a sterile 15 ml Falcon tube (Thermo Fisher Scientific, USA). The tube was immediately placed on ice after collection and subsequently stored at 4° C overnight.</ns0:p><ns0:p>The preparation of the Lonepinella koalarum culturing media was modified from methods developed by <ns0:ref type='bibr' target='#b82'>Osawa et al. (1995)</ns0:ref>. A 2 % agarose (Fisher BioReagents, USA) solution of Bacto TM Brain Heart Infusion (BHI; BD Biosciences, USA) was prepared following manufacturer protocols. After the media had solidified in petri dishes, a 2 % tannic acid solution was prepared by combining 1 g of tannic acid powder per 50 ml of sterile Nanopure TM water (Spectrum Chemical MFG CORP, USA). The solution was vortexed for 1 min until PeerJ reviewing PDF | (2020:03:47130:1:1:NEW 9 Aug 2020)</ns0:p><ns0:p>Manuscript to be reviewed homogenized, resulting in a brown, transparent liquid. Using a sterile serological pipette, 5 ml of the 2 % tannic acid solution was gently added to each BHI media plate and left for 20 min. After incubation, the remaining liquid on the plate was decanted. No antibiotic compounds were added to the medium.</ns0:p></ns0:div>
<ns0:div><ns0:head>Culturing of isolates and DNA extraction</ns0:head><ns0:p>The koala fecal pellet was cut in half with sterile tweezers. Tweezers were re-sterilized and used to move approximately 300 mg of material from the center of the pellet to a sterile 2 ml Eppendorf tube containing 1 ml of sterile, Nanopure TM water. The tube was vortexed for 3 min, intermittently checking until the solution was homogenized into a slurry. One hundred µl of the homogenized fecal slurry was micro-pipetted onto to a BHI+tannin plate and stored in an anaerobic chamber (BD GasPak TM EZ anaerobe chamber system; BD Biosciences, USA) at 37° C for 3 days. Each individual colony that grew was plated onto a freshly made BHI+tannin plate using standard dilution streaking techniques. The new plates were stored in an anaerobic chamber at 37 °C for another 3 days. This step was repeated two more times to decrease the probability of contamination or co-culture.</ns0:p><ns0:p>An individual colony from each of the plates from the third round of dilution streaking was moved to a sterile 30 ml glass culture tube containing 5 ml of sterile Bacto TM BHI liquid media (prepared following manufacturer protocol; BD Biosciences, USA). Each tube was then capped with a sterile rubber stopper and purged with nitrogen gas in order to create an anaerobic environment. The tubes were placed in an incubated orbital shaker (ThermoFisher Scientific MaxQ TM 4450) for 3 days at 37 °C at 250 rpm.</ns0:p><ns0:p>Using a sterile serological pipette, 1.8 ml of each liquid culture was transferred to a sterile 2 ml Eppendorf tube. The tubes were spun at 13,000 g for 2 min and the supernatant was carefully decanted. The DNA was extracted from the pellet in each sample with the Promega Wizard Genomic DNA Purification Kit (Promega, USA) according to the manufacturer's protocol. DNA was eluted in a final volume of 100 μl and stored at 4 °C.</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:03:47130:1:1:NEW 9 Aug 2020)</ns0:p><ns0:p>Manuscript to be reviewed</ns0:p></ns0:div>
<ns0:div><ns0:head>PCR and Sanger sequencing</ns0:head><ns0:p>PCR amplification of the 16S rRNA gene was performed on each of the eluted DNA samples. PCR reactions were prepared using the bacteria-specific 'universal' primer pair 27F (5′-AGAGTTTGATCMTGGCTCAG-3′; <ns0:ref type='bibr'>Stackebrandt &</ns0:ref><ns0:ref type='bibr' target='#b101'>Goodfellow, 1991) and</ns0:ref><ns0:ref type='bibr'>1391R (5'-GACGGGCGGTGTGTRCA-3';</ns0:ref><ns0:ref type='bibr' target='#b104'>Turner et al., 1999)</ns0:ref>. PCR amplifications were performed in a BioRad T100 TM Thermal Cycler in 50 μl reactions. Each reaction contained 2 μl of the eluted DNA from the aforementioned extraction, 5 μl of 10x Taq buffer (Qiagen, USA), 10 μl of Q buffer (Qiagen), 1.25 μl of 10mM dNTPs (Qiagen), 2.5 μl of 10mM 27F primer, 2.5 μl of 10mM 1391R primer, 0.3 μl of Taq polymerase (Qiagen), and 26.45 μl of sterile water. The cycling conditions were: (1) 95 °C for 3 min, (2) 40 cycles of 15 sec at 95 °C, 30 sec at 54 °C, and 1 min at 72 °C, (3) a final incubation at 72 °C for 5 min, and (4) holding at 12 °C upon completion.</ns0:p><ns0:p>The PCR product for each sample was purified and concentrated by following the manufacturer's protocol for the NucleoSpin Gel and PCR Clean-up kit (Macherey-Nagel, USA).</ns0:p><ns0:p>The purified PCR product for each sample was quantified by following the manufacturer's protocol for the Qubit dsDNA HS Assay Kit (Thermo Fisher Scientific, USA). The PCR product for each sample was then diluted to 26 ng/μl and submitted for forward and reverse Sanger sequencing at the University of California Davis DNA Sequencing Facility. The program SeqTrace version 0.9.0 <ns0:ref type='bibr' target='#b102'>(Stucky, 2012)</ns0:ref> was used to edit and create consensus sequences of the reads received from the sequencing facility, following the protocol detailed in <ns0:ref type='bibr' target='#b30'>Dunitz et al. (2015)</ns0:ref>. The consensus sequence for each sample was uploaded to the NCBI blast website for organism identification <ns0:ref type='bibr' target='#b65'>(Madden, 2003)</ns0:ref>. The DNA of one of the isolates that had been identified as L. koalarum was used for whole-genome sequencing, as described below. We refer to this isolate as L. koalarum strain UCD-LQP1.</ns0:p></ns0:div>
<ns0:div><ns0:head>Whole genome sequencing and assembly</ns0:head><ns0:p>DNA from one sample identified as L. koalarum strain UCD-LQP1 was submitted for whole genome PacBio sequencing at SNPsaurus. After sequencing, the demultiplexed bam file was tested for reads that contained palindromic sequences since a preliminary assembly with Canu version 1.8 <ns0:ref type='bibr' target='#b57'>(Koren et al., 2017)</ns0:ref> indicated the presence of adapter sequences. Palindromic reads were split in half, aligned with minimap2 (an executable in Canu), and those palindromic reads Manuscript to be reviewed that aligned over at least two-thirds of the split read were reduced to the first part of the palindrome <ns0:ref type='bibr' target='#b57'>(Koren et al., 2017)</ns0:ref>. This procedure efficiently removed adapter sequences. These adapter-free reads were used in the hybrid assembly described below.</ns0:p><ns0:p>The same DNA that had been used for PacBio sequencing was also submitted for Illumina sequencing. Ten ng of genomic DNA were used in a 1:10 reaction of the Nextera DNA Flex Library preparation protocol (Illumina, USA). Fragmented DNA was amplified with Phusion DNA polymerase (New England Biolabs) in 12 PCR cycles with 1 min extension time. Samples were sequenced on a HiSeq4000 instrument (University of Oregon GC3F) with paired-end 150 bp reads. The 10,309,488 raw reads were quality controlled and filtered for adaptors and PhiX using the Joint Genome Institute's BBDuk tool version 37.68 <ns0:ref type='bibr' target='#b15'>(Bushnell, 2014)</ns0:ref>, resulting in 10,302,312 reads. The 308 cleaned PacBio reads and 10,302,312 filtered Illumina reads were combined with all default parameters of Unicycler version 0.4.5, a tool used to assemble bacterial genomes from both long and short reads <ns0:ref type='bibr' target='#b107'>(Wick et al., 2017)</ns0:ref>.</ns0:p></ns0:div>
<ns0:div><ns0:head>Genome annotation</ns0:head><ns0:p>Completeness and contamination of the L. koalarum strain UCD-LQP1 assembly were determined with CheckM version 1.0.8 <ns0:ref type='bibr' target='#b85'>(Parks et al., 2015)</ns0:ref>, number of contigs, total length, GC%, N50, N75, L50, and L75 were determined with QUAST (Quality Assessment Tool for Genome Assemblies; <ns0:ref type='bibr' target='#b40'>Gurevich et al., 2013)</ns0:ref>, and the assembly was annotated with PROKKA version 1.12 <ns0:ref type='bibr' target='#b96'>(Seemann, 2014)</ns0:ref>. The L. koalarum strain UCD-LQP1 genome assembly was uploaded to the Rapid Annotation using Subsystem Technology online tool (RAST), a genome annotation program for bacterial and archaeal genomes <ns0:ref type='bibr'>(Aziz et al., 2008)</ns0:ref>. The SEED viewer in RAST was used to browse features of the genome <ns0:ref type='bibr' target='#b83'>(Overbeek et al., 2014)</ns0:ref>. To screen the L. koalarum strain UCD-LQP1 assembly for genes putatively involved in tannin degradation and xenobiotic metabolisms; i.e., the degradation of plant secondary metabolites, coding regions in the assembly were identified using Prodigal version 2.6.3 <ns0:ref type='bibr' target='#b45'>(Hyatt et al., 2010)</ns0:ref>. Each identified coding region was annotated using eggNOG (a database of orthologous groups and functional annotation that is updated more regularly than PROKKA) mapper version 4.5.1 <ns0:ref type='bibr' target='#b46'>(Jensen et al., 2008)</ns0:ref>. Then, KEGG (Kyoto Encyclopedia of Genes and Genomes) pathways putatively involved in xenobiotics biodegradation and metabolism, were extracted from the eggNOG annotations PeerJ reviewing <ns0:ref type='table'>PDF | (2020:03:47130:1:1:NEW 9 Aug 2020)</ns0:ref> Manuscript to be reviewed <ns0:ref type='bibr'>ko00362, ko00627, ko00364, ko00625, ko00361, ko00623, ko00622, ko00633, ko00642, ko00643, ko00791, ko00930, ko00363, ko00621, ko00626, ko00624, ko00365, ko00984, ko00980, ko00982, and ko00983 (Kanehisa & Goto, 2000)</ns0:ref>), and the corresponding nucleotide sequences from the L. koalarum genome assemblies were saved. Individual genes with hits in KEGG pathways were manually mapped onto KEGG reference maps using the KEGG webtool <ns0:ref type='bibr' target='#b52'>(Kanehisa & Goto, 2000)</ns0:ref>.</ns0:p></ns0:div>
<ns0:div><ns0:head>16S rRNA gene based phylogenetic placement of genome</ns0:head><ns0:p>The 16S rRNA gene sequence within the genome assembly was extracted from RAST by searching for 'ssu rRNA' in the function search of the SEED genome browser <ns0:ref type='bibr'>(Aziz et al., 2008;</ns0:ref><ns0:ref type='bibr' target='#b83'>Overbeek et al., 2014)</ns0:ref>. Following the protocol outlined in <ns0:ref type='bibr' target='#b30'>Dunitz et al. (2015)</ns0:ref>, the sequence was uploaded to the Ribosomal Database Project (RDP; <ns0:ref type='bibr' target='#b22'>Cole et al., 2014)</ns0:ref> and grouped with all sequences in the Pasteurellaceae family and one chosen outgroup, Agarivoran spp., to root the tree. The taxon names from the RDP output file were manually cleaned up and their 16S sequences were used to build a phylogenetic tree with the program FastTree <ns0:ref type='bibr' target='#b89'>(Price, Dehal & Arkin, 2009)</ns0:ref>. Nodes and tip labels were manually edited for Figure <ns0:ref type='figure'>1</ns0:ref> in iTOL (interactive tree of life; web tool; <ns0:ref type='bibr' target='#b61'>Letunic & Bork, 2019)</ns0:ref>. The 16S rRNA gene alignment <ns0:ref type='bibr' target='#b117'>(Wilkins & Coil, 2020a)</ns0:ref> and its resulting phylogenetic tree are available on Figshare <ns0:ref type='bibr' target='#b118'>(Wilkins & Coil, 2020b)</ns0:ref>. During the preparation of this manuscript, two more L. koalarum type strains had their genomes sequenced: one by the DOE Joint Genome Institute, USA (GenBank accession number GCA_004339625.1; 2,486,773 bp long) and one by the Maclean Lab in Australia (GenBank accession number GCA_004565475.1; 2,509,358 bp). Both assemblies were based on type strains originating from the same isolation of L. koalarum in 1995 <ns0:ref type='bibr' target='#b82'>(Osawa et al., 1995)</ns0:ref>, DSM 10053 and ATCC 700131, respectively. These two L. koalarum genome assemblies were henceforth included in our analysis. When we refer to all three L. koalarum genome assemblies, we simply say 'in L.</ns0:p><ns0:p>koalarum' and when we refer to the strain sequenced in this study, we use 'the assembly of L. koalarum strain UCD-LQP1'. <ns0:ref type='table'>PDF | (2020:03:47130:1:1:NEW 9 Aug 2020)</ns0:ref> Manuscript to be reviewed</ns0:p></ns0:div>
<ns0:div><ns0:head>PeerJ reviewing</ns0:head></ns0:div>
<ns0:div><ns0:head>Comparative genomics</ns0:head><ns0:p>The GTDB-Tk software toolkit version 0.3.0 <ns0:ref type='bibr'>(Chaumeil, Hugenholtz & Parks, 2018)</ns0:ref> of the Genome Taxonomy Database (GTDB) project was chosen to place L. koalarum into a pregenerated conserved marker gene tree using 120 marker genes <ns0:ref type='bibr' target='#b84'>(Parks et al., 2018)</ns0:ref>. After placing the assembly into the GTDB tree, a clade in the tree was extracted that contained L. koalarum strain UCD-LQP1 and 55 other taxa, of which all members belonged to the order Pasteurellales.</ns0:p><ns0:p>This clade contained all sequenced genomes of the closest neighboring taxa (n = 55) to L. koalarum in the GTDB tree at the time of this analysis (3 rd of August 2019). All of these 55 genomes were downloaded from GenBank (using the accession numbers in the GTDB) to perform a comparative genomic analysis in Anvi'o version 5.5 <ns0:ref type='bibr' target='#b33'>(Eren et al., 2015)</ns0:ref>. The two other L. koalarum genomes from GenBank were included in the following analysis as well. Accession numbers of all genome assemblies included can be found in Supplementary Table <ns0:ref type='table'>S1</ns0:ref> (n = 58).</ns0:p><ns0:p>The Anvi'o workflow for microbial pangenomics was followed <ns0:ref type='bibr' target='#b26'>(Delmont & Eren, 2018)</ns0:ref>. The blastp program from NCBI was used for a gene search <ns0:ref type='bibr' target='#b4'>(Altschul et al., 1990)</ns0:ref>, the Markov Cluster algorithm (MCL) version 14.137 (van Dongen & Abreu-Goodger, 2012) was used for clustering, and the program MUSCLE was used for alignment <ns0:ref type='bibr' target='#b32'>(Edgar, 2004</ns0:ref>). An inflation parameter of 6 was chosen to identify clusters in amino acid sequences. Genomes in the pangenome of Anvi'o were ordered based on a genomic marker gene tree. This tree was built in PhyloSift version 1.0.1 <ns0:ref type='bibr' target='#b25'>(Darling et al., 2014)</ns0:ref> with its updated markers database (version 4, posted on 12th of February 2018; <ns0:ref type='bibr' target='#b51'>Jospin, 2018)</ns0:ref> for the alignment. We used RAxML version 8.2.10 on the CIPRES web server for the tree inference <ns0:ref type='bibr' target='#b71'>(Miller, Pfeiffer & Schwartz, 2010)</ns0:ref> following the analysis in <ns0:ref type='bibr' target='#b119'>(Wilkins et al., 2019)</ns0:ref>. Gene clusters in Anvi'o were ordered based on presence/absence. We also used Anvi'o to compute average nucleotide identities across the genomes with PyANI <ns0:ref type='bibr' target='#b90'>(Pritchard et al., 2016)</ns0:ref>. In the heatmap, ANI values > 95% (and >70% for a separate figure, respectively) were colored in red.</ns0:p><ns0:p>Gene clusters from the Anvi'o microbial pangenomics analysis that could only be found in the three L. koalarum genome assemblies were extracted. Then, we also extracted all gene clusters that could only be found in the assembly of L. koalarum strain UCD-LQP1. Partial sequences were removed. A literature search of the remaining genes was conducted to identify possible roles L. koalarum might play in the gut microbiome of koalas. Tables were summarized in R version 3.4.0 (R Development Core Team, 2013).</ns0:p></ns0:div>
<ns0:div><ns0:head>Carbohydrate metabolism</ns0:head><ns0:p>Since the majority of gene clusters unique to L. koalarum genome assemblies fell into the COG (Clusters of Orthologous Groups) category 'Carbohydrate metabolism', we decided to screen all three assemblies against the Carbohydrate-Active Enzymes Database (CAZy), an expert resource for glycogenomics <ns0:ref type='bibr' target='#b18'>(Cantarel et al., 2009;</ns0:ref><ns0:ref type='bibr' target='#b63'>Lombard et al., 2014)</ns0:ref>. In brief, CAZy domains were identified based on CAZy family HMMs (Hidden Markov Models) with a coverage of >95% and an e-value < 1e-15. Searches were done through dbCAN, a web resource for automated carbohydrate-active enzyme annotation <ns0:ref type='bibr' target='#b122'>(Yin et al., 2012)</ns0:ref> and CAZy hits were only retained if they had been found with all three search tools. The three search tools included (i) HMMER version 3.3 <ns0:ref type='bibr' target='#b31'>(Eddy, 1998)</ns0:ref>, (ii) DIAMOND version 0.9.29 for fast blast hits in the CAZy database <ns0:ref type='bibr' target='#b14'>(Buchfink, Xie & Huson, 2015;</ns0:ref><ns0:ref type='bibr'>default parameters;</ns0:ref><ns0:ref type='bibr'>i.e</ns0:ref>., e-value < 1e-102, hits per query (-k) = 1), and (iii) Hotpep version 1 for short, conserved motifs in the PPR (Peptide Pattern Recognition) library <ns0:ref type='bibr' target='#b16'>(Busk et al., 2017;</ns0:ref><ns0:ref type='bibr'>default parameters;</ns0:ref><ns0:ref type='bibr'>i.e., frequency > 2.6, hits > 6)</ns0:ref>. For a detailed walk-through of the assembly, annotation, search for KEGG pathways, and comparative genomics analyses, please refer to the associated Jupyter notebook <ns0:ref type='bibr' target='#b109'>(Wilkins, 2020a)</ns0:ref>.</ns0:p></ns0:div>
<ns0:div><ns0:head>Identification of antibiotic resistance genes</ns0:head><ns0:p>All three L. koalarum genome assemblies were uploaded to the Comprehensive Antibiotic Resistance Database (CARD version 3.0.7; <ns0:ref type='bibr' target='#b47'>Jia et al., 2017)</ns0:ref> and the ResFinder database version 5.1.0 <ns0:ref type='bibr' target='#b123'>(Zankari et al., 2012)</ns0:ref> to screen them for putative antibiotic resistance genes and their variants using blastn searches against CARD 2020 reference sequences using default parameters.</ns0:p><ns0:p>The Resistance Gene Identifier (RGI) search pipeline was used to detect SNPs (single nucleotide polymorphisms) using the 'perfect, strict, complete genes only' criterion on their website. Manuscript to be reviewed that had been previously associated with antibiotic resistance in other bacterial species <ns0:ref type='bibr' target='#b1'>(Alcock et al., 2020)</ns0:ref>.</ns0:p></ns0:div>
<ns0:div><ns0:head>Results and discussion</ns0:head></ns0:div>
<ns0:div><ns0:head>Identification of isolates</ns0:head><ns0:p>Besides the isolates identified as L. koalarum, we had several other colonies growing on the BHI+tannin plates, including isolates with 16S rRNA gene sequences that matched Bacillus cereus, Bacillus nealsonii, Bacillus sonorensis, and Escherichia coli. E. coli was the most common species isolated.</ns0:p></ns0:div>
<ns0:div><ns0:head>Assembly taxonomy and gene annotation</ns0:head><ns0:p>The hybrid assembly generated was 2,608,483 bp in length with an N50 of 2,299,135 bp and a coverage of 672. According to the marker gene analysis in CheckM, the assembly was 99.21% complete and less than 1% contaminated with a GC content of 39.02% (see Table <ns0:ref type='table'>1</ns0:ref> for additional details). One contig in the assembly appears to be a 3,899 bp long plasmid. This is indicated by circularity of that contig and positive matches to plasmids in related taxa when uploaded to the NCBI blast website for organism identification <ns0:ref type='bibr' target='#b65'>(Madden, 2003)</ns0:ref>. The two most similar sequences on GenBank were a 71 percent similar sequence of Pasteurella multocida strain U-B411 plasmid pCCK411 (accession number FR798946.1) and a 70 percent similar sequence of Mannheimia haemolytica strain 48 plasmid pKKM48 (accession number MH316128.1). The putative plasmid sequence was deposited on FigShare <ns0:ref type='bibr' target='#b121'>(Wilkins & Jospin, 2020)</ns0:ref>.</ns0:p><ns0:p>The taxonomy of L. koalarum strain UCD-LQP1 was confirmed in three ways. First, a phylogenetic tree was built based on the 16S rRNA gene extracted from the new assembly. This 16S rRNA gene sequence was aligned with other closely related 16S rRNA gene sequences on the RDP website where 16S rRNA gene sequences of type strains are curated and sequences of the closest relatives of a taxon are usually readily available <ns0:ref type='bibr' target='#b30'>(Dunitz et al., 2015)</ns0:ref>. The phylogenetically closest sequence to L. koalarum strain UCD-LQP1 in the 16S rRNA gene tree was one from Lonepinella koalarum Y17189 (Fig. <ns0:ref type='figure'>1</ns0:ref>). Second, a whole genome concatenated gene marker tree was inferred using the Genome Taxonomy Database (GTDB), as well as using PeerJ reviewing PDF | (2020:03:47130:1:1:NEW 9 Aug 2020)</ns0:p><ns0:p>Manuscript to be reviewed PhyloSift, in parallel. In the GTDB tree, L. koalarum UCD-LQP1 was placed closest to Actinobacillus succinogenes (GenBank accession number GCA_000017245.1). Note that as of February 3, 2020, GTDB did not include any of the L. koalarum genome assemblies. In the PhyloSift marker gene tree, all three L. koalarum assemblies clustered together, and A. succinogenes was their phylogenetically closest neighbor (Fig. <ns0:ref type='figure' target='#fig_5'>2</ns0:ref>). Third, the average nucleotide identity (ANI) between the genome of the L. koalarum type strain (DSM 10053; GenBank accession number GCA_004339625.1) and the assembly of L. koalarum UCD-LQP1 was estimated at 98.91 percent (standard deviation 0.17%). The ANI value between L. koalarum UCD-LQP1 and GCA_004565475.1 was 98.99 percent (SD 0.15%) and the ANI value between GCA_004339625.1 and GCA_004565475.1 was 99.99 percent (SD 0.08%). Both of these genome assemblies are based on the type strain of L. koalarum that originated in 1995 <ns0:ref type='bibr' target='#b82'>(Osawa et al., 1995)</ns0:ref>. All three approaches confirmed the taxonomy of strain UCD-LQP1 as Lonepinella koalarum. Interestingly, A. succinogenes (GenBank accession number GCA_000017245.1) belongs now to a different taxonomic group based on GTDB taxonomy, namely Basfia succinogenes. <ns0:ref type='bibr' target='#b84'>Parks et al. (2018)</ns0:ref>, among others (e.g., <ns0:ref type='bibr' target='#b44'>Hug et al., 2016;</ns0:ref><ns0:ref type='bibr' target='#b19'>Castelle & Banfield, 2018)</ns0:ref>, have suggested relying on whole genome sequencing to reorganize the microbial tree of life, which will result in a majority of changes in classification and naming, and ultimately reflect a more accurate evolutionary relationship among groups <ns0:ref type='bibr' target='#b84'>(Parks et al., 2018)</ns0:ref>.</ns0:p><ns0:p>There were no positive hits for any annotations associated with tannin degradation in the RAST SEED viewer. This negative result is in contrast to the experimentally verified tannin-degrading functions reported for this bacterium <ns0:ref type='bibr' target='#b82'>(Osawa et al., 1995)</ns0:ref>. Moreover, tannic acid powder had been used to prepare the culturing medium and was expected to help select for bacterial tannin degraders. There are several potential explanations for the absence of any positive hits for tannins in the RAST database including (1) the genes responsible for tannin degradation in the assembly of L. koalarum UCD-LQP1 are not labeled as such, or (2) L. koalarum does not have any tannin-degradation functionality. We thus carried out additional sequence-based analyses searching for possible toxin degrading genes in the new assembly.</ns0:p><ns0:p>According to the annotation with PROKKA, there were 2,551 predicted genes and 2,479 protein coding genes. In comparison, eggNOG predicted 2,370 protein coding genes. Neither annotation PeerJ reviewing PDF | (2020:03:47130:1:1:NEW 9 Aug 2020)</ns0:p><ns0:p>Manuscript to be reviewed included any genes annotated as 'tannase'. However, among the eggNOG predictions, there were 79 genes putatively involved in Class 1.11 Xenobiotics biodegradation and the degradation of plant secondary metabolites (Table <ns0:ref type='table'>2</ns0:ref>). There are 20 KEGG pathways included in this group.</ns0:p><ns0:p>We searched for all twenty pathways in the assembly of L. koalarum strain UCD-LQP1 and found positive hits in 13 pathways (Table <ns0:ref type='table'>2</ns0:ref>). Each hit represents a translated amino acid sequence from the assembly of L. koalarum UCD-LQP1 that is encoded by an individual gene in a pathway. The largest proportion of hits (n=15) comprised putative enzymes that are members in this KEGG class, but do not fall into a particular pathway (KEGG pathway ko00983: Drug metabolism -other enzymes). Potential tannin-degrading genes might be found in this group but have not been labeled as tannase genes because their sequences are not similar enough to any known tannase genes or because these tannase genes are not annotated in any database. The second largest KEGG pathway was ko00362 benzoate degradation, followed by pathway ko00980 metabolism of xenobiotics by Cytochrome P450 and pathway ko00625 chloroalkane and chloroalkene degradation. KEGG pathways with fewer hits included the degradation compounds such as aminobenzoate, xylene, naphthalene, dioxin, and chlorocyclohexane.</ns0:p><ns0:p>Mapping individual genes onto KEGG pathways revealed continuous degradation chains for the following compounds: Azathioprine (pro-drug) to 6-Thioguanine (Supplementary Fig. <ns0:ref type='figure'>S1);</ns0:ref> Aminobenzoate degradation; i.e., 4-Carboxy-2-hydroxymuconate semialdehyde to Pyruvate and Oxaloacetate, which can then be fed into the Cytrate cycle (Supplementary Fig. <ns0:ref type='figure' target='#fig_5'>S2</ns0:ref>); 2-Aminobenzene-sulfonate to Pyruvate, which, again, can be fed directly into Glycolysis or with another enzyme that was present (1.2.1.10) can be converted into Acetaldehyde, then Acetyl-CoA , and then fed into the Cytrate cycle (Supplementary Fig. <ns0:ref type='figure' target='#fig_5'>S2</ns0:ref>). In the group of xenobiotics metabolized by cytochrome P450 there were seven complete chains (Supplementary Fig. <ns0:ref type='figure'>S3</ns0:ref>): degradation of (i) benzo(a)pyrene, (ii) Aflatoxin B1, (iii) 1-Nitronaphtalene, (iv) 1,1-Dichloroethylene, (v) Trichloroethylene, (vi) Bromobenzene, and (vii) 1,2-Dibromoethane. All of these complete, putative conversion chains present in L. koalarum might explain further how this member of the koala gut microbiome contributes to koala gastro-physiology (see discussion below). Amino acid sequences encoded by putative toxin degrading genes in L. koalarum strain UCD-LQP1 can be downloaded from FigShare <ns0:ref type='bibr' target='#b110'>(Wilkins, 2020b)</ns0:ref>. A table linking eggNOG annotations to positions in individual assemblies and translated amino acid sequences can be found in Supplementary Table <ns0:ref type='table'>S2</ns0:ref>. A complete table of all eggNOG annotations in the assembly of L. koalarum strain UCD-LQP1 can be found in Supplementary Table <ns0:ref type='table'>S3</ns0:ref>.</ns0:p><ns0:p>Eucalyptus spp. leaves contain more than 100 different chemical compounds including phenolics, terpenoids and lipids that are harmful for herbivores, even at low concentration <ns0:ref type='bibr' target='#b66'>(Maghsoodlou et al., 2015)</ns0:ref>. Koalas are highly specialized folivores feeding on these leaves. We assumed that L. koalarum plays a beneficial role for koala hosts because some strains have shown experimentally to be able to degrade tannins <ns0:ref type='bibr' target='#b79'>(Osawa, 1990;</ns0:ref><ns0:ref type='bibr' target='#b82'>Osawa et al., 1995)</ns0:ref>, and tannic acid was used to isolate L. koalarum strain UCD-LQP1. Alas, we did not find any direct evidence for tannase genes in the assembly of L. koalarum UCD-LQP1. However, genes encoding several putative pathways involved in plant secondary metabolite degradation were found in the assembly of L. koalarum UCD-LQP1. The predicted pathways included those for degradation of compounds that had been extracted from Eucalyptus leaves (e.g., benzoate, aminobenzoate, and chlorocyclohexane; <ns0:ref type='bibr' target='#b92'>Quinlivan et al., 2003;</ns0:ref><ns0:ref type='bibr' target='#b70'>Marzoug et al., 2011;</ns0:ref><ns0:ref type='bibr' target='#b95'>Sebei et al., 2015;</ns0:ref><ns0:ref type='bibr' target='#b66'>Maghsoodlou et al., 2015;</ns0:ref><ns0:ref type='bibr' target='#b97'>Shiffman et al., 2017)</ns0:ref>. Degradation of these toxic compounds might explain the beneficial role that L. koalarum plays in the koala gut microbiome.</ns0:p></ns0:div>
<ns0:div><ns0:head>Comparative genomics and unique genes in L. koalarum</ns0:head><ns0:p>The GTDB tree clade used to extract related genomes of L. koalarum strain UCD-LQP1 consisted mostly of Haemophilus spp. (n = 28), followed by Rodentibacter spp. (n = 13), Pasteurella spp. (n = 5), Aggregibacter spp. (n = 4), and seven other genera (Table <ns0:ref type='table'>S1</ns0:ref>). Whole genome marker phylogenetic trees showed that not all genera were monophyletic. This can be seen in Figure <ns0:ref type='figure' target='#fig_5'>2</ns0:ref> in the way the coloring based on genus name does not group perfectly when taxa are ordered according to their phylogenetic relationship. This was especially the case for Haemophilus spp., which is shown in light purple. Some of the Haemophilus genomes were grouped together, whereas others grouped with genomes labeled as Pasteurella spp., Necropsobacter spp. and Avibacterium spp. One species of Rodentibacter (R. Manuscript to be reviewed taxonomy to Basfia succinogenes, most probably the most closely related taxon to L. koalarum that has its genome sequenced to date. Twelve out of the 55 NCBI microbial genome assemblies have different taxonomic names in the new GTDB taxonomy (Supplementary Table <ns0:ref type='table'>S1</ns0:ref>). For a discussion of the re-organization and re-naming of the microbial tree of life based on whole genome sequencing see above Assembly taxonomy and gene annotation. The whole genome marker gene tree used to order genomes in Anvio's visualization can be downloaded from FigShare <ns0:ref type='bibr' target='#b111'>(Wilkins, 2020c)</ns0:ref>, as well as its corresponding amino acid alignment <ns0:ref type='bibr' target='#b112'>(Wilkins, 2020d)</ns0:ref>.</ns0:p><ns0:p>Average nucleotide identities have been put forward as a measure of genomic relatedness among bacteria that could help designate genera and be used besides the 16S rRNA gene as a taxonomic marker <ns0:ref type='bibr' target='#b7'>(Barco et al., 2020)</ns0:ref>. Moreover, it has been suggested to use an ANI threshold of larger than 95% to delineate bacterial species <ns0:ref type='bibr' target='#b39'>(Goris et al., 2007)</ns0:ref>. Based on this definition, the genomes used for the comparative genomic analysis with L. koalarum are all distinct species (heatmap in Fig. <ns0:ref type='figure' target='#fig_5'>2</ns0:ref>). We created a second heatmap visualizing genomic relatedness at the 70 % level (Supplementary Fig. <ns0:ref type='figure'>S4</ns0:ref>). This heatmap revealed several distinct clusters of closely related genomes vs. singleton genomes (i.e., taxa that did not group together with anything else at the 70 percent threshold): Cluster 1) Aggregatibacter spp., 2) first main Haemophilus spp. group, 3) Rodentibacter spp., 4) second main Haemophilus spp. group, 5) L. koalarum genome assemblies, and 6) two Necropsobacter spp. and another Haemophilus spp. Notably, Rodentibacter heylii, all Pasteurella spp., and Avibacterium paragallinarum did not cluster with anything. The heatmap is a way of visualizing sequence similarity groups and overall, it showed that the genera Haemophilus, Pasteurella and Rodentibacter do not represent coherent groups of species or genera. These three genera were found in several sub-groups (clusters in the ANI heatmap in Supplementary Fig. <ns0:ref type='figure'>S4</ns0:ref>) that have been described previously based on a much larger sample size and a few marker genes <ns0:ref type='bibr' target='#b74'>(Naushad et al., 2015)</ns0:ref>. Even some of the same singleton genomes were reported as their own branches in previous phylogenetic trees <ns0:ref type='bibr' target='#b21'>(Christensen et al., 2003)</ns0:ref>. L. koalarum was placed in the middle of a group containing mostly Haemophilus, Pasteurella and Basfia species. Pasteurellaceae, the single constituent family of the order Pasteurellales hosts a diverse group of mostly pathogenic bacteria that had been assigned to this group based on phenotypic traits, often related to their pathology, and GC content <ns0:ref type='bibr' target='#b67'>(Mannheim, Pohl & Holländer, 1980)</ns0:ref>. For example, the genus Haemophilus includes a plethora of taxa that cause pneumonia and meningitis in humans, and Pasteurella have been associated with a range of infectious diseases in cattle, fowl and pigs <ns0:ref type='bibr' target='#b74'>(Naushad et al., 2015)</ns0:ref>. Moreover, since sequencebased taxonomies have become more common, new genera have been created within each genus, such as for example Aggregatibacter <ns0:ref type='bibr' target='#b77'>(Norskov-Lauritsen, 2006)</ns0:ref> or Avibacterium <ns0:ref type='bibr' target='#b9'>(Blackall et al., 2005)</ns0:ref>. We believe that a work-over of the evolutionary genetic relationship of the Pasteurellales is overdue.</ns0:p><ns0:p>The proportion of gene clusters that were unique to the three L. koalarum genome assemblies, relative to 55 of their most closely related genomes, was large relative to the size of genes that were unique to other genera in Anvio's pangenome analysis (Fig. <ns0:ref type='figure' target='#fig_5'>2</ns0:ref>). There were 282 gene clusters that could exclusively be found in the three L. koalarum genome assemblies. Among them, there were 136 gene clusters with complete sequences and COG annotation (Supplementary Table <ns0:ref type='table'>S4</ns0:ref>). There were 36 gene clusters unique to L. koalarum strain UCD-LQP1 and 19 of these had complete sequences and COG annotations (Supplementary Table <ns0:ref type='table'>S5</ns0:ref>).</ns0:p><ns0:p>Out of the 136 gene clusters with known COG functions that were unique to the three L. koalarum genome assemblies, 22 different gene clusters fell into the COG category 'Carbohydrate metabolism/transport'. This was the largest category, followed by 'Inorganic ion transport' (n = 15), 'Cell wall', 'Transcription', and 'Energy production' (n = 11, each), and 'Defense' (n = 7; Table <ns0:ref type='table'>3</ns0:ref> and Supplementary Table <ns0:ref type='table'>S4</ns0:ref>). The translated amino acid sequences for these gene clusters, extracted from L. koalarum strain UCD-LQP1, can be found in Supplementary Table <ns0:ref type='table'>S6</ns0:ref>.</ns0:p><ns0:p>Gene clusters in the category 'Carbohydrate metabolism and transport' are discussed in detail below. It is worth mentioning that several putative components of the phosphotranspherase system were unique to L. koalarum. This system transports sugars into bacteria including glucose, mannose, fructose, and cellobiose. It can differ among bacterial species, mirroring the most suitable carbon sources available in the environment where a species evolved <ns0:ref type='bibr' target='#b103'>(Tchieu et al., 2001)</ns0:ref>. L. koalarum also stood out in terms of genes coding for cell wall components including for example teichoic acid and other outer membrane proteins (Table <ns0:ref type='table'>3</ns0:ref> and Supplementary Table <ns0:ref type='table'>S4</ns0:ref>). These outer membrane proteins are diverse and can significantly differ among bacterial species <ns0:ref type='bibr' target='#b94'>(Schleifer & Kandler, 1972)</ns0:ref>. A few other potentially unique gene clusters included genes coding for type IV pilus assembly proteins for species-specific pili and fimbria <ns0:ref type='bibr' target='#b91'>(Proft & Baker, 2009)</ns0:ref>; defense mechanisms, such as putative bacteriophage resistance proteins, phage repressor proteins; and drug transport and efflux pumps. Several of these factors are characteristic for pathogenic bacteria <ns0:ref type='bibr' target='#b23'>(Craig, Pique & Tainer, 2004</ns0:ref>). Here it is worth noting that a gram-negative bacterium that was assigned to the genus Lonepinella based on 16S rRNA gene sequences caused a human wound infection after a wildlife worker had been bitten by a koala <ns0:ref type='bibr' target='#b99'>(Sinclair et al., 2019)</ns0:ref>.</ns0:p></ns0:div>
<ns0:div><ns0:head>Carbohydrate metabolism</ns0:head><ns0:p>Since the majority of unique gene clusters in all three L. koalarum genome assemblies were related to carbohydrate metabolism and transport, we decided to screen all three L. koalarum assemblies for potential enzymes that assemble, modify, and breakdown oligo-and polysaccharides. Using very stringent selection thresholds of the CAZy database where genes coding for carbohydrate-active enzymes have to be identified by three different methods, we found evidence for the presence of genes encoding 15 different glycoside hydrolase families, three different carbohydrate esterase families, and nine different glycosyltransferase families (Table <ns0:ref type='table'>4</ns0:ref>). Note, gene families in L. koalarum are predicted to have these activities in carbohydrate metabolism and transport based on characterized other members in the CAZy database, but we do not provide experimental evidence that L. koalarum performs these activities. All 28 identified CAZy gene families had also been annotated in the 2,370 eggNOG annotations (Supplementary Table <ns0:ref type='table'>S3</ns0:ref>). Glycoside hydrolase families, GH2, GH31, GH32, GH43, and GH77 were only found in the three L. koalarum genome assemblies relative to the other taxa in the comparative genomic analysis (see also Table <ns0:ref type='table'>3</ns0:ref> and Supplementary Table <ns0:ref type='table'>S7</ns0:ref>).</ns0:p><ns0:p>These five glycoside hydrolases are responsible for the hydrolysis of glycosidic bonds. Notably, when Lonepinella koalarum was isolated and described the first time as a phylogenetically and phenotypically novel group within the family Pasteurellaceae, enzyme activities were determined using commercially available oxidase/catalase tests as well as high-pressure liquid chromatography <ns0:ref type='bibr' target='#b82'>(Osawa et al., 1995)</ns0:ref>. The new taxon in 1995 (first described L. koalarum) showed positive results for beta-galactosidase (putatively enzyme family GH2) and alphaamylase (putatively enzyme family GH77) and negative results for urease, arginine dihydrolase, lysine decarboxylase, and tryptophane desaminase in congruence with the sequence-based results here.</ns0:p><ns0:p>Genes coding for oligosaccharide-degrading enzymes in the families GH1, GH2, GH3, GH42, and GH43 have also been found in another study that was investigating koala and wombat metagenomes <ns0:ref type='bibr' target='#b97'>(Shiffman et al., 2017)</ns0:ref>. Especially GH2, GH3 and GH43 were relatively common in koala metagenomes, relative to wallaby foregut <ns0:ref type='bibr' target='#b88'>(Pope et al., 2010)</ns0:ref>, cow rumen <ns0:ref type='bibr'>(Brulc et al., 2009)</ns0:ref>, and termite hindgut <ns0:ref type='bibr' target='#b42'>(He et al., 2013)</ns0:ref> metagenomes, where these enzymes had also been characterized. These five glycoside hydrolase families comprise mostly oligosaccharidedegrading enzymes <ns0:ref type='bibr' target='#b3'>(Allgaier et al., 2010)</ns0:ref>; i.e., they are able to break down a specific group of monosaccharide sugars in other bacteria that had been characterized for the CAZy database. However, presumably the major components of koala diet that are difficult to digest for the host are plant secondary metabolites and plant cell walls in Eucalyptus leaves, and oligosaccharidedegrading enzymes only play a significant role in a koala's diet after other enzymes have already degraded cellulose in leaf plant cell walls <ns0:ref type='bibr'>(Moore et al., 2005)</ns0:ref>. Oligosaccharides in Eucalyptus leaves will be absorbed by the koala in the small intestine and only a small fraction enter the caecum and colon. This means that the bacteria in the hindgut are most likely using their metabolic pathways to process the products of the degradation of complex carbohydrates with cross-feeding among microbiome members. The benefit of this activity to koala nutrition is not well understood. Interestingly, among the genes that code for the three carbohydrate-active enzyme families that were found exclusively in the assembly of L. koalarum strain UCD-LQP1, two were actual lignocellulases; i.e., microbial enzymes that hydrolyze the beta-1,4 linkages in cellulose <ns0:ref type='bibr' target='#b3'>(Allgaier et al., 2010)</ns0:ref>: Enzyme family GH42 and CE4. GH42 enzymes have mostly been described in cellulose-degrading bacteria, archaea and fungi <ns0:ref type='bibr' target='#b59'>(Kosugi, Murashima & Doi, 2002;</ns0:ref><ns0:ref type='bibr' target='#b98'>Shipkowski & Brenchley, 2006;</ns0:ref><ns0:ref type='bibr' target='#b28'>Di Lauro et al., 2008)</ns0:ref>. CE4 is a member of the carbohydrate esterase family, which groups enzymes that catalyze the de-acetylation of plant cell wall polysaccharides <ns0:ref type='bibr' target='#b8'>(Biely, 2012)</ns0:ref>. Digestion of plant cell walls, (i.e., cellulose, hemicellulose, and lignin), could be a second explanation (besides toxin degradation) of how L. koalarum plays a beneficial role in the koala gut microbiome.</ns0:p></ns0:div>
<ns0:div><ns0:head>Antibiotic resistance genes</ns0:head><ns0:p>Screening the three L. koalarum genome assemblies against the ResFinder database did not result in any detection of antibiotic resistance variants. However, there were three hits in the CARD database. First, all three L. koalarum assemblies contained a gene coding for a translated amino acid variant at a specific position (SNP R234F) that had been shown to confer resistance to pulvomycin in other bacterial species based on CARD predictions. Secondly, a variant was found to be encoded in all three L. koalarum genome assemblies that had been described before in Haemophilus influenza mutant PBP3, conferring resistance to beta-lactam antibiotics (cephalosporin, cephamycin, and penam) with SNPs D350N and S357N. The third result was an amino acid position with reference to a protein homolog model in a Klebsiella pneumoniae mutant, conferring resistance to the antibiotic efflux pump KpnH (including macrolide antibiotics, fluoroquinolone, aminoglycoside, carbapenem, cephalosporin, penam, and penem).</ns0:p><ns0:p>These results are based on predictions from the CARD 2020 database. All three hits are nucleotide sequences in the L. koalarum assemblies that are predicted to encode proteins that showed the same amino acid variants as other bacterial species in the CARD database. We do not know whether these variants confer antibiotic resistance in L. koalarum. Additional experiments are necessary to confirm that these CARD predictions work for L. koalarum. The corresponding nucleotide sequences and CARD output files are deposited on FigShare (UCD-LQP1: Wilkins, 2020e; ATCC 700131: <ns0:ref type='bibr' target='#b114'>Wilkins, 2020f;</ns0:ref><ns0:ref type='bibr'>and DSM 10053: Wilkins, 2020e)</ns0:ref>.</ns0:p></ns0:div>
<ns0:div><ns0:head>Recommendations for future koala management strategies</ns0:head><ns0:p>In previous work, we identified L. koalarum as the most predictive taxon of koala survival during antibiotic treatment and we suggested that this bacterium is important for koala health <ns0:ref type='bibr' target='#b24'>(Dahlhausen et al., 2018)</ns0:ref>. Here, we isolated a L. koalarum strain from the feces of a healthy koala and sequenced and characterized its genome. We found several putative detoxification pathways in L. koalarum strain UCD-LQP1 that could explain its potential beneficial role in the koala gut for koala survival and fitness. Besides detoxification of plant secondary metabolites, we found several putative genes involved in carbohydrate metabolism, particularly cellulose degradation. Some of these genes were only found in L. koalarum assemblies and not in 55 of their closely related genomes. Based on CARD predictions, the L. koalarum assemblies contain some sequences that are similar to antibiotic resistance genes in other bacterial species. We suggest confirming these antibiotic resistances in L. koalarum experimentally and testing the PeerJ reviewing PDF | (2020:03:47130:1:1:NEW 9 Aug 2020)</ns0:p><ns0:p>Manuscript to be reviewed efficiency of these antibiotic compounds against Chlamydia infections in koalas. In light of the various threats that koalas face, from chlamydia infection to wildfires <ns0:ref type='bibr' target='#b87'>(Polkinghorne, Hanger & Timms, 2013)</ns0:ref>, and the growing interest in rescuing and treating them in sanctuaries and zoos, it is important to identify beneficial members of their microbiome. This could (i) help decide which antibiotic compounds to choose during chlamydia treatment in order to maximize persistence of beneficial members in the koala gut microbiome, and (ii) guide the development of probiotic cocktails during recovery (Jin <ns0:ref type='bibr' target='#b48'>Song et al., 2019)</ns0:ref>.</ns0:p><ns0:p>Figure <ns0:ref type='figure'>1</ns0:ref> 16S rRNA gene phylogenetic placement of Lonepinella koalarum strain UCD-LQP1</ns0:p><ns0:p>The 16S rRNA gene was extracted from the L. koalarum genome assembly by searching for 'ssu rRNA' in the RAST function search of the SEED genome browser <ns0:ref type='bibr'>(Aziz et al., 2008;</ns0:ref><ns0:ref type='bibr' target='#b83'>Overbeek et al., 2014)</ns0:ref>. Included are all known 16S rRNA sequences in the Pasteurellaceae According to GTDB taxonomy, those two genomes are now Basfia species. See Discussion section). Each wedge represents a gene cluster. Gene clusters were grouped into mostly shared, shared, private, and in red: exclusively found in Lonepinella koalarum genome assemblies: 'LK', and exclusively found in L. koalarum strain UCD-LQP1. The gene marker tree was created in PhyloSift version 1.0.1 <ns0:ref type='bibr' target='#b25'>(Darling et al., 2014)</ns0:ref> with its updated markers database <ns0:ref type='bibr'>(version 4, posted on 12th of February 2018;</ns0:ref><ns0:ref type='bibr' target='#b51'>Jospin, 2018)</ns0:ref> for the alignment and RAxML version 8.2.10 on the CIPRES web server for the tree inference <ns0:ref type='bibr' target='#b71'>(Miller, Pfeiffer & Schwartz, 2010)</ns0:ref>. Gene clusters were ordered based on presence/absence. Also shown is GC content in light brown, number of genes per kilo base pairs in light grey, number of gene clusters in dark grey, and number of singleton gene clusters in orange, for each assembly, respectively. The heatmap shows ANI (Average nucleotide identity) values > 95%. The ANI heatmap is aligned with the Anvi'o profile, leading to the genome IDs on the y-axis. The Anvi'o database <ns0:ref type='bibr' target='#b115'>(Wilkins, 2020g)</ns0:ref> and profile <ns0:ref type='bibr' target='#b116'>(Wilkins, 2020h)</ns0:ref> </ns0:p></ns0:div><ns0:figure xml:id='fig_0'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:03:47130:1:1:NEW 9 Aug 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_1'><ns0:head>(</ns0:head><ns0:label /><ns0:figDesc>Class 1.11 Xenobiotics biodegradation and metabolism includes the following KEGG pathways:</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_2'><ns0:head /><ns0:label /><ns0:figDesc>Briefly, antibiotic resistance genes were searched with nucleotide sequences as input. RGI first predicts complete open reading frames (ORFs) using Prodigal version 2.6.3. To find protein homologs in the CARD references, DIAMOND version 0.9.29 is used. The 'perfect' algorithm detects perfect matches of individual amino acids to positions in the curated reference sequences PeerJ reviewing PDF | (2020:03:47130:1:1:NEW 9 Aug 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_3'><ns0:head /><ns0:label /><ns0:figDesc>heylii) was closest to Aggregatibacter spp. (yellow and pink in Fig. 2). Actinobacillus succinogenes and Mannheimia succiniproducens grouped with Pasteurella spp., while the former was the most closely related non-Lonepinella genome to L. koalarum strain UCD-LQP1. Here it is worth noting that both A. succinogenes and M. succiniproducens have been renamed in the new GTDB PeerJ reviewing PDF | (2020:03:47130:1:1:NEW 9 Aug 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_4'><ns0:head /><ns0:label /><ns0:figDesc>family and one outgroup, Agarivoran spp. Nodes and tip labels are colored corresponding to the Anvi'o profile in Figure2; i.e., red: Lonepinella koalarum (Unicycler: assembly of L. koalarum strain UCD-LQP1, in bold and marked with a star), dark purple: Pasteurella spp., light purple: Haemophilus spp., orange: Actinobacillus spp., pink: Aggregatibacter spp., and green: Mannheimia spp. Black are genera that were not used in Figure2, and brown depicts the outgroup Agarivoran spp.</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_5'><ns0:head>Figure 2</ns0:head><ns0:label>2</ns0:label><ns0:figDesc>Figure 2</ns0:figDesc></ns0:figure>
<ns0:figure type='table' xml:id='tab_0'><ns0:head /><ns0:label /><ns0:figDesc>are accessible on FigShare.COG2337 mRNA-degrading endonuclease, toxin component of the MazEF toxinantitoxin module V COG0845 Multidrug efflux pump subunit AcrA (membrane-fusion protein) V COG3093 Plasmid maintenance system antidote protein VapI, contains XRE-type HTH domain V COG2828 2-Methylaconitate cis-trans-isomerase PrpF (2-methyl citrate pathway)</ns0:figDesc><ns0:table><ns0:row><ns0:cell>1</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell cols='2'>Statistic</ns0:cell><ns0:cell /><ns0:cell>Value</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell cols='2'>Completeness</ns0:cell><ns0:cell /><ns0:cell>99.205 %</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell cols='3'>Contamination COG1048 Aconitase A</ns0:cell><ns0:cell>0.705 %</ns0:cell><ns0:cell>C C</ns0:cell></ns0:row><ns0:row><ns0:cell cols='4'>Number of Contigs COG1454 Alcohol dehydrogenase, class IV 29</ns0:cell><ns0:cell>C</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell cols='3'>COG3312 FoF1-type ATP synthase assembly protein I</ns0:cell><ns0:cell>C</ns0:cell></ns0:row><ns0:row><ns0:cell cols='4'>Total Length COG0435 Glutathionyl-hydroquinone reductase 2,608,483 bp</ns0:cell><ns0:cell>C</ns0:cell></ns0:row><ns0:row><ns0:cell>GC%</ns0:cell><ns0:cell cols='3'>COG0371 Glycerol dehydrogenase or related enzyme, iron-containing ADH family 39.02 COG0778 Nitroreductase</ns0:cell><ns0:cell>C C</ns0:cell></ns0:row><ns0:row><ns0:cell>N50</ns0:cell><ns0:cell cols='3'>2,299,135 bp COG1053 Succinate dehydrogenase/fumarate reductase, flavoprotein subunit</ns0:cell><ns0:cell>C</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell cols='3'>COG4972 Tfp pilus assembly protein, ATPase PilM</ns0:cell><ns0:cell>W</ns0:cell></ns0:row><ns0:row><ns0:cell>N75</ns0:cell><ns0:cell cols='3'>2,299,135 bp COG1116 ABC-type nitrate/sulfonate/bicarbonate transport system, ATPase component</ns0:cell><ns0:cell>P</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell cols='2'>ABC-type nitrate/sulfonate/bicarbonate transport system, periplasmic</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>L50</ns0:cell><ns0:cell>COG0715</ns0:cell><ns0:cell>component</ns0:cell><ns0:cell>1</ns0:cell><ns0:cell>P</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell cols='3'>COG0600 ABC-type nitrate/sulfonate/bicarbonate transport system, permease component</ns0:cell><ns0:cell>P</ns0:cell></ns0:row><ns0:row><ns0:cell>L75</ns0:cell><ns0:cell cols='2'>COG2807 Cyanate permease</ns0:cell><ns0:cell>1</ns0:cell><ns0:cell>P</ns0:cell></ns0:row><ns0:row><ns0:cell cols='4'>Number of Predicted Genes COG2382 Enterochelin esterase or related enzyme 2,551 COG3301 Formate-dependent nitrite reductase, membrane component NrfD</ns0:cell><ns0:cell>P P</ns0:cell></ns0:row><ns0:row><ns0:cell cols='3'>Number of Protein Coding COG3230 Heme oxygenase</ns0:cell><ns0:cell>2,479</ns0:cell><ns0:cell>P</ns0:cell></ns0:row><ns0:row><ns0:cell>Genes</ns0:cell><ns0:cell cols='3'>COG0672 High-affinity Fe2+/Pb2+ permease</ns0:cell><ns0:cell>P</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell cols='2'>Iron uptake system EfeUOB, periplasmic (or lipoprotein) component</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>2</ns0:cell><ns0:cell>COG2822</ns0:cell><ns0:cell>EfeO/EfeM</ns0:cell><ns0:cell /><ns0:cell>P</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell cols='2'>NADPH-dependent ferric siderophore reductase, contains FAD-binding and</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell /><ns0:cell>COG2375</ns0:cell><ns0:cell>SIP domains</ns0:cell><ns0:cell /><ns0:cell>P</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell cols='3'>COG2223 Nitrate/nitrite transporter NarK</ns0:cell><ns0:cell>P</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell cols='3'>COG2223 Nitrate/nitrite transporter NarK</ns0:cell><ns0:cell>P</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell cols='3'>COG4771 Outer membrane receptor for ferrienterochelin and colicins</ns0:cell><ns0:cell>P</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell cols='3'>COG2837 Periplasmic deferrochelatase/peroxidase EfeB</ns0:cell><ns0:cell>P</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell cols='3'>COG0659 Sulfate permease or related transporter, MFS superfamily</ns0:cell><ns0:cell>P</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell cols='3'>COG4388 Mu-like prophage I protein</ns0:cell><ns0:cell>X</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell cols='2'>Phage repressor protein C, contains Cro/C1-type HTH and peptisase s24</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell /><ns0:cell>COG2932</ns0:cell><ns0:cell>domains</ns0:cell><ns0:cell /><ns0:cell>X</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell cols='3'>COG5412 Phage-related protein</ns0:cell><ns0:cell>X</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell cols='3'>COG1943 REP element-mobilizing transposase RayT</ns0:cell><ns0:cell>X</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell cols='3'>COG2189 Adenine specific DNA methylase Mod</ns0:cell><ns0:cell>L</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell cols='2'>ATP-dependent exoDNAse (exonuclease V) beta subunit (contains helicase</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell /><ns0:cell>COG1074</ns0:cell><ns0:cell cols='2'>and exonuclease domains)</ns0:cell><ns0:cell>L</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell cols='2'>ATP-dependent exoDNAse (exonuclease V), alpha subunit, helicase</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell /><ns0:cell>COG0507</ns0:cell><ns0:cell>superfamily I</ns0:cell><ns0:cell /><ns0:cell>L</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell cols='3'>COG3057 Negative regulator of replication initiation</ns0:cell><ns0:cell>L</ns0:cell></ns0:row><ns0:row><ns0:cell>1</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row></ns0:table></ns0:figure>
</ns0:body>
" | "Response letter - Isolation and sequence-based characterization of a koala symbiont: Lonepinella koalarum
Dear Dr. Dahlhausen and colleagues:
Thanks for submitting your manuscript to PeerJ. I have now received three independent reviews of your work, and as you will see, the reviewers raised some minor concerns about the research. Despite this, these reviewers are optimistic about your work and the potential impact it will lend to research on koala biology. Thus, I encourage you to revise your manuscript, accordingly, taking into account all of the concerns raised by the reviewers.
While the concerns of the reviewers are relatively minor, this is a major revision to ensure that the original reviewers have a chance to evaluate your responses to their concerns.
The concerns of Reviewer 3 will require the most attention. The reviewer has provided many ideas for improving the structure of your work and delivery of results.
I look forward to seeing your revision, and thanks again for submitting your work to PeerJ.
-joe
>>>> Dear Dr. Gillespie, thank you for your consideration and short turnaround time. We have responded to each comment we received. They proved very constructive and helpful.
Reviewer 1 (Raphael Eisenhofer)
Basic reporting
The English used throughout this manuscript is clear and the literature cited is relevant.
>>>> Dear Dr. Eisenhofer, thank you for taking the time to review our manuscript.
I would, however, recommend incorporating this recent and highly relevant paper into the introduction (perhaps in the paragraph of lines 62-74):
Faecal inoculations alter the gastrointestinal microbiome and allow dietary expansion in a wild specialist herbivore, the koala (Blyton et al. 2019).
>>>> This reference has been added.
Additionally, on line 80, I would replace Brice 2019 with this reference:
Gut microbes of mammalian herbivores facilitate intake of plant toxins. Kohl KD, Weiss RB, Cox J, Dale C, Dearing MD
Justification being that Kohl et al. 2014 experimentally demonstrated the point you’re making.
>>>> Thank you for noticing. We have changed those references.
In figure 1, I would recommend bolding and placing an asterisk on the UCD-LQP1 leaf for clarity.
>>>> This is done and also mentioned in the figure legend.
In figure 2, I would again recommend annotation of the UCD-LQP1 track to help quickly orient the reader. The ANI heatmap confused for a while until I realised that perhaps it is incorrectly oriented? It makes sense to me for it to be rotated 90 ° counterclockwise.
Actually, I think the issue is that the y-axis of the ANI heatmap is missing from the provided figure (was present when viewing through Anvi’o).
>>>> The ANI heatmap is aligned with the Anvi’o profile, that is why there are no labels at the y-axis. The samples are the same in the profile and in the heatmap and the two figures melt into each other. The labels are only visible if you zoom into the heatmap at high resolution (which might have been impossible with the low-resolution figure that was provided with the manuscript for review). UCD-LQP1 is now bold and marked with a star as in Fig. 1. We also clarified the figure legend.
Do the authors know the history of the koala they sampled? When was it brought to the San Francisco Zoo? Was it brought over from a Zoo (or the wild) in Australia? If so, which state/population, etc? What species of Eucalyptus is it being fed? Such information could be useful for future investigations comparing the author’s L. koalarum assembly to ones from different populations.
>>>> The SF Zoo was private about the history of their koalas. We assume that the koala in our study was born at the zoo, as the zookeeper pointed out to us that the koala’s mother lives in the same zoo. Koalas at the SF Zoo are fed blue gum leaves (Eucalyptus globulus), which grow quite abundantly in California. This information is now also added to the manuscript (lines 129-133).
Overall, the level of reporting and availability of resources in this manuscript is commendable. I love that a Jupyter notebook was uploaded containing extra background and the scripts ran. I can also confirm that the Anvi’o databases they provided are working. However, I could not find the raw sequencing data anywhere. I would request that the authors upload this both for reproducibility, and for future use with improved assembly software/tools.
>>>> We are glad that others also care about reproducibility and sharing of original scripts and data. The NCBI accession numbers for raw sequencing data, as well as genome assembly were somehow removed from the manuscript when we submitted to PeerJ by some automatic manuscript processing. Here is the part that was removed. We will make sure it will be included in the final manuscript.
Data deposition
NCBI Bioproject ID: PRJNA560698
Biosample Accession Number: SAMN12598050
Isolate Name: Lonepinella koalarum Strain UCD-LQP1
NCBI Assembly Accession Number: GCA_008723255.1
FigShare: https://figshare.com/projects/Lonepinella_koalarum_genome_assembly/74613
Scripts: https://doi.org/10.6084/m9.figshare.11678262
Experimental design
I see no issues with the methods and analyses used by the authors, and they are cited appropriately.
Validity of the findings
No comment.
Comments for the Authors
This paper is a logical extension from the authors’ previous work (Dahlhausen et al., 2018), and I think it’s great to see more work being done on the functional characterisation of host-associated microbes. Overall, I’m happy to recommend acceptance of the manuscript for publication on the provision that my comments are addressed.
>>>> Thank you.
Reviewer 2 (anonymous)
Basic reporting
The authors isolated a strain of Lonepinella koalarum and investigated genes that may be used for plant secondary metabolite degradation, carbohydrate metabolism, and antibiotic resistance. They compared the isolate genome to two other L. koalarum genomes from Genbank.
The article is well written and includes a sufficient introduction and background. Figures and tables are relevant to the article, appropriately described and labeled. Data is available from NCBI and permits for sample collection is provided. However, the NCBI accession number is missing from the manuscript.
>>>> We would also like to thank this reviewer for the time and effort and comments.
The NCBI accession numbers for raw sequencing data, as well as genome assembly were somehow removed from the manuscript when we submitted to PeerJ by some automatic manuscript processing. Here is the part that was removed. We will make sure it will be included in the final manuscript.
Data deposition
NCBI Bioproject ID: PRJNA560698
Biosample Accession Number: SAMN12598050
Isolate Name: Lonepinella koalarum Strain UCD-LQP1
NCBI Assembly Accession Number: GCA_008723255.1
FigShare: https://figshare.com/projects/Lonepinella_koalarum_genome_assembly/74613
Scripts: https://doi.org/10.6084/m9.figshare.11678262
Experimental design
The primary research is within the scope of the journal. A rigorous investigation was performed with well-designed, robust methods that could be reproduced.
Validity of the findings
All underlying data has been provided. Conclusions are well stated, linked to original research question and limited to supporting results.
Comments for the Authors
The fact that you have an isolate means that you could test for antibiotic resistance but it is out of the scope of the manuscript and could be a future project.
>>>> Yes, we agree. This would be an exciting follow-up study - experimenting with the isolated strain.
The manuscript follows on from previous work that showed that L. koalarum could be used as an indicator of koala health. How common is L. koalarum?
>>>> While occurring in low relative abundances in the koala gut, limited previous findings suggest that L. koalarum can be found in a majority of koalas (approximately 80%), although more research is needed to improve the accuracy of this statistic.
Minor edits are included below:
Line 102: brackets around Phascolarctos cinereus.
>>>> This has been added.
Line 389: Supplementary Table S2 doesn’t refer to taxonomic names.
>>>> Thank you for noticing. This should be Supplementary Table S1.
Figure 2: I find ‘private’ confusing. Maybe it would be better to refer to these gene clusters as ‘unique’.
>>>> We have changed the labeling accordingly.
Table 3 title: genome assemblies unitalicised.
>>>> Thank you for noticing. This has been changed now.
Table 4: I’m not sure what the “no approved entry” means. It maybe best to remove it.
>>>> This has been removed.
Reviewer 3 (anonymous)
Basic reporting
Clear and unambiguous English is used throughout and the literature cited is appropriate. The background and context presented in the introduction could be improved (see my point by point responses). The structure is generally appropriate although a section of the results is presented in the conclusions. The article is self-contained.
Experimental design
The research is original, performed to a high technical standard and the methods are well described. The research question is well defined, however, the knowledge gap that it fills could be better expressed. I suggest couching the genetic characterisation of lonepinella around what it tells us about its functional niche and how that related to koala nutrition and digestion. Tannin degradation may be one aspect but it is only a small part.
Validity of the findings
The overall findings are sound. I would, however, suggest a more detailed presentation of some of the outcomes. See my general comments for more details.
Comments for the Authors
The manuscript by Dahlhausen aims to genetically characterise Lonepinella koalarum, a member of the koala gut microbiome that has previously been identified as able to degrade tannin-protein complexes and may improve koala survival during antibiotic treatment. The methods and analyses are appropriate to meet these aims and their presentation is refreshingly detailed. However, I think that the major findings of the study are not well described or elaborated upon.
There is a strong focus on the detection (or the lack there of) of tannin-degrading genes. Yet such genes are poorly characterised and not well represented in annotation databases. Additionally, the benefit of tannin-protein complex degradation to koala nutrition is questionable as it is thought that the liberated protein cannot be absorbed by the koala in the hindgut and may instead only serve to benefit the microbial community. Furthermore, phenotypic studies have already shown that lonepinella is capable of degrading tannin complexes and thus genetic determination of this function provides little additional information about the species.
Instead, I feel that the main benefit of this work would be to identify lonepinella’s functional niche within the microbiome and describe how its functions could benefit koala nutrition, digestion and health. While many of the analyses need to achieve such a characterisation have already been performed, they have not been presented in sufficient detail to build an overall picture of lonepinella’s role in the microbiome. The analyses also mainly describe how lonepinella differs from closely related bacteria but what about its overall all metabolic functions? More details and discussion is needed around specific degradation pathways, what section of these pathways are present in lonepinella and the functions these enzymes and pathways have. The manuscript would also benefit from placing these functions and pathways in the context of koala gastro-physiology and lonepinella’s proposed functional niche within the microbiome. See my point-by point responses for further details.
>>>> We would also like to thank this reviewer for their time and effort and comments that significantly improved our manuscript.
Point-by Point comments:
Abstract
L21: write species name in full at first mention
>>>> This has been done.
L23: What is meant by a “common member”. L. koalarum is found at very low relative abundance in the faecal microbiomes of koalas and is not detected in the majority of koalas (potentially due to it falling below the detection threshold in most cases, see (Alfano 2013, Brice 2019 and Blyton 2019).
>>>> This part of the sentence was removed (lines 23-24).
L18-34: It would be good if the abstract included a summary of the studies major findings. What unique genes were found? Where did it fall in the phylogenetic tree? What plant secondary metabolite degrading genes were found? etc. It would also be useful to have a conclusion to indicate what these findings suggest for the role of L. koalarum in koala digestion and health.
>>>> We added major findings and conclusions to the abstract (lines 27-42). The previous abstract was vague, indeed.
Introduction
In general the introduction is very focused on PBMCs and their detoxification, however, the koala microbiome as a whole also plays an important role in macro nutrient digestion and fibre degradation (see Brice 2019 and Blyton 2019). This should also be covered in the introduction and be flagged as potential ways that lonepinella could contribute to koala digestion/health. These functions are covered in the manuscript’s analysis of the lonepinella genome and could be nicely given context in the introduction.
>>>> Macronutrient digestion and fiber degradation are now also added into the introduction at lines 78-81.
It would also be useful to indicate why the genomes were screened for genes associated with antibiotic resistance.
>>>> This has been added at lines 103-115.
L43: They can also cause negative post-digestive feedback by making the koala feel “sick” as discussed in Lawler, Foley and Eschler. Also many of these really are not toxins but rather anti-nutrient compounds. I think it is inappropriate to simply refer to them as toxins. I suggest referring to them as PBMCs.
>>>> We agree that the term ‘toxin’ is very specific and does not include many anti-nutrient compounds. We changed the term to ‘PCDs’ for plant chemical defenses throughout the manuscript. None of the authors was familiar with the term ‘PBMCs’.
L70: It is worth briefly outlining what tannin-protein complexes are and how they influence koala nutrition.
>>>> This information has been added at lines 87-90.
L72: You should give a bit more detail on this finding. Was it the abundance of L. koalarum before treatment that was predictive? The presence/absence of the bacteria? Or was it how well L. koalarum was maintained through the treatment that predicted koala survival?
In a previous study, a co-occurrence network analysis identified four bacterial taxa, including one of the genus Lonepinella, that could be found in feces of koalas that lived at both the beginning and end of their antibiotic treatment after Chlamydia infection. However, these four taxa were absent from feces of koalas that died. Furthermore, in the same study a random forest analysis revealed that the most predictive taxon of whether a koala would live or die during their antibiotic treatment was identified as Lonepinella koalarum. These findings show that koalas that died after antibiotic treatment had much lower relative abundance (sometimes even zero), of L. koalarum compared to koalas that didn’t die. We added this information to lines 90-99.
L69-71: I think it is important here to provide a bit more background about what is known about L. koalarum given it is the subject of this study. For example, Osawa showed that L. koalarum may by mucosal associated in the caecum. If doing this extends to introduction too much then some of the detail on the Eucalypt toxins could be condensed.
>>>> This information has been added at lines 85-87. It is also discussed in more detail in the discussion section (lines 530-537).
L80: “well-being” isn’t a scientific term. I suggest revising.
>>>> This has been removed.
Materials and methods
L207-226: Why was the phylogenetic tree assembled from the 16S gene sequences instead of constructing a full genome tree? What was the benefit of constructing this tree compared with the GTDB tree?
>>>> As mentioned in the text at lines 30-31 and 346-358, the GTDB reference database does not contain any L. koalarum assemblies to date. This taxonomic database was mostly used to identify the most closely related genomes for the comparative genomic analysis. A 16S gene tree was built to confirm that our isolate was indeed L. koalarum (lines 346-352 and Fig. 1).
Results and discussion
L289: It is of interest what proportion of isolates from the culturing reported in the methods were found to be L. koalarum by 16s Sanger sequencing and what (if any) other taxa were isolated. Perhaps a short summary could be given in the results?
>>>> We added this information to lines 327-332.
L290: What was the genome coverage?
>>>> 672. This number has also been added to the manuscript at line 329.
L328: How many tannin-degrading genes are present in the database used for the genome annotation? In general tannases are not well characterised. For example, only one KEGG orthalog was present last time I looked.
>>>> There were four confirmed tannase genes in RAST the last time we looked (Sunday, July 12th 2020).
L331: Was a clear zone present around isolate UCD-LQP1 on the plate? Bacteria without tannin-protein complex degrading capability are capable of growing on those plates if they are tannin tolerant. Only the presence of a clear zone indicates tannin degradation.
>>>> Yes, we did observe a clear zone around the colonies. However, we did not include all of the appropriate controls to test for tannase activity, so we do not report on them in this manuscript.
L338-340: Again how many tannases are present in the annotation databases?
>>>> There were seven confirmed tannase references in the eggNOG database (Sunday, June 7th 2020).
L346-347: These enzymes can nonetheless play important roles in degradation. The identified enzyme and function of these genes should be summarised in a supplementary table. It would also be worth interrogating these functions and describing them in the text.
>>>> We mapped these genes onto individual pathways on the KEGG website and report complete pathways in the current version of the amended manuscript (lines 394-406).
L348-349: Or because the tannases are not in the database.
>>>> This has been added.
L350-L354: It would be useful to map the identified genes onto these pathways (there are tools for this on the KEGG website) and determine if they form a continuous degradation/synthesis chain (i.e. are all enzymes necessary for conversion between two compounds present?). These are large pathways, what sections of the pathway are represented? What compounds are degraded by the genes identified in L. koalarum? This will give a much more specific indication of what functions lonepinella is performing.
>>>> We performed this additional analysis. The results are mentioned at lines 394-406 and in the new Supplementary Figures S1 to S3.
L491: oligosaccharides in Eucalyptus leaves will be absorbed by the koala in the small intestine and only a small fraction would enter the caecum and colon. As you say, this means that the bacteria in the hindgut are most likely using these pathways to process the products of the degradation of complex carbohydrates with cross-feeding among microbiome members. The benefit of such activity to koala nutrition is not well understood.
>>>> We have clarified this sentence (lines 543-555).
L492-501: This is an interesting and important finding. Was there evidence that L. koalarum is capable of fibre fermentation and the production of short-chain fatty acids?
>>>> No culturing experiments have been performed in this direction. The only evidence is the presence of a GH42 homolog and a CE4 homolog in the UCD-LQP1 assembly.
L502: reference required
>>>> This sentence has been removed.
Conclusions
L506-523: These are not conclusions and should be presented as a section on antibiotic resistance in the results section.
>>>> This section has been renamed (line 562).
L528: This sentence does not fit with the previous sentence as in the first you say beneficial members need to be identified but then explicitly talk about L. koalarum without first justifying that it is one of these beneficial members. I suggest revising.
>>>> This sentence has been revised (lines 585-600).
" | Here is a paper. Please give your review comments after reading it. |
9,826 | "<ns0:abstract xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:p>Koalas (Phascolarctos cinereus) are highly specialized herbivorous marsupials that feed almost exclusively on Eucalyptus leaves, which are known to contain varying concentrations of many different toxic chemical compounds. The literature suggests that Lonepinella koalarum, a bacterium in the Pasteurellaceae family, can break down some of these toxic chemical compounds. Furthermore, in a previous study, we identified L. koalarum as the most predictive taxon of koala survival during antibiotic treatment. Therefore, we believe that this bacterium may be important for koala health. Here, we isolated a strain of L. koalarum from a healthy koala female and sequenced its genome using a combination of short-read and long-read sequencing. We placed the genome assembly into a phylogenetic tree based on 120 genome markers using the Genome Taxonomy Database (GTDB), which currently does not include any L. koalarum assemblies.</ns0:p><ns0:p>Our genome assembly fell in the middle of a group of Haemophilus, Pasteurella and Basfia species. According to average nucleotide identity and a 16S rRNA gene tree, the closest relative of our isolate is L. koalarum strain Y17189. Then, we annotated the gene sequences and compared them to 55 closely related, publicly available genomes. Several genes that are known to be involved in carbohydrate metabolism could exclusively be found in L. koalarum relative to the other taxa in the pangenome, including glycoside hydrolase families GH2, GH31, GH32, GH43 and GH77. Among the predicted genes of L. koalarum were 79 candidates putatively involved in the degradation of plant secondary metabolites. Additionally, several genes coding for amino acid variants were found that had been shown to confer antibiotic resistance in other bacterial species against pulvomycin, beta-lactam antibiotics and the antibiotic efflux pump KpnH. In summary, this genetic characterization allows us to build hypotheses to explore the potentially beneficial role that L. koalarum might play in the koala intestinal microbiome. Characterizing and understanding beneficial symbionts at the whole genome level is important for the</ns0:p></ns0:div>
</ns0:abstract>
<ns0:body xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:head>Introduction</ns0:head><ns0:p>Koalas (Phascolarctos cinereus) are arboreal marsupials that are highly specialized herbivores in that they feed almost exclusively on the foliage of select Eucalyptus species <ns0:ref type='bibr'>(Moore & Foley, 2005;</ns0:ref><ns0:ref type='bibr' target='#b17'>Callaghan et al., 2011)</ns0:ref>. All Eucalyptus species contain chemical defenses against isolates associated with the intestinal microbial communities of koalas have been characterized in the context of degradation of PCDs found in Eucalyptus leaves <ns0:ref type='bibr' target='#b79'>(Osawa, 1990</ns0:ref><ns0:ref type='bibr' target='#b80'>(Osawa, , 1992;;</ns0:ref><ns0:ref type='bibr' target='#b81'>Osawa et al., 1993</ns0:ref><ns0:ref type='bibr' target='#b82'>Osawa et al., , 1995;;</ns0:ref><ns0:ref type='bibr' target='#b64'>Looft, Levine & Stanton, 2013)</ns0:ref>. One of these cultured isolates is a bacterium known as Lonepinella koalarum. It had been first isolated from the mucus around the caecum in koalas and was shown to degrade tannin-protein complexes that can be found in Eucalyptus leaves <ns0:ref type='bibr' target='#b82'>(Osawa et al., 1995;</ns0:ref><ns0:ref type='bibr' target='#b37'>Goel et al., 2005)</ns0:ref>. Briefly, tannin-protein complexes are extremely diverse and result from the reaction between plant defense secondary metabolites; i.e., tannins, and proteins. Tannins bind proteins followed by the formation of a precipitate, which cannot be digested by koalas or utilized by microbes <ns0:ref type='bibr' target='#b0'>(Adamczyk et al., 2017)</ns0:ref>. In our previous work, L. koalarum was identified as the most predictive taxon of koala survival during antibiotic treatment <ns0:ref type='bibr' target='#b24'>(Dahlhausen et al., 2018)</ns0:ref>. Briefly, a co-occurrence network analysis identified four bacterial taxa, including one of the genus Lonepinella, that could be found in feces of koalas that survived their antibiotic treatment after Chlamydia infection. However, these four taxa were absent from feces of koalas that died. Furthermore, in the same study a random forest analysis revealed that the most predictive taxon of whether a koala would live or die during their antibiotic treatment was identified as L. koalarum. This finding suggests that L. koalarum could be important for koala health, but the study did not present any evidence relating to PCD degradation in the highly specialized diet of koalas.</ns0:p><ns0:p>It is well understood that animals with highly specialized diets also are likely to have highly specialized intestinal microbial communities <ns0:ref type='bibr' target='#b43'>(Higgins et al., 2011;</ns0:ref><ns0:ref type='bibr' target='#b54'>Kohl et al., 2014;</ns0:ref><ns0:ref type='bibr' target='#b2'>Alfano et al., 2015;</ns0:ref><ns0:ref type='bibr' target='#b56'>Kohl, Stengel & Denise Dearing, 2016)</ns0:ref>. Disturbances of a specialized microbial community, such as the introduction of antibiotics, can have profound effects on the host's health <ns0:ref type='bibr'>(Kohl & Denise Dearing, 2016;</ns0:ref><ns0:ref type='bibr' target='#b11'>Brice et al., 2019)</ns0:ref>. Yet, koalas are regularly treated with antibiotics due to the high prevalence of Chlamydia infections in many populations <ns0:ref type='bibr' target='#b87'>(Polkinghorne, Hanger & Timms, 2013)</ns0:ref>. While recent advances in Chlamydia pecorum vaccines for koalas are a promising alternative for managing koala populations, antibiotics are still the current treatment method for bacterial infections in koalas <ns0:ref type='bibr' target='#b106'>(Waugh et al., 2016;</ns0:ref><ns0:ref type='bibr' target='#b27'>Desclozeaux et al., 2017;</ns0:ref><ns0:ref type='bibr' target='#b78'>Nyari et al., 2018)</ns0:ref>. The antibiotics used in practice might not only target Chlamydia pecorum but also beneficial koala gut symbionts as a side effect. Therefore, it is important to learn about bacteria associated with koala health, such as L. koalarum, in order to further the development of alternative treatments for bacterial infections in koalas and to recommend antibiotic compounds that are potentially less disruptive to members of the koala gut microbiome.</ns0:p><ns0:p>Here we isolated a strain of L. koalarum (hereafter called strain UCD-LQP1) from the feces of a healthy koala (P. cinereus) female at the San Francisco Zoo. We sequenced the genome of L. koalarum UCD-LQP1 using a combination of long-and short-read sequencing, and then assembled and annotated the genome. We compared the genome assembly to the most closely related genomes that are currently publicly available. The genome assembly of L. koalarum UCD-LQP1 was placed in a phylogenetic tree and screened for genes putatively involved in the degradation of plant secondary metabolites, carbohydrate metabolism, and antibiotic resistance.</ns0:p><ns0:p>Additionally, we identified and characterized putative genes that were unique to this strain and two recently sequenced genomes of L. koalarum from Australia.</ns0:p></ns0:div>
<ns0:div><ns0:head>Materials and methods</ns0:head></ns0:div>
<ns0:div><ns0:head>Sampling of koala feces and preparation of culturing media</ns0:head><ns0:p>A koala fecal pellet was collected, with permission from the San Francisco Zoo, from a healthy, adult, captive, female koala (Phascolarctos cinereus). We do not have any information on the geographical origin of this koala. Koalas at the SF Zoo are fed blue gum leaves (Eucalyptus globulus), which grow quite abundantly in California. Jim Nappi and Graham Crawford of the San Francisco Zoo organized and permitted koala fecal sample collection. The fresh fecal pellet was collected from the floor with sterilized tweezers and stored in a sterile 15 ml Falcon tube (Thermo Fisher Scientific, USA). The tube was immediately placed on ice after collection and subsequently stored at 4° C overnight.</ns0:p><ns0:p>The preparation of the Lonepinella koalarum culturing media was modified from methods developed by <ns0:ref type='bibr' target='#b82'>Osawa et al. (1995)</ns0:ref>. A 2 % agarose (Fisher BioReagents, USA) solution of Bacto TM Brain Heart Infusion (BHI; BD Biosciences, USA) was prepared following manufacturer protocols. After the media had solidified in petri dishes, a 2 % tannic acid solution was prepared by combining 1 g of tannic acid powder per 50 ml of sterile Nanopure TM water (Spectrum Chemical MFG CORP, USA). The solution was vortexed for 1 min until Manuscript to be reviewed homogenized, resulting in a brown, transparent liquid. Using a sterile serological pipette, 5 ml of the 2 % tannic acid solution was gently added to each BHI media plate and left for 20 min. After incubation, the remaining liquid on the plate was decanted. No antibiotic compounds were added to the medium.</ns0:p></ns0:div>
<ns0:div><ns0:head>Culturing of isolates and DNA extraction</ns0:head><ns0:p>The koala fecal pellet was cut in half with sterile tweezers. Tweezers were re-sterilized and used to move approximately 300 mg of material from the center of the pellet to a sterile 2 ml Eppendorf tube containing 1 ml of sterile, Nanopure TM water. The tube was vortexed for 3 min, intermittently checking until the solution was homogenized into a slurry. One hundred µl of the homogenized fecal slurry was micro-pipetted onto to a BHI+tannin plate and stored in an anaerobic chamber (BD GasPak TM EZ anaerobe chamber system; BD Biosciences, USA) at 37° C for 3 days. Each individual colony that grew was plated onto a freshly made BHI+tannin plate using standard dilution streaking techniques. The new plates were stored in an anaerobic chamber at 37 °C for another 3 days. This step was repeated two more times to decrease the probability of contamination or co-culture.</ns0:p><ns0:p>An individual colony from each of the plates from the third round of dilution streaking was moved to a sterile 30 ml glass culture tube containing 5 ml of sterile Bacto TM BHI liquid media (prepared following manufacturer protocol; BD Biosciences, USA). Each tube was then capped with a sterile rubber stopper and purged with nitrogen gas in order to create an anaerobic environment. The tubes were placed in an incubated orbital shaker (ThermoFisher Scientific MaxQ TM 4450) for 3 days at 37 °C at 250 rpm.</ns0:p><ns0:p>Using a sterile serological pipette, 1.8 ml of each liquid culture was transferred to a sterile 2 ml Eppendorf tube. The tubes were spun at 13,000 g for 2 min and the supernatant was carefully decanted. The DNA was extracted from the pellet in each sample with the Promega Wizard Genomic DNA Purification Kit (Promega, USA) according to the manufacturer's protocol. DNA was eluted in a final volume of 100 μl and stored at 4 °C. PCR reactions were prepared using the bacteria-specific 'universal' primer pair 27F (5′-AGAGTTTGATCMTGGCTCAG-3′; <ns0:ref type='bibr'>Stackebrandt &</ns0:ref><ns0:ref type='bibr' target='#b101'>Goodfellow, 1991) and</ns0:ref><ns0:ref type='bibr'>1391R (5'-GACGGGCGGTGTGTRCA-3';</ns0:ref><ns0:ref type='bibr' target='#b104'>Turner et al., 1999)</ns0:ref>. PCR amplifications were performed in a BioRad T100 TM Thermal Cycler in 50 μl reactions. Each reaction contained 2 μl of the eluted DNA from the aforementioned extraction, 5 μl of 10x Taq buffer (Qiagen, USA), 10 μl of Q buffer (Qiagen), 1.25 μl of 10mM dNTPs (Qiagen), 2.5 μl of 10mM 27F primer, 2.5 μl of 10mM 1391R primer, 0.3 μl of Taq polymerase (Qiagen), and 26.45 μl of sterile water. The cycling conditions were: (1) 95 °C for 3 min, (2) 40 cycles of 15 sec at 95 °C, 30 sec at 54 °C, and 1 min at 72 °C, (3) a final incubation at 72 °C for 5 min, and (4) holding at 12 °C upon completion.</ns0:p><ns0:p>The PCR product for each sample was purified and concentrated by following the manufacturer's protocol for the NucleoSpin Gel and PCR Clean-up kit (Macherey-Nagel, USA).</ns0:p><ns0:p>The purified PCR product for each sample was quantified by following the manufacturer's protocol for the Qubit dsDNA HS Assay Kit (Thermo Fisher Scientific, USA). The PCR product for each sample was then diluted to 26 ng/μl and submitted for forward and reverse Sanger sequencing at the University of California Davis DNA Sequencing Facility. The program SeqTrace version 0.9.0 <ns0:ref type='bibr' target='#b102'>(Stucky, 2012)</ns0:ref> was used to edit and create consensus sequences of the reads received from the sequencing facility, following the protocol detailed in <ns0:ref type='bibr' target='#b30'>Dunitz et al. (2015)</ns0:ref>. The consensus sequence for each sample was uploaded to the NCBI blast website for organism identification <ns0:ref type='bibr' target='#b65'>(Madden, 2003)</ns0:ref>. The DNA of one of the isolates that had been identified as L. koalarum was used for whole-genome sequencing, as described below. We refer to this isolate as L. koalarum strain UCD-LQP1.</ns0:p></ns0:div>
<ns0:div><ns0:head>Whole genome sequencing and assembly</ns0:head><ns0:p>DNA from one sample identified as L. koalarum strain UCD-LQP1 was submitted for whole genome PacBio sequencing at SNPsaurus. After sequencing, the demultiplexed bam file was tested for reads that contained palindromic sequences since a preliminary assembly with Canu version 1.8 <ns0:ref type='bibr' target='#b57'>(Koren et al., 2017)</ns0:ref> indicated the presence of adapter sequences. Palindromic reads were split in half, aligned with minimap2 (an executable in Canu), and those palindromic reads Manuscript to be reviewed that aligned over at least two-thirds of the split read were reduced to the first part of the palindrome <ns0:ref type='bibr' target='#b57'>(Koren et al., 2017)</ns0:ref>. This procedure efficiently removed adapter sequences. These adapter-free reads were used in the hybrid assembly described below.</ns0:p><ns0:p>The same DNA that had been used for PacBio sequencing was also submitted for Illumina sequencing. Ten ng of genomic DNA were used in a 1:10 reaction of the Nextera DNA Flex Library preparation protocol (Illumina, USA). Fragmented DNA was amplified with Phusion DNA polymerase (New England Biolabs) in 12 PCR cycles with 1 min extension time. Samples were sequenced on a HiSeq4000 instrument (University of Oregon GC3F) with paired-end 150 bp reads. The 10,309,488 raw reads were quality controlled and filtered for adaptors and PhiX using the Joint Genome Institute's BBDuk tool version 37.68 <ns0:ref type='bibr' target='#b15'>(Bushnell, 2014)</ns0:ref>, resulting in 10,302,312 reads. The 308 cleaned PacBio reads and 10,302,312 filtered Illumina reads were combined with all default parameters of Unicycler version 0.4.5, a tool used to assemble bacterial genomes from both long and short reads <ns0:ref type='bibr' target='#b107'>(Wick et al., 2017)</ns0:ref>.</ns0:p></ns0:div>
<ns0:div><ns0:head>Genome annotation</ns0:head><ns0:p>Completeness and contamination of the L. koalarum strain UCD-LQP1 assembly were determined with CheckM version 1.0.8 <ns0:ref type='bibr' target='#b85'>(Parks et al., 2015)</ns0:ref>, number of contigs, total length, GC%, N50, N75, L50, and L75 were determined with QUAST (Quality Assessment Tool for Genome Assemblies; <ns0:ref type='bibr' target='#b40'>Gurevich et al., 2013)</ns0:ref>, and the assembly was annotated with PROKKA version 1.12 <ns0:ref type='bibr' target='#b96'>(Seemann, 2014)</ns0:ref>. The L. koalarum strain UCD-LQP1 genome assembly was uploaded to the Rapid Annotation using Subsystem Technology online tool (RAST), a genome annotation program for bacterial and archaeal genomes <ns0:ref type='bibr'>(Aziz et al., 2008)</ns0:ref>. The SEED viewer in RAST was used to browse features of the genome <ns0:ref type='bibr' target='#b83'>(Overbeek et al., 2014)</ns0:ref>. To screen the L. koalarum strain UCD-LQP1 assembly for genes putatively involved in tannin degradation and xenobiotic metabolisms; i.e., the degradation of plant secondary metabolites, coding regions in the assembly were identified using Prodigal version 2.6.3 <ns0:ref type='bibr' target='#b45'>(Hyatt et al., 2010)</ns0:ref>. Each identified coding region was annotated using eggNOG (a database of orthologous groups and functional annotation that is updated more regularly than PROKKA) mapper version 4.5.1 <ns0:ref type='bibr' target='#b46'>(Jensen et al., 2008)</ns0:ref>. Then, KEGG (Kyoto Encyclopedia of Genes and Genomes) pathways putatively involved in xenobiotics biodegradation and metabolism, were extracted from the eggNOG annotations PeerJ reviewing <ns0:ref type='table'>PDF | (2020:03:47130:2:0:NEW 11 Sep 2020)</ns0:ref> Manuscript to be reviewed <ns0:ref type='bibr'>ko00362, ko00627, ko00364, ko00625, ko00361, ko00623, ko00622, ko00633, ko00642, ko00643, ko00791, ko00930, ko00363, ko00621, ko00626, ko00624, ko00365, ko00984, ko00980, ko00982, and ko00983 (Kanehisa & Goto, 2000)</ns0:ref>), and the corresponding nucleotide sequences from the L. koalarum genome assemblies were saved. Individual genes with hits in KEGG pathways were manually mapped onto KEGG reference maps using the KEGG webtool <ns0:ref type='bibr' target='#b52'>(Kanehisa & Goto, 2000)</ns0:ref>.</ns0:p></ns0:div>
<ns0:div><ns0:head>16S rRNA gene based phylogenetic placement of genome</ns0:head><ns0:p>The 16S rRNA gene sequence within the genome assembly was extracted from RAST by searching for 'ssu rRNA' in the function search of the SEED genome browser <ns0:ref type='bibr'>(Aziz et al., 2008;</ns0:ref><ns0:ref type='bibr' target='#b83'>Overbeek et al., 2014)</ns0:ref>. Following the protocol outlined in <ns0:ref type='bibr' target='#b30'>Dunitz et al. (2015)</ns0:ref>, the 16S rRNA gene sequence was uploaded to the Ribosomal Database Project (RDP; <ns0:ref type='bibr' target='#b22'>Cole et al., 2014)</ns0:ref> and grouped with all 16S rRNA gene sequences in the Pasteurellaceae family and one chosen outgroup, Agarivoran spp., to root the tree. The taxon names from the RDP output file were manually cleaned up and their 16S rRNA gene sequences were used to build a phylogenetic tree with the program FastTree <ns0:ref type='bibr' target='#b89'>(Price, Dehal & Arkin, 2009)</ns0:ref>. Nodes and tip labels were manually edited for Figure <ns0:ref type='figure'>1</ns0:ref> in iTOL (interactive tree of life; web tool; <ns0:ref type='bibr' target='#b61'>Letunic & Bork, 2019)</ns0:ref>. The 16S rRNA gene sequence alignment <ns0:ref type='bibr' target='#b117'>(Wilkins & Coil, 2020a)</ns0:ref> and its resulting phylogenetic tree are available on Figshare <ns0:ref type='bibr' target='#b118'>(Wilkins & Coil, 2020b)</ns0:ref>. During the preparation of this manuscript, two more L. koalarum type strains had their genomes sequenced: one by the DOE Joint Genome Institute, USA (GenBank accession number GCA_004339625.1; 2,486,773 bp long) and one by the Maclean Lab in Australia (GenBank accession number GCA_004565475.1; 2,509,358 bp).</ns0:p><ns0:p>Both assemblies were based on type strains originating from the same isolation of L. koalarum in 1995 <ns0:ref type='bibr' target='#b82'>(Osawa et al., 1995)</ns0:ref>, DSM 10053 and ATCC 700131, respectively. These two L. koalarum genome assemblies were henceforth included in our analysis. When we refer to all three L. koalarum genome assemblies, we simply say 'in L. koalarum' and when we refer to the strain sequenced in this study, we use 'the assembly of L. koalarum strain UCD-LQP1'. <ns0:ref type='table'>PDF | (2020:03:47130:2:0:NEW 11 Sep 2020)</ns0:ref> Manuscript to be reviewed</ns0:p></ns0:div>
<ns0:div><ns0:head>PeerJ reviewing</ns0:head></ns0:div>
<ns0:div><ns0:head>Comparative genomics</ns0:head><ns0:p>The GTDB-Tk software toolkit version 0.3.0 <ns0:ref type='bibr'>(Chaumeil, Hugenholtz & Parks, 2018)</ns0:ref> of the Genome Taxonomy Database (GTDB) project was chosen to place L. koalarum into a pregenerated conserved marker gene tree using 120 marker genes <ns0:ref type='bibr' target='#b84'>(Parks et al., 2018)</ns0:ref>. After placing the assembly into the GTDB tree, a clade in the tree was extracted that contained L. koalarum strain UCD-LQP1 and 55 other taxa, of which all members belonged to the order Pasteurellales.</ns0:p><ns0:p>This clade contained all sequenced genomes of the closest neighboring taxa (n = 55) to L. koalarum in the GTDB tree at the time of this analysis (3 rd of August 2019). All of these 55 genomes were downloaded from GenBank (using the accession numbers in the GTDB) to perform a comparative genomic analysis in Anvi'o version 5.5 <ns0:ref type='bibr' target='#b33'>(Eren et al., 2015)</ns0:ref>. The two other L. koalarum genomes from GenBank were included in the following analysis as well. Accession numbers of all genome assemblies included can be found in Supplementary Table <ns0:ref type='table'>S1</ns0:ref> (n = 58).</ns0:p><ns0:p>The Anvi'o workflow for microbial pangenomics was followed <ns0:ref type='bibr' target='#b26'>(Delmont & Eren, 2018)</ns0:ref>. The blastp program from NCBI was used for a gene search <ns0:ref type='bibr' target='#b4'>(Altschul et al., 1990)</ns0:ref>, the Markov Cluster algorithm (MCL) version 14.137 (van Dongen & Abreu-Goodger, 2012) was used for clustering, and the program MUSCLE was used for alignment <ns0:ref type='bibr' target='#b32'>(Edgar, 2004</ns0:ref>). An inflation parameter of 6 was chosen to identify clusters in amino acid sequences. Genomes in the pangenome of Anvi'o were ordered based on a genomic marker gene tree. This tree was built in PhyloSift version 1.0.1 <ns0:ref type='bibr' target='#b25'>(Darling et al., 2014)</ns0:ref> with its updated markers database (version 4, posted on 12th of February 2018; <ns0:ref type='bibr' target='#b51'>Jospin, 2018)</ns0:ref> for the alignment. We used RAxML version 8.2.10 on the CIPRES web server for the tree inference <ns0:ref type='bibr' target='#b71'>(Miller, Pfeiffer & Schwartz, 2010)</ns0:ref> following the analysis in <ns0:ref type='bibr' target='#b119'>(Wilkins et al., 2019)</ns0:ref>. Gene clusters in Anvi'o were ordered based on presence/absence. We also used Anvi'o to compute average nucleotide identities across the genomes with PyANI <ns0:ref type='bibr' target='#b90'>(Pritchard et al., 2016)</ns0:ref>. In the heatmap, ANI values > 95% (and >70% for a separate figure, respectively) were colored in red.</ns0:p><ns0:p>Gene clusters from the Anvi'o microbial pangenomics analysis that could only be found in the three L. koalarum genome assemblies were extracted. Then, we also extracted all gene clusters that could only be found in the assembly of L. koalarum strain UCD-LQP1. Partial sequences were removed. A literature search of the remaining genes was conducted to identify possible roles L. koalarum might play in the gut microbiome of koalas. Tables were summarized in R version 3.4.0 (R Development Core Team, 2013).</ns0:p></ns0:div>
<ns0:div><ns0:head>Carbohydrate metabolism</ns0:head><ns0:p>Since the majority of gene clusters unique to L. koalarum genome assemblies fell into the COG (Clusters of Orthologous Groups) category 'Carbohydrate metabolism', we decided to screen all three assemblies against the Carbohydrate-Active Enzymes Database (CAZy), an expert resource for glycogenomics <ns0:ref type='bibr' target='#b18'>(Cantarel et al., 2009;</ns0:ref><ns0:ref type='bibr' target='#b63'>Lombard et al., 2014)</ns0:ref>. In brief, CAZy domains were identified based on CAZy family HMMs (Hidden Markov Models) with a coverage of >95% and an e-value < 1e-15. Searches were done through dbCAN, a web resource for automated carbohydrate-active enzyme annotation <ns0:ref type='bibr' target='#b122'>(Yin et al., 2012)</ns0:ref> and CAZy hits were only retained if they had been found with all three search tools. The three search tools included (i) HMMER version 3.3 <ns0:ref type='bibr' target='#b31'>(Eddy, 1998)</ns0:ref>, (ii) DIAMOND version 0.9.29 for fast blast hits in the CAZy database <ns0:ref type='bibr' target='#b14'>(Buchfink, Xie & Huson, 2015;</ns0:ref><ns0:ref type='bibr'>default parameters;</ns0:ref><ns0:ref type='bibr'>i.e</ns0:ref>., e-value < 1e-102, hits per query (-k) = 1), and (iii) Hotpep version 1 for short, conserved motifs in the PPR (Peptide Pattern Recognition) library <ns0:ref type='bibr' target='#b16'>(Busk et al., 2017;</ns0:ref><ns0:ref type='bibr'>default parameters;</ns0:ref><ns0:ref type='bibr'>i.e., frequency > 2.6, hits > 6)</ns0:ref>. For a detailed walk-through of the assembly, annotation, search for KEGG pathways, and comparative genomics analyses, please refer to the associated Jupyter notebook <ns0:ref type='bibr' target='#b109'>(Wilkins, 2020a)</ns0:ref>.</ns0:p></ns0:div>
<ns0:div><ns0:head>Identification of antibiotic resistance genes</ns0:head><ns0:p>All three L. koalarum genome assemblies were uploaded to the Comprehensive Antibiotic Resistance Database (CARD version 3.0.7; <ns0:ref type='bibr' target='#b47'>Jia et al., 2017)</ns0:ref> and the ResFinder database version 5.1.0 <ns0:ref type='bibr' target='#b123'>(Zankari et al., 2012)</ns0:ref> to screen them for putative antibiotic resistance genes and their variants using blastn searches against CARD 2020 reference sequences using default parameters.</ns0:p><ns0:p>The Resistance Gene Identifier (RGI) search pipeline was used to detect SNPs (single nucleotide polymorphisms) using the 'perfect, strict, complete genes only' criterion on their website. Manuscript to be reviewed that had been previously associated with antibiotic resistance in other bacterial species <ns0:ref type='bibr' target='#b1'>(Alcock et al., 2020)</ns0:ref>.</ns0:p></ns0:div>
<ns0:div><ns0:head>Results and discussion</ns0:head></ns0:div>
<ns0:div><ns0:head>Identification of isolates</ns0:head><ns0:p>Besides the isolates identified as L. koalarum, we had several other colonies growing on the BHI+tannin plates, including isolates with 16S rRNA gene sequences that matched Bacillus cereus, Bacillus nealsonii, Bacillus sonorensis, and Escherichia coli. E. coli was the most common species isolated.</ns0:p></ns0:div>
<ns0:div><ns0:head>Assembly taxonomy and gene annotation</ns0:head><ns0:p>The hybrid assembly generated was 2,608,483 bp in length with an N50 of 2,299,135 bp and a coverage of 672. According to the marker gene analysis in CheckM, the assembly was 99.21% complete and less than 1% contaminated with a GC content of 39.02% (see Table <ns0:ref type='table'>1</ns0:ref> for additional details). One contig in the assembly appears to be a 3,899 bp long plasmid. This is indicated by circularity of that contig and positive matches to plasmids in related taxa when uploaded to the NCBI blast website for organism identification <ns0:ref type='bibr' target='#b65'>(Madden, 2003)</ns0:ref>. The two most similar sequences on GenBank were a 71 percent similar sequence of Pasteurella multocida strain U-B411 plasmid pCCK411 (accession number FR798946.1) and a 70 percent similar sequence of Mannheimia haemolytica strain 48 plasmid pKKM48 (accession number MH316128.1). The putative plasmid sequence was deposited on FigShare <ns0:ref type='bibr' target='#b121'>(Wilkins & Jospin, 2020)</ns0:ref>.</ns0:p><ns0:p>The taxonomy of L. koalarum strain UCD-LQP1 was confirmed in three ways. First, a phylogenetic tree was built based on the 16S rRNA gene extracted from the new assembly. This 16S rRNA gene sequence was aligned with other closely related 16S rRNA gene sequences on the RDP website where 16S rRNA gene sequences of type strains are curated and sequences of the closest relatives of a taxon are usually readily available <ns0:ref type='bibr' target='#b30'>(Dunitz et al., 2015)</ns0:ref>. The phylogenetically closest sequence to L. koalarum strain UCD-LQP1 in the 16S rRNA gene tree was one from Lonepinella koalarum Y17189 (Fig. <ns0:ref type='figure'>1</ns0:ref>). Second, a whole genome concatenated gene marker tree was inferred using the Genome Taxonomy Database (GTDB), as well as using Manuscript to be reviewed PhyloSift, in parallel. In the GTDB tree, L. koalarum UCD-LQP1 was placed closest to Actinobacillus succinogenes (GenBank accession number GCA_000017245.1). Note that as of February 3, 2020, GTDB did not include any of the L. koalarum genome assemblies. In the PhyloSift marker gene tree, all three L. koalarum assemblies clustered together, and A. succinogenes was their phylogenetically closest neighbor (Fig. <ns0:ref type='figure' target='#fig_10'>2</ns0:ref>). Third, the average nucleotide identity (ANI) between the genome of the L. koalarum type strain (DSM 10053; GenBank accession number GCA_004339625.1) and the assembly of L. koalarum UCD-LQP1 was estimated at 98.91 percent (standard deviation 0.17%). The ANI value between L. koalarum UCD-LQP1 and GCA_004565475.1 was 98.99 percent (SD 0.15%) and the ANI value between GCA_004339625.1 and GCA_004565475.1 was 99.99 percent (SD 0.08%). Both of these genome assemblies are based on the type strain of L. koalarum that originated in 1995 <ns0:ref type='bibr' target='#b82'>(Osawa et al., 1995)</ns0:ref>. All three approaches confirmed the taxonomy of strain UCD-LQP1 as Lonepinella koalarum. Interestingly, A. succinogenes (GenBank accession number GCA_000017245.1) belongs now to a different taxonomic group based on GTDB taxonomy, namely Basfia succinogenes. <ns0:ref type='bibr' target='#b84'>Parks et al. (2018)</ns0:ref>, among others (e.g., <ns0:ref type='bibr' target='#b44'>Hug et al., 2016;</ns0:ref><ns0:ref type='bibr' target='#b19'>Castelle & Banfield, 2018)</ns0:ref>, have suggested relying on whole genome sequencing to reorganize the microbial tree of life, which will result in a majority of changes in classification and naming, and ultimately reflect a more accurate evolutionary relationship among groups <ns0:ref type='bibr' target='#b84'>(Parks et al., 2018)</ns0:ref>.</ns0:p><ns0:p>There were no positive hits for any annotations associated with tannin degradation in the RAST SEED viewer. This negative result is in contrast to the experimentally verified tannin-degrading functions reported for this bacterium <ns0:ref type='bibr' target='#b82'>(Osawa et al., 1995)</ns0:ref>. Moreover, tannic acid powder had been used to prepare the culturing medium and was expected to help select for bacterial tannin degraders. There are several potential explanations for the absence of any positive hits for tannins in the RAST database including (1) the genes responsible for tannin degradation in the assembly of L. koalarum UCD-LQP1 are not labeled as such, or (2) L. koalarum does not have any tannin-degradation functionality. We thus carried out additional sequence-based analyses searching for possible PCD degrading genes in the new assembly.</ns0:p><ns0:p>According to the annotation with PROKKA, there were 2,551 predicted genes and 2,479 protein coding genes. In comparison, eggNOG predicted 2,370 protein coding genes. Neither annotation Manuscript to be reviewed included any genes annotated as 'tannase'. However, among the eggNOG predictions, there were 79 genes putatively involved in Class 1.11 Xenobiotics biodegradation and the degradation of plant secondary metabolites (Table <ns0:ref type='table'>2</ns0:ref>). There are 20 KEGG pathways included in this group.</ns0:p><ns0:p>We searched for all twenty pathways in the assembly of L. koalarum strain UCD-LQP1 and found positive hits in 13 pathways (Table <ns0:ref type='table'>2</ns0:ref>). Each hit represents a translated amino acid sequence from the assembly of L. koalarum UCD-LQP1 that is encoded by an individual gene in a pathway. The largest proportion of hits (n=15) comprised putative enzymes that are members in this KEGG class, but do not fall into a particular pathway (KEGG pathway ko00983: Drug metabolism -other enzymes). Potential tannin-degrading genes might be found in this group but have not been labeled as tannase genes because their sequences are not similar enough to any known tannase genes or because these tannase genes are not annotated in any database. The second largest KEGG pathway was ko00362 benzoate degradation, followed by pathway ko00980 metabolism of xenobiotics by Cytochrome P450 and pathway ko00625 chloroalkane and chloroalkene degradation. KEGG pathways with fewer hits included the degradation compounds such as aminobenzoate, xylene, naphthalene, dioxin, and chlorocyclohexane.</ns0:p><ns0:p>Mapping individual genes onto KEGG pathways revealed continuous degradation chains for the following compounds: Azathioprine (pro-drug) to 6-Thioguanine (Supplementary Fig. <ns0:ref type='figure'>S1);</ns0:ref> Aminobenzoate degradation; i.e., 4-Carboxy-2-hydroxymuconate semialdehyde to Pyruvate and Oxaloacetate, which can then be fed into the citric acid cycle (Supplementary Fig. <ns0:ref type='figure' target='#fig_10'>S2</ns0:ref>); 2-Aminobenzene-sulfonate to Pyruvate, which, again, can be fed directly into Glycolysis or with another enzyme that was present (1.2.1.10) can be converted into Acetaldehyde, then Acetyl-CoA , and then fed into the Cytrate cycle (Supplementary Fig. <ns0:ref type='figure' target='#fig_10'>S2</ns0:ref>). In the group of xenobiotics metabolized by cytochrome P450 there were seven complete chains (Supplementary Fig. <ns0:ref type='figure'>S3</ns0:ref>): degradation of (i) benzo(a)pyrene, (ii) Aflatoxin B1, (iii) 1-Nitronaphtalene, (iv) 1,1-Dichloroethylene, (v) Trichloroethylene, (vi) Bromobenzene, and (vii) 1,2-Dibromoethane. All of these complete, putative conversion chains present in L. koalarum might explain further how this member of the koala gut microbiome contributes to koala gastro-physiology (see discussion below). Amino acid sequences encoded by putative PCD degrading genes in L. koalarum strain UCD-LQP1 can be downloaded from FigShare <ns0:ref type='bibr' target='#b110'>(Wilkins, 2020b)</ns0:ref>. A table linking eggNOG annotations to positions in individual assemblies and translated amino acid sequences can be found in Supplementary Table <ns0:ref type='table'>S2</ns0:ref>. A complete table of all eggNOG annotations in the assembly of L. koalarum strain UCD-LQP1 can be found in Supplementary Table <ns0:ref type='table'>S3</ns0:ref>.</ns0:p><ns0:p>Eucalyptus spp. leaves contain more than 100 different chemical compounds including phenolics, terpenoids and lipids that are harmful for herbivores, even at low concentration <ns0:ref type='bibr' target='#b66'>(Maghsoodlou et al., 2015)</ns0:ref>. Koalas are highly specialized folivores feeding on these leaves. We assumed that L. koalarum plays a beneficial role for koala hosts because some strains have shown experimentally to be able to degrade tannins <ns0:ref type='bibr' target='#b79'>(Osawa, 1990;</ns0:ref><ns0:ref type='bibr' target='#b82'>Osawa et al., 1995)</ns0:ref>, and tannic acid was used to isolate L. koalarum strain UCD-LQP1. Alas, we did not find any direct evidence for tannase genes in the assembly of L. koalarum UCD-LQP1. However, genes encoding several putative pathways involved in plant secondary metabolite degradation were found in the assembly of L. koalarum UCD-LQP1. The predicted pathways included those for degradation of compounds that had been extracted from Eucalyptus leaves (e.g., benzoate, aminobenzoate, and chlorocyclohexane; <ns0:ref type='bibr' target='#b92'>Quinlivan et al., 2003;</ns0:ref><ns0:ref type='bibr' target='#b70'>Marzoug et al., 2011;</ns0:ref><ns0:ref type='bibr' target='#b95'>Sebei et al., 2015;</ns0:ref><ns0:ref type='bibr' target='#b66'>Maghsoodlou et al., 2015;</ns0:ref><ns0:ref type='bibr' target='#b97'>Shiffman et al., 2017)</ns0:ref>. Degradation of these PCDs might explain the beneficial role that L. koalarum plays in the koala gut microbiome.</ns0:p></ns0:div>
<ns0:div><ns0:head>Comparative genomics and unique genes in L. koalarum</ns0:head><ns0:p>The GTDB tree clade used to extract related genomes of L. koalarum strain UCD-LQP1 consisted mostly of Haemophilus spp. (n = 28), followed by Rodentibacter spp. (n = 13), Pasteurella spp. (n = 5), Aggregibacter spp. (n = 4), and seven other genera (Table <ns0:ref type='table'>S1</ns0:ref>). Whole genome marker phylogenetic trees showed that not all genera were monophyletic. This can be seen in Figure <ns0:ref type='figure' target='#fig_10'>2</ns0:ref> in the way the coloring based on genus name does not group perfectly when taxa are ordered according to their phylogenetic relationship. This was especially the case for Haemophilus spp., which is shown in light purple. Some of the Haemophilus genomes were grouped together, whereas others grouped with genomes labeled as Pasteurella spp., Necropsobacter spp. and Avibacterium spp. One species of Rodentibacter (R. Manuscript to be reviewed taxonomy to Basfia succinogenes, most probably the most closely related taxon to L. koalarum that has its genome sequenced to date. Twelve out of the 55 NCBI microbial genome assemblies have different taxonomic names in the new GTDB taxonomy (Supplementary Table <ns0:ref type='table'>S1</ns0:ref>). For a discussion of the re-organization and re-naming of the microbial tree of life based on whole genome sequencing see above Assembly taxonomy and gene annotation. The whole genome marker gene tree used to order genomes in Anvio's visualization can be downloaded from FigShare <ns0:ref type='bibr' target='#b111'>(Wilkins, 2020c)</ns0:ref>, as well as its corresponding amino acid alignment <ns0:ref type='bibr' target='#b112'>(Wilkins, 2020d)</ns0:ref>.</ns0:p><ns0:p>Average nucleotide identities have been put forward as a measure of genomic relatedness among bacteria that could help designate genera and be used besides the 16S rRNA gene as a taxonomic marker <ns0:ref type='bibr' target='#b7'>(Barco et al., 2020)</ns0:ref>. Moreover, it has been suggested to use an ANI threshold of larger than 95% to delineate bacterial species <ns0:ref type='bibr' target='#b39'>(Goris et al., 2007)</ns0:ref>. Based on this definition, the genomes used for the comparative genomic analysis with L. koalarum are all distinct species (heatmap in Fig. <ns0:ref type='figure' target='#fig_10'>2</ns0:ref>). We created a second heatmap visualizing genomic relatedness at the 70 % level (Supplementary Fig. <ns0:ref type='figure'>S4</ns0:ref>). This heatmap revealed several distinct clusters of closely related genomes vs. singleton genomes (i.e., taxa that did not group together with anything else at the 70 percent threshold): Cluster 1) Aggregatibacter spp., 2) first main Haemophilus spp. group, 3) Rodentibacter spp., 4) second main Haemophilus spp. group, 5) L. koalarum genome assemblies, and 6) two Necropsobacter spp. and another Haemophilus spp. Notably, Rodentibacter heylii, all Pasteurella spp., and Avibacterium paragallinarum did not cluster with anything. The heatmap is a way of visualizing sequence similarity groups and overall, it showed that the genera Haemophilus, Pasteurella and Rodentibacter do not represent coherent groups of species or genera. These three genera were found in several sub-groups (clusters in the ANI heatmap in Supplementary Fig. <ns0:ref type='figure'>S4</ns0:ref>) that have been described previously based on a much larger sample size and a few marker genes <ns0:ref type='bibr' target='#b74'>(Naushad et al., 2015)</ns0:ref>. Even some of the same singleton genomes were reported as their own branches in previous phylogenetic trees <ns0:ref type='bibr' target='#b21'>(Christensen et al., 2003)</ns0:ref>. L. koalarum was placed in the middle of a group containing mostly Haemophilus, Pasteurella and Basfia species. Pasteurellaceae, the single constituent family of the order Pasteurellales hosts a diverse group of mostly pathogenic bacteria that had been assigned to this group based on phenotypic traits, often related to their pathology, and GC content <ns0:ref type='bibr' target='#b67'>(Mannheim, Pohl & Holländer, 1980)</ns0:ref>. For example, the genus Haemophilus includes a plethora of taxa that cause pneumonia and meningitis in humans, and Pasteurella have been associated with a range of infectious diseases in cattle, fowl and pigs <ns0:ref type='bibr' target='#b74'>(Naushad et al., 2015)</ns0:ref>. Moreover, since sequencebased taxonomies have become more common, new genera have been created within each genus, such as for example Aggregatibacter <ns0:ref type='bibr' target='#b77'>(Norskov-Lauritsen, 2006)</ns0:ref> or Avibacterium <ns0:ref type='bibr' target='#b9'>(Blackall et al., 2005)</ns0:ref>. We believe that a work-over of the evolutionary genetic relationship of the Pasteurellales is overdue.</ns0:p><ns0:p>The proportion of gene clusters that were unique to the three L. koalarum genome assemblies, relative to 55 of their most closely related genomes, was large relative to the size of genes that were unique to other genera in Anvio's pangenome analysis (Fig. <ns0:ref type='figure' target='#fig_10'>2</ns0:ref>). There were 282 gene clusters that could exclusively be found in the three L. koalarum genome assemblies. Among them, there were 136 gene clusters with complete sequences and COG annotation (Supplementary Table <ns0:ref type='table'>S4</ns0:ref>). There were 36 gene clusters unique to L. koalarum strain UCD-LQP1 and 19 of these had complete sequences and COG annotations (Supplementary Table <ns0:ref type='table'>S5</ns0:ref>).</ns0:p><ns0:p>Out of the 136 gene clusters with known COG functions that were unique to the three L. koalarum genome assemblies, 22 different gene clusters fell into the COG category 'Carbohydrate metabolism/transport'. This was the largest category, followed by 'Inorganic ion transport' (n = 15), 'Cell wall', 'Transcription', and 'Energy production' (n = 11, each), and 'Defense' (n = 7; Table <ns0:ref type='table'>3</ns0:ref> and Supplementary Table <ns0:ref type='table'>S4</ns0:ref>). The translated amino acid sequences for these gene clusters, extracted from L. koalarum strain UCD-LQP1, can be found in Supplementary Table <ns0:ref type='table'>S6</ns0:ref>.</ns0:p><ns0:p>Gene clusters in the category 'Carbohydrate metabolism and transport' are discussed in detail below. It is worth mentioning that several putative components of the phosphotransferase system were unique to L. koalarum. This system transports sugars into bacteria including glucose, mannose, fructose, and cellobiose. It can differ among bacterial species, mirroring the most suitable carbon sources available in the environment where a species evolved <ns0:ref type='bibr' target='#b103'>(Tchieu et al., 2001)</ns0:ref>. L. koalarum also stood out in terms of genes coding for cell wall components including for example teichoic acid and other outer membrane proteins (Table <ns0:ref type='table'>3</ns0:ref> and Supplementary Table <ns0:ref type='table'>S4</ns0:ref>). These outer membrane proteins are diverse and can significantly differ among bacterial species <ns0:ref type='bibr' target='#b94'>(Schleifer & Kandler, 1972)</ns0:ref>. A few other potentially unique gene clusters included genes coding for type IV pilus assembly proteins for species-specific pili and fimbria <ns0:ref type='bibr' target='#b91'>(Proft & Baker, 2009)</ns0:ref>; defense mechanisms, such as putative bacteriophage resistance proteins, phage repressor proteins; and drug transport and efflux pumps. Several of these factors are characteristic for pathogenic bacteria <ns0:ref type='bibr' target='#b23'>(Craig, Pique & Tainer, 2004</ns0:ref>). Here it is worth noting that a gram-negative bacterium that was assigned to the genus Lonepinella based on 16S rRNA gene sequences caused a human wound infection after a wildlife worker had been bitten by a koala <ns0:ref type='bibr' target='#b99'>(Sinclair et al., 2019)</ns0:ref>.</ns0:p></ns0:div>
<ns0:div><ns0:head>Carbohydrate metabolism</ns0:head><ns0:p>Since the majority of unique gene clusters in all three L. koalarum genome assemblies were related to carbohydrate metabolism and transport, we decided to screen all three L. koalarum assemblies for potential enzymes that assemble, modify, and breakdown oligo-and polysaccharides. Using very stringent selection thresholds of the CAZy database where genes coding for carbohydrate-active enzymes have to be identified by three different methods, we found evidence for the presence of genes encoding 15 different glycoside hydrolase families, three different carbohydrate esterase families, and nine different glycosyltransferase families (Table <ns0:ref type='table'>4</ns0:ref>). Note, gene families in L. koalarum are predicted to have these activities in carbohydrate metabolism and transport based on characterized other members in the CAZy database, but we do not provide experimental evidence that L. koalarum performs these activities. All 28 identified CAZy gene families had also been annotated in the 2,370 eggNOG annotations (Supplementary Table <ns0:ref type='table'>S3</ns0:ref>). Glycoside hydrolase families, GH2, GH31, GH32, GH43, and GH77 were only found in the three L. koalarum genome assemblies relative to the other taxa in the comparative genomic analysis (see also Table <ns0:ref type='table'>3</ns0:ref> and Supplementary Table <ns0:ref type='table'>S7</ns0:ref>).</ns0:p><ns0:p>These five glycoside hydrolases are responsible for the hydrolysis of glycosidic bonds. Notably, when Lonepinella koalarum was isolated and described the first time as a phylogenetically and phenotypically novel group within the family Pasteurellaceae, enzyme activities were determined using commercially available oxidase/catalase tests as well as high-pressure liquid chromatography <ns0:ref type='bibr' target='#b82'>(Osawa et al., 1995)</ns0:ref>. The new taxon in 1995 (first described L. koalarum) showed positive results for beta-galactosidase (putatively enzyme family GH2) and alphaamylase (putatively enzyme family GH77) and negative results for urease, arginine dihydrolase, lysine decarboxylase, and tryptophane desaminase in congruence with the sequence-based results here.</ns0:p><ns0:p>Genes coding for oligosaccharide-degrading enzymes in the families GH1, GH2, GH3, GH42, and GH43 have also been found in another study that was investigating koala and wombat metagenomes <ns0:ref type='bibr' target='#b97'>(Shiffman et al., 2017)</ns0:ref>. Especially GH2, GH3 and GH43 were relatively common in koala metagenomes, relative to wallaby foregut <ns0:ref type='bibr' target='#b88'>(Pope et al., 2010)</ns0:ref>, cow rumen <ns0:ref type='bibr'>(Brulc et al., 2009)</ns0:ref>, and termite hindgut <ns0:ref type='bibr' target='#b42'>(He et al., 2013)</ns0:ref> metagenomes, where these enzymes had also been characterized. These five glycoside hydrolase families comprise mostly oligosaccharidedegrading enzymes <ns0:ref type='bibr' target='#b3'>(Allgaier et al., 2010)</ns0:ref>; i.e., they are able to break down a specific group of monosaccharide sugars in other bacteria that had been characterized for the CAZy database. However, presumably the major components of koala diet that are difficult to digest for the host are plant secondary metabolites and plant cell walls in Eucalyptus leaves, and oligosaccharidedegrading enzymes only play a significant role in a koala's diet after other enzymes have already degraded cellulose in leaf plant cell walls <ns0:ref type='bibr'>(Moore et al., 2005)</ns0:ref>. Oligosaccharides in Eucalyptus leaves will be absorbed by the koala in the small intestine and only a small fraction enter the caecum and colon. This means that the bacteria in the hindgut are most likely using their metabolic pathways to process the products of the degradation of complex carbohydrates with cross-feeding among microbiome members. The benefit of this activity to koala nutrition is not well understood. Interestingly, among the genes that code for the three carbohydrate-active enzyme families that were found exclusively in the assembly of L. koalarum strain UCD-LQP1, two were actual lignocellulases; i.e., microbial enzymes that hydrolyze the beta-1,4 linkages in cellulose <ns0:ref type='bibr' target='#b3'>(Allgaier et al., 2010)</ns0:ref>: Enzyme family GH42 and CE4. GH42 enzymes have mostly been described in cellulose-degrading bacteria, archaea and fungi <ns0:ref type='bibr' target='#b59'>(Kosugi, Murashima & Doi, 2002;</ns0:ref><ns0:ref type='bibr' target='#b98'>Shipkowski & Brenchley, 2006;</ns0:ref><ns0:ref type='bibr' target='#b28'>Di Lauro et al., 2008)</ns0:ref>. CE4 is a member of the carbohydrate esterase family, which groups enzymes that catalyze the de-acetylation of plant cell wall polysaccharides <ns0:ref type='bibr' target='#b8'>(Biely, 2012)</ns0:ref>. Digestion of plant cell walls, (i.e., cellulose, hemicellulose, and lignin), could be a second explanation (besides PCD degradation) of how L. koalarum plays a beneficial role in the koala gut microbiome.</ns0:p></ns0:div>
<ns0:div><ns0:head>Antibiotic resistance genes</ns0:head><ns0:p>Screening the three L. koalarum genome assemblies against the ResFinder database did not result in any detection of antibiotic resistance variants. However, there were three hits in the CARD database. First, all three L. koalarum assemblies contained a gene coding for a translated amino acid variant at a specific position (SNP R234F) that had been shown to confer resistance to pulvomycin in other bacterial species based on CARD predictions. Secondly, a variant was found to be encoded in all three L. koalarum genome assemblies that had been described before in Haemophilus influenza mutant PBP3, conferring resistance to beta-lactam antibiotics (cephalosporin, cephamycin, and penam) with SNPs D350N and S357N. The third result was an amino acid position with reference to a protein homolog model in a Klebsiella pneumoniae mutant, conferring resistance to the antibiotic efflux pump KpnH (including macrolide antibiotics, fluoroquinolone, aminoglycoside, carbapenem, cephalosporin, penam, and penem).</ns0:p><ns0:p>These results are based on predictions from the CARD 2020 database. All three hits are nucleotide sequences in the L. koalarum assemblies that are predicted to encode proteins that showed the same amino acid variants as other bacterial species in the CARD database. We do not know whether these variants confer antibiotic resistance in L. koalarum. Additional experiments are necessary to confirm that these CARD predictions work for L. koalarum. The corresponding nucleotide sequences and CARD output files are deposited on FigShare (UCD-LQP1: Wilkins, 2020e; ATCC 700131: <ns0:ref type='bibr' target='#b114'>Wilkins, 2020f;</ns0:ref><ns0:ref type='bibr'>and DSM 10053: Wilkins, 2020e)</ns0:ref>.</ns0:p></ns0:div>
<ns0:div><ns0:head>Recommendations for future koala management strategies</ns0:head><ns0:p>In previous work, we identified L. koalarum as the most predictive taxon of koala survival during antibiotic treatment and we suggested that this bacterium is important for koala health <ns0:ref type='bibr' target='#b24'>(Dahlhausen et al., 2018)</ns0:ref>. Here, we isolated a L. koalarum strain from the feces of a healthy koala and sequenced and characterized its genome. We found several putative detoxification pathways in L. koalarum strain UCD-LQP1 that could explain its potential beneficial role in the koala gut for koala survival and fitness. Besides detoxification of plant secondary metabolites, we found several putative genes involved in carbohydrate metabolism, particularly cellulose degradation. Some of these genes were only found in L. koalarum assemblies and not in 55 of their closely related genomes. Based on CARD predictions, the L. koalarum assemblies contain some sequences that are similar to antibiotic resistance genes in other bacterial species. We suggest confirming these antibiotic resistances in L. koalarum experimentally and testing the Manuscript to be reviewed efficiency of these antibiotic compounds against Chlamydia infections in koalas. In light of the various threats that koalas face, from chlamydia infection to wildfires <ns0:ref type='bibr' target='#b87'>(Polkinghorne, Hanger & Timms, 2013)</ns0:ref>, and the growing interest in rescuing and treating them in sanctuaries and zoos, it is important to identify beneficial members of their microbiome. This could (i) help decide which antibiotic compounds to choose during chlamydia treatment in order to maximize persistence of beneficial members in the koala gut microbiome, and (ii) guide the development of probiotic cocktails during recovery (Jin <ns0:ref type='bibr' target='#b48'>Song et al., 2019)</ns0:ref>.</ns0:p><ns0:p>Figure <ns0:ref type='figure'>1</ns0:ref> 16S rRNA gene phylogenetic placement of Lonepinella koalarum strain UCD-LQP1</ns0:p><ns0:p>The 16S rRNA gene was extracted from the L. koalarum genome assembly by searching for 'ssu rRNA' in the RAST function search of the SEED genome browser <ns0:ref type='bibr'>(Aziz et al., 2008;</ns0:ref><ns0:ref type='bibr' target='#b83'>Overbeek et al., 2014)</ns0:ref>. Included are all known 16S rRNA sequences in the Pasteurellaceae According to GTDB taxonomy, those two genomes are now Basfia species. See Discussion section). Each wedge represents a gene cluster. Gene clusters were grouped into mostly shared, shared, private, and in red: exclusively found in Lonepinella koalarum genome assemblies: 'LK', and exclusively found in L. koalarum strain UCD-LQP1. The gene marker tree was created in PhyloSift version 1.0.1 <ns0:ref type='bibr' target='#b25'>(Darling et al., 2014)</ns0:ref> with its updated markers database <ns0:ref type='bibr'>(version 4, posted on 12th of February 2018;</ns0:ref><ns0:ref type='bibr' target='#b51'>Jospin, 2018)</ns0:ref> for the alignment and RAxML version 8.2.10 on the CIPRES web server for the tree inference <ns0:ref type='bibr' target='#b71'>(Miller, Pfeiffer & Schwartz, 2010)</ns0:ref>. Gene clusters were ordered based on presence/absence. Also shown is GC content in light brown, number of genes per kilo base pairs in light grey, number of gene clusters in dark grey, and number of singleton gene clusters in orange, for each assembly, respectively. The heatmap shows ANI (Average nucleotide identity) values > 95%. The ANI heatmap is aligned with the Anvi'o profile, leading to the genome IDs on the y-axis. The Anvi'o database <ns0:ref type='bibr' target='#b115'>(Wilkins, 2020g)</ns0:ref> and profile <ns0:ref type='bibr' target='#b116'>(Wilkins, 2020h)</ns0:ref> </ns0:p></ns0:div><ns0:figure xml:id='fig_0'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:03:47130:2:0:NEW 11 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_1'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:03:47130:2:0:NEW 11 Sep 2020) Manuscript to be reviewed PCR and Sanger sequencing PCR amplification of the 16S rRNA gene was performed on each of the eluted DNA samples.</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_2'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:03:47130:2:0:NEW 11 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_3'><ns0:head>(</ns0:head><ns0:label /><ns0:figDesc>Class 1.11 Xenobiotics biodegradation and metabolism includes the following KEGG pathways:</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_4'><ns0:head /><ns0:label /><ns0:figDesc>Briefly, antibiotic resistance genes were searched with nucleotide sequences as input. RGI first predicts complete open reading frames (ORFs) using Prodigal version 2.6.3. To find protein homologs in the CARD references, DIAMOND version 0.9.29 is used. The 'perfect' algorithm detects perfect matches of individual amino acids to positions in the curated reference sequences PeerJ reviewing PDF | (2020:03:47130:2:0:NEW 11 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_5'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:03:47130:2:0:NEW 11 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_6'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:03:47130:2:0:NEW 11 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_7'><ns0:head /><ns0:label /><ns0:figDesc>heylii) was closest to Aggregatibacter spp. (yellow and pink in Fig. 2). Actinobacillus succinogenes and Mannheimia succiniproducens grouped with Pasteurella spp., while the former was the most closely related non-Lonepinella genome to L. koalarum strain UCD-LQP1. Here it is worth noting that both A. succinogenes and M. succiniproducens have been renamed in the new GTDB PeerJ reviewing PDF | (2020:03:47130:2:0:NEW 11 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_8'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:03:47130:2:0:NEW 11 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_9'><ns0:head /><ns0:label /><ns0:figDesc>family and one outgroup, Agarivoran spp. Nodes and tip labels are colored corresponding to the Anvi'o profile in Figure2; i.e., red: Lonepinella koalarum (Unicycler: assembly of L. koalarum strain UCD-LQP1, in bold and marked with a star), dark purple: Pasteurella spp., light purple: Haemophilus spp., orange: Actinobacillus spp., pink: Aggregatibacter spp., and green: Mannheimia spp. Black are genera that were not used in Figure2, and brown depicts the outgroup Agarivoran spp.</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_10'><ns0:head>Figure 2</ns0:head><ns0:label>2</ns0:label><ns0:figDesc>Figure 2</ns0:figDesc></ns0:figure>
<ns0:figure type='table' xml:id='tab_0'><ns0:head /><ns0:label /><ns0:figDesc>are accessible on FigShare.COG2337 mRNA-degrading endonuclease, toxin component of the MazEF toxinantitoxin module V COG0845 Multidrug efflux pump subunit AcrA (membrane-fusion protein) V COG3093 Plasmid maintenance system antidote protein VapI, contains XRE-type HTH domain V COG2828 2-Methylaconitate cis-trans-isomerase PrpF (2-methyl citrate pathway)</ns0:figDesc><ns0:table><ns0:row><ns0:cell>1</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell cols='2'>Statistic</ns0:cell><ns0:cell /><ns0:cell>Value</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell cols='2'>Completeness</ns0:cell><ns0:cell /><ns0:cell>99.205 %</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell cols='3'>Contamination COG1048 Aconitase A</ns0:cell><ns0:cell>0.705 %</ns0:cell><ns0:cell>C C</ns0:cell></ns0:row><ns0:row><ns0:cell cols='4'>Number of Contigs COG1454 Alcohol dehydrogenase, class IV 29</ns0:cell><ns0:cell>C</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell cols='3'>COG3312 FoF1-type ATP synthase assembly protein I</ns0:cell><ns0:cell>C</ns0:cell></ns0:row><ns0:row><ns0:cell cols='4'>Total Length COG0435 Glutathionyl-hydroquinone reductase 2,608,483 bp</ns0:cell><ns0:cell>C</ns0:cell></ns0:row><ns0:row><ns0:cell>GC%</ns0:cell><ns0:cell cols='3'>COG0371 Glycerol dehydrogenase or related enzyme, iron-containing ADH family 39.02 COG0778 Nitroreductase</ns0:cell><ns0:cell>C C</ns0:cell></ns0:row><ns0:row><ns0:cell>N50</ns0:cell><ns0:cell cols='3'>2,299,135 bp COG1053 Succinate dehydrogenase/fumarate reductase, flavoprotein subunit</ns0:cell><ns0:cell>C</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell cols='3'>COG4972 Tfp pilus assembly protein, ATPase PilM</ns0:cell><ns0:cell>W</ns0:cell></ns0:row><ns0:row><ns0:cell>N75</ns0:cell><ns0:cell cols='3'>2,299,135 bp COG1116 ABC-type nitrate/sulfonate/bicarbonate transport system, ATPase component</ns0:cell><ns0:cell>P</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell cols='2'>ABC-type nitrate/sulfonate/bicarbonate transport system, periplasmic</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>L50</ns0:cell><ns0:cell>COG0715</ns0:cell><ns0:cell>component</ns0:cell><ns0:cell>1</ns0:cell><ns0:cell>P</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell cols='3'>COG0600 ABC-type nitrate/sulfonate/bicarbonate transport system, permease component</ns0:cell><ns0:cell>P</ns0:cell></ns0:row><ns0:row><ns0:cell>L75</ns0:cell><ns0:cell cols='2'>COG2807 Cyanate permease</ns0:cell><ns0:cell>1</ns0:cell><ns0:cell>P</ns0:cell></ns0:row><ns0:row><ns0:cell cols='4'>Number of Predicted Genes COG2382 Enterochelin esterase or related enzyme 2,551 COG3301 Formate-dependent nitrite reductase, membrane component NrfD</ns0:cell><ns0:cell>P P</ns0:cell></ns0:row><ns0:row><ns0:cell cols='3'>Number of Protein Coding COG3230 Heme oxygenase</ns0:cell><ns0:cell>2,479</ns0:cell><ns0:cell>P</ns0:cell></ns0:row><ns0:row><ns0:cell>Genes</ns0:cell><ns0:cell cols='3'>COG0672 High-affinity Fe2+/Pb2+ permease</ns0:cell><ns0:cell>P</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell cols='2'>Iron uptake system EfeUOB, periplasmic (or lipoprotein) component</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>2</ns0:cell><ns0:cell>COG2822</ns0:cell><ns0:cell>EfeO/EfeM</ns0:cell><ns0:cell /><ns0:cell>P</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell cols='2'>NADPH-dependent ferric siderophore reductase, contains FAD-binding and</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell /><ns0:cell>COG2375</ns0:cell><ns0:cell>SIP domains</ns0:cell><ns0:cell /><ns0:cell>P</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell cols='3'>COG2223 Nitrate/nitrite transporter NarK</ns0:cell><ns0:cell>P</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell cols='3'>COG2223 Nitrate/nitrite transporter NarK</ns0:cell><ns0:cell>P</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell cols='3'>COG4771 Outer membrane receptor for ferrienterochelin and colicins</ns0:cell><ns0:cell>P</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell cols='3'>COG2837 Periplasmic deferrochelatase/peroxidase EfeB</ns0:cell><ns0:cell>P</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell cols='3'>COG0659 Sulfate permease or related transporter, MFS superfamily</ns0:cell><ns0:cell>P</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell cols='3'>COG4388 Mu-like prophage I protein</ns0:cell><ns0:cell>X</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell cols='2'>Phage repressor protein C, contains Cro/C1-type HTH and peptisase s24</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell /><ns0:cell>COG2932</ns0:cell><ns0:cell>domains</ns0:cell><ns0:cell /><ns0:cell>X</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell cols='3'>COG5412 Phage-related protein</ns0:cell><ns0:cell>X</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell cols='3'>COG1943 REP element-mobilizing transposase RayT</ns0:cell><ns0:cell>X</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell cols='3'>COG2189 Adenine specific DNA methylase Mod</ns0:cell><ns0:cell>L</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell cols='2'>ATP-dependent exoDNAse (exonuclease V) beta subunit (contains helicase</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell /><ns0:cell>COG1074</ns0:cell><ns0:cell cols='2'>and exonuclease domains)</ns0:cell><ns0:cell>L</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell cols='2'>ATP-dependent exoDNAse (exonuclease V), alpha subunit, helicase</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell /><ns0:cell>COG0507</ns0:cell><ns0:cell>superfamily I</ns0:cell><ns0:cell /><ns0:cell>L</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell cols='3'>COG3057 Negative regulator of replication initiation</ns0:cell><ns0:cell>L</ns0:cell></ns0:row><ns0:row><ns0:cell>1</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row></ns0:table></ns0:figure>
</ns0:body>
" | "Second response letter - Isolation and sequence-based characterization of a koala symbiont: Lonepinella koalarum
Dear Dr. Dahlhausen and colleagues:
Thanks for revising your manuscript. The reviewers are very satisfied with your revision (as am I). Great! However, there are a couple issues to still address and a few minor edits to make. Please address these ASAP so we may move towards acceptance of your work.
-joe
>>>> Dear Dr. Gillespie, again, thank you for your time and consideration. We have responded as soon as possible to each comment we received, and we have addressed all issues.
Reviewer 1 (Raphael Eisenhofer)
Basic reporting
No comment.
Experimental design
No comment.
Validity of the findings
No comment.
Comments for the Authors
The authors have addressed the reviewer comments and suggestions. I'm happy to recommend the manuscript for acceptance.
I'd like to thank the other reviewers for their valuable feedback and comments.
>>>> Dear Dr. Eisenhofer, thank you for reviewing our manuscript.
Reviewer 2 (Anonymous)
Basic reporting
No comment.
Experimental design
No comment.
Validity of the findings
No comment.
Comments for the Authors
Just some minor comments:
>>>> We would like to thank this reviewer for their time and consideration.
Line 267: 16S rRNA gene sequences
>>>> This has been added (page 9 lines 239-258).
Line 420 and Line 423: Citric acid cycle?
>>>> This has been corrected (page 14 line 397).
Line 527: phosphotransferase
>>>> This has been corrected (page 17 line 495).
Reviewer 3 (Anonymous)
Basic reporting
No comment.
Experimental design
No comment.
Validity of the findings
No comment.
Comments for the Authors
I reviewed a previous version of this manuscript. I find the current version of the maiscript much improved and the authors have adequately addressed my previous concerns. I am particulary pleased to see that they have taken many of my suggestions on board.
>>>> We would like to thank this reviewer for their valuable comments.
I have only a few minor comments/suggestions on the current version and would recommend acceptance of the article subject to them being addressed.
point by point comments:
Introduction
L: 87-90: I think the salient point here is that tannin bound protein is not able to be digested by the koala or utilised by the microbes. Maybe simplify/clarify the current explination?
>>>> We simplified and clarified this sentence (page 4 lines 87-92).
L94: Well I would assume they were alive at the beginning of the antibiotic treament. I suggest revising this sentence.
>>>> We changed this sentence (page 4 line 94).
Materials and methods
L226: I don’t think the comma is needed in “i.e.,”
>>>>We think the comma is needed and we used it consistently throughout the manuscript.
Results and discussion
L375: Here and elsewhere in the results and discussion the word toxin is still used when it is probably not the appropriate descriptor. I suggest carefully editing the manuscript to ensure that the word toxin is only used when referring to a compound that has a toxic effect in vivo.
>>>> We edited the manuscript and avoided using this term throughout the manuscript.
Table 3: This table is quite large and could be moved to the supplimenry material if space limited.
>>>> We left this table in the manuscript because space is not considerably limited.
First response letter - Isolation and sequence-based characterization of a koala symbiont: Lonepinella koalarum
Dear Dr. Dahlhausen and colleagues:
Thanks for submitting your manuscript to PeerJ. I have now received three independent reviews of your work, and as you will see, the reviewers raised some minor concerns about the research. Despite this, these reviewers are optimistic about your work and the potential impact it will lend to research on koala biology. Thus, I encourage you to revise your manuscript, accordingly, taking into account all of the concerns raised by the reviewers.
While the concerns of the reviewers are relatively minor, this is a major revision to ensure that the original reviewers have a chance to evaluate your responses to their concerns.
The concerns of Reviewer 3 will require the most attention. The reviewer has provided many ideas for improving the structure of your work and delivery of results.
I look forward to seeing your revision, and thanks again for submitting your work to PeerJ.
-joe
>>>> Dear Dr. Gillespie, thank you for your consideration and short turnaround time. We have responded to each comment we received. They proved very constructive and helpful.
Reviewer 1 (Raphael Eisenhofer)
Basic reporting
The English used throughout this manuscript is clear and the literature cited is relevant.
>>>> Dear Dr. Eisenhofer, thank you for taking the time to review our manuscript.
I would, however, recommend incorporating this recent and highly relevant paper into the introduction (perhaps in the paragraph of lines 62-74):
Faecal inoculations alter the gastrointestinal microbiome and allow dietary expansion in a wild specialist herbivore, the koala (Blyton et al. 2019).
>>>> This reference has been added.
Additionally, on line 80, I would replace Brice 2019 with this reference:
Gut microbes of mammalian herbivores facilitate intake of plant toxins. Kohl KD, Weiss RB, Cox J, Dale C, Dearing MD
Justification being that Kohl et al. 2014 experimentally demonstrated the point you’re making.
>>>> Thank you for noticing. We have changed those references.
In figure 1, I would recommend bolding and placing an asterisk on the UCD-LQP1 leaf for clarity.
>>>> This is done and also mentioned in the figure legend.
In figure 2, I would again recommend annotation of the UCD-LQP1 track to help quickly orient the reader. The ANI heatmap confused for a while until I realised that perhaps it is incorrectly oriented? It makes sense to me for it to be rotated 90 ° counterclockwise.
Actually, I think the issue is that the y-axis of the ANI heatmap is missing from the provided figure (was present when viewing through Anvi’o).
>>>> The ANI heatmap is aligned with the Anvi’o profile, that is why there are no labels at the y-axis. The samples are the same in the profile and in the heatmap and the two figures melt into each other. The labels are only visible if you zoom into the heatmap at high resolution (which might have been impossible with the low-resolution figure that was provided with the manuscript for review). UCD-LQP1 is now bold and marked with a star as in Fig. 1. We also clarified the figure legend.
Do the authors know the history of the koala they sampled? When was it brought to the San Francisco Zoo? Was it brought over from a Zoo (or the wild) in Australia? If so, which state/population, etc? What species of Eucalyptus is it being fed? Such information could be useful for future investigations comparing the author’s L. koalarum assembly to ones from different populations.
>>>> The SF Zoo was private about the history of their koalas. We assume that the koala in our study was born at the zoo, as the zookeeper pointed out to us that the koala’s mother lives in the same zoo. Koalas at the SF Zoo are fed blue gum leaves (Eucalyptus globulus), which grow quite abundantly in California. This information is now also added to the manuscript (lines 129-133).
Overall, the level of reporting and availability of resources in this manuscript is commendable. I love that a Jupyter notebook was uploaded containing extra background and the scripts ran. I can also confirm that the Anvi’o databases they provided are working. However, I could not find the raw sequencing data anywhere. I would request that the authors upload this both for reproducibility, and for future use with improved assembly software/tools.
>>>> We are glad that others also care about reproducibility and sharing of original scripts and data. The NCBI accession numbers for raw sequencing data, as well as genome assembly were somehow removed from the manuscript when we submitted to PeerJ by some automatic manuscript processing. Here is the part that was removed. We will make sure it will be included in the final manuscript.
Data deposition
NCBI Bioproject ID: PRJNA560698
Biosample Accession Number: SAMN12598050
Isolate Name: Lonepinella koalarum Strain UCD-LQP1
NCBI Assembly Accession Number: GCA_008723255.1
FigShare: https://figshare.com/projects/Lonepinella_koalarum_genome_assembly/74613
Scripts: https://doi.org/10.6084/m9.figshare.11678262
Experimental design
I see no issues with the methods and analyses used by the authors, and they are cited appropriately.
Validity of the findings
No comment.
Comments for the Authors
This paper is a logical extension from the authors’ previous work (Dahlhausen et al., 2018), and I think it’s great to see more work being done on the functional characterisation of host-associated microbes. Overall, I’m happy to recommend acceptance of the manuscript for publication on the provision that my comments are addressed.
>>>> Thank you.
Reviewer 2 (anonymous)
Basic reporting
The authors isolated a strain of Lonepinella koalarum and investigated genes that may be used for plant secondary metabolite degradation, carbohydrate metabolism, and antibiotic resistance. They compared the isolate genome to two other L. koalarum genomes from Genbank.
The article is well written and includes a sufficient introduction and background. Figures and tables are relevant to the article, appropriately described and labeled. Data is available from NCBI and permits for sample collection is provided. However, the NCBI accession number is missing from the manuscript.
>>>> We would also like to thank this reviewer for the time and effort and comments.
The NCBI accession numbers for raw sequencing data, as well as genome assembly were somehow removed from the manuscript when we submitted to PeerJ by some automatic manuscript processing. Here is the part that was removed. We will make sure it will be included in the final manuscript.
Data deposition
NCBI Bioproject ID: PRJNA560698
Biosample Accession Number: SAMN12598050
Isolate Name: Lonepinella koalarum Strain UCD-LQP1
NCBI Assembly Accession Number: GCA_008723255.1
FigShare: https://figshare.com/projects/Lonepinella_koalarum_genome_assembly/74613
Scripts: https://doi.org/10.6084/m9.figshare.11678262
Experimental design
The primary research is within the scope of the journal. A rigorous investigation was performed with well-designed, robust methods that could be reproduced.
Validity of the findings
All underlying data has been provided. Conclusions are well stated, linked to original research question and limited to supporting results.
Comments for the Authors
The fact that you have an isolate means that you could test for antibiotic resistance but it is out of the scope of the manuscript and could be a future project.
>>>> Yes, we agree. This would be an exciting follow-up study - experimenting with the isolated strain.
The manuscript follows on from previous work that showed that L. koalarum could be used as an indicator of koala health. How common is L. koalarum?
>>>> While occurring in low relative abundances in the koala gut, limited previous findings suggest that L. koalarum can be found in a majority of koalas (approximately 80%), although more research is needed to improve the accuracy of this statistic.
Minor edits are included below:
Line 102: brackets around Phascolarctos cinereus.
>>>> This has been added.
Line 389: Supplementary Table S2 doesn’t refer to taxonomic names.
>>>> Thank you for noticing. This should be Supplementary Table S1.
Figure 2: I find ‘private’ confusing. Maybe it would be better to refer to these gene clusters as ‘unique’.
>>>> We have changed the labeling accordingly.
Table 3 title: genome assemblies unitalicised.
>>>> Thank you for noticing. This has been changed now.
Table 4: I’m not sure what the “no approved entry” means. It maybe best to remove it.
>>>> This has been removed.
Reviewer 3 (anonymous)
Basic reporting
Clear and unambiguous English is used throughout and the literature cited is appropriate. The background and context presented in the introduction could be improved (see my point by point responses). The structure is generally appropriate although a section of the results is presented in the conclusions. The article is self-contained.
Experimental design
The research is original, performed to a high technical standard and the methods are well described. The research question is well defined, however, the knowledge gap that it fills could be better expressed. I suggest couching the genetic characterisation of lonepinella around what it tells us about its functional niche and how that related to koala nutrition and digestion. Tannin degradation may be one aspect but it is only a small part.
Validity of the findings
The overall findings are sound. I would, however, suggest a more detailed presentation of some of the outcomes. See my general comments for more details.
Comments for the Authors
The manuscript by Dahlhausen aims to genetically characterise Lonepinella koalarum, a member of the koala gut microbiome that has previously been identified as able to degrade tannin-protein complexes and may improve koala survival during antibiotic treatment. The methods and analyses are appropriate to meet these aims and their presentation is refreshingly detailed. However, I think that the major findings of the study are not well described or elaborated upon.
There is a strong focus on the detection (or the lack there of) of tannin-degrading genes. Yet such genes are poorly characterised and not well represented in annotation databases. Additionally, the benefit of tannin-protein complex degradation to koala nutrition is questionable as it is thought that the liberated protein cannot be absorbed by the koala in the hindgut and may instead only serve to benefit the microbial community. Furthermore, phenotypic studies have already shown that lonepinella is capable of degrading tannin complexes and thus genetic determination of this function provides little additional information about the species.
Instead, I feel that the main benefit of this work would be to identify lonepinella’s functional niche within the microbiome and describe how its functions could benefit koala nutrition, digestion and health. While many of the analyses need to achieve such a characterisation have already been performed, they have not been presented in sufficient detail to build an overall picture of lonepinella’s role in the microbiome. The analyses also mainly describe how lonepinella differs from closely related bacteria but what about its overall all metabolic functions? More details and discussion is needed around specific degradation pathways, what section of these pathways are present in lonepinella and the functions these enzymes and pathways have. The manuscript would also benefit from placing these functions and pathways in the context of koala gastro-physiology and lonepinella’s proposed functional niche within the microbiome. See my point-by point responses for further details.
>>>> We would also like to thank this reviewer for their time and effort and comments that significantly improved our manuscript.
Point-by Point comments:
Abstract
L21: write species name in full at first mention
>>>> This has been done.
L23: What is meant by a “common member”. L. koalarum is found at very low relative abundance in the faecal microbiomes of koalas and is not detected in the majority of koalas (potentially due to it falling below the detection threshold in most cases, see (Alfano 2013, Brice 2019 and Blyton 2019).
>>>> This part of the sentence was removed (lines 23-24).
L18-34: It would be good if the abstract included a summary of the studies major findings. What unique genes were found? Where did it fall in the phylogenetic tree? What plant secondary metabolite degrading genes were found? etc. It would also be useful to have a conclusion to indicate what these findings suggest for the role of L. koalarum in koala digestion and health.
>>>> We added major findings and conclusions to the abstract (lines 27-42). The previous abstract was vague, indeed.
Introduction
In general the introduction is very focused on PBMCs and their detoxification, however, the koala microbiome as a whole also plays an important role in macro nutrient digestion and fibre degradation (see Brice 2019 and Blyton 2019). This should also be covered in the introduction and be flagged as potential ways that lonepinella could contribute to koala digestion/health. These functions are covered in the manuscript’s analysis of the lonepinella genome and could be nicely given context in the introduction.
>>>> Macronutrient digestion and fiber degradation are now also added into the introduction at lines 78-81.
It would also be useful to indicate why the genomes were screened for genes associated with antibiotic resistance.
>>>> This has been added at lines 103-115.
L43: They can also cause negative post-digestive feedback by making the koala feel “sick” as discussed in Lawler, Foley and Eschler. Also many of these really are not toxins but rather anti-nutrient compounds. I think it is inappropriate to simply refer to them as toxins. I suggest referring to them as PBMCs.
>>>> We agree that the term ‘toxin’ is very specific and does not include many anti-nutrient compounds. We changed the term to ‘PCDs’ for plant chemical defenses throughout the manuscript. None of the authors was familiar with the term ‘PBMCs’.
L70: It is worth briefly outlining what tannin-protein complexes are and how they influence koala nutrition.
>>>> This information has been added at lines 87-90.
L72: You should give a bit more detail on this finding. Was it the abundance of L. koalarum before treatment that was predictive? The presence/absence of the bacteria? Or was it how well L. koalarum was maintained through the treatment that predicted koala survival?
In a previous study, a co-occurrence network analysis identified four bacterial taxa, including one of the genus Lonepinella, that could be found in feces of koalas that lived at both the beginning and end of their antibiotic treatment after Chlamydia infection. However, these four taxa were absent from feces of koalas that died. Furthermore, in the same study a random forest analysis revealed that the most predictive taxon of whether a koala would live or die during their antibiotic treatment was identified as Lonepinella koalarum. These findings show that koalas that died after antibiotic treatment had much lower relative abundance (sometimes even zero), of L. koalarum compared to koalas that didn’t die. We added this information to lines 90-99.
L69-71: I think it is important here to provide a bit more background about what is known about L. koalarum given it is the subject of this study. For example, Osawa showed that L. koalarum may by mucosal associated in the caecum. If doing this extends to introduction too much then some of the detail on the Eucalypt toxins could be condensed.
>>>> This information has been added at lines 85-87. It is also discussed in more detail in the discussion section (lines 530-537).
L80: “well-being” isn’t a scientific term. I suggest revising.
>>>> This has been removed.
Materials and methods
L207-226: Why was the phylogenetic tree assembled from the 16S gene sequences instead of constructing a full genome tree? What was the benefit of constructing this tree compared with the GTDB tree?
>>>> As mentioned in the text at lines 30-31 and 346-358, the GTDB reference database does not contain any L. koalarum assemblies to date. This taxonomic database was mostly used to identify the most closely related genomes for the comparative genomic analysis. A 16S gene tree was built to confirm that our isolate was indeed L. koalarum (lines 346-352 and Fig. 1).
Results and discussion
L289: It is of interest what proportion of isolates from the culturing reported in the methods were found to be L. koalarum by 16s Sanger sequencing and what (if any) other taxa were isolated. Perhaps a short summary could be given in the results?
>>>> We added this information to lines 327-332.
L290: What was the genome coverage?
>>>> 672. This number has also been added to the manuscript at line 329.
L328: How many tannin-degrading genes are present in the database used for the genome annotation? In general tannases are not well characterised. For example, only one KEGG orthalog was present last time I looked.
>>>> There were four confirmed tannase genes in RAST the last time we looked (Sunday, July 12th 2020).
L331: Was a clear zone present around isolate UCD-LQP1 on the plate? Bacteria without tannin-protein complex degrading capability are capable of growing on those plates if they are tannin tolerant. Only the presence of a clear zone indicates tannin degradation.
>>>> Yes, we did observe a clear zone around the colonies. However, we did not include all of the appropriate controls to test for tannase activity, so we do not report on them in this manuscript.
L338-340: Again how many tannases are present in the annotation databases?
>>>> There were seven confirmed tannase references in the eggNOG database (Sunday, June 7th 2020).
L346-347: These enzymes can nonetheless play important roles in degradation. The identified enzyme and function of these genes should be summarised in a supplementary table. It would also be worth interrogating these functions and describing them in the text.
>>>> We mapped these genes onto individual pathways on the KEGG website and report complete pathways in the current version of the amended manuscript (lines 394-406).
L348-349: Or because the tannases are not in the database.
>>>> This has been added.
L350-L354: It would be useful to map the identified genes onto these pathways (there are tools for this on the KEGG website) and determine if they form a continuous degradation/synthesis chain (i.e. are all enzymes necessary for conversion between two compounds present?). These are large pathways, what sections of the pathway are represented? What compounds are degraded by the genes identified in L. koalarum? This will give a much more specific indication of what functions lonepinella is performing.
>>>> We performed this additional analysis. The results are mentioned at lines 394-406 and in the new Supplementary Figures S1 to S3.
L491: oligosaccharides in Eucalyptus leaves will be absorbed by the koala in the small intestine and only a small fraction would enter the caecum and colon. As you say, this means that the bacteria in the hindgut are most likely using these pathways to process the products of the degradation of complex carbohydrates with cross-feeding among microbiome members. The benefit of such activity to koala nutrition is not well understood.
>>>> We have clarified this sentence (lines 543-555).
L492-501: This is an interesting and important finding. Was there evidence that L. koalarum is capable of fibre fermentation and the production of short-chain fatty acids?
>>>> No culturing experiments have been performed in this direction. The only evidence is the presence of a GH42 homolog and a CE4 homolog in the UCD-LQP1 assembly.
L502: reference required
>>>> This sentence has been removed.
Conclusions
L506-523: These are not conclusions and should be presented as a section on antibiotic resistance in the results section.
>>>> This section has been renamed (line 562).
L528: This sentence does not fit with the previous sentence as in the first you say beneficial members need to be identified but then explicitly talk about L. koalarum without first justifying that it is one of these beneficial members. I suggest revising.
>>>> This sentence has been revised (lines 585-600).
" | Here is a paper. Please give your review comments after reading it. |
9,827 | "<ns0:abstract xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:p>Since COVID-19, caused by severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2), was declared as a pandemic disease by the World Health Organization (WHO) in early 2020, many countries, organizations, and companies have tried to find the best way to diagnose the virus and contain its spreading. SARS-CoV-2 is a positive-sense single RNA (+ssRNA) coronavirus and mainly spreads through droplets, respiratory secretions, and direct contact. The early detection of the virus plays a central role in lowering COVID19 incidents and mortality rates. Thus, finding a simple, accurate, cheap, and quick detection approach for SARS-CoV-2 at early stage of the viral infection is urgent and at high demand all around the world. The Food and Drug Administration (FDA) and other health agencies have declared Emergency Use Authorization (EUA) to develop diagnostic methods for COVID-19 and fulfill the demand. However, not all developed methods are appropriate and selecting a suitable method is challenging. Among all detection methods, rRT-PCR is the gold standard method. Unlike molecular methods, serological methods lack the ability of early detection with low accuracy. In this review, we summarized the current knowledge about COVID-19 detection methods aiming to highlight the advantages and disadvantages of molecular and serological methods (Fig. <ns0:ref type='figure'>1</ns0:ref>).</ns0:p></ns0:div>
</ns0:abstract>
<ns0:body xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:head>Introduction</ns0:head><ns0:p>In January 2020, WHO initially named a newly identified β-coronavirus that caused many pneumonia cases in December 2019 in Wuhan, China as the 2019-novel coronavirus (2019-nCoV) <ns0:ref type='bibr' target='#b0'>[1,</ns0:ref><ns0:ref type='bibr' target='#b1'>2]</ns0:ref>. Eventually, WHO and Coronavirus Study Group (CSG) of International committee officially named the virus as SARS-CoV-2 and the disease as coronavirus disease 2019 (COVID-19) <ns0:ref type='bibr' target='#b2'>[3]</ns0:ref>. SARS-CoV-2 is a member of the coronaviruses (CoV) family and it is an enveloped, non-segmented, positive-sense single RNA (+ssRNA) coronavirus <ns0:ref type='bibr' target='#b0'>[1]</ns0:ref>. In early 2020, the whole genome sequence of SARS-CoV-2 was revealed which was 29.9kb <ns0:ref type='bibr' target='#b3'>[4]</ns0:ref> and 96.2% and 79.5% identical to a bat CoV RaTG13 and SARS-CoV genome sequences, respectively <ns0:ref type='bibr' target='#b1'>[2,</ns0:ref><ns0:ref type='bibr' target='#b4'>5]</ns0:ref>. CoVs genome includes six to twelve open reading frames (ORFs) <ns0:ref type='bibr' target='#b5'>[6]</ns0:ref>. The first and largest ORF (ORF1a/b) occupies approximately two-thirds of the viral RNA <ns0:ref type='bibr' target='#b6'>[7,</ns0:ref><ns0:ref type='bibr' target='#b7'>8]</ns0:ref> and the remaining one-third of the genome encodes the four main structural proteins which includes spike (S), envelope (E), membrane (M), and nucleocapsid (N) protein and other accessory proteins <ns0:ref type='bibr' target='#b7'>[8]</ns0:ref><ns0:ref type='bibr' target='#b8'>[9]</ns0:ref><ns0:ref type='bibr' target='#b9'>[10]</ns0:ref><ns0:ref type='bibr' target='#b10'>[11]</ns0:ref>. The S protein plays a major role in SARS-CoV-2 infectious process and it is a promising target for vaccine and therapeutic development <ns0:ref type='bibr' target='#b11'>[12,</ns0:ref><ns0:ref type='bibr' target='#b12'>13]</ns0:ref>.</ns0:p><ns0:p>COVID-19 virus is a highly contagious and spreads through droplets, respiratory secretions, and direct contact <ns0:ref type='bibr' target='#b13'>[14,</ns0:ref><ns0:ref type='bibr' target='#b14'>15]</ns0:ref>. Recent studies reported that the virus was isolated from fecal swabs and blood samples of COVID-19 patients <ns0:ref type='bibr' target='#b15'>[16,</ns0:ref><ns0:ref type='bibr' target='#b16'>17]</ns0:ref> suggesting that the virus may have different routes to transmit between humans. The number of SARS-CoV-2 virus that causes ill to human is not clearly defined yet; however, a large hospitalized cohort (n=1145) was analyzed and the overall mean log10 viral load was Manuscript to be reviewed 5•6 copies per mL <ns0:ref type='bibr' target='#b17'>[18]</ns0:ref>. Elderly people and whom has chronic underlying diseases, such as but not limited to hypertension <ns0:ref type='bibr' target='#b18'>[19,</ns0:ref><ns0:ref type='bibr' target='#b19'>20]</ns0:ref>, diabetes <ns0:ref type='bibr' target='#b20'>[21,</ns0:ref><ns0:ref type='bibr' target='#b21'>22]</ns0:ref>, and chronic obstructive pulmonary disease <ns0:ref type='bibr' target='#b22'>[23]</ns0:ref>, are the most vulnerable <ns0:ref type='bibr' target='#b23'>[24,</ns0:ref><ns0:ref type='bibr' target='#b24'>25]</ns0:ref>. Current studies showed that the median age of COVID-19 patients was 47-59 years and females were the minority, less than 46% <ns0:ref type='bibr' target='#b25'>[26]</ns0:ref><ns0:ref type='bibr' target='#b26'>[27]</ns0:ref><ns0:ref type='bibr' target='#b27'>[28]</ns0:ref>. While children and youth have lower rates of COVID-19 infection compared to elder people <ns0:ref type='bibr' target='#b28'>[29]</ns0:ref><ns0:ref type='bibr' target='#b29'>[30]</ns0:ref><ns0:ref type='bibr' target='#b30'>[31]</ns0:ref>. The incubation period of the virus is one to fourteen days with 3-7 days being the most <ns0:ref type='bibr' target='#b31'>[32]</ns0:ref> It has been reported that the clinical symptoms of confirmed COVID-19 patients were varied from mild flu-like symptoms to very severe respiratory symptoms and even respiratory and kidney failures and death <ns0:ref type='bibr' target='#b32'>[33,</ns0:ref><ns0:ref type='bibr' target='#b33'>34]</ns0:ref>. According to WHO and other sources fever, dry cough, and tiredness are the most common symptoms while sore throat, diarrhea, headache, conjunctivitis, rash on skin, and discoloration of fingers or toes are less common symptoms of COVID-19 patients <ns0:ref type='bibr' target='#b34'>[35]</ns0:ref><ns0:ref type='bibr' target='#b35'>[36]</ns0:ref><ns0:ref type='bibr'>[37]</ns0:ref>. A recent study used an appbased symptom tracker showed that people who had COVID-19 loss of smell and taste and those with a positive test result (65.03%) intended to have anosmia higher than those with a negative test result (21.71%) <ns0:ref type='bibr' target='#b36'>[38,</ns0:ref><ns0:ref type='bibr' target='#b37'>39]</ns0:ref>. Although COVID-19 became a pandemic disease, the mortality rate is low (3.4%) compared to SARS and MERS patients, 9.6% and 35% respectively <ns0:ref type='bibr' target='#b38'>[40]</ns0:ref>. Manuscript to be reviewed <ns0:ref type='bibr' target='#b39'>[41]</ns0:ref>. Having a rapid and accurate diagnostic method at early stage of infection can help to contain the pandemic. Thus, many companies and laboratories were given authority under EUA restrictions to develop diagnostic methods. Consequently, hundreds of diagnostic kits based on different methods are available now, but selecting the proper method requires further investigation. In this review, the standard and current molecular and serological detection methods for SARS-CoV-2 will be discussed and highlighted.</ns0:p><ns0:p>As of today, among all detection methods rRT-PCR is the gold standard method. Unlike molecular methods, serological methods lack the ability of early detection with low accuracy. This review intends to help health care providers and related branches to choose the appropriate method for battling the COVID-19 pandemic and rise the public knowledge about the methods that could be used to detect the virus.</ns0:p></ns0:div>
<ns0:div><ns0:head>Survey Methodology</ns0:head><ns0:p>This literature review explored the peer-reviewed and preprint literatures with mainly focusing on COVID-19 disease and its molecular and serological detection methods. We searched the following databases and websites from March to July 2020: Google Scholar, PubMed, bioRxiv, medRxiv, I-TASSER, CDC, WHO, Coronavirus Manuscript to be reviewed</ns0:p></ns0:div>
<ns0:div><ns0:head>Diagnostic methods</ns0:head><ns0:p>Under the pressure of the pandemic, COVID-19 test demand is sharply increased which pushes a lot of biotech companies/ inventors to produce different kits based on variant approaches to detect SARS-CoV-2. The molecular and serological methods are the main methods to detect the virus.</ns0:p></ns0:div>
<ns0:div><ns0:head>Molecular methods</ns0:head><ns0:p>Based on how viral RNA be processed and detected, there are three major </ns0:p></ns0:div>
<ns0:div><ns0:head>rRT-PCR method</ns0:head><ns0:p>It is the gold standard and reliable molecular method to diagnose SARS-CoV-2 with high sensitivity (positive agreement) and specificity (negative agreement) <ns0:ref type='bibr' target='#b41'>[43]</ns0:ref>. This method has been developed by several laboratories to detect COVID-19 virus <ns0:ref type='bibr' target='#b42'>[44]</ns0:ref><ns0:ref type='bibr' target='#b43'>[45]</ns0:ref><ns0:ref type='bibr' target='#b44'>[46]</ns0:ref>.</ns0:p><ns0:p>In this method (Fig. <ns0:ref type='figure' target='#fig_12'>2B</ns0:ref>), cDNA is generated from the extracted RNA of COVID-19 virus with specific primers for the following genes 2019nCoV-N1 (N1), 2019nCoV-N2 (N2), and RNAse P (RP; internal control) as recommended by U.S. CDC (Table <ns0:ref type='table'>1</ns0:ref>) and other health agencies <ns0:ref type='bibr' target='#b41'>[43,</ns0:ref><ns0:ref type='bibr' target='#b45'>47,</ns0:ref><ns0:ref type='bibr' target='#b46'>48]</ns0:ref> (Table <ns0:ref type='table'>S1</ns0:ref>). The upper respiratory system's swabs are the main specimens that are used to detect COVID-19 virus; however, serum, ocular secretions, and stool can be used as well <ns0:ref type='bibr' target='#b47'>[49]</ns0:ref><ns0:ref type='bibr' target='#b48'>[50]</ns0:ref><ns0:ref type='bibr' target='#b49'>[51]</ns0:ref>. If both genes (N1 and N2) were positive, it is considered as a positive sample as shown in Table <ns0:ref type='table'>2</ns0:ref>. The positive result confirms the presence of viral RNA in the specimen, but not necessarily the virus viability <ns0:ref type='bibr' target='#b50'>[52]</ns0:ref>. Besides the internal control (RP), there are three controls that must be run to make sure the result is legitimate (Table <ns0:ref type='table'>S2</ns0:ref>). These controls are 2019-nCoV Positive Control (nCoVPC), No Template Control (NTC), and Human Specimen Control (HSC) <ns0:ref type='bibr'>[42]</ns0:ref>. Even though rRT-RPC is the gold standard method and the most widely used for diagnosing COVID-19 virus in clinic and research laboratories, it has some limitations. Beside highly costed, professional skills needed, it is time-consuming (requires 2-5 days from collecting a sample till getting the result) and must be done in a laboratory.</ns0:p></ns0:div>
<ns0:div><ns0:head>Isothermal amplification-based method</ns0:head><ns0:p>It is another molecular approach where a nucleic acid is rapidly and specifically amplified by a polymerase with high strand displacement activity (e.g. optimized Bst polymerase) and different sets of primers at constant temperature (60-65°C) without the need of thermal cycler <ns0:ref type='bibr' target='#b51'>[53]</ns0:ref>. ID NOW COVID-19 (Abbott) is a recent example of using </ns0:p></ns0:div>
<ns0:div><ns0:head>CRISPR-Cas12 based method</ns0:head><ns0:p>In this method (e.g. SARS-CoV-2 DETECTR), the RNA virus is extracted from a specimen and designated regions of N2, E, RP genes are amplified at 62°C for 20 min by specific primes through Reverse Transcription Loop-mediated Isothermal Amplification (RT-LAMP) approach <ns0:ref type='bibr' target='#b53'>[56]</ns0:ref><ns0:ref type='bibr' target='#b54'>[57]</ns0:ref><ns0:ref type='bibr' target='#b55'>[58]</ns0:ref>. Then, designed Cas12 gRNAs direct Cas12 protein to specific areas of the above amplified genes where a reporter molecule (a single stranded DNA (ssDNA) probe) is cleaved. This reaction occurs at 37°C for 10 min and the result is visualized by a fluorescent reader or a lateral flow strip (Fig. <ns0:ref type='figure' target='#fig_12'>2C</ns0:ref>). Both genes N2 and E must be positive to consider the sample is positive (Table <ns0:ref type='table'>3</ns0:ref>).</ns0:p><ns0:p>Broughton et al. showed that SARS-CoV-2 DETECTR was reliable to detect coronavirus in respiratory swab samples with 90% sensitivity and 100% specificity <ns0:ref type='bibr' target='#b56'>[59]</ns0:ref>.</ns0:p><ns0:p>Unlike rRT-PCR, this method is fast (<50 minutes), cheap, and point-of-care test (POCT). It requires less equipment and the result can be visualized by naked eyes. However, it requires troubleshooting and specific design of all enzymes, primers, and reporters that are used in this method.</ns0:p><ns0:p>In addition to the above molecular methods, Recombinase polymerase amplification (RPA) <ns0:ref type='bibr' target='#b57'>[60]</ns0:ref>has been developed and/or integrated with other methods to detect COVID-19 virus. This method does not require thermal cycler and can be used as POCT with low cost and high sensitivity and specificity. The drawback is that it requires several specific designed primers which could be difficult to obtain and the result of this method could be interfered by virus quantification and debris <ns0:ref type='bibr' target='#b58'>[61]</ns0:ref>. Manuscript to be reviewed</ns0:p></ns0:div>
<ns0:div><ns0:head>Serological methods</ns0:head><ns0:p>Unlike molecular methods, serological methods (also called antibody tests) can be applied to detect past and current SARS-CoV-2 infection and monitor the progress of the disease periods and immune response. They can detect the presence of antibodies (e.g. IgG, IgM, and IgA) in a COVID-19 patient's serum and plasma. Other biological fluids such as but not limited to saliva and sputum could be used as well. Antibodies are produced as a defense mechanism by the immune system against SARS-CoV-2. First, IgM is produced after a few days of infection and last for approximately two weeks which followed by IgG production that is last longer <ns0:ref type='bibr' target='#b59'>[62,</ns0:ref><ns0:ref type='bibr' target='#b61'>63]</ns0:ref>. Thus, detecting IgM in a patient's sample indicates early-stage infection while detecting IgG indicates a current or prior infection <ns0:ref type='bibr' target='#b50'>[52]</ns0:ref>. In addition to lacking an early detection, accuracy is the main challenge of these approaches where crossover could occur with other antibodies that produced as a result of infection of other coronavirus family members such as SARS-CoV <ns0:ref type='bibr' target='#b62'>[64,</ns0:ref><ns0:ref type='bibr' target='#b63'>65]</ns0:ref>.</ns0:p></ns0:div>
<ns0:div><ns0:head>Lateral flow assay (LFA)</ns0:head><ns0:p>It is one of the most popular serological method that has been applied in clinics to detect antigens <ns0:ref type='bibr' target='#b64'>[66]</ns0:ref>, antibodies <ns0:ref type='bibr' target='#b65'>[67]</ns0:ref>, and amplified nucleic acids <ns0:ref type='bibr' target='#b66'>[68,</ns0:ref><ns0:ref type='bibr' target='#b67'>69]</ns0:ref> in variant biological samples such as blood (serum or plasma) <ns0:ref type='bibr' target='#b68'>[70,</ns0:ref><ns0:ref type='bibr' target='#b69'>71]</ns0:ref> , urine <ns0:ref type='bibr' target='#b70'>[72]</ns0:ref>, and saliva <ns0:ref type='bibr' target='#b71'>[73]</ns0:ref>.</ns0:p><ns0:p>LFA is a paper-like membrane strip that is coated with two lines. The first line, the test line, contains anti-human IgG/IgM antibodies, while the second line, the control line, contains anti-rabbit IgG antibodies. After adding a patients specimen (e.g. blood) into the sample well, IgG/IgM antibodies are moved by capillary action toward the lines crossing through the conjugated pad where a specific conjugated antigen (e.g. gold Manuscript to be reviewed COVID-19 antigen conjugate) and rabbit-gold conjugated antibodies are impeded <ns0:ref type='bibr' target='#b72'>[74]</ns0:ref>.</ns0:p><ns0:p>IgG/IgM antibodies are interacted and made a complex with gold COVID-19 antigen conjugate. The complex binds anti-human IgG/IgM antibodies and immobilizes at the test line, while the rabbit-gold conjugate antibodies bind anti-rabbit IgG antibody and immobilized at the control line. The result will be visible as a red line due to the accumulation of gold particles. If both test and control lines appear red, the result is positive and negative when only the control line appears red. If both lines disappear or only the test line appears, the result is invalid (Fig. <ns0:ref type='figure' target='#fig_9'>3</ns0:ref>). The advantages of FLA are rapid (10-30 minutes), cheap, no need for professional skills, and portable (POCT). It can be done by 1-2 blood drops and the result is visualized by naked eyes without an expensive equipment. The drawback of FLA is a qualitative method, tells the presence or absence of antibodies against the virus without telling how much they were in a patient's sample, and it less accurate compared to rRT-PCR. It was showed that FLA has clinical 57% sensitivity and 100% specificity for IgM and 81% sensitivity and 100% specificity for IgG <ns0:ref type='bibr' target='#b73'>[75]</ns0:ref>.</ns0:p></ns0:div>
<ns0:div><ns0:head>Enzyme-linked immunosorbent assay (ELISA)</ns0:head><ns0:p>It is another serological method and called enzyme immunoassay (EIA). ELISA is a plate-based method that has been used for detecting and quantifying soluble substances such as proteins and antibodies in clinic and research laboratories. It includes direct and indirect formats <ns0:ref type='bibr' target='#b74'>[76]</ns0:ref>. The indirect ELISA, the most popular and more sensitive than the direct ELISA, an antigen (e.g. a recombinant protein (N protein) of SARS-CoV-2 virus) is coated onto the inner surface of 96-well or 384-well polystyrene plates <ns0:ref type='bibr' target='#b75'>[77]</ns0:ref>. A diluted patient's plasma which may have anti-SARS-CoV-2 IgG/IgM is added to the wells. The plate is incubated for one hour to allow the antibodies to interact with coated antigens. After washing the plate to eliminate unspecific interactions, a conjugated antibody with a reported enzyme such as horseradish peroxidase (HRP) or alkaline phosphatase (AP) is added to form sandwich complexes <ns0:ref type='bibr' target='#b76'>[78,</ns0:ref><ns0:ref type='bibr' target='#b77'>79]</ns0:ref>. These complexes are detected and quantified by adding a substrate (e.g. 3,3′,5,5′tetramethylbenzidine) that is utilized by the report enzyme and leads to change in the reaction color <ns0:ref type='bibr' target='#b78'>[80,</ns0:ref><ns0:ref type='bibr' target='#b79'>81]</ns0:ref>. The color is detected and measured by a plate reader (Fig. <ns0:ref type='figure' target='#fig_10'>4</ns0:ref>).</ns0:p><ns0:p>ELISA is relatively fast (2-5 hours) and cheap compared to rRT-PCR, and it is similar to FLA regard to accuracy. It has been reported that ELISA results were 50% (IgG) and 81% (IgM) for patients on day zero and became 81% (IgG) and 100%(IgM) on day five of SARS-CoV-2 infection <ns0:ref type='bibr' target='#b15'>[16]</ns0:ref>. Another study accomplished by Xiang et al. showed that using ELISA to detect IgM and IgG on day four of symptom onsite revealed a sensitivity of 77.3% and specificity of 100% for IgM while those were 83.3% and 95% respectively for IgG <ns0:ref type='bibr' target='#b61'>[63]</ns0:ref>.</ns0:p><ns0:p>Worth mention that there are other serological methods that are less common than FLA and ELISA. A colloidal gold immunochromatography assay (GICA), and Chemiluminescent immunoassay (CLIA) were developed to diagnose COVID-19; however, they have low sensitivity at the beginning of the infection <ns0:ref type='bibr' target='#b80'>[82,</ns0:ref><ns0:ref type='bibr' target='#b81'>83]</ns0:ref>. Pan et al.</ns0:p><ns0:p>reported that the sensitivity of GICA were 11.1% on the first week and 92.9% on the second weeks after the onset of symptoms <ns0:ref type='bibr'>[84]</ns0:ref>. Neutralization assays, on the other hands, are standard methods for determining antibody efficacy (e.g. serum virus neutralization (SVN) assay). They are used to check whether a patient has active antibodies that can neutralize the SARS-CoV-2 infection <ns0:ref type='bibr' target='#b82'>[85,</ns0:ref><ns0:ref type='bibr' target='#b83'>86]</ns0:ref>. These assays play a Manuscript to be reviewed key role in determining if an individual is eligible to donate his/her convalescent plasma as a treatment for seriously ill people although such treatment has not been fully validated <ns0:ref type='bibr' target='#b84'>[87]</ns0:ref>.</ns0:p><ns0:p>Both molecular and serological methods are not perfect in terms of detecting COVID-19 virus and each method has its own limitations. Though molecular methods are more reliable than serological methods, both methods could give false results due to various reasons. For instance, incorrect sampling, inadequate viral material in the specimen, improper RNA extraction, cross-reactions with other viral species, contamination, and technical issues could lead to positive and negative false results. To overcome such issues and increase the certainty of given results, these methods can be followed by secondary diagnostic methods such as a chest CT scan and x-ray imaging <ns0:ref type='bibr' target='#b85'>[88]</ns0:ref><ns0:ref type='bibr' target='#b86'>[89]</ns0:ref><ns0:ref type='bibr' target='#b87'>[90]</ns0:ref>.</ns0:p></ns0:div>
<ns0:div><ns0:head>Conclusion</ns0:head><ns0:p>Scientists have made significant progress in the characterization of the COVID-19 virus and how to limit its spreading. Also, they are working hard on diagnostic methods and finding therapies and vaccines against the virus. Currently, neither an approved vaccine nor a specific antiviral treatment is available for COVID-19 disease. Thus, detecting SARS-CoV-2 at the early infectious stage by a rapid and accurate diagnostic method could save thousands of lives. In this review we have discussed and summarized the current knowledge about molecular and serological methods that have been used to detect SARS-CoV-2. Though the molecular methods are more expensive, slower, and less available than serological methods, they are more accurate and rRT-PCR is the gold standard method among them (Table <ns0:ref type='table'>4</ns0:ref>). Further research and collaboration between scientists and companies are needed to overcome some Manuscript to be reviewed limitations of current methods and might find a new and better avenue to detect the virus. For instance, standardized the methods, produce new and high-quality kits and make them available at low cost will make the current methods more reliable. A sample is loaded in the sample well (1) and incubated to allow the capillary action to move sample antibodies (IgG/IgM) forward <ns0:ref type='bibr' target='#b1'>(2)</ns0:ref>. Gold COVID-19 antigen conjugates from the conjugate pad recognize and interact with sample antibodies forming complexes (3) that are immobilized by anti-human IgG/IgM antibodies and display the test red line (4). Control antibodies (rabbit-gold conjugates) are immobilized by anti-rabbit IgG antibodies and show the control red line. ( <ns0:ref type='formula'>6</ns0:ref>) FLA results possibilities are illustrated. A coated SARS-CoV-2 protein (antigen) onto wells of ELISA plate (1) interacts with the first antibody (anti-SARS-CoV-2 antibody) that is in a patient's sample. (3) After adding a secondary antibody (a conjugated antibody), it recognizes and interacts with the first antibodies. The reaction is developed by adding a substrate (4) which is cleaved by the conjugated enzyme and changes the reaction color after incubation (4) and ( <ns0:ref type='formula'>5</ns0:ref>), respectively. ( <ns0:ref type='formula'>6</ns0:ref>) results are read by ELISA plate reader. <ns0:ref type='table'>Table 1:</ns0:ref> Primers and probes that have been recommended by the U.S.CDC to detect SARS-CoV-2 by rRT-PCR.</ns0:p></ns0:div>
<ns0:div><ns0:head>List of Figures</ns0:head></ns0:div>
<ns0:div><ns0:head>List of Tables</ns0:head></ns0:div>
<ns0:div><ns0:head>Table 2:</ns0:head><ns0:p>Expected results and their interpretations of rRT-PCR method for COVID-19 specimens.</ns0:p></ns0:div>
<ns0:div><ns0:head>Table 3:</ns0:head><ns0:p>Expected results and their interpretations of SARS-CoV-2 DETECTR method for COVID-19 specimens.</ns0:p></ns0:div>
<ns0:div><ns0:head>Table 4:</ns0:head><ns0:p>Comparison between molecular and serological methods for detecting COVID-19 virus. <ns0:ref type='table'>Table S1:</ns0:ref> Primers and probes that have been recommended by other institutions to perform rRT-PCR and detect SARS-CoV-2.</ns0:p></ns0:div>
<ns0:div><ns0:head>Supplementary Tables</ns0:head></ns0:div>
<ns0:div><ns0:head>Table S2:</ns0:head><ns0:p>The rRT-PCR controls with expected results and interpretations.</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:04:47565:1:0:NEW 25 Aug 2020)</ns0:p><ns0:p>Manuscript to be reviewed </ns0:p></ns0:div><ns0:figure xml:id='fig_0'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:04:47565:1:0:NEW 25 Aug 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_1'><ns0:head /><ns0:label /><ns0:figDesc>Resource Center (John Hopkins University), Chinese Center for Disease Control and Prevention (CCDC), and National Institute of Infectious Diseases (NIID). And the top keywords that searched were: COVID-19, SARS-CoV-2, coronavirus, genomic RNA, protein structure, ACE2, transmission, symptoms, molecular detection methods, serological detection methods, rRT-PCR, ID NOW COVID-19, isothermal amplification, CRISPR, SARS-CoV-2 DETECTR, LAMP, recombinase polymerase amplification (RPA), LFA, and ELISA. PeerJ reviewing PDF | (2020:04:47565:1:0:NEW 25 Aug 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_2'><ns0:head /><ns0:label /><ns0:figDesc>molecular methods which are: real-time reverse transcription polymerase chain reaction (rRT-PCR), isothermal amplification, and clustered regularly interspaced short palindromic repeats (CRISPR) based methods. All these methods follow the same protocol that have been recommended by the Centers for Disease Control and Prevention (CDC) for collecting specimens from COVID-19 patients [42].</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_3'><ns0:head /><ns0:label /><ns0:figDesc>isothermal amplification technique to detect COVID-19 virus in clinics. It is molecular point-of-care platform in the United States of America and used under an Emergency Use Authorization (EUA) only to diagnose SARS-CoV-2. In this test a certain region of RdRp gene of SARS-CoV-2 is amplified by specific primers and results are displayed in a short time compared to rRT-PCR (Fig. 2A). It can show a positive result as little as 5 minutes and a negative result in 13 minutes. It has a performance of ≥94 % sensitivity and ≥98% specificity compared to lab-based PCR reference tests as it is advertised by the manufacturers [54]. However, Harrington et. al. results that published in a peer-reviewed journal showed that the overall sensitivity and specificity were 74.73% and 99.41%, respectively [55].</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_4'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:04:47565:1:0:NEW 25 Aug 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_5'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:04:47565:1:0:NEW 25 Aug 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_6'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:04:47565:1:0:NEW 25 Aug 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_7'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:04:47565:1:0:NEW 25 Aug 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_8'><ns0:head>Figure1: 19 Figure 2 :</ns0:head><ns0:label>192</ns0:label><ns0:figDesc>Figure1: Workflow summary of molecular and serological detection methods of</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_9'><ns0:head>Figure 3 :</ns0:head><ns0:label>3</ns0:label><ns0:figDesc>Figure 3: Schematic flowchart of FLA.</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_10'><ns0:head>Figure 4 :</ns0:head><ns0:label>4</ns0:label><ns0:figDesc>Figure 4: Schematic flowchart of indirect ELISA.</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_11'><ns0:head>Figure 1 Workflow</ns0:head><ns0:label>1</ns0:label><ns0:figDesc>Figure 1</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_12'><ns0:head>Figure 2 Schematic</ns0:head><ns0:label>2</ns0:label><ns0:figDesc>Figure 2</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_13'><ns0:head>Figure 3 Schematic</ns0:head><ns0:label>3</ns0:label><ns0:figDesc>Figure 3</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_14'><ns0:head>Figure 4 Schematic</ns0:head><ns0:label>4</ns0:label><ns0:figDesc>Figure 4</ns0:figDesc></ns0:figure>
<ns0:figure type='table' xml:id='tab_0'><ns0:head /><ns0:label /><ns0:figDesc>The first incidents of COVID-19 were diagnosed in Wuhan, China, in December 2019. After a few months, WHO announced COVID-19 as a pandemic disease across the whole world. As of July 2 ed 2020, a total of 10,716,063 confirmed cases globally, 2,686,587 confirmed in USA and 8,029,476 outside of USA, with 516,726 globally deaths were reported by the Coronavirus Resource Center, John Hopkins University</ns0:figDesc><ns0:table /><ns0:note>PeerJ reviewing PDF | (2020:04:47565:1:0:NEW 25 Aug 2020)</ns0:note></ns0:figure>
</ns0:body>
" | "Rebuttal Letter
First, we would like to thank the editor and reviewers for taking the time to review our manuscript.
Editor’s comments
I think that you should pay attention to all reviewers' comments. Especially, these two ones:
'Language is at times imprecise' and '...parts on biology of SARS-COV-2 coronavirus, which are
commonplace, should be removed... this review is too general, and should be shortened'.
We are grateful to the editor for the suggestions.
-
We have checked and enhanced the language in the revised manuscript.
-
We removed the commonplace parts on the biology of SARS-COV-2 coronavirus and
corresponding figure (Fig. 2).
-
We made the revised manuscript shorter and more specific.
Reviewer 1
Basic reporting
Is the review of broad and cross-disciplinary interest and within the scope of the journal?
-
The review proposes a description of the SARS-CoV-2 virus genome, a summary of transmission
mechanisms, and an overview of direct and indirect methods of viral infection detection.
-
The topic is of great interest, broad and interdisciplinary, and within the scope of the journal.
We thank the reviewer for appreciating our work.
Has the field been reviewed recently? If so, is there a good reason for this review (different point of
view, accessible to a different audience, etc.)?
-
Reviews are of great interest in order to help readers to make sense of the large information
content present in the scientific literature. Since the number of publications related to COVID-19
is extremely high and growing fast, reviews published within few months have a large amount of
novel information to discuss. That said, an example of a review that discusses diagnostics
approaches for COVID-19 is https://doi.org/10.3389/fcell.2020.00468. The value of this published
review is that it clearly lists advantages and disadvantages of every approach discussed and
clearly differentiates test used in the clinics from experimental assays that are under
development.
We appreciate the positive feedback from the reviewer.
Does the Introduction adequately introduce the subject and make it clear who the audience is/what
the motivation is?
-
The audience is clearly identified. Both the abstract and the introduction do not set a clear
expectation for the following sections: “SARS-CoV-2 structure” and “Transmission and symptoms
of COVID-19”. If the sections are to be retained, they should be announced in the introduction
and in the abstract.
We are grateful to the reviewer for this suggestion. To address this suggestion and make
the review shorter and more focused as suggested by the editor and another reviewer, we
included short paragraphs about “SARS-CoV-2 structure” and “Transmission and
symptoms of COVID-19” in the introduction (pages 3 and 4) and announced them in the
abstract. The original sections and fig.2 were removed.
Experimental design
Is the Survey Methodology consistent with a comprehensive, unbiased coverage of the subject? If
not, what is missing?
-
Survey Methodology clearly describes the databases used for literature research and the
keywords used in the search. It would be useful to state the date the searches were conducted. A
nice addition to the diagnostic methods might have been a discussion on neutralization methods.
Thank you for pointing this out. We stated the time when the reaches were done on the
revised manuscript (page5, line111). Also, we added neutralization methods to the revised
manuscript (pages 11 and 12, lines 253-259).
Are sources adequately cited? Quoted or paraphrased as appropriate?
-
Sources are adequately cited and paraphrased. The bibliography includes 104 references.
We appreciate the reviewer’s evaluation.
Is the review organized logically into coherent paragraphs/subsections?
-
The review is logically organized and easy to follow. More emphasis of advantages and
disadvantages of every testing approach and a clear delineation of what tests are currently used
in the clinics and what assays are experimental would have been very helpful.
We would like to thank the reviewer for the suggestion. We emphasized the advantages
and disadvantages of the methods more in the revised manuscript. And we mentioned
which method is currently used in clinic or experimental laboratories.
-
The paragraphs on the structure of the virus and on transmission and symptoms need to be
announced in the Introduction and Abstract. On transmission, current estimation of the number of
viruses needed to infect a human host might be relevant and interesting to the readership. A
stratification of symptoms according to the WHO might be helpful.
the structure of the virus and on transmission and symptoms
We are grateful to the reviewer for these valuable suggestions. With regards to the
structure of the virus and its transmission and symptoms, we have announced them in the
abstract and introduction of the revised manuscript.
With regards to “current estimation of the number of viruses needed to infect a human
host might be relevant and interesting to the readership” we have added it to the revised
manuscript (page 3, lines 71-74).
In the revised manuscript, we have revised and rewritten the symptoms mainly according
to WHO recommendations (page 4, lines 83-86).
Validity of the findings
Is there a well developed and supported argument that meets the goals set out in the Introduction?
-
The introduction states that the goal of diagnostic efforts is to achieve a rapid and accurate
diagnostic method. Nowhere in the review it is stated that serological methods are not fulfilling the
scope of early detection because seroconversion happens weeks after infection. Antibody
measurement indicates exposure to the virus. It is also very important to clarify the concept of
“serological method” that in the present review appears to be detection of proteins in serum.
Antibody measurement indicates exposure, viral antigen testing indicates presence of the virus
and likely active infection. This very important concept needs to be clarified throughout the
manuscript and in the abstract. If a discussion of serology diagnostic methods is retained in the
manuscript, the goal in the abstract and introduction of rapid testing needs to be rephrased and
expanded.
We thank the reviewer for the helpful comments. We clarified the concept of the
“serological method” and highlighted and expanded the goal of the rapid tests in the
revised manuscript. To eliminate any confusion that may occur and make the review
shorter as suggested by the editor and another reviewer, we removed the part that
addressed detecting viral protein in serum.
Does the Conclusion identify unresolved questions / gaps / future directions?
-
The conclusion is that further research is needed to overcome present limitations of diagnostic
approaches but it is not very specific nor lays out future directions. Interesting points might be:
including need for standardization, need of coordination with regulatory agencies, accessibility,
reagent quality are not discussed. Additionally, the use of testing in a scenario where a vaccine is
available would be interesting.
We thank the reviewer for suggesting such valuable points. We edited the conclusion and
made it more specific and has clear future directions. Also, we included the suggested
points in the conclusion of the revised manuscript.
It is a very interesting suggestion “the use of testing in a scenario where a vaccine is
available” and it would have been curious to explore it. However, it is difficult to properly
address the scenario at this time for many reasons, for example: 1) there is no a proved
vaccine yet 2) we don’t know how a proved vaccine would work especially with the
immune system of a patient which may interfere with serological tests, and 3) Having no
information about the efficiency of a proved vaccine and how much it could eliminate the
presence of the virus makes it hard to predict how the vaccine would affect the use of
diagnostic tests.
Comments for the author
General comments.
The review covers recent information about the SARS-CoV-2 structure, COVID-19 symptoms and
transmission, and diagnostic tests including 1) PCR and other nucleic acid amplification
technologies, 2) serology, and 3) direct antigen test. The intended audience is clearly defined. The
authors compiled a good amount of information on the testing approaches currently available.
We thank the reviewer for appreciating our work.
Major issues with the manuscript (issues are listed in order of importance): tell
-
Serology tests do not include direct antigen tests in serum. The concept of serological tests
should be clarified.
Thank you for pointing this out and we agreed. We clarified the concept of the serological
test in the revised manuscript.
-
The need for COVID-19 testing should be clarified in the abstract and introduction. If the goal is to
achieve early diagnosis, antibody testing is not the preferred test.
We thank the reviewer for the comment. We clarified the need for COVID-19 testing in the
abstract and introduction of the revised manuscript.
-
Tests should be critically discussed and advantages and disadvantages including costs,
turnaround time, sensitivity, specificity should be presented. The fact that PCR is the gold
standard for virus detection should be stated.
We are grateful to the reviewer for this comment. In the revised manuscript we discussed
the tests more critically. Although we summarized costs, turnaround time, sensitivity,
specificity, and other aspects of the tests in table 4, we highlighted them more in the text
of the revised manuscript.
We stated that rRT-PCR is a gold standard method in different places of the revised
manuscript.
-
Introduction should mention that a description of the virus and its symptoms is present.
We agreed with the reviewer. We added short paragraphs that descript the virus and its
symptoms in the introduction of the revised manuscript (pages 3 and 4).
-
A distinction between clinically used tests and research assays should be clearly defined.
We thank the reviewer for this helpful suggestion. We clearly distinguished between a clinic
and research use of the assays in the revised manuscript.
-
Language is at times imprecise.
We are grateful to the reviewer for this suggestion. In the revised manuscript the language
has been checked and improved.
Reviewer 2
Basic reporting
this paper represents good academic work on summarizing existing techniques to detect SARSCoV-2 in real life samples collected form patients or prospective patients. This field has been
extensively reviewed recently, for example in examples below. Many of the published review go into
much further details on technologies reviewed, and provide a very detailed landscape of futuristic
trends, which current review lacks. Also, this review is too general, and contain parts on biology of
SARS-COV-2 coronavirus, which are commonplace, and should be removed
1. PMID: 32729549, 2. PMID: 32729494, 3. PMID: 32641875, 4. PMID: 32609256, and 5. PMID:
32607246
We thank the reviewer for appreciating our work and raising some concerns. We agreed
with the reviewer that the field has been extensively studied and reviewed. For example,
most of the above reviews that listed by the reviewer were published after we submitted our
review. Nevertheless, each article and review have its own scope and audience. In our
review, we tried to discuss and summarize the current molecular and serological methods
in simple and informative ways that highlight the advantages and disadvantages of these
methods. Therefore, not only the scientific community but also other communities could get
benefit from this review. Giving a lot of details on each detection method would make the
review extremely long and might cause confusion.
We improved the revised manuscript and made it shorter and more precise. Also, we
removed the commonplace parts on the biology of SARS-COV-2 and the corresponding
fig.2.
Experimental design
Also, this review is too general, and contain parts on biology of SARS-COV-2 coronavirus, which are
commonplace, and should be removed. Sources are cited adequately, but some parts are present in
incorrect sections, i.e. LAMP PCR is reviewed in two different sections, depending on subtype of the
method
We are grateful to the reviewer for these comments. We improved the revised manuscript
and made it more specific. Also, the commonplace parts on the biology of SARS-COV-2
coronavirus were removed.
We apologize for presenting some parts in incorrect sections. Accordingly, we have
corrected that in the revised manuscript (pages 7 and 8).
Validity of the findings
I think this review is too general, and, because of that, short of expectations.
We thank the reviewer for the comment. We have improved the revised manuscript and
made it more specific.
Comments for the author
nice effort, but not specific (focused) enough for PeerJ
We thank the reviewer for appreciating our work. We have made the revised manuscript
more focused and fit to the PeeJ scope.
" | Here is a paper. Please give your review comments after reading it. |
9,828 | "<ns0:abstract xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:p>Since COVID-19, caused by severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2), was declared as a pandemic disease by the World Health Organization (WHO) in early 2020, many countries, organizations, and companies have tried to find the best way to diagnose the virus and contain its spreading. SARS-CoV-2 is a positive-sense single RNA (+ssRNA) coronavirus and mainly spreads through droplets, respiratory secretions, and direct contact. The early detection of the virus plays a central role in lowering COVID19 incidents and mortality rates. Thus, finding a simple, accurate, cheap, and quick detection approach for SARS-CoV-2 at early stage of the viral infection is urgent and at high demand all around the world. The Food and Drug Administration (FDA) and other health agencies have declared Emergency Use Authorization (EUA) to develop diagnostic methods for COVID-19 and fulfill the demand. However, not all developed methods are appropriate and selecting a suitable method is challenging. Among all detection methods, rRT-PCR is the gold standard method. Unlike molecular methods, serological methods lack the ability of early detection with low accuracy. In this review, we summarized the current knowledge about COVID-19 detection methods aiming to highlight the advantages and disadvantages of molecular and serological methods (Fig. <ns0:ref type='figure'>1</ns0:ref>).</ns0:p></ns0:div>
</ns0:abstract>
<ns0:body xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:head>Introduction</ns0:head><ns0:p>In January 2020, WHO initially named a newly identified β-coronavirus that caused many pneumonia cases in December 2019 in Wuhan, China as the 2019-novel coronavirus (2019-nCoV) <ns0:ref type='bibr' target='#b0'>[1,</ns0:ref><ns0:ref type='bibr' target='#b1'>2]</ns0:ref>. Eventually, WHO and Coronavirus Study Group (CSG) of International committee officially named the virus as SARS-CoV-2 and the disease as coronavirus disease 2019 (COVID-19) <ns0:ref type='bibr' target='#b3'>[3]</ns0:ref>. SARS-CoV-2 is a member of the coronaviruses (CoV) family and it is an enveloped, non-segmented, positive-sense single RNA (+ssRNA) coronavirus <ns0:ref type='bibr' target='#b0'>[1]</ns0:ref>. In early 2020, the whole genome sequence of SARS-CoV-2 was revealed which was 29.9kb <ns0:ref type='bibr' target='#b4'>[4]</ns0:ref> and 96.2% and 79.5% identical to a bat CoV RaTG13 and SARS-CoV genome sequences, respectively <ns0:ref type='bibr' target='#b1'>[2,</ns0:ref><ns0:ref type='bibr' target='#b6'>5]</ns0:ref>. CoVs genome includes six to twelve open reading frames (ORFs) <ns0:ref type='bibr' target='#b7'>[6]</ns0:ref>. The first and largest ORF (ORF1a/b) occupies approximately two-thirds of the viral RNA <ns0:ref type='bibr' target='#b8'>[7,</ns0:ref><ns0:ref type='bibr' target='#b9'>8]</ns0:ref> and the remaining one-third of the genome encodes the four main structural proteins which includes spike (S), envelope (E), membrane (M), and nucleocapsid (N) protein and other accessory proteins <ns0:ref type='bibr' target='#b9'>[8]</ns0:ref><ns0:ref type='bibr' target='#b11'>[9]</ns0:ref><ns0:ref type='bibr' target='#b12'>[10]</ns0:ref><ns0:ref type='bibr' target='#b13'>[11]</ns0:ref>. The S protein plays a major role in SARS-CoV-2 infectious process and it is a promising target for vaccine and therapeutic development <ns0:ref type='bibr' target='#b14'>[12,</ns0:ref><ns0:ref type='bibr' target='#b15'>13]</ns0:ref>.</ns0:p><ns0:p>COVID-19 virus is a highly contagious and spreads through droplets, respiratory secretions, and direct contact <ns0:ref type='bibr' target='#b16'>[14,</ns0:ref><ns0:ref type='bibr' target='#b17'>15]</ns0:ref>. Recent studies reported that the virus was isolated from fecal swabs and blood samples of COVID-19 patients <ns0:ref type='bibr' target='#b18'>[16,</ns0:ref><ns0:ref type='bibr' target='#b19'>17]</ns0:ref> suggesting that the virus may have different routes to transmit between humans. The number of Manuscript to be reviewed SARS-CoV-2 virus that causes ill to human is not clearly defined yet; however, a large hospitalized cohort (n=1145) was analyzed and the overall mean log10 viral load was 5•6 copies per mL <ns0:ref type='bibr' target='#b20'>[18]</ns0:ref>. Elderly people and whom has chronic underlying diseases, such as but not limited to hypertension <ns0:ref type='bibr' target='#b21'>[19,</ns0:ref><ns0:ref type='bibr' target='#b22'>20]</ns0:ref>, diabetes <ns0:ref type='bibr' target='#b23'>[21,</ns0:ref><ns0:ref type='bibr' target='#b24'>22]</ns0:ref>, and chronic obstructive pulmonary disease <ns0:ref type='bibr' target='#b25'>[23]</ns0:ref>, are the most vulnerable <ns0:ref type='bibr' target='#b26'>[24,</ns0:ref><ns0:ref type='bibr' target='#b27'>25]</ns0:ref>. Current studies showed that the median age of COVID-19 patients was 47-59 years and females were the minority, less than 46% <ns0:ref type='bibr' target='#b28'>[26]</ns0:ref><ns0:ref type='bibr' target='#b29'>[27]</ns0:ref><ns0:ref type='bibr' target='#b31'>[28]</ns0:ref>. While children and youth have lower rates of COVID-19 infection compared to elder people <ns0:ref type='bibr' target='#b32'>[29]</ns0:ref><ns0:ref type='bibr' target='#b33'>[30]</ns0:ref><ns0:ref type='bibr' target='#b34'>[31]</ns0:ref>. The incubation period of the virus is one to fourteen days with 3-7 days being the most <ns0:ref type='bibr' target='#b35'>[32]</ns0:ref> It has been reported that the clinical symptoms of confirmed COVID-19 patients were varied from mild flu-like symptoms to very severe respiratory symptoms and even respiratory and kidney failures and death <ns0:ref type='bibr' target='#b36'>[33,</ns0:ref><ns0:ref type='bibr' target='#b37'>34]</ns0:ref>. According to WHO and other sources fever, dry cough, and tiredness are the most common symptoms while sore throat, diarrhea, headache, conjunctivitis, rash on skin, and discoloration of fingers or toes are less common symptoms of COVID-19 patients <ns0:ref type='bibr' target='#b38'>[35]</ns0:ref><ns0:ref type='bibr' target='#b39'>[36]</ns0:ref><ns0:ref type='bibr'>[37]</ns0:ref>. A recent study used an appbased symptom tracker showed that people who had COVID-19 loss of smell and taste and those with a positive test result (65.03%) intended to have anosmia higher than those with a negative test result (21.71%) <ns0:ref type='bibr' target='#b40'>[38,</ns0:ref><ns0:ref type='bibr' target='#b41'>39]</ns0:ref>. Although COVID-19 became a pandemic disease, the mortality rate is low (3.4%) compared to SARS and MERS patients, 9.6% and 35% respectively <ns0:ref type='bibr' target='#b42'>[40]</ns0:ref>.</ns0:p><ns0:p>The first incidents of COVID-19 were diagnosed in Wuhan, China, in December 2019. After a few months, WHO announced COVID-19 as a pandemic disease across the whole world. As of July 2 ed 2020, a total of 10,716,063 confirmed cases globally, 2,686,587 confirmed in USA and 8,029,476 outside of USA, with 516,726 globally deaths were reported by the Coronavirus Resource Center, John Hopkins University <ns0:ref type='bibr' target='#b43'>[41]</ns0:ref>. Having a rapid and accurate diagnostic method at early stage of infection can help to contain the pandemic. Thus, many companies and laboratories were given authority under EUA restrictions to develop diagnostic methods. Consequently, hundreds of diagnostic kits based on different methods are available now, but selecting the proper method requires further investigation. In this review, the standard and current molecular and serological detection methods for SARS-CoV-2 will be discussed and highlighted.</ns0:p><ns0:p>As of today, among all detection methods rRT-PCR is the gold standard method. Unlike molecular methods, serological methods lack the ability of early detection with low accuracy. This review intends to help health care providers and related branches to choose the appropriate method for battling the COVID-19 pandemic and rise the public knowledge about the methods that could be used to detect the virus.</ns0:p></ns0:div>
<ns0:div><ns0:head>Survey Methodology</ns0:head><ns0:p>This literature review explored the peer-reviewed and preprint literatures with mainly focusing on COVID-19 disease and its molecular and serological detection methods. We searched the following databases and websites from March to July 2020: </ns0:p></ns0:div>
<ns0:div><ns0:head>Diagnostic methods</ns0:head><ns0:p>Under the pressure of the pandemic, COVID-19 test demand is sharply increased which pushes a lot of biotech companies/ inventors to produce different kits based on variant approaches to detect SARS-CoV-2. The molecular and serological methods are the main methods to detect the virus.</ns0:p></ns0:div>
<ns0:div><ns0:head>Molecular methods</ns0:head><ns0:p>Based on how viral RNA be processed and detected, there are three major molecular methods which are: real-time reverse transcription polymerase chain reaction </ns0:p></ns0:div>
<ns0:div><ns0:head>rRT-PCR method</ns0:head><ns0:p>It is the gold standard and reliable molecular method to diagnose SARS-CoV-2 with high sensitivity (positive agreement) and specificity (negative agreement) <ns0:ref type='bibr'>[43]</ns0:ref>. This method has been developed by several laboratories to detect COVID-19 virus <ns0:ref type='bibr'>[44]</ns0:ref><ns0:ref type='bibr'>[45]</ns0:ref><ns0:ref type='bibr'>[46]</ns0:ref>.</ns0:p><ns0:p>In this method (Fig. <ns0:ref type='figure' target='#fig_14'>2B</ns0:ref>), cDNA is generated from the extracted RNA of COVID-19 virus with specific primers for the following genes 2019nCoV-N1 (N1), 2019nCoV-N2 (N2), and RNAse P (RP; internal control) as recommended by U.S. CDC (Table <ns0:ref type='table'>1</ns0:ref>) and other health agencies <ns0:ref type='bibr'>[43,</ns0:ref><ns0:ref type='bibr'>47,</ns0:ref><ns0:ref type='bibr'>48]</ns0:ref> (Table <ns0:ref type='table'>S1</ns0:ref>). The upper respiratory system's swabs are the main specimens that are used to detect COVID-19 virus; however, serum, ocular secretions, and stool can be used as well [49-51]. If both genes (N1 and N2) were positive, it is considered as a positive sample as shown in Table <ns0:ref type='table'>2</ns0:ref>. The positive result confirms the presence of viral RNA in the specimen, but not necessarily the virus viability <ns0:ref type='bibr'>[52]</ns0:ref>. Besides the internal control (RP), there are three controls that must be run to make sure the result is legitimate (Table <ns0:ref type='table'>S2</ns0:ref>). These controls are 2019-nCoV Positive Control (nCoVPC), No Template Control (NTC), and Human Specimen Control (HSC) <ns0:ref type='bibr'>[42]</ns0:ref>. Even though rRT-RPC is the gold standard method and the most widely used for diagnosing COVID-19 virus in clinic and research laboratories, it has some limitations <ns0:ref type='bibr'>[53]</ns0:ref>. Beside highly costed, professional skills needed, it is time-consuming (requires 2-5 days from collecting a sample till getting the result) and must be done in a laboratory.</ns0:p></ns0:div>
<ns0:div><ns0:head>Isothermal amplification-based method</ns0:head><ns0:p>It is another molecular approach where a nucleic acid is rapidly and specifically amplified by a polymerase with high strand displacement activity (e.g. optimized Bst </ns0:p></ns0:div>
<ns0:div><ns0:head>CRISPR-Cas12 based method</ns0:head><ns0:p>In this method (e.g. SARS-CoV-2 DETECTR), the RNA virus is extracted from a specimen and designated regions of N2, E, RP genes are amplified at 62°C for 20 min by specific primes through Reverse Transcription Loop-mediated Isothermal Amplification (RT-LAMP) approach[57-59]. Then, designed Cas12 gRNAs direct Cas12 protein to specific areas of the above amplified genes where a reporter molecule (a single stranded DNA (ssDNA) probe) is cleaved. This reaction occurs at 37°C for 10 min and the result is visualized by a fluorescent reader or a lateral flow strip (Fig. <ns0:ref type='figure' target='#fig_14'>2C</ns0:ref>). Both genes N2 and E must be positive to consider the sample is positive (Table <ns0:ref type='table'>3</ns0:ref>).</ns0:p><ns0:p>Broughton et al. showed that SARS-CoV-2 DETECTR was reliable to detect coronavirus in respiratory swab samples with 90% sensitivity and 100% specificity <ns0:ref type='bibr' target='#b46'>[60]</ns0:ref>.</ns0:p><ns0:p>Unlike rRT-PCR, this method is fast (<50 minutes), cheap, and point-of-care test (POCT). It requires less equipment and the result can be visualized by naked eyes. However, it requires troubleshooting and specific design of all enzymes, primers, and reporters that are used in this method.</ns0:p><ns0:p>In addition to the above molecular methods, Recombinase polymerase amplification (RPA) <ns0:ref type='bibr' target='#b47'>[61]</ns0:ref>has been developed and/or integrated with other methods to detect COVID-19 virus. This method does not require thermal cycler and can be used as POCT with low cost and high sensitivity and specificity. The drawback is that it requires several Manuscript to be reviewed specific designed primers which could be difficult to obtain and the result of this method could be interfered by virus quantification and debris <ns0:ref type='bibr' target='#b48'>[62]</ns0:ref>.</ns0:p></ns0:div>
<ns0:div><ns0:head>Serological methods</ns0:head><ns0:p>Unlike molecular methods, serological methods (also called antibody tests) can be applied to detect past and current SARS-CoV-2 infection and monitor the progress of the disease periods and immune response. They can detect the presence of antibodies (e.g. IgG, IgM, and IgA) in a COVID-19 patient's serum and plasma. Other biological fluids such as but not limited to saliva and sputum could be used as well. Antibodies are produced as a defense mechanism by the immune system against SARS-CoV-2. First, IgM is produced after a few days of infection and last for approximately two weeks which followed by IgG production that is last longer <ns0:ref type='bibr' target='#b49'>[63,</ns0:ref><ns0:ref type='bibr' target='#b50'>64]</ns0:ref>. Thus, detecting IgM in a patient's sample indicates early-stage infection while detecting IgG indicates a current or prior infection <ns0:ref type='bibr'>[52]</ns0:ref>. In addition to lacking an early detection, accuracy is the main challenge of these approaches where crossover could occur with other antibodies that produced as a result of infection of other coronavirus family members such as SARS-CoV <ns0:ref type='bibr' target='#b51'>[65,</ns0:ref><ns0:ref type='bibr' target='#b52'>66]</ns0:ref>.</ns0:p></ns0:div>
<ns0:div><ns0:head>Lateral flow assay (LFA)</ns0:head><ns0:p>It is one of the most popular serological method that has been applied in clinics to detect antigens <ns0:ref type='bibr' target='#b53'>[67]</ns0:ref>, antibodies <ns0:ref type='bibr' target='#b54'>[68]</ns0:ref>, and amplified nucleic acids <ns0:ref type='bibr' target='#b55'>[69,</ns0:ref><ns0:ref type='bibr' target='#b56'>70]</ns0:ref> in variant biological samples such as blood (serum or plasma) <ns0:ref type='bibr' target='#b57'>[71,</ns0:ref><ns0:ref type='bibr' target='#b58'>72]</ns0:ref> , urine <ns0:ref type='bibr' target='#b59'>[73]</ns0:ref>, and saliva <ns0:ref type='bibr' target='#b60'>[74]</ns0:ref>.</ns0:p><ns0:p>LFA is a paper-like membrane strip that is coated with two lines. The first line, the test line, contains anti-human IgG/IgM antibodies, while the second line, the control line, Manuscript to be reviewed contains anti-rabbit IgG antibodies. After adding a patients specimen (e.g. blood) into the sample well, IgG/IgM antibodies are moved by capillary action toward the lines crossing through the conjugated pad where a specific conjugated antigen (e.g. gold COVID-19 antigen conjugate) and rabbit-gold conjugated antibodies are impeded <ns0:ref type='bibr' target='#b61'>[75]</ns0:ref>.</ns0:p><ns0:p>IgG/IgM antibodies are interacted and made a complex with gold COVID-19 antigen conjugate. The complex binds anti-human IgG/IgM antibodies and immobilizes at the test line, while the rabbit-gold conjugate antibodies bind anti-rabbit IgG antibody and immobilized at the control line. The result will be visible as a red line due to the accumulation of gold particles. If both test and control lines appear red, the result is positive and negative when only the control line appears red. If both lines disappear or only the test line appears, the result is invalid (Fig. <ns0:ref type='figure' target='#fig_11'>3</ns0:ref>). The advantages of FLA are rapid (10-30 minutes), cheap, no need for professional skills, and portable (POCT). It can be done by 1-2 blood drops and the result is visualized by naked eyes without an expensive equipment. The drawback of FLA is a qualitative method, tells the presence or absence of antibodies against the virus without telling how much they were in a patient's sample, and it less accurate compared to rRT-PCR. It was showed that FLA has clinical 57% sensitivity and 100% specificity for IgM and 81% sensitivity and 100% specificity for IgG <ns0:ref type='bibr' target='#b62'>[76]</ns0:ref>.</ns0:p></ns0:div>
<ns0:div><ns0:head>Enzyme-linked immunosorbent assay (ELISA)</ns0:head><ns0:p>It is another serological method and called enzyme immunoassay (EIA). ELISA is a plate-based method that has been used for detecting and quantifying soluble substances such as proteins and antibodies in clinic and research laboratories. It includes direct and indirect formats <ns0:ref type='bibr' target='#b63'>[77]</ns0:ref>. The indirect ELISA, the most popular and more Manuscript to be reviewed sensitive than the direct ELISA, an antigen (e.g. a recombinant protein (N protein) of SARS-CoV-2 virus) is coated onto the inner surface of 96-well or 384-well polystyrene plates <ns0:ref type='bibr' target='#b64'>[78]</ns0:ref>. A diluted patient's plasma which may have anti-SARS-CoV-2 IgG/IgM is added to the wells. The plate is incubated for one hour to allow the antibodies to interact with coated antigens. After washing the plate to eliminate unspecific interactions, a conjugated antibody with a reported enzyme such as horseradish peroxidase (HRP) or alkaline phosphatase (AP) is added to form sandwich complexes <ns0:ref type='bibr' target='#b65'>[79,</ns0:ref><ns0:ref type='bibr' target='#b66'>80]</ns0:ref>. These complexes are detected and quantified by adding a substrate (e.g. 3,3′,5,5′tetramethylbenzidine) that is utilized by the report enzyme and leads to change in the reaction color <ns0:ref type='bibr' target='#b67'>[81,</ns0:ref><ns0:ref type='bibr' target='#b68'>82]</ns0:ref>. The color is detected and measured by a plate reader (Fig. <ns0:ref type='figure' target='#fig_12'>4</ns0:ref>).</ns0:p><ns0:p>ELISA is relatively fast (2-5 hours) and cheap compared to rRT-PCR, and it is similar to FLA regard to accuracy. It has been reported that ELISA results were 50% (IgG) and 81% (IgM) for patients on day zero and became 81% (IgG) and 100%(IgM) on day five of SARS-CoV-2 infection <ns0:ref type='bibr' target='#b18'>[16]</ns0:ref>. Another study accomplished by Xiang et al. showed that using ELISA to detect IgM and IgG on day four of symptom onsite revealed a sensitivity of 77.3% and specificity of 100% for IgM while those were 83.3% and 95% respectively for IgG <ns0:ref type='bibr' target='#b50'>[64]</ns0:ref>.</ns0:p><ns0:p>Worth mention that there are other serological methods that are less common than FLA and ELISA <ns0:ref type='bibr' target='#b69'>[83]</ns0:ref>. A colloidal gold immunochromatography assay (GICA), and Chemiluminescent immunoassay (CLIA) were developed to diagnose COVID-19; however, they have low sensitivity at the beginning of the infection <ns0:ref type='bibr' target='#b70'>[84,</ns0:ref><ns0:ref type='bibr' target='#b71'>85]</ns0:ref>. <ns0:ref type='bibr'>Pan et al.</ns0:ref> reported that the sensitivity of GICA were 11.1% on the first week and 92.9% on the second weeks after the onset of symptoms <ns0:ref type='bibr' target='#b72'>[86]</ns0:ref>. Neutralization assays, on the other Manuscript to be reviewed hands, are standard methods for determining antibody efficacy (e.g. serum virus neutralization (SVN) assay). They are used to check whether a patient has active antibodies that can neutralize the SARS-CoV-2 infection <ns0:ref type='bibr' target='#b73'>[87,</ns0:ref><ns0:ref type='bibr' target='#b74'>88]</ns0:ref>. These assays play a key role in determining if an individual is eligible to donate his/her convalescent plasma as a treatment for seriously ill people although such treatment has not been fully validated <ns0:ref type='bibr' target='#b75'>[89]</ns0:ref>.</ns0:p><ns0:p>Both molecular and serological methods are not perfect in terms of detecting COVID-19 virus and each method has its own limitations <ns0:ref type='bibr' target='#b76'>[90]</ns0:ref>. Though molecular methods are more reliable than serological methods, both methods could give false results due to various reasons. For instance, incorrect sampling, inadequate viral material in the specimen, improper RNA extraction, cross-reactions with other viral species, contamination, and technical issues could lead to positive and negative false results. To overcome such issues and increase the certainty of given results, these methods can be followed by secondary diagnostic methods such as a chest CT scan and x-ray imaging <ns0:ref type='bibr' target='#b77'>[91]</ns0:ref><ns0:ref type='bibr' target='#b78'>[92]</ns0:ref><ns0:ref type='bibr' target='#b79'>[93]</ns0:ref>.</ns0:p></ns0:div>
<ns0:div><ns0:head>Conclusion</ns0:head><ns0:p>Scientists have made significant progress in the characterization of the COVID-19 virus and how to limit its spreading. Also, they are working hard on diagnostic methods and finding therapies and vaccines against the virus. Currently, neither an approved vaccine nor a specific antiviral treatment is available for COVID-19 disease. Thus, detecting SARS-CoV-2 at the early infectious stage by a rapid and accurate diagnostic method could save thousands of lives. In this review we have discussed and summarized the current knowledge about molecular and serological methods that have Manuscript to be reviewed been used to detect SARS-CoV-2. Though the molecular methods are more expensive, slower, and less available than serological methods, they are more accurate and rRT-PCR is the gold standard method among them (Table <ns0:ref type='table'>4</ns0:ref>). Further research and collaboration between scientists and companies are needed to overcome some limitations of current methods and might find a new and better avenue to detect the virus. For instance, standardized the methods, produce new and high-quality kits and make them available at low cost will make the current methods more reliable. <ns0:ref type='table'>Table 1:</ns0:ref> Primers and probes that have been recommended by the U.S.CDC to detect SARS-CoV-2 by rRT-PCR.</ns0:p></ns0:div>
<ns0:div><ns0:head>42.</ns0:head></ns0:div>
<ns0:div><ns0:head>Resources for Laboratories Working on Coronavirus (COVID-19).</ns0:head></ns0:div>
<ns0:div><ns0:head>List of Tables</ns0:head></ns0:div>
<ns0:div><ns0:head>Table 2:</ns0:head><ns0:p>Expected results and their interpretations of rRT-PCR method for COVID-19 specimens.</ns0:p><ns0:p>Table <ns0:ref type='table'>3</ns0:ref>:</ns0:p><ns0:p>Expected results and their interpretations of SARS-CoV-2 DETECTR method for COVID-19 specimens. Manuscript to be reviewed </ns0:p></ns0:div><ns0:figure xml:id='fig_0'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:04:47565:2:1:NEW 22 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_1'><ns0:head /><ns0:label /><ns0:figDesc>Google Scholar, PubMed, bioRxiv, medRxiv, I-TASSER, CDC, WHO, Coronavirus Resource Center (John Hopkins University), Chinese Center for Disease Control and Prevention (CCDC), and National Institute of Infectious Diseases (NIID). And the top keywords that searched were: COVID-19, SARS-CoV-2, coronavirus, genomic RNA, protein structure, ACE2, transmission, symptoms, molecular detection methods, serological detection methods, rRT-PCR, ID NOW COVID-19, isothermal amplification, PeerJ reviewing PDF | (2020:04:47565:2:1:NEW 22 Sep 2020) Manuscript to be reviewed CRISPR, SARS-CoV-2 DETECTR, LAMP, recombinase polymerase amplification (RPA), LFA, and ELISA.</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_2'><ns0:head>(</ns0:head><ns0:label /><ns0:figDesc>rRT-PCR), isothermal amplification, and clustered regularly interspaced short palindromic repeats (CRISPR) based methods. All these methods follow the same protocol that have been recommended by the Centers for Disease Control and Prevention (CDC) for collecting specimens from COVID-19 patients [42].</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_3'><ns0:head /><ns0:label /><ns0:figDesc>polymerase) and different sets of primers at constant temperature (60-65°C) without the need of thermal cycler [54]. ID NOW COVID-19 (Abbott) is a recent example of using isothermal amplification technique to detect COVID-19 virus in clinics. It is molecular point-of-care platform in the United States of America and used under an Emergency Use Authorization (EUA) only to diagnose SARS-CoV-2. In this test a certain region of RdRp gene of SARS-CoV-2 is amplified by specific primers and results are displayed in a short time compared to rRT-PCR (Fig. 2A). It can show a positive result as little as 5 minutes and a negative result in 13 minutes. It has a performance of ≥94 % sensitivity PeerJ reviewing PDF | (2020:04:47565:2:1:NEW 22 Sep 2020) Manuscript to be reviewed and ≥98% specificity compared to lab-based PCR reference tests as it is advertised by the manufacturers [55]. However, Harrington et. al. results that published in a peerreviewed journal showed that the overall sensitivity and specificity were 74.73% and 99.41%, respectively [56].</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_4'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:04:47565:2:1:NEW 22 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_5'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:04:47565:2:1:NEW 22 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_6'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:04:47565:2:1:NEW 22 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_7'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:04:47565:2:1:NEW 22 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_8'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:04:47565:2:1:NEW 22 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_9'><ns0:head /><ns0:label /><ns0:figDesc>Figure1: Workflow summary of molecular and serological detection methods of</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_10'><ns0:head>Figure 2 :</ns0:head><ns0:label>2</ns0:label><ns0:figDesc>Figure 2: Schematic flowchart of molecular detection methods for COVID-19 virus. (A) A viral RNA is directly amplified and detected from the sample by isothermal amplification method (e.g. ID NOW COVID-19). (B) The standard rRT-PCR method is illustrated where a viral RNA is extracted and converted into cDNA. Specific areas of cDNA (target genes) are amplified and detected by rRT-PCR. (C) demonstration of CRISPR-Cas12 based method. After amplifying specific areas of extracted viral RNA, the Cas12 enzyme recognizes specific sequences and then cleaves the ssDNA probe. The cleavage probes are visualized as a red line in the lateral flow strip. The figure is created with BioRender.com.</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_11'><ns0:head>Figure 3 :</ns0:head><ns0:label>3</ns0:label><ns0:figDesc>Figure 3: Schematic flowchart of FLA. A sample is loaded in the sample well (A) and incubated to allow the capillary action to move sample antibodies (IgG/IgM) forward (B). Gold COVID-19 antigen conjugates from the conjugate pad recognize and interact with sample antibodies forming complexes (C) that are immobilized by anti-human IgG/IgM antibodies and display the test red line (D). Control antibodies (rabbit-gold conjugates) are immobilized by antirabbit IgG antibodies and show the control red line (E). (F) FLA results possibilities are illustrated. The figure is created with BioRender.com.</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_12'><ns0:head>Figure 4 :</ns0:head><ns0:label>4</ns0:label><ns0:figDesc>Figure 4: Schematic flowchart of indirect ELISA.A coated SARS-CoV-2 protein (antigen) onto wells of ELISA plate (A) interacts with the first antibody (anti-SARS-CoV-2 antibody) that is in a patient's sample (B). (C) After adding a secondary antibody (a conjugated antibody), it recognizes and interacts with the first antibodies. The reaction is developed by adding a substrate (D) which is cleaved by the conjugated enzyme and changes the reaction color after incubation (E) and (F), respectively. (G) Results are read by ELISA plate reader. The figure is created with BioRender.com.</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_13'><ns0:head>Figure 1 Workflow</ns0:head><ns0:label>1</ns0:label><ns0:figDesc>Figure 1</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_14'><ns0:head>Figure 2 Schematic</ns0:head><ns0:label>2</ns0:label><ns0:figDesc>Figure 2</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_15'><ns0:head>Figure 3 Schematic</ns0:head><ns0:label>3</ns0:label><ns0:figDesc>Figure 3</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_16'><ns0:head>Figure 4 Schematic</ns0:head><ns0:label>4</ns0:label><ns0:figDesc>Figure 4</ns0:figDesc></ns0:figure>
<ns0:figure type='table' xml:id='tab_0'><ns0:head>Table 4 : Comparison between molecular and serological methods for detecting COVID-19 virus. Supplementary TablesTable S1 :</ns0:head><ns0:label>4S1</ns0:label><ns0:figDesc /><ns0:table /></ns0:figure>
</ns0:body>
" | "Rebuttal letter
Editor’s comments
I must apologize for so long period of time that our reviewers are taking to provide me with their decisions. I am sorry but because of so long waiting time I must ask you to include in your literature review several additional publications. I hope that you will be able to make these changes quickly.
• There is no need for apologize and we appreciate your time and hard work.
• We are grateful to the editor for suggesting below publications.
Molecular and Serological Tests for COVID-19 a Comparative Review of SARS-CoV-2 Coronavirus Laboratory and Point-of-Care Diagnostics. R Kubina, A Dziedzic - Diagnostics, 2020
• We included this reference in the revised manuscript (Reference #53)
Antonio La Marca, et al., Testing for SARS-CoV-2 (COVID-19): a systematic review and clinical guide to molecular and serological in-vitro diagnostic assays. Reproductive BioMedicine 41, 483-499, 2020
• We included this reference in the revised manuscript (Reference #83)
Bisoffi et al. Sensitivity, Specificity and Predictive Values of Molecular and Serological Tests for COVID-19: A Longitudinal Study in Emergency Room - Diagnostics 2020
• We included this reference in the revised manuscript (Reference #90)
" | Here is a paper. Please give your review comments after reading it. |
9,829 | "<ns0:abstract xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:p>SARS-CoV-2 is a betacoronavirus responsible for human cases of COVID-19, a pandemic with global impact that first emerged in late 2019. Since then, the viral genome has shown considerable variance as the disease spread across the world, in part due to the zoonotic origins of the virus and the human host adaptation process. As a virus with an RNA genome that codes for its own genomic replication proteins, mutations in these proteins can significantly impact the variance rate of the genome, affecting both the survival and infection rate of the virus, and attempts at combating the disease. In this study, we analyzed the mutation densities of viral isolates carrying frequently observed mutations for four proteins in the RNA synthesis complex over time in comparison to wildtype isolates. Our observations suggest mutations in nsp14, an error-correcting exonuclease protein, have the strongest association with increased mutation load in both regions without selective pressure and across the genome, compared to nsp7, 8, and 12, which form the core polymerase complex. We propose nsp14 as a priority research target for understanding genomic variance rate in SARS-CoV-2 isolates, and nsp14 mutations as potential predictors for high mutability strains.</ns0:p></ns0:div>
</ns0:abstract>
<ns0:body xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:head>Introduction</ns0:head><ns0:p>COVID-19 is an ongoing global pandemic characterized by long-term respiratory system damage in patients, and caused by the SARS-CoV-2 betacoronavirus. It is likely of zoonotic origin, but capable of human-to-human transmission, and since the first observed cases in the Wuhan province of China <ns0:ref type='bibr'>(Chan et al., 2020;</ns0:ref><ns0:ref type='bibr' target='#b5'>Riou & Althaus, 2020)</ns0:ref>, it has infected over 14 million people, with 612,054 recorded deaths (as of 22 July 2020). In addition to its immediate PeerJ reviewing PDF | (2020:08:51800:1:0:CHECK 8 Sep 2020)</ns0:p><ns0:p>Manuscript to be reviewed effects on the respiratory system, its long term effects are still being researched, including symptoms such as neuroinvasion <ns0:ref type='bibr'>(Li, Bai & Hashikawa, 2020;</ns0:ref><ns0:ref type='bibr' target='#b11'>Wu et al., 2020</ns0:ref><ns0:ref type='bibr'>), cardiovascular complications (Kochi et al., 2020;</ns0:ref><ns0:ref type='bibr' target='#b14'>Zhu et al., 2020)</ns0:ref>, and gastrointestinal and liver damage <ns0:ref type='bibr'>(Lee, Huo & Huang, 2020;</ns0:ref><ns0:ref type='bibr' target='#b13'>Xu et al., 2020)</ns0:ref>. Due to its high transmissibility, and capacity for asymptomatic transmission <ns0:ref type='bibr' target='#b10'>(Wong et al., 2020)</ns0:ref>, study of COVID-19 and its underlying pathogen remain a high priority. As a result, the high amount of frequently updated data on viral genomes on databases such as GISAID <ns0:ref type='bibr'>(Elbe & Buckland-Merrett, 2017)</ns0:ref> and <ns0:ref type='bibr'>NextStrain (Hadfield et al., 2018)</ns0:ref> provides researchers with invaluable resources to track the evolution of the virus as it spreads across the world. SARS-CoV-2 has a linear, single-stranded RNA genome, and does not depend on host proteins for genomic replication, instead using an RNA synthesis complex formed from nonstructural proteins (nsp) coded by its own genome. Four of the key proteins involved in the complex are nsp7, nsp8, nsp12, and nsp14, all of which are formed from cleavage of the polyprotein Orf1ab into mature peptides. Nsp12, also known as RdRp (RNA-dependent RNA polymerase), is responsible for synthesizing new strands of RNA using the viral genome as a template. Nsp7 and nsp8 act as essential co-factors for the polymerase unit, together creating the core polymerase complex <ns0:ref type='bibr'>(Kirchdoerfer & Ward, 2019;</ns0:ref><ns0:ref type='bibr'>Peng et al., 2020)</ns0:ref>, while nsp14 is an exonuclease which provides error-correcting capability to the RNA synthesis complex, therefore allowing the SARS-CoV-2 to maintain its large size genome <ns0:ref type='bibr' target='#b8'>(Subissi et al., 2014;</ns0:ref><ns0:ref type='bibr'>Ma et al., 2015;</ns0:ref><ns0:ref type='bibr'>Ogando et al., 2019;</ns0:ref><ns0:ref type='bibr' target='#b6'>Romano et al., 2020)</ns0:ref>. Owing to their role in maintaining replication fidelity and directly affecting the mutation-selection equilibrium of RNA viruses, these proteins are key targets of study in understanding the mutation accumulation and adaptive evolution of the virus <ns0:ref type='bibr'>(Eckerle et al., 2010;</ns0:ref><ns0:ref type='bibr'>Peng et al., 2020)</ns0:ref>.</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:08:51800:1:0:CHECK 8 Sep 2020)</ns0:p></ns0:div>
<ns0:div><ns0:head>Manuscript to be reviewed</ns0:head><ns0:p>In our previous study, we examined the top 10 most frequent mutations in the SARS-CoV-2 nsp12, and identified that four of them are associated with an increase in mutation density in two genes, the membrane glycoprotein (M) and the envelope glycoprotein (E) (the combination of which is hereafter referred to as MoE, as we previously described), which are not under selective pressure, and mutations in these genes are potential markers of reduced replication fidelity <ns0:ref type='bibr'>(Eskier et al., 2020a)</ns0:ref>. In this study, we follow up on our previous findings and analyze the mutations in nsps 7, 8, and 14, in addition to nsp12, to identify whether the mutations are associated with a nonselective increase in mutation load or not. We then examine whole genome mutation densities in mutant isolates in comparison to wildtype isolates using linear regression models, in order to understand whether the mutations are associated with potential functional impact. Our findings indicate that mutations in nsp14 are most likely to be predictors of accelerated mutation load increase.</ns0:p></ns0:div>
<ns0:div><ns0:head>Materials and Methods</ns0:head></ns0:div>
<ns0:div><ns0:head>Genome sequence filtering, retrieval, and preprocessing</ns0:head><ns0:p>As previously described <ns0:ref type='bibr'>(Eskier et al., 2020a)</ns0:ref>, SARS-CoV-2 isolate genome sequences and the corresponding metadata were obtained from the GISAID EpiCoV database (date of accession: 17 June 2020). We applied further quality filters, including selecting only isolates obtained from human hosts (excluding environmental samples and animal hosts), those sequenced for the full length of the genome (sequence size of 29 kb or greater), and those with high coverage for the reference genome (< 1% N content, < 0.05% unique mutations, no unverified indel mutations).</ns0:p><ns0:p>To ensure alignment accuracy, all nonstandard unverified nucleotide masking was changed to N due to the specifications of the alignment software, using the Linux sed command, and the isolates were aligned against the SARS-CoV-2 reference genome (NCBI Reference Sequence PeerJ reviewing PDF | (2020:08:51800:1:0:CHECK 8 Sep 2020)</ns0:p><ns0:p>Manuscript to be reviewed NC_045512.2, available at https://www.ncbi.nlm.nih.gov/nuccore/NC_045512.2) using the MAFFT (v7.450) alignment software <ns0:ref type='bibr'>(Katoh et al., 2002)</ns0:ref>, using the parameters outlined in the software manual for aligning closely related viral genomes (available at https://mafft.cbrc.jp/alignment/software/closelyrelatedviralgenomes.html). Variant sites in the isolates were annotated using snp-sites (2.5.1), bcftools (1.10.2), and ANNOVAR (release date 24 October 2019) software <ns0:ref type='bibr' target='#b9'>(Wang, Li & Hakonarson, 2010;</ns0:ref><ns0:ref type='bibr'>Page et al., 2016)</ns0:ref>, to identify whether a given mutation was synonymous or nonsynonymous. In addition, the 5' untranslated region of the genome (bases 1-265) and the 100 nucleotides at the 3' end were removed from the alignment and annotation files due to a high number of gaps and unidentified nucleotides. We further removed any sequences with incomplete sequencing location or date data in order to avoid complications in downstream analyses. Following the filters, 29,600 genomes were used for the analyses.</ns0:p></ns0:div>
<ns0:div><ns0:head>Mutation density calculation</ns0:head><ns0:p>Variants were categorized as synonymous and nonsynonymous following annotation by ANNOVAR, with intergenic or terminal mutations being considered synonymous. Gene mutation densities were calculated separately for synonymous and nonsynonymous mutations, as well as the total of SNVs, for each isolate, using a non-reference nucleotides per kilobase of region metric. Mutation densities were calculated for the combined membrane glycoprotein (M) and envelope glycoprotein (E) genes (MoE), the surface glycoprotein gene (S), and the whole genome.</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:08:51800:1:0:CHECK 8 Sep 2020)</ns0:p><ns0:p>Manuscript to be reviewed</ns0:p></ns0:div>
<ns0:div><ns0:head>Statistical Analysis</ns0:head><ns0:p>Descriptive statistics for continuous variable days were calculated with mean, standard deviation, median, and interquartile range. Kolmogorov-Smirnov test was used to check the normality assumption of the continuous variables. In cases of non-normally distributed data, the Wilcoxon rank-sum (Mann-Whitney U) test was performed to determine whether the difference between the two MoE status groups was statistically significant. The Fisher's exact test and the Pearson chi-square test were used for the analysis of categorical variables. The univariate logistic regression method was utilized to assess the mutations associated with MoE status in single variables, and then multiple logistic regression method was performed. The final multiple logistic regression model was executed with the backward stepwise method. The relationship between mutation density and time in isolates with mutations of interest, as well as in the group comprising all isolates, was examined via non-polynomial linear regression model and Spearman's rank correlation. A p-value of less than 0.05 was considered statistically significant.</ns0:p><ns0:p>All statistical analyses were performed using IBM SPSS version 25.0 (Chicago, IL, USA).</ns0:p></ns0:div>
<ns0:div><ns0:head>Results and Discussion</ns0:head></ns0:div>
<ns0:div><ns0:head>Increases in the mutation load of SARS-CoV-2 are unevenly distributed across its genome</ns0:head><ns0:p>To identify the trends in SARS-CoV-2 mutation load over time, we calculated the average mutation density per day for all isolates for whole genome, S gene, and MoE regions, capping outliers at the 95 th and 5 th percentile values to minimize the potential effects of sequencing errors (Fig. <ns0:ref type='figure' target='#fig_4'>1</ns0:ref>). Our results show that both at the genome level and the S gene, a very strong positive correlation between average mutation density and time. In comparison, MoE has a weak positive correlation, with a wider spread of mean density in early and late periods compared to the genome and the S gene. This is consistent with reduced selective pressure on the M and E genes, <ns0:ref type='table'>PDF | (2020:08:51800:1:0:CHECK 8 Sep 2020)</ns0:ref> Manuscript to be reviewed as has previously been described <ns0:ref type='bibr'>(Dilucca et al., 2020b)</ns0:ref>. The top nonsynonymous mutation is 23403A>G (in 22271 isolates), responsible for the D614G substitution in the spike protein, followed by the 14408C>T mutation (in 22226 isolates) in the nsp12 region of the Orf1ab gene, causing P323L substitution in the RdRp protein, and the 28144C>T mutation (in 3081 isolates), responsible for the L84S substitution in the Orf8 protein. The most common synonymous mutation is the 8782C>T mutation (in 3047 isolates), and is found on the nsp4 coding region of the Orf1ab gene. For the S gene, the most frequent synonymous mutation is the 23731C>T mutation (in 622 isolates), and the second most common nonsynonymous mutation, after the aforementioned D614G mutation, is 25350C>T (in 215 isolates), responsible for the P1263L substitution. For MoE, the most common synonymous and nonsynonymous mutations are 26735C>T (in 341 isolates) and 27046C>T (in 530 isolates), respectively, both of which are found in the M gene, and the latter of which causes T175M amino acid substitution. Other than the D614G mutation, all of the mentioned mutations are C>T substitutions, the prevalence of which in T-or A-rich regions of the SARS-CoV-2 genome have been previously documented <ns0:ref type='bibr' target='#b7'>(Simmonds, 2020)</ns0:ref>.</ns0:p><ns0:formula xml:id='formula_0'>PeerJ reviewing</ns0:formula></ns0:div>
<ns0:div><ns0:head>Mutations in RNA synthesis complex proteins are associated with higher mutation load</ns0:head><ns0:p>After identifying the increase in mutation load over time, which was more prominent in genes with high functional impact (S, Orf1ab) compared to other structural genes (M, E, N), as seen in Figure <ns0:ref type='figure' target='#fig_4'>1</ns0:ref> and Supplementary Figures <ns0:ref type='figure' target='#fig_5'>1 and 2</ns0:ref>, we sought to examine possible associations of variants in proteins involved in SARS-CoV-2 genome replication with the increase. We first identified the five most frequently observed mutations for nsps 7, 8, 12 (also known as RdRp) and 14, four of the proteins cleaved from the Orf1ab polyprotein and are involved in the RNA polymerization, followed by analyzing the association of each mutation with the presence of PeerJ reviewing PDF | (2020:08:51800:1:0:CHECK 8 Sep 2020)</ns0:p><ns0:p>Manuscript to be reviewed MoE mutations (hereafter referred to as MoE status) using the chi-square test. 12 out of the 20 mutations were found to have a significant association with MoE status (p-value < 0.05) (Table <ns0:ref type='table'>1</ns0:ref>). Compared to our previous findings on the top 10 nsp12 mutations (Eskier et al. 2020), which was based on an analysis of 11,208 samples as of 5 May 2020, 13536C>T and 13862C>T have increased in rank of appearance, from 6 th and 7 th to 4 th and 5 th , respectively, and decreased in pvalue to show statistically significant associations. In addition, the 13730C>T mutation have increased in rank of appearance from 4 th to 3 rd . Out of the other nsps tested, nsp14 was found to have four significant mutations, while nsp7 had two and nsp8 had one.</ns0:p></ns0:div>
<ns0:div><ns0:head>Effects of geographical location on MoE status</ns0:head><ns0:p>In addition to time and genotype, we also examined the potential association between the location of isolates and MoE status as a possible confounding factor. We first examined whether there is a significant association between location, defined here as continent the isolate was originally obtained, and MoE status. Our results indicate that there is a strong association between location and MoE status, with the highest percentage of MoE present isolates in Asia (14.5%), and the percentage ratio in South America (6.5%) (p-value <0.001). In comparison to our previous findings, South America had a dramatic decrease in MoE present isolate percentage, likely as a result of the increased sequencing efforts (from 118 isolates to 416) removing potential sampling biases or localized founder effects. Africa, Asia, and North America had an increase in MoE present proportion, while Europe, Oceania, and South America showed lowered percentages (Table <ns0:ref type='table'>2</ns0:ref>).</ns0:p><ns0:p>After observing the potential confounding effect of location on MoE status, we sought to understand whether a location is more or less likely to predict MoE status, using a logistic regression model (Table <ns0:ref type='table' target='#tab_1'>3</ns0:ref>). Comparing each individual region (1) to the other five (0), we found Manuscript to be reviewed that Asia, Europe, and North and South America are all possible predictors of MoE status (pvalue < 0.05), with Asia and Europe 1.697 and 1.184 times as likely to be MoE present as the other regions, and North and South America 0.589 and 0.650 times as likely, respectively.</ns0:p><ns0:p>Using these findings, we created different logistic regression models to identify which of the 12 mutations are likely to be independent predictors of MoE status (Table <ns0:ref type='table' target='#tab_2'>4</ns0:ref>). In the single variable model, all 12 mutations we previously identified and location were found to be potential predictors (p-value < 0.05). Forming final models including the 12 mutations (Final Model A) and the mutations as well as locations (Final Model B), we observed that the predictor effect of two of the mutations nsp8 12478G>A and nsp14 18998C>T do not appear to be sufficiently independent of the other mutations in Final Model A. After adding the location variable to the Final Model A, location remains a significant predictor, with all five non-reference locations less likely to predict MoE than Asia, the reference location, and nsp12 14805C>T is found to not have a predictor effect independent of location (p-value = 0.073). Following Final Model B, nine mutations appear to have a significant association with MoE status, independent of other variables: 11916C>T, 12073C>T, 13536C>T, 13730C>T, 13862C>T, 14408C>T, 18060C>T, 18736T>C, and 18877C>T (p-value < 0.05).</ns0:p></ns0:div>
<ns0:div><ns0:head>Nsp14 mutations have significant impact on increased genomic mutation density</ns0:head><ns0:p>We then examined the effects of each mutation on genomic mutation density to see whether the Manuscript to be reviewed which have a significant association with MoE status also show a similar relationship with genomic mutation density (Fig. <ns0:ref type='figure' target='#fig_5'>2</ns0:ref>). 18060C>T (L7L) has the lowest odds ratio for MoE status (Table <ns0:ref type='table' target='#tab_2'>4</ns0:ref>), and while it shows a slower increase in synonymous mutation density compared to wildtype isolates (Fig. <ns0:ref type='figure' target='#fig_5'>2A</ns0:ref>), it has a significant impact on faster mutation density increase in nonsynonymous mutations (Fig. <ns0:ref type='figure' target='#fig_5'>2B</ns0:ref>). In comparison, 18877C>T (L270L) (Fig. <ns0:ref type='figure' target='#fig_5'>2C-D</ns0:ref>) and 18736T>C (F233L) (Fig. <ns0:ref type='figure' target='#fig_5'>2E-F</ns0:ref>) both show a high prediction capacity for MoE and an increased mutation density. In comparison, mutations in nsp7 (Supp. Figs. <ns0:ref type='figure'>3-4</ns0:ref>) and nsp12 (Supp. Figs. <ns0:ref type='figure'>5-8</ns0:ref>)</ns0:p><ns0:p>show much lower impact on altered mutation density increase rate. 12073C>T, an nsp7 mutation, displays high divergence from wildtype isolate patterns; however, its low sample size (n = 16) creates a skewed distribution of isolates across time, complicating any potential inference.</ns0:p></ns0:div>
<ns0:div><ns0:head>Conclusions</ns0:head><ns0:p>Our previous work identified RdRp mutations as contributors to the evolution of the SARS-CoV-2 genome and this study confirmed those findings. Furthermore, we hypothesized that mutations of the other critical components of the viral replication and transcription machinery may have similar effects. Our results implicate nsp14 as a source of increased mutation rate in SARS-CoV-2 genomes. Three of the five most common nsp14 mutations, namely 18060C>T, 18736T>C and 18877C>T are associated with increases in both genome-wide mutational load, as well as MoE status, an alternative indicator of mutational rate and virus evolution. Interestingly all three are located within the ExoN domain, which is responsible for the proofreading activity of nsp14; however, only 18736T>C mutation is non-synonymous (F233L), while 18060C>T and PeerJ reviewing PDF | (2020:08:51800:1:0:CHECK 8 Sep 2020)</ns0:p><ns0:p>Manuscript to be reviewed 18877C>T are synonymous mutations and therefore, only after functional studies it will be possible to understand their effects on viral replication processes.</ns0:p><ns0:p>The origins and fates of the three nsp14 mutations are also quite different: Being present in the first case detected in the Washington state of the US in mid-January, 18060C>T mutation has been almost completely confined to the US, as 1,657 of 2,007 isolates (82.6%) originating from the US (https://bigd.big.ac.cn/ncov/variation/annotation/variant/18060, accessed 6 September 2020). On the other hand, 18877C>T mutation arising around at the end of January likely in Saudi Arabia and being detected in much less cases (n=893), is still present in many isolates, most frequently in Saudi Arabia (54.1%) and Turkey (37.4%). 18736T>C mutation was first detected in the US at the beginning of March and like the 18060C>T mutation, has almost completely been limited to the US (281/362 or 77.6%). Unlike the other two, this mutation has been detected in only two isolates since 27 May, and not after 1 July 2020. However, it should be noted that 18877C>T mutation arose within the dominant 23403A>G / 14408C>T lineage, while the other two nsp14 mutations are in different lineages. Therefore, dominance or disappearance of different nsp14 mutations may have less to do with these particular mutations and more with the co-mutations. Yet, we cannot rule out possible effects of these nsp14 mutations on the fitness of SARS-CoV-2.</ns0:p><ns0:p>Previous studies on alphacoronavirus nsp14 protein had shown that nsp14, via its exonuclease activity, can modulate host-virus interactions, degrading double-stranded RNA produced during genome replication to suppress immune response, thus increasing viral viability <ns0:ref type='bibr' target='#b0'>(Becares et al., 2016)</ns0:ref>. SARS-CoV-2 nsp14, due to similar exonuclease activity, is therefore a potential modulator of host interactions, independent of its link to increased mutation load. However, the exact effect of the mutations we identified, two of which are synonymous and may only Manuscript to be reviewed indirectly affect protein structure, have to be studied experimentally to show any possible changes in viral property that they might affect. Of note, a recent study where codon usage of SARS-CoV-2 was analyzed in terms of temporal evolution of the virus genome revealed that nsp14 is one of three genes (together with S and N genes) that display the highest Codon Adaptation Index (CAI) values <ns0:ref type='bibr' target='#b3'>(Dilucca et al., 2020a)</ns0:ref>. CAI is a measure of optimal codon usage and indicates how well codons adapt to the host. Based on higher CAI values in nsp14, one could speculate that such mutations have been accumulating preferentially to reach the optimal mutation rate that allows the most advantageous mutation-selection equilibrium for SARS-CoV-2. Indeed, our previous results <ns0:ref type='bibr'>(Eskier et al., 2020b)</ns0:ref> indicated that the mutation densities of SARS-CoV-2 genomes are closely related to the pandemic stage and population dynamics directly affects the average mutational load of the viral genome. During the rapid growth stages, such as those observed in March in the UK and the US, replication fidelity can be traded off to gain higher replication rates and broader mutational diversity. However, mutations in the replication machinery that result in too high mutation rates would likely be detrimental and eliminated. On the other hand, a small percentage of the resulting mutations could possibly be advantageous, including those that could confer resistance to antiviral drugs. So far, we or others have not been able to detect such mutations advantageous for the virus, however, higher mutation rates make appearance of such a mutation more likely.</ns0:p><ns0:p>We believe that the mutations discussed in this study can be of help to future studies, in both fighting the COVID-19 pandemic, and better understanding of how mutations in coronavirus replication proteins can affect viral viability and replication fidelity in hosts. Also, it is yet to be determined whether COVID-19 cases infected with SARS-CoV-2 that has mutation(s) that are Wildtype isolates in all graphs carry the reference nucleotide for the nine positions of interest <ns0:ref type='bibr'>(11916,</ns0:ref><ns0:ref type='bibr'>12073,</ns0:ref><ns0:ref type='bibr'>13536,</ns0:ref><ns0:ref type='bibr'>13730,</ns0:ref><ns0:ref type='bibr'>13862,</ns0:ref><ns0:ref type='bibr'>14408,</ns0:ref><ns0:ref type='bibr'>18060,</ns0:ref><ns0:ref type='bibr'>18736,</ns0:ref><ns0:ref type='bibr'>18877)</ns0:ref> (n = 5910). Correlation scores are calculated using Spearman rank correlation. Manuscript to be reviewed Manuscript to be reviewed </ns0:p><ns0:note type='other'>Figure Legends</ns0:note><ns0:note type='other'>Figure 2</ns0:note><ns0:p>The distribution of synonymous and nonsynonymous mutations in isolates carrying nsp14 mutations compared to wildtype isolates. <ns0:ref type='bibr'>(11916,</ns0:ref><ns0:ref type='bibr'>12073,</ns0:ref><ns0:ref type='bibr'>13536,</ns0:ref><ns0:ref type='bibr'>13730,</ns0:ref><ns0:ref type='bibr'>13862,</ns0:ref><ns0:ref type='bibr'>14408,</ns0:ref><ns0:ref type='bibr'>18060,</ns0:ref><ns0:ref type='bibr'>18736,</ns0:ref><ns0:ref type='bibr'>18877)</ns0:ref> (n = 5910). Correlation scores are calculated using Spearman rank correlation.</ns0:p></ns0:div><ns0:figure xml:id='fig_0'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:08:51800:1:0:CHECK 8 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_1'><ns0:head /><ns0:label /><ns0:figDesc>relationship between the mutations and MoE status are indicative of a genome-wide trend. Due to selection potentially effecting nonsynonymous mutations differentially, we separated the mutations in the two categories and calculated mutation density separately for each category. Our results show that nsp14 mutations show the most consistent association with mutations between MoE and the whole genome. All three nsp14 mutations (18060C>T, 18736T>C, and 18877C>T) PeerJ reviewing PDF | (2020:08:51800:1:0:CHECK 8 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_2'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:08:51800:1:0:CHECK 8 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_3'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:08:51800:1:0:CHECK 8 Sep 2020)Manuscript to be reviewed associated with higher mutation rate respond better to nucleoside analogs, such as remdesivir or ribavirin.Dilucca M, Forcelloni S, Georgakilas AG, Giansanti A, Pavlopoulou A. 2020b.Codon Usage and Phenotypic Divergences of SARS-CoV-2 Genes. Viruses 12:498. DOI: 10.3390/v12050498. Eckerle LD, Becker MM, Halpin RA, Li K, Venter E, Lu X, Scherbakova S, Graham RL, Baric RS, Stockwell TB, Spiro DJ, Denison MR. 2010. Infidelity of SARS-CoV Nsp14exonuclease mutant virus replication is revealed by complete genome sequencing. PLoS pathogens 6:e1000896. DOI: 10.1371/journal.ppat.1000896. Elbe S, Buckland-Merrett G. 2017. Data, disease and diplomacy: GISAID's innovative contribution to global health. Global Challenges 1:33-46. DOI: 10.1002/gch2.1018.</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_4'><ns0:head>Figure 1 .</ns0:head><ns0:label>1</ns0:label><ns0:figDesc>Figure 1. The average mutation density per day for genome, S gene, and M and E genes.</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_5'><ns0:head>Figure 2 .</ns0:head><ns0:label>2</ns0:label><ns0:figDesc>Figure 2. The distribution of synonymous and nonsynonymous mutations in isolates</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_6'><ns0:head>(</ns0:head><ns0:label /><ns0:figDesc>A-B) Isolates carrying the synonymous 18060C>T mutation (n = 1585). (C-D) Isolates carrying the synonymous 18877C>T mutation (n = 893). (E-F) Isolates carrying the nonsynonymous 18736C>T mutation (n=236). Wildtype isolates in all graphs carry the reference nucleotide for the nine positions of interest</ns0:figDesc></ns0:figure>
<ns0:figure type='table' xml:id='tab_0'><ns0:head>1 Table 1 .</ns0:head><ns0:label>11</ns0:label><ns0:figDesc>Comparisons of MoE and nsp mutations.</ns0:figDesc><ns0:table><ns0:row><ns0:cell>NSP</ns0:cell><ns0:cell>Mutations</ns0:cell><ns0:cell>Values</ns0:cell><ns0:cell cols='2'>MoE Absent n %</ns0:cell><ns0:cell cols='2'>MoE Present n %</ns0:cell><ns0:cell>n</ns0:cell><ns0:cell>Total</ns0:cell><ns0:cell>%</ns0:cell><ns0:cell>p</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>11916C>T S3884L</ns0:cell><ns0:cell>Absent Present</ns0:cell><ns0:cell>26326 433</ns0:cell><ns0:cell>98.4 1.6</ns0:cell><ns0:cell>2833 8</ns0:cell><ns0:cell>99.7 0.3</ns0:cell><ns0:cell cols='2'>441</ns0:cell><ns0:cell>98.5 1.5</ns0:cell><ns0:cell><0.001*</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>12076C>T N3937N</ns0:cell><ns0:cell>Absent Present</ns0:cell><ns0:cell>26735 24</ns0:cell><ns0:cell>99.9 0.1</ns0:cell><ns0:cell>2837 4</ns0:cell><ns0:cell>99.9 0.1</ns0:cell><ns0:cell>28</ns0:cell><ns0:cell /><ns0:cell>99.9 0.1</ns0:cell><ns0:cell>0.339</ns0:cell></ns0:row><ns0:row><ns0:cell>nsp7</ns0:cell><ns0:cell>11919C>T S3885F</ns0:cell><ns0:cell>Absent Present</ns0:cell><ns0:cell>26738 21</ns0:cell><ns0:cell>99.9 0.1</ns0:cell><ns0:cell cols='2'>2840 100.0 1 -</ns0:cell><ns0:cell>22</ns0:cell><ns0:cell /><ns0:cell>99.9 0.1</ns0:cell><ns0:cell>0.717</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>12073C>T D3936D</ns0:cell><ns0:cell>Absent Present</ns0:cell><ns0:cell cols='3'>26750 100.0 2834 9 -7</ns0:cell><ns0:cell>99.8 0.2</ns0:cell><ns0:cell>16</ns0:cell><ns0:cell /><ns0:cell>99.9 0.1</ns0:cell><ns0:cell><0.001*</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>11962C>T L3899L</ns0:cell><ns0:cell>Absent Present</ns0:cell><ns0:cell cols='4'>26746 100.0 2840 100.0 13 -1 -</ns0:cell><ns0:cell>14</ns0:cell><ns0:cell cols='2'>100.0 -</ns0:cell><ns0:cell>1.000</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>12478G>A M4071I</ns0:cell><ns0:cell>Absent Present</ns0:cell><ns0:cell cols='3'>26757 100.0 2750 2 -91</ns0:cell><ns0:cell>96.8 3.2</ns0:cell><ns0:cell>93</ns0:cell><ns0:cell /><ns0:cell>99.7 0.3</ns0:cell><ns0:cell><0.001*</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>12550G>A L4095L</ns0:cell><ns0:cell>Absent Present</ns0:cell><ns0:cell>26697 62</ns0:cell><ns0:cell>99.8 0.2</ns0:cell><ns0:cell cols='2'>2841 100.0 --</ns0:cell><ns0:cell>62</ns0:cell><ns0:cell /><ns0:cell>99.8 0.2</ns0:cell><ns0:cell>-</ns0:cell></ns0:row><ns0:row><ns0:cell>nsp8</ns0:cell><ns0:cell>12415C>T N4050N</ns0:cell><ns0:cell>Absent Present</ns0:cell><ns0:cell>26725 34</ns0:cell><ns0:cell>99.9 0.1</ns0:cell><ns0:cell cols='2'>2841 100.0 --</ns0:cell><ns0:cell>34</ns0:cell><ns0:cell /><ns0:cell>99.9 0.1</ns0:cell><ns0:cell>-</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>12557A>G I4098V</ns0:cell><ns0:cell>Absent Present</ns0:cell><ns0:cell>26729 30</ns0:cell><ns0:cell>99.9 0.1</ns0:cell><ns0:cell cols='2'>2841 100.0 --</ns0:cell><ns0:cell>30</ns0:cell><ns0:cell /><ns0:cell>99.9 0.1</ns0:cell><ns0:cell>-</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>12400C>T L4045L</ns0:cell><ns0:cell>Absent Present</ns0:cell><ns0:cell>26734 25</ns0:cell><ns0:cell>99.9 0.1</ns0:cell><ns0:cell cols='2'>2840 100.0 1 -</ns0:cell><ns0:cell>26</ns0:cell><ns0:cell /><ns0:cell>99.9 0.1</ns0:cell><ns0:cell>0.508</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>14408C>T P4715L</ns0:cell><ns0:cell cols='2'>Absent Present 19261 7498</ns0:cell><ns0:cell>28.0 72.0</ns0:cell><ns0:cell>702 2139</ns0:cell><ns0:cell>24.7 75.3</ns0:cell><ns0:cell cols='2'>8200</ns0:cell><ns0:cell>27.7 72.3</ns0:cell><ns0:cell><0.001*</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>14805C>T Y4847Y</ns0:cell><ns0:cell>Absent Present</ns0:cell><ns0:cell>24397 2362</ns0:cell><ns0:cell>91.2 8.8</ns0:cell><ns0:cell>2704 137</ns0:cell><ns0:cell>95.2 4.8</ns0:cell><ns0:cell cols='2'>2499</ns0:cell><ns0:cell>91.6 8.4</ns0:cell><ns0:cell><0.001*</ns0:cell></ns0:row><ns0:row><ns0:cell>nsp12</ns0:cell><ns0:cell>13730C>T A4489V</ns0:cell><ns0:cell>Absent Present</ns0:cell><ns0:cell>26238 521</ns0:cell><ns0:cell>98.1 1.9</ns0:cell><ns0:cell>2820 21</ns0:cell><ns0:cell>99.3 0.7</ns0:cell><ns0:cell cols='2'>542</ns0:cell><ns0:cell>98.2 1.8</ns0:cell><ns0:cell><0.001*</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>13536C>T Y4424Y</ns0:cell><ns0:cell>Absent Present</ns0:cell><ns0:cell>26469 290</ns0:cell><ns0:cell>98.9 1.1</ns0:cell><ns0:cell>2823 18</ns0:cell><ns0:cell>99.4 0.6</ns0:cell><ns0:cell cols='2'>308</ns0:cell><ns0:cell>99.0 1.0</ns0:cell><ns0:cell>0.025*</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>13862C>T T4533I</ns0:cell><ns0:cell>Absent Present</ns0:cell><ns0:cell>26535 224</ns0:cell><ns0:cell>99.2 0.8</ns0:cell><ns0:cell>2833 8</ns0:cell><ns0:cell>99.7 0.3</ns0:cell><ns0:cell cols='2'>232</ns0:cell><ns0:cell>99.2 0.8</ns0:cell><ns0:cell>0.001*</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>18060C>T L5932L</ns0:cell><ns0:cell>Absent Present</ns0:cell><ns0:cell>25247 1512</ns0:cell><ns0:cell>94.3 5.7</ns0:cell><ns0:cell>2768 73</ns0:cell><ns0:cell>97.4 2.6</ns0:cell><ns0:cell cols='2'>1585</ns0:cell><ns0:cell>94.6 5.4</ns0:cell><ns0:cell>0.001*</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>18877C>T L6205L</ns0:cell><ns0:cell>Absent Present</ns0:cell><ns0:cell>26185 574</ns0:cell><ns0:cell>97.9 2.1</ns0:cell><ns0:cell>2522 319</ns0:cell><ns0:cell>88.8 11.2</ns0:cell><ns0:cell cols='2'>893</ns0:cell><ns0:cell>97.0 3.0</ns0:cell><ns0:cell>0.001*</ns0:cell></ns0:row><ns0:row><ns0:cell>nsp14</ns0:cell><ns0:cell>18998C>T A6245V</ns0:cell><ns0:cell>Absent Present</ns0:cell><ns0:cell>26454 305</ns0:cell><ns0:cell>98.9 1.1</ns0:cell><ns0:cell>2836 5</ns0:cell><ns0:cell>99.8 0.2</ns0:cell><ns0:cell cols='2'>310</ns0:cell><ns0:cell>99.0 1.0</ns0:cell><ns0:cell>0.001*</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>18736T>C F6158L</ns0:cell><ns0:cell>Absent Present</ns0:cell><ns0:cell cols='3'>26751 100.0 2613 8 -228</ns0:cell><ns0:cell>92.0 8.0</ns0:cell><ns0:cell cols='2'>236</ns0:cell><ns0:cell>99.2 0.8</ns0:cell><ns0:cell>0.001*</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>19524C>T L6420L</ns0:cell><ns0:cell>Absent Present</ns0:cell><ns0:cell>26530 229</ns0:cell><ns0:cell>99.1 0.9</ns0:cell><ns0:cell>2825 16</ns0:cell><ns0:cell>99.4 0.6</ns0:cell><ns0:cell cols='2'>245</ns0:cell><ns0:cell>99.2 0.8</ns0:cell><ns0:cell>0.102</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>Total</ns0:cell><ns0:cell /><ns0:cell cols='4'>26759 100.0 2840 100.0</ns0:cell><ns0:cell /><ns0:cell cols='2'>100.0</ns0:cell></ns0:row></ns0:table></ns0:figure>
<ns0:figure type='table' xml:id='tab_1'><ns0:head>Table 3 .</ns0:head><ns0:label>3</ns0:label><ns0:figDesc>Logistic regression model of MoE and location on single variables. Each location was represented as itself (1) and others (0).</ns0:figDesc><ns0:table><ns0:row><ns0:cell>Locations</ns0:cell><ns0:cell>p</ns0:cell><ns0:cell>OR</ns0:cell><ns0:cell>95% C.I.</ns0:cell></ns0:row><ns0:row><ns0:cell>Asia</ns0:cell><ns0:cell cols='3'><0.001* 1.697 1.513 to 1.903</ns0:cell></ns0:row><ns0:row><ns0:cell>Africa</ns0:cell><ns0:cell>0.937</ns0:cell><ns0:cell cols='2'>1.015 0.703 to 1.465</ns0:cell></ns0:row><ns0:row><ns0:cell>South America</ns0:cell><ns0:cell>0.032*</ns0:cell><ns0:cell cols='2'>0.650 0.439 to 0.963</ns0:cell></ns0:row><ns0:row><ns0:cell>Europe</ns0:cell><ns0:cell cols='3'><0.001* 1.184 1.095 to 1.281</ns0:cell></ns0:row><ns0:row><ns0:cell cols='4'>North America <0.001* 0.589 0.533 to 0.650</ns0:cell></ns0:row><ns0:row><ns0:cell>Oceania</ns0:cell><ns0:cell>0.330</ns0:cell><ns0:cell cols='2'>1.085 0.921 to 1.278</ns0:cell></ns0:row><ns0:row><ns0:cell cols='4'>OR, Odds-Ratio; C.I.: confidence interval, *p-value<0.05 was statistically significant.</ns0:cell></ns0:row></ns0:table><ns0:note>PeerJ reviewing PDF | (2020:08:51800:1:0:CHECK 8 Sep 2020)</ns0:note></ns0:figure>
<ns0:figure type='table' xml:id='tab_2'><ns0:head>Table 4 (on next page)</ns0:head><ns0:label>4</ns0:label><ns0:figDesc>Logistic regression model of MoE on single variables and a final model.</ns0:figDesc><ns0:table><ns0:row><ns0:cell>(Final Model A) Logistic regression model of ten mutations on final model. (Final Model B)</ns0:cell></ns0:row><ns0:row><ns0:cell>Logistic regression model of four mutations and location on final model. OR: Odds-Ratio; C.I.:</ns0:cell></ns0:row><ns0:row><ns0:cell>confidence interval; Multiple logistic regression final model was executed on all these</ns0:cell></ns0:row><ns0:row><ns0:cell>statistically significant variables, included together in the model, and selected with backward</ns0:cell></ns0:row><ns0:row><ns0:cell>stepwise method; *p-value<0.05 was statistically significant.</ns0:cell></ns0:row></ns0:table><ns0:note>PeerJ reviewing PDF | (2020:08:51800:1:0:CHECK 8 Sep 2020)</ns0:note></ns0:figure>
<ns0:figure type='table' xml:id='tab_3'><ns0:head>Table 4 .</ns0:head><ns0:label>4</ns0:label><ns0:figDesc>Logistic regression model of MoE on single variables and a final model. (Final Model A) Logistic regression model of ten 2 mutations on final model. (Final Model B) Logistic regression model of four mutations and location on final model.</ns0:figDesc><ns0:table><ns0:row><ns0:cell cols='4'>1 Single Variables</ns0:cell><ns0:cell /><ns0:cell cols='2'>Final Model A</ns0:cell><ns0:cell /><ns0:cell>Final Model B</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>Mutations</ns0:cell><ns0:cell>p</ns0:cell><ns0:cell>OR</ns0:cell><ns0:cell>95% C.I.</ns0:cell><ns0:cell>p</ns0:cell><ns0:cell>OR</ns0:cell><ns0:cell>95% C.I.</ns0:cell><ns0:cell>p</ns0:cell><ns0:cell>OR</ns0:cell><ns0:cell>95% C.I.</ns0:cell></ns0:row><ns0:row><ns0:cell>Nsp7.11916</ns0:cell><ns0:cell><0.001*</ns0:cell><ns0:cell>0.172</ns0:cell><ns0:cell>0.085 to 0.346</ns0:cell><ns0:cell><0.001*</ns0:cell><ns0:cell>0.180</ns0:cell><ns0:cell>0.089 to 0.363</ns0:cell><ns0:cell>0.001*</ns0:cell><ns0:cell>0.314</ns0:cell><ns0:cell>0.154 to 0.641</ns0:cell></ns0:row><ns0:row><ns0:cell>Nsp7.12076</ns0:cell><ns0:cell>0.403</ns0:cell><ns0:cell>1.571</ns0:cell><ns0:cell>0.545 to 4.530</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell></ns0:row><ns0:row><ns0:cell>Nsp7.11919</ns0:cell><ns0:cell>0.433</ns0:cell><ns0:cell>0.448</ns0:cell><ns0:cell>0.060 to 3.334</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell></ns0:row><ns0:row><ns0:cell>Nsp7.12073</ns0:cell><ns0:cell><0.001*</ns0:cell><ns0:cell>7.341</ns0:cell><ns0:cell>2.732 to 19.728</ns0:cell><ns0:cell><0.001*</ns0:cell><ns0:cell>8.108</ns0:cell><ns0:cell>3.009 to 21.847</ns0:cell><ns0:cell><0.001*</ns0:cell><ns0:cell>9.164</ns0:cell><ns0:cell>3.311 to 25.361</ns0:cell></ns0:row><ns0:row><ns0:cell>Nsp7.11962</ns0:cell><ns0:cell>0.756</ns0:cell><ns0:cell>0.724</ns0:cell><ns0:cell>0.095 to 5.540</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell></ns0:row><ns0:row><ns0:cell>Nsp8.12478</ns0:cell><ns0:cell cols='2'><0.001* 442.707</ns0:cell><ns0:cell>108.996 to 1798.139</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell></ns0:row><ns0:row><ns0:cell>Nsp8.12550</ns0:cell><ns0:cell>0.997</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell></ns0:row><ns0:row><ns0:cell>Nsp8.12415</ns0:cell><ns0:cell>0.998</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell></ns0:row><ns0:row><ns0:cell>Nsp8.12557</ns0:cell><ns0:cell>0.998</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell></ns0:row><ns0:row><ns0:cell>Nsp8.12400</ns0:cell><ns0:cell>0.388</ns0:cell><ns0:cell>0.377</ns0:cell><ns0:cell>0.051 to 2.780</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell></ns0:row><ns0:row><ns0:cell>Nsp12.14408</ns0:cell><ns0:cell><0.001*</ns0:cell><ns0:cell>1.186</ns0:cell><ns0:cell>1.085 to 1.297</ns0:cell><ns0:cell><0.001*</ns0:cell><ns0:cell>1.310</ns0:cell><ns0:cell>1.144 to 1.500</ns0:cell><ns0:cell><0.001*</ns0:cell><ns0:cell>1.662</ns0:cell><ns0:cell>1.435 to 1.926</ns0:cell></ns0:row><ns0:row><ns0:cell>Nsp12.14805</ns0:cell><ns0:cell><0.001*</ns0:cell><ns0:cell>0.523</ns0:cell><ns0:cell>0.439 to 0.625</ns0:cell><ns0:cell>0.007*</ns0:cell><ns0:cell>0.746</ns0:cell><ns0:cell>0.603 to 0.923</ns0:cell><ns0:cell>0.073</ns0:cell><ns0:cell>0.817</ns0:cell><ns0:cell>0.655 to 1.019</ns0:cell></ns0:row><ns0:row><ns0:cell>Nsp12.13730</ns0:cell><ns0:cell><0.001*</ns0:cell><ns0:cell>0.375</ns0:cell><ns0:cell>0.242 to 0.581</ns0:cell><ns0:cell>0.002*</ns0:cell><ns0:cell>0.497</ns0:cell><ns0:cell>0.317 to 0.778</ns0:cell><ns0:cell><0.001*</ns0:cell><ns0:cell>0.393</ns0:cell><ns0:cell>0.250 to 0.619</ns0:cell></ns0:row><ns0:row><ns0:cell>Nsp12.13536</ns0:cell><ns0:cell>0.026*</ns0:cell><ns0:cell>0.582</ns0:cell><ns0:cell>0.361 to 0.938</ns0:cell><ns0:cell>0.044*</ns0:cell><ns0:cell>0.611</ns0:cell><ns0:cell>0.379 to 0.987</ns0:cell><ns0:cell>0.009*</ns0:cell><ns0:cell>0.528</ns0:cell><ns0:cell>0.327 to 0.855</ns0:cell></ns0:row><ns0:row><ns0:cell>Nsp12.13862</ns0:cell><ns0:cell>0.002*</ns0:cell><ns0:cell>0.335</ns0:cell><ns0:cell>0.165 to 0.678</ns0:cell><ns0:cell>0.004*</ns0:cell><ns0:cell>0.355</ns0:cell><ns0:cell>0.175 to 0.720</ns0:cell><ns0:cell>0.001*</ns0:cell><ns0:cell>0.293</ns0:cell><ns0:cell>0.144 to 0.594</ns0:cell></ns0:row><ns0:row><ns0:cell>Nsp14.18060</ns0:cell><ns0:cell><0.001*</ns0:cell><ns0:cell>0.440</ns0:cell><ns0:cell>0.347 to 0.559</ns0:cell><ns0:cell>0.001*</ns0:cell><ns0:cell>0.625</ns0:cell><ns0:cell>0.479 to 0.816</ns0:cell><ns0:cell>0.001*</ns0:cell><ns0:cell>1.658</ns0:cell><ns0:cell>1.244 to 2.209</ns0:cell></ns0:row><ns0:row><ns0:cell>Nsp14.18877</ns0:cell><ns0:cell><0.001*</ns0:cell><ns0:cell>5.770</ns0:cell><ns0:cell>5.002 to 6.656</ns0:cell><ns0:cell><0.001*</ns0:cell><ns0:cell>5.543</ns0:cell><ns0:cell>4.793 to 6.409</ns0:cell><ns0:cell><0.001*</ns0:cell><ns0:cell>6.437</ns0:cell><ns0:cell>5.483 to 7.557</ns0:cell></ns0:row><ns0:row><ns0:cell>Nsp14.18998</ns0:cell><ns0:cell><0.001*</ns0:cell><ns0:cell>0.153</ns0:cell><ns0:cell>0.063 to 0.370</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell></ns0:row><ns0:row><ns0:cell>Nsp14.18736</ns0:cell><ns0:cell cols='2'><0.001* 291.773</ns0:cell><ns0:cell>144.002 to 591.182</ns0:cell><ns0:cell cols='6'><0.001* 368.884 180.195 to 755.153 <0.001* 970.884 469.324 to 2008.453</ns0:cell></ns0:row><ns0:row><ns0:cell>Nsp14.19524</ns0:cell><ns0:cell>0.104</ns0:cell><ns0:cell>0.656</ns0:cell><ns0:cell>0.395 to 1.091</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell></ns0:row><ns0:row><ns0:cell>Location</ns0:cell><ns0:cell><0.001*</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell><0.001*</ns0:cell><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>Africa</ns0:cell><ns0:cell>0.019*</ns0:cell><ns0:cell>0.634</ns0:cell><ns0:cell>0.434-0.927</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>0.017*</ns0:cell><ns0:cell>0.580</ns0:cell><ns0:cell>0.391-0.860</ns0:cell></ns0:row><ns0:row><ns0:cell>South America</ns0:cell><ns0:cell><0.001*</ns0:cell><ns0:cell>0.409</ns0:cell><ns0:cell>0.273-0.612</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell><0.001*</ns0:cell><ns0:cell>0.302</ns0:cell><ns0:cell>0.198-0.461</ns0:cell></ns0:row><ns0:row><ns0:cell>Europe</ns0:cell><ns0:cell><0.001*</ns0:cell><ns0:cell>0.671</ns0:cell><ns0:cell>0.597-0.755</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell><0.001*</ns0:cell><ns0:cell>0.681</ns0:cell><ns0:cell>0.591-0.785</ns0:cell></ns0:row><ns0:row><ns0:cell>North America</ns0:cell><ns0:cell><0.001*</ns0:cell><ns0:cell>0.415</ns0:cell><ns0:cell>0.361-0.477</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell><0.001*</ns0:cell><ns0:cell>0.228</ns0:cell><ns0:cell>0.192-0.271</ns0:cell></ns0:row><ns0:row><ns0:cell>Oceania</ns0:cell><ns0:cell><0.001*</ns0:cell><ns0:cell>0.675</ns0:cell><ns0:cell>0.557-0.817</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell><0.001*</ns0:cell><ns0:cell>0.536</ns0:cell><ns0:cell>0.428-0.670</ns0:cell></ns0:row></ns0:table></ns0:figure>
<ns0:note place='foot'>*p-value<0.05 was statistically significant.</ns0:note>
</ns0:body>
" | "Editor's Decision
Both reviewers are very positive and their suggestions can substantially improve your manuscript. I hope that you can perform necessary changes very quickly and re-submit your paper soon.
Thank you for the opportunity to revise our manuscript. We are delighted with the constructive comments of the reviewers. We have revised the manuscript accordingly. Below is the detailed response to the reviewers’ comments and we hope our manuscript will now be acceptable for publication in PeerJ.
Reviewer 1: Alexandros Georgakilas
Basic reporting
In this bioinformatics work the authors analyze current SARS-COV-2 libraries and identify that the most frequently observed mutations for four proteins in the RNA synthesis complex over time in comparison to wildtype isolates. Specifically, their results suggest mutations in nsp14, an error-correcting exonuclease protein, have the strongest association with increased mutation load. The findings are important but one can wonder the following :
We thank the reviewer for their helpful contributions, and hope that our additions to the manuscript, as well as our responses here, can allay their concerns about the validity and potential impact of our findings.
1. All exonucleases have a high error rate and vary significantly from organism to organism. Therefore one wonders why to expect different in viruses.
DNA and RNA polymerases have high error rates during replication or transcription, but such errors are often corrected via exonucleases, who help reduce such error rate and therefore increase polymerase fidelity. However, many viruses lack such error correction capacity, leading to increased mutation rates over generations that restrict their genome size. Viruses that possess exonuclease-mediated error correction capacity, in the order Nidovirales which includes the coronavirus family, can increase their genome size (Ogando et al., 2019, https://www.frontiersin.org/articles/10.3389/fmicb.2019.01813/full). While diversity generated by increased mutation rate is essential for adaptive evolution and virulence of the viruses, too high mutation rates can be disadvantageous. Therefore, mutation-selection balance is essential for RNA viruses and replication fidelity plays a central role in control of this balance. Based on our previous findings on RdRp mutations showing correlation with increased mutation load over time, we thought it important to study mutations in other crucial members of the RNA synthesis complex, as we described in lines 57-69. For increased clarity, we added the following statements with the accompanying references in the introduction section:
“[…] therefore allowing the SARS-CoV-2 to maintain its large size genome (Subissi et al., 2014; Ma et al., 2015; Ogando et al., 2019; Romano et al., 2020)” and “[…] directly affecting the mutation-selection equilibrium of RNA viruses”.
Moreover, the importance of nsp14 in the evolution and adaptation was recently highlighted by a study where nsp14 was one of the three genes (N and S are the other two) with the highest Codon Adaptation Index (CAI) values. Higher CAI scores indicate that nsp14 has been accumulating preferential mutations that result in better adaptation of SARS-CoV-2 to humans (Dilucca et al. 2020 https://www.biorxiv.org/content/10.1101/2020.05.29.123976v1.full.pdf).
A paragraph that discusses this point has been added in the Discussion section of the revised manuscript.
Which is the mutation rate per Kbp or Mbp?
As described in lines 89-90 of the materials and methods section, the mutation density is calculated as per kilobase of region of interest.
2. How these findings compare with current findings for D614G mutation in SARS-CoV-2 which is becoming infamous for its rising dominance worldwide?
We thank the reviewer for their insightful question. Our previous findings and examination of other SARS-CoV-2 databases revealed that the D614G mutation in the surface glycoprotein is found almost exclusively co-ocurring with the 14408C>T mutation (in RdRp), therefore, it was not feasible to identify its own impact on the increased mutation density separately from the impact of the 14408C>T mutation, which we have already discussed in our study. Furthermore, as the surface glycoprotein is not involved in genome replication, and only indirectly affects replication fidelity, such as through altered immune response, we did not include its correlation with MoE or mutation density in our findings.
In terms of dominance, the three nsp14 mutations are observed in a limited number of isolates, actually 18060C>T mutation has not been observed after 1 July, and only twice after May 27. Also, 18060C>T and 18736T>C mutations are mostly confined to single country (US). These points are discussed in a paragraph added to the Discussion section of the revised manuscript.
Experimental design
Methodologies
The authors in the design and description of their methods need to be more analytical and exact.
I list below some points
'... obtained from the GISAID
77 EpiCoV database (date of accession: 17 June, 2020) were filtered to remove low-coverage or incomplete genomes, aligned against the reference genomic sequence for SARS-CoV-2'.
Which is the reference genomic sequence? How alignment was made.
In general in many cases, their methods section is not complete and self-explanatory.
We have taken the reviewer’s helpful comment into consideration, and revised the “Genome sequence filtering, retrieval, and preprocessing” in the methods section of the revised manuscript.
Validity of the findings
The validity of the findings and results will be better established after the above corrections. In addition, it is not really clear this increased mutation load of it offers to the virus evolutionary advantage also? What about interactions with host?
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4934755/
We thank the reviewer for their suggestion, and have added the following information regarding studies on other coronavirus nsp14 proteins in the discussion section:
“Previous studies on alphacoronavirus nsp14 protein had shown that nsp14, via its exonuclease activity, can modulate host-virus interactions, degrading double-stranded RNA produced during genome replication to suppress immune response, thus increasing viral viability (Becares et al., 2016). SARS-CoV-2 nsp14, due to similar exonuclease activity, is therefore a potential modulator of host interactions, independent of its link to increased mutation load. However, the exact effect of the mutations we identified, two of which are synonymous and may only indirectly affect protein structure, have to be studied experimentally to show any possible changes in viral viability that they might effect.”
Comments for the Author
The authors need to explain better some aspects of their work as described above and provide better solid evidence for the original nature of their findings and comparison with current advances almost every day in the field.
For example, see also for other coronavirus
https://www.pnas.org/content/112/30/9436
Unfortunately, modeling the effect of the mutations on the protein structure and its interactions with host factors, in a manner similar to the study suggested by the reviewer, is not within our capacity to perform in a timely fashion, especially given that two of the mutations are synonymous mutations, and have to be considered in a wider scope, such as folding kinetics, or RNA motifs or tertiary structures. Such experimental findings are outside of the predictive capacity of computational studies.
However, we did make some amendments to the text to highlight the importance and relevance of our findings:
1- The importance of nsp14 and replication fidelity for adaptive evolution is better explained in the texts added to the Introduction and Discussion sections (see above), with references,
2- A reference that provided evidence for preferential accumulation of mutations in SARS-CoV-2 nsp14 has been added, as it supports our finding of certain nsp14 mutations being associated with higher mutation densities.
Reviewer 2
Basic reporting
The COVID-19 pandemic has mobilized researchers worldwide in a somewhat unprecedented way. This article uses currently available literature to provide information that can help elucidate the continuing evolution of the SARS-CoV-2 virus and assessing viral strains' mutability. The paper's structure adheres to the current journal's standards and uses concise and professional language.
Experimental design
The methodology followed throughout the paper is precise, and the consideration of calculating mutation densities for non-synonymous and synonymous mutations separately is a positive feature. Moreover, the idea of geographical location as a factor to consider while studying the association of each mutation with the presence of MoE mutations is an interesting one, but as stated, briefly in the paper itself, sequencing efforts may influence the results.
Validity of the findings
In summary, though, the information produced by the current article is of high scientific interest, and after some minor revisions, the paper is recommended for publication.
We sincerely thank the reviewer for their kind review, and have revised the manuscript as per their suggestions.
Comments for the author
The above mentioned minor revisions include:
Line 40: As of which date are the numbers regarding death and infection relevant? Is it the same as the day the data was obtained from the database?
We added the relevant information in line 40, as follows:
“[…] (as of 22 July 2020) […]”
Lines 132-133: Additional information would be helpful, especially since the N structural protein’s gene is now firstly mentioned in the article. Maybe the authors can specify the differences in the mutational load?
We thank the reviewer for their helpful comment. The mutation load differences for S, M, and E genes were already shown in Figure 1. We further added the same plots for Orf1ab and N genes as Supplementary Figures 1 and 2, respectively, and renumbered the other supplementary figures accordingly, both in the Supplementary Figures material and in the text. We also added the following to line 133:
“[…] as seen in Figure 1 and Supplementary Figures 1 and 2 […]”
The geographical location data could be addressed more in the conclusions segment, while restating the biases they mentioned in the text before.
We thank the reviewer for their suggestion. The skewed geographical distribution of the MoE variable is not something we have studied as part of our hypothesis, but rather merely a confounding factor we sought to eliminate while evaluating the data. As we have only discussed parts of our findings that are relevant to our hypothesis of RNA replication complex mutations being potentially linked to increased mutation load, while the information might be of interest to other researchers, we do not feel they merit further mentioning outside of the results and discussion section.
" | Here is a paper. Please give your review comments after reading it. |
9,830 | "<ns0:abstract xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:p>SARS-CoV-2 is a betacoronavirus responsible for human cases of COVID-19, a pandemic with global impact that first emerged in late 2019. Since then, the viral genome has shown considerable variance as the disease spread across the world, in part due to the zoonotic origins of the virus and the human host adaptation process. As a virus with an RNA genome that codes for its own genomic replication proteins, mutations in these proteins can significantly impact the variance rate of the genome, affecting both the survival and infection rate of the virus, and attempts at combating the disease. In this study, we analyzed the mutation densities of viral isolates carrying frequently observed mutations for four proteins in the RNA synthesis complex over time in comparison to wildtype isolates. Our observations suggest mutations in nsp14, an error-correcting exonuclease protein, have the strongest association with increased mutation load in both regions without selective pressure and across the genome, compared to nsp7, 8, and 12, which form the core polymerase complex. We propose nsp14 as a priority research target for understanding genomic variance rate in SARS-CoV-2 isolates, and nsp14 mutations as potential predictors for high mutability strains.</ns0:p></ns0:div>
</ns0:abstract>
<ns0:body xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:head>Introduction</ns0:head><ns0:p>COVID-19 is an ongoing global pandemic characterized by long-term respiratory system damage in patients, and caused by the SARS-CoV-2 betacoronavirus. It is likely of zoonotic origin, but capable of human-to-human transmission, and since the first observed cases in the Wuhan province of China <ns0:ref type='bibr'>(Chan et al., 2020;</ns0:ref><ns0:ref type='bibr' target='#b5'>Riou & Althaus, 2020)</ns0:ref>, it has infected over 14 million people, with 612,054 recorded deaths (as of 22 July 2020). In addition to its immediate PeerJ reviewing PDF | (2020:08:51800:2:0:NEW 21 Sep 2020)</ns0:p><ns0:p>Manuscript to be reviewed effects on the respiratory system, its long term effects are still being researched, including symptoms such as neuroinvasion <ns0:ref type='bibr'>(Li, Bai & Hashikawa, 2020;</ns0:ref><ns0:ref type='bibr' target='#b11'>Wu et al., 2020</ns0:ref><ns0:ref type='bibr'>), cardiovascular complications (Kochi et al., 2020;</ns0:ref><ns0:ref type='bibr' target='#b14'>Zhu et al., 2020)</ns0:ref>, and gastrointestinal and liver damage <ns0:ref type='bibr'>(Lee, Huo & Huang, 2020;</ns0:ref><ns0:ref type='bibr' target='#b13'>Xu et al., 2020)</ns0:ref>. Due to its high transmissibility, and capacity for asymptomatic transmission <ns0:ref type='bibr' target='#b10'>(Wong et al., 2020)</ns0:ref>, study of COVID-19 and its underlying pathogen remain a high priority. As a result, the high amount of frequently updated data on viral genomes on databases such as GISAID <ns0:ref type='bibr'>(Elbe & Buckland-Merrett, 2017)</ns0:ref> and <ns0:ref type='bibr'>NextStrain (Hadfield et al., 2018)</ns0:ref> provides researchers with invaluable resources to track the evolution of the virus as it spreads across the world. SARS-CoV-2 has a linear, single-stranded RNA genome, and does not depend on host proteins for genomic replication, instead using an RNA synthesis complex formed from nonstructural proteins (nsp) coded by its own genome. Four of the key proteins involved in the complex are nsp7, nsp8, nsp12, and nsp14, all of which are formed from cleavage of the polyprotein Orf1ab into mature peptides. Nsp12, also known as RdRp (RNA-dependent RNA polymerase), is responsible for synthesizing new strands of RNA using the viral genome as a template. Nsp7 and nsp8 act as essential co-factors for the polymerase unit, together creating the core polymerase complex <ns0:ref type='bibr'>(Kirchdoerfer & Ward, 2019;</ns0:ref><ns0:ref type='bibr'>Peng et al., 2020)</ns0:ref>, while nsp14 is an exonuclease which provides error-correcting capability to the RNA synthesis complex, therefore allowing the SARS-CoV-2 to maintain its large size genome <ns0:ref type='bibr' target='#b8'>(Subissi et al., 2014;</ns0:ref><ns0:ref type='bibr'>Ma et al., 2015;</ns0:ref><ns0:ref type='bibr'>Ogando et al., 2019;</ns0:ref><ns0:ref type='bibr' target='#b6'>Romano et al., 2020)</ns0:ref>. Owing to their role in maintaining replication fidelity and directly affecting the mutation-selection equilibrium of RNA viruses, these proteins are key targets of study in understanding the mutation accumulation and adaptive evolution of the virus <ns0:ref type='bibr'>(Eckerle et al., 2010;</ns0:ref><ns0:ref type='bibr'>Peng et al., 2020)</ns0:ref>.</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:08:51800:2:0:NEW 21 Sep 2020)</ns0:p></ns0:div>
<ns0:div><ns0:head>Manuscript to be reviewed</ns0:head><ns0:p>In our previous study, we examined the top 10 most frequent mutations in the SARS-CoV-2 nsp12, and identified that four of them are associated with an increase in mutation density in two genes, the membrane glycoprotein (M) and the envelope glycoprotein (E) (the combination of which is hereafter referred to as MoE, as we previously described), which are not under selective pressure, and mutations in these genes are potential markers of reduced replication fidelity <ns0:ref type='bibr'>(Eskier et al., 2020a)</ns0:ref>. In this study, we follow up on our previous findings and analyze the mutations in nsps 7, 8, and 14, in addition to nsp12, to identify whether the mutations are associated with a nonselective increase in mutation load or not. We then examine whole genome mutation densities in mutant isolates in comparison to wildtype isolates using linear regression models, in order to understand whether the mutations are associated with potential functional impact. Our findings indicate that mutations in nsp14 are most likely to be predictors of accelerated mutation load increase.</ns0:p></ns0:div>
<ns0:div><ns0:head>Materials and Methods</ns0:head></ns0:div>
<ns0:div><ns0:head>Genome sequence filtering, retrieval, and preprocessing</ns0:head><ns0:p>As previously described <ns0:ref type='bibr'>(Eskier et al., 2020a)</ns0:ref>, SARS-CoV-2 isolate genome sequences and the corresponding metadata were obtained from the GISAID EpiCoV database (date of accession: 17 June 2020). We applied further quality filters, including selecting only isolates obtained from human hosts (excluding environmental samples and animal hosts), those sequenced for the full length of the genome (sequence size of 29 kb or greater), and those with high coverage for the reference genome (< 1% N content, < 0.05% unique mutations, no unverified indel mutations).</ns0:p><ns0:p>To ensure alignment accuracy, all nonstandard unverified nucleotide masking was changed to N due to the specifications of the alignment software, using the Linux sed command, and the isolates were aligned against the SARS-CoV-2 reference genome (NCBI Reference Sequence Manuscript to be reviewed NC_045512.2, available at https://www.ncbi.nlm.nih.gov/nuccore/NC_045512.2) using the MAFFT (v7.450) alignment software <ns0:ref type='bibr'>(Katoh et al., 2002)</ns0:ref>, using the parameters outlined in the software manual for aligning closely related viral genomes (available at https://mafft.cbrc.jp/alignment/software/closelyrelatedviralgenomes.html). Variant sites in the isolates were annotated using snp-sites (2.5.1), bcftools (1.10.2), and ANNOVAR (release date 24 October 2019) software <ns0:ref type='bibr' target='#b9'>(Wang, Li & Hakonarson, 2010;</ns0:ref><ns0:ref type='bibr'>Page et al., 2016)</ns0:ref>, to identify whether a given mutation was synonymous or nonsynonymous. In addition, the 5' untranslated region of the genome (bases 1-265) and the 100 nucleotides at the 3' end were removed from the alignment and annotation files due to a high number of gaps and unidentified nucleotides. We further removed any sequences with incomplete sequencing location or date data in order to avoid complications in downstream analyses. Following the filters, 29,600 genomes were used for the analyses.</ns0:p></ns0:div>
<ns0:div><ns0:head>Mutation density calculation</ns0:head><ns0:p>Variants were categorized as synonymous and nonsynonymous following annotation by ANNOVAR, with intergenic or terminal mutations being considered synonymous. Gene mutation densities were calculated separately for synonymous and nonsynonymous mutations, as well as the total of SNVs, for each isolate, using a non-reference nucleotides per kilobase of region metric. Mutation densities were calculated for the combined membrane glycoprotein (M) and envelope glycoprotein (E) genes (MoE), the surface glycoprotein gene (S), and the whole genome.</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:08:51800:2:0:NEW 21 Sep 2020)</ns0:p><ns0:p>Manuscript to be reviewed</ns0:p></ns0:div>
<ns0:div><ns0:head>Statistical Analysis</ns0:head><ns0:p>Descriptive statistics for continuous variable days were calculated with mean, standard deviation, median, and interquartile range. Kolmogorov-Smirnov test was used to check the normality assumption of the continuous variables. In cases of non-normally distributed data, the Wilcoxon rank-sum (Mann-Whitney U) test was performed to determine whether the difference between the two MoE status groups was statistically significant. The Fisher's exact test and the Pearson chi-square test were used for the analysis of categorical variables. The univariate logistic regression method was utilized to assess the mutations associated with MoE status in single variables, and then multiple logistic regression method was performed. The final multiple logistic regression model was executed with the backward stepwise method. The relationship between mutation density and time in isolates with mutations of interest, as well as in the group comprising all isolates, was examined via non-polynomial linear regression model and Spearman's rank correlation. A p-value of less than 0.05 was considered statistically significant.</ns0:p><ns0:p>All statistical analyses were performed using IBM SPSS version 25.0 (Chicago, IL, USA).</ns0:p></ns0:div>
<ns0:div><ns0:head>Results and Discussion</ns0:head></ns0:div>
<ns0:div><ns0:head>Increases in the mutation load of SARS-CoV-2 are unevenly distributed across its genome</ns0:head><ns0:p>To identify the trends in SARS-CoV-2 mutation load over time, we calculated the average mutation density per day for all isolates for whole genome, S gene, and MoE regions, capping outliers at the 95 th and 5 th percentile values to minimize the potential effects of sequencing errors (Fig. <ns0:ref type='figure' target='#fig_7'>1</ns0:ref>). Our results show that both at the genome level and the S gene, a very strong positive correlation between average mutation density and time. In comparison, MoE has a weak positive correlation, with a wider spread of mean density in early and late periods compared to the genome and the S gene. This is consistent with reduced selective pressure on the M and E genes, <ns0:ref type='table'>PDF | (2020:08:51800:2:0:NEW 21 Sep 2020)</ns0:ref> Manuscript to be reviewed as has previously been described <ns0:ref type='bibr'>(Dilucca et al., 2020b)</ns0:ref>. The top nonsynonymous mutation is 23403A>G (in 22271 isolates), responsible for the D614G substitution in the spike protein, followed by the 14408C>T mutation (in 22226 isolates) in the nsp12 region of the Orf1ab gene, causing P323L substitution in the RdRp protein, and the 28144C>T mutation (in 3081 isolates), responsible for the L84S substitution in the Orf8 protein. The most common synonymous mutation is the 8782C>T mutation (in 3047 isolates), and is found on the nsp4 coding region of the Orf1ab gene. For the S gene, the most frequent synonymous mutation is the 23731C>T mutation (in 622 isolates), and the second most common nonsynonymous mutation, after the aforementioned D614G mutation, is 25350C>T (in 215 isolates), responsible for the P1263L substitution. For MoE, the most common synonymous and nonsynonymous mutations are 26735C>T (in 341 isolates) and 27046C>T (in 530 isolates), respectively, both of which are found in the M gene, and the latter of which causes T175M amino acid substitution. Other than the D614G mutation, all of the mentioned mutations are C>T substitutions, the prevalence of which in T-or A-rich regions of the SARS-CoV-2 genome have been previously documented <ns0:ref type='bibr' target='#b7'>(Simmonds, 2020)</ns0:ref>.</ns0:p><ns0:formula xml:id='formula_0'>PeerJ reviewing</ns0:formula></ns0:div>
<ns0:div><ns0:head>Mutations in RNA synthesis complex proteins are associated with higher mutation load</ns0:head><ns0:p>After identifying the increase in mutation load over time, which was more prominent in genes with high functional impact (S, Orf1ab) compared to other structural genes (M, E, N), as seen in Figure <ns0:ref type='figure' target='#fig_7'>1</ns0:ref> and Supplementary Figures <ns0:ref type='figure' target='#fig_8'>1 and 2</ns0:ref>, we sought to examine possible associations of variants in proteins involved in SARS-CoV-2 genome replication with the increase. We first identified the five most frequently observed mutations for nsps 7, 8, 12 (also known as RdRp) and 14, four of the proteins cleaved from the Orf1ab polyprotein and are involved in the RNA polymerization, followed by analyzing the association of each mutation with the presence of Manuscript to be reviewed MoE mutations (hereafter referred to as MoE status) using the chi-square test. 12 out of the 20 mutations were found to have a significant association with MoE status (p-value < 0.05) (Table <ns0:ref type='table'>1</ns0:ref>). Compared to our previous findings on the top 10 nsp12 mutations (Eskier et al. 2020), which was based on an analysis of 11,208 samples as of 5 May 2020, 13536C>T and 13862C>T have increased in rank of appearance, from 6 th and 7 th to 4 th and 5 th , respectively, and decreased in pvalue to show statistically significant associations. In addition, the 13730C>T mutation have increased in rank of appearance from 4 th to 3 rd . Out of the other nsps tested, nsp14 was found to have four significant mutations, while nsp7 had two and nsp8 had one.</ns0:p></ns0:div>
<ns0:div><ns0:head>Effects of geographical location on MoE status</ns0:head><ns0:p>In addition to time and genotype, we also examined the potential association between the location of isolates and MoE status as a possible confounding factor. We first examined whether there is a significant association between location, defined here as continent the isolate was originally obtained, and MoE status. Our results indicate that there is a strong association between location and MoE status, with the highest percentage of MoE present isolates in Asia (14.5%), and the percentage ratio in South America (6.5%) (p-value <0.001). In comparison to our previous findings, South America had a dramatic decrease in MoE present isolate percentage, likely as a result of the increased sequencing efforts (from 118 isolates to 416) removing potential sampling biases or localized founder effects. Africa, Asia, and North America had an increase in MoE present proportion, while Europe, Oceania, and South America showed lowered percentages (Table <ns0:ref type='table'>2</ns0:ref>).</ns0:p><ns0:p>After observing the potential confounding effect of location on MoE status, we sought to understand whether a location is more or less likely to predict MoE status, using a logistic regression model (Table <ns0:ref type='table' target='#tab_1'>3</ns0:ref>). Comparing each individual region (1) to the other five (0), we found Manuscript to be reviewed that Asia, Europe, and North and South America are all possible predictors of MoE status (pvalue < 0.05), with Asia and Europe 1.697 and 1.184 times as likely to be MoE present as the other regions, and North and South America 0.589 and 0.650 times as likely, respectively.</ns0:p><ns0:p>Using these findings, we created different logistic regression models to identify which of the 12 mutations are likely to be independent predictors of MoE status (Table <ns0:ref type='table' target='#tab_2'>4</ns0:ref>). In the single variable model, all 12 mutations we previously identified and location were found to be potential predictors (p-value < 0.05). Forming final models including the 12 mutations (Final Model A) and the mutations as well as locations (Final Model B), we observed that the predictor effect of two of the mutations nsp8 12478G>A and nsp14 18998C>T do not appear to be sufficiently independent of the other mutations in Final Model A. After adding the location variable to the Final Model A, location remains a significant predictor, with all five non-reference locations less likely to predict MoE than Asia, the reference location, and nsp12 14805C>T is found to not have a predictor effect independent of location (p-value = 0.073). Following Final Model B, nine mutations appear to have a significant association with MoE status, independent of other variables: 11916C>T, 12073C>T, 13536C>T, 13730C>T, 13862C>T, 14408C>T, 18060C>T, 18736T>C, and 18877C>T (p-value < 0.05).</ns0:p></ns0:div>
<ns0:div><ns0:head>Nsp14 mutations have significant impact on increased genomic mutation density</ns0:head><ns0:p>We then examined the effects of each mutation on genomic mutation density to see whether the Manuscript to be reviewed which have a significant association with MoE status also show a similar relationship with genomic mutation density (Fig. <ns0:ref type='figure' target='#fig_8'>2</ns0:ref>). 18060C>T (L7L) has the lowest odds ratio for MoE status (Table <ns0:ref type='table' target='#tab_2'>4</ns0:ref>), and while it shows a slower increase in synonymous mutation density compared to wildtype isolates (Fig. <ns0:ref type='figure' target='#fig_8'>2A</ns0:ref>), it has a significant impact on faster mutation density increase in nonsynonymous mutations (Fig. <ns0:ref type='figure' target='#fig_8'>2B</ns0:ref>). In comparison, 18877C>T (L270L) (Fig. <ns0:ref type='figure' target='#fig_8'>2C-D</ns0:ref>) and 18736T>C (F233L) (Fig. <ns0:ref type='figure' target='#fig_8'>2E-F</ns0:ref>) both show a high prediction capacity for MoE and an increased mutation density. In comparison, mutations in nsp7 (Supp. Figs. <ns0:ref type='figure'>3-4</ns0:ref>) and nsp12 (Supp. Figs. <ns0:ref type='figure'>5-8</ns0:ref>)</ns0:p><ns0:p>show much lower impact on altered mutation density increase rate. 12073C>T, an nsp7 mutation, displays high divergence from wildtype isolate patterns; however, its low sample size (n = 16) creates a skewed distribution of isolates across time, complicating any potential inference.</ns0:p></ns0:div>
<ns0:div><ns0:head>Conclusions</ns0:head><ns0:p>Our previous work identified RdRp mutations as contributors to the evolution of the SARS-CoV-2 genome and this study confirmed those findings. Furthermore, we hypothesized that mutations of the other critical components of the viral replication and transcription machinery may have similar effects. Our results implicate nsp14 as a source of increased mutation rate in SARS-CoV-2 genomes. Three of the five most common nsp14 mutations, namely 18060C>T, 18736T>C and 18877C>T are associated with increases in both genome-wide mutational load, as well as MoE status, an alternative indicator of mutational rate and virus evolution. Interestingly all three are located within the ExoN domain, which is responsible for the proofreading activity of nsp14; however, only 18736T>C mutation is non-synonymous (F233L), while 18060C>T and Manuscript to be reviewed 18877C>T are synonymous mutations and therefore, only after functional studies it will be possible to understand their effects on viral replication processes.</ns0:p><ns0:p>The origins and fates of the three nsp14 mutations are also quite different: Being present in the first case detected in the Washington state of the US in mid-January, 18060C>T mutation has been almost completely confined to the US, as 1,657 of 2,007 isolates (82.6%) originating from the US (https://bigd.big.ac.cn/ncov/variation/annotation/variant/18060, accessed 6 September 2020). On the other hand, 18877C>T mutation arising around at the end of January likely in Saudi Arabia and being detected in much less cases (n=893), is still present in many isolates, most frequently in Saudi Arabia (54.1%) and Turkey (37.4%). 18736T>C mutation was first detected in the US at the beginning of March and like the 18060C>T mutation, has almost completely been limited to the US (281/362 or 77.6%). Unlike the other two, this mutation has been detected in only two isolates since 27 May, and not after 1 July 2020. However, it should be noted that 18877C>T mutation arose within the dominant 23403A>G / 14408C>T lineage, while the other two nsp14 mutations are in different lineages. Therefore, dominance or disappearance of different nsp14 mutations may have less to do with these particular mutations and more with the co-mutations. Yet, we cannot rule out possible effects of these nsp14 mutations on the fitness of SARS-CoV-2.</ns0:p><ns0:p>Previous studies on alphacoronavirus nsp14 protein had shown that nsp14, via its exonuclease activity, can modulate host-virus interactions, degrading double-stranded RNA produced during genome replication to suppress immune response, thus increasing viral viability <ns0:ref type='bibr' target='#b0'>(Becares et al., 2016)</ns0:ref>. SARS-CoV-2 nsp14, due to similar exonuclease activity, is therefore a potential modulator of host interactions, independent of its link to increased mutation load. However, the exact effect of the mutations we identified, two of which are synonymous and may only Manuscript to be reviewed indirectly affect protein structure, have to be studied experimentally to show any possible changes in viral property that they might affect. Of note, a recent study where codon usage of SARS-CoV-2 was analyzed in terms of temporal evolution of the virus genome revealed that nsp14 is one of three genes (together with S and N genes) that display the highest Codon Adaptation Index (CAI) values <ns0:ref type='bibr' target='#b3'>(Dilucca et al., 2020a)</ns0:ref>. CAI is a measure of optimal codon usage and indicates how well codons adapt to the host. Based on higher CAI values in nsp14, one could speculate that such mutations have been accumulating preferentially to reach the optimal mutation rate that allows the most advantageous mutation-selection equilibrium for SARS-CoV-2. Indeed, our previous results <ns0:ref type='bibr'>(Eskier et al., 2020b)</ns0:ref> indicated that the mutation densities of SARS-CoV-2 genomes are closely related to the pandemic stage and population dynamics directly affects the average mutational load of the viral genome. During the rapid growth stages, such as those observed in March in the UK and the US, replication fidelity can be traded off to gain higher replication rates and broader mutational diversity. However, mutations in the replication machinery that result in too high mutation rates would likely be detrimental and eliminated. On the other hand, a small percentage of the resulting mutations could possibly be advantageous, including those that could confer resistance to antiviral drugs. So far, we or others have not been able to detect such mutations advantageous for the virus, however, higher mutation rates make appearance of such a mutation more likely.</ns0:p><ns0:p>We believe that the mutations discussed in this study can be of help to future studies, in both fighting the COVID-19 pandemic, and better understanding of how mutations in coronavirus replication proteins can affect viral viability and replication fidelity in hosts. Also, it is yet to be determined whether COVID-19 cases infected with SARS-CoV-2 that has mutation(s) that are Manuscript to be reviewed Manuscript to be reviewed Manuscript to be reviewed OR, Odds-Ratio; C.I.: confidence interval; Multiple logistic regression final model was executed on all these statistically significant variables, included together in the model, and selected with backward stepwise method; *p-value<0.05 was statistically significant.</ns0:p><ns0:note type='other'>Figure Legends</ns0:note><ns0:p>PeerJ reviewing PDF | (2020:08:51800:2:0:NEW 21 Sep 2020)</ns0:p><ns0:note type='other'>Figure 1</ns0:note><ns0:p>The average mutation density per day for genome, S gene, and M and E genes. </ns0:p><ns0:note type='other'>Figure 2</ns0:note><ns0:p>The distribution of synonymous and nonsynonymous mutations in isolates carrying nsp14 mutations compared to wildtype isolates. <ns0:ref type='bibr'>(11916,</ns0:ref><ns0:ref type='bibr'>12073,</ns0:ref><ns0:ref type='bibr'>13536,</ns0:ref><ns0:ref type='bibr'>13730,</ns0:ref><ns0:ref type='bibr'>13862,</ns0:ref><ns0:ref type='bibr'>14408,</ns0:ref><ns0:ref type='bibr'>18060,</ns0:ref><ns0:ref type='bibr'>18736,</ns0:ref><ns0:ref type='bibr'>18877)</ns0:ref> (n = 5910). Correlation scores are calculated using Spearman rank correlation.</ns0:p></ns0:div><ns0:figure xml:id='fig_0'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:08:51800:2:0:NEW 21 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_1'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:08:51800:2:0:NEW 21 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_2'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:08:51800:2:0:NEW 21 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_3'><ns0:head /><ns0:label /><ns0:figDesc>relationship between the mutations and MoE status are indicative of a genome-wide trend. Due to selection potentially effecting nonsynonymous mutations differentially, we separated the mutations in the two categories and calculated mutation density separately for each category. Our results show that nsp14 mutations show the most consistent association with mutations between MoE and the whole genome. All three nsp14 mutations (18060C>T, 18736T>C, and 18877C>T) PeerJ reviewing PDF | (2020:08:51800:2:0:NEW 21 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_4'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:08:51800:2:0:NEW 21 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_5'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:08:51800:2:0:NEW 21 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_6'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:08:51800:2:0:NEW 21 Sep 2020)Manuscript to be reviewed associated with higher mutation rate respond better to nucleoside analogs, such as remdesivir or ribavirin.Dilucca M, Forcelloni S, Georgakilas AG, Giansanti A, Pavlopoulou A. 2020b.Codon Usage and Phenotypic Divergences of SARS-CoV-2 Genes. Viruses 12:498. DOI: 10.3390/v12050498. Eckerle LD, Becker MM, Halpin RA, Li K, Venter E, Lu X, Scherbakova S, Graham RL, Baric RS, Stockwell TB, Spiro DJ, Denison MR. 2010. Infidelity of SARS-CoV Nsp14exonuclease mutant virus replication is revealed by complete genome sequencing. PLoS pathogens 6:e1000896. DOI: 10.1371/journal.ppat.1000896. Elbe S, Buckland-Merrett G. 2017. Data, disease and diplomacy: GISAID's innovative contribution to global health. Global Challenges 1:33-46. DOI: 10.1002/gch2.1018.</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_7'><ns0:head>Figure 1 .</ns0:head><ns0:label>1</ns0:label><ns0:figDesc>Figure 1. The average mutation density per day for genome, S gene, and M and E genes.</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_8'><ns0:head>Figure 2 .</ns0:head><ns0:label>2</ns0:label><ns0:figDesc>Figure 2. The distribution of synonymous and nonsynonymous mutations in isolates</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_9'><ns0:head>(</ns0:head><ns0:label /><ns0:figDesc>A) The mutation density vs. time for the whole SARS-CoV-2 genome. (B) The mutation density vs. time for the S gene. (C) The combined mutation density vs. time for the M and E genes. Values in y-axis represent the average number of SNVs in the corresponding day, normalized by kilobase of region of interest. SNV counts of genomes are normalized by capping at the 25-and 75-percentile values to minimize the effects of potential sequencing or assembly artifacts. Correlation scores are calculated using Spearman rank correlation.</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_10'><ns0:head>(</ns0:head><ns0:label /><ns0:figDesc>A-B) Isolates carrying the synonymous 18060C>T mutation (n = 1585). (C-D) Isolates carrying the synonymous 18877C>T mutation (n = 893). (E-F) Isolates carrying the nonsynonymous 18736T>C mutation (n=236). Wildtype isolates in all graphs carry the reference nucleotide for the nine positions of interest</ns0:figDesc></ns0:figure>
<ns0:figure type='table' xml:id='tab_0'><ns0:head>1 Table 1 .</ns0:head><ns0:label>11</ns0:label><ns0:figDesc>Comparisons of MoE and nsp mutations.</ns0:figDesc><ns0:table><ns0:row><ns0:cell>NSP</ns0:cell><ns0:cell>Mutations</ns0:cell><ns0:cell>Values</ns0:cell><ns0:cell cols='2'>MoE Absent n %</ns0:cell><ns0:cell cols='2'>MoE Present n %</ns0:cell><ns0:cell>n</ns0:cell><ns0:cell>Total</ns0:cell><ns0:cell>%</ns0:cell><ns0:cell>p</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>11916C>T S3884L</ns0:cell><ns0:cell>Absent Present</ns0:cell><ns0:cell>26326 433</ns0:cell><ns0:cell>98.4 1.6</ns0:cell><ns0:cell>2833 8</ns0:cell><ns0:cell>99.7 0.3</ns0:cell><ns0:cell cols='2'>441</ns0:cell><ns0:cell>98.5 1.5</ns0:cell><ns0:cell><0.001*</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>12076C>T N3937N</ns0:cell><ns0:cell>Absent Present</ns0:cell><ns0:cell>26735 24</ns0:cell><ns0:cell>99.9 0.1</ns0:cell><ns0:cell>2837 4</ns0:cell><ns0:cell>99.9 0.1</ns0:cell><ns0:cell>28</ns0:cell><ns0:cell /><ns0:cell>99.9 0.1</ns0:cell><ns0:cell>0.339</ns0:cell></ns0:row><ns0:row><ns0:cell>nsp7</ns0:cell><ns0:cell>11919C>T S3885F</ns0:cell><ns0:cell>Absent Present</ns0:cell><ns0:cell>26738 21</ns0:cell><ns0:cell>99.9 0.1</ns0:cell><ns0:cell cols='2'>2840 100.0 1 -</ns0:cell><ns0:cell>22</ns0:cell><ns0:cell /><ns0:cell>99.9 0.1</ns0:cell><ns0:cell>0.717</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>12073C>T D3936D</ns0:cell><ns0:cell>Absent Present</ns0:cell><ns0:cell cols='3'>26750 100.0 2834 9 -7</ns0:cell><ns0:cell>99.8 0.2</ns0:cell><ns0:cell>16</ns0:cell><ns0:cell /><ns0:cell>99.9 0.1</ns0:cell><ns0:cell><0.001*</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>11962C>T L3899L</ns0:cell><ns0:cell>Absent Present</ns0:cell><ns0:cell cols='4'>26746 100.0 2840 100.0 13 -1 -</ns0:cell><ns0:cell>14</ns0:cell><ns0:cell cols='2'>100.0 -</ns0:cell><ns0:cell>1.000</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>12478G>A M4071I</ns0:cell><ns0:cell>Absent Present</ns0:cell><ns0:cell cols='3'>26757 100.0 2750 2 -91</ns0:cell><ns0:cell>96.8 3.2</ns0:cell><ns0:cell>93</ns0:cell><ns0:cell /><ns0:cell>99.7 0.3</ns0:cell><ns0:cell><0.001*</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>12550G>A L4095L</ns0:cell><ns0:cell>Absent Present</ns0:cell><ns0:cell>26697 62</ns0:cell><ns0:cell>99.8 0.2</ns0:cell><ns0:cell cols='2'>2841 100.0 --</ns0:cell><ns0:cell>62</ns0:cell><ns0:cell /><ns0:cell>99.8 0.2</ns0:cell><ns0:cell>-</ns0:cell></ns0:row><ns0:row><ns0:cell>nsp8</ns0:cell><ns0:cell>12415C>T N4050N</ns0:cell><ns0:cell>Absent Present</ns0:cell><ns0:cell>26725 34</ns0:cell><ns0:cell>99.9 0.1</ns0:cell><ns0:cell cols='2'>2841 100.0 --</ns0:cell><ns0:cell>34</ns0:cell><ns0:cell /><ns0:cell>99.9 0.1</ns0:cell><ns0:cell>-</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>12557A>G I4098V</ns0:cell><ns0:cell>Absent Present</ns0:cell><ns0:cell>26729 30</ns0:cell><ns0:cell>99.9 0.1</ns0:cell><ns0:cell cols='2'>2841 100.0 --</ns0:cell><ns0:cell>30</ns0:cell><ns0:cell /><ns0:cell>99.9 0.1</ns0:cell><ns0:cell>-</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>12400C>T L4045L</ns0:cell><ns0:cell>Absent Present</ns0:cell><ns0:cell>26734 25</ns0:cell><ns0:cell>99.9 0.1</ns0:cell><ns0:cell cols='2'>2840 100.0 1 -</ns0:cell><ns0:cell>26</ns0:cell><ns0:cell /><ns0:cell>99.9 0.1</ns0:cell><ns0:cell>0.508</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>14408C>T P4715L</ns0:cell><ns0:cell cols='2'>Absent Present 19261 7498</ns0:cell><ns0:cell>28.0 72.0</ns0:cell><ns0:cell>702 2139</ns0:cell><ns0:cell>24.7 75.3</ns0:cell><ns0:cell cols='2'>8200</ns0:cell><ns0:cell>27.7 72.3</ns0:cell><ns0:cell><0.001*</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>14805C>T Y4847Y</ns0:cell><ns0:cell>Absent Present</ns0:cell><ns0:cell>24397 2362</ns0:cell><ns0:cell>91.2 8.8</ns0:cell><ns0:cell>2704 137</ns0:cell><ns0:cell>95.2 4.8</ns0:cell><ns0:cell cols='2'>2499</ns0:cell><ns0:cell>91.6 8.4</ns0:cell><ns0:cell><0.001*</ns0:cell></ns0:row><ns0:row><ns0:cell>nsp12</ns0:cell><ns0:cell>13730C>T A4489V</ns0:cell><ns0:cell>Absent Present</ns0:cell><ns0:cell>26238 521</ns0:cell><ns0:cell>98.1 1.9</ns0:cell><ns0:cell>2820 21</ns0:cell><ns0:cell>99.3 0.7</ns0:cell><ns0:cell cols='2'>542</ns0:cell><ns0:cell>98.2 1.8</ns0:cell><ns0:cell><0.001*</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>13536C>T Y4424Y</ns0:cell><ns0:cell>Absent Present</ns0:cell><ns0:cell>26469 290</ns0:cell><ns0:cell>98.9 1.1</ns0:cell><ns0:cell>2823 18</ns0:cell><ns0:cell>99.4 0.6</ns0:cell><ns0:cell cols='2'>308</ns0:cell><ns0:cell>99.0 1.0</ns0:cell><ns0:cell>0.025*</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>13862C>T T4533I</ns0:cell><ns0:cell>Absent Present</ns0:cell><ns0:cell>26535 224</ns0:cell><ns0:cell>99.2 0.8</ns0:cell><ns0:cell>2833 8</ns0:cell><ns0:cell>99.7 0.3</ns0:cell><ns0:cell cols='2'>232</ns0:cell><ns0:cell>99.2 0.8</ns0:cell><ns0:cell>0.001*</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>18060C>T L5932L</ns0:cell><ns0:cell>Absent Present</ns0:cell><ns0:cell>25247 1512</ns0:cell><ns0:cell>94.3 5.7</ns0:cell><ns0:cell>2768 73</ns0:cell><ns0:cell>97.4 2.6</ns0:cell><ns0:cell cols='2'>1585</ns0:cell><ns0:cell>94.6 5.4</ns0:cell><ns0:cell>0.001*</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>18877C>T L6205L</ns0:cell><ns0:cell>Absent Present</ns0:cell><ns0:cell>26185 574</ns0:cell><ns0:cell>97.9 2.1</ns0:cell><ns0:cell>2522 319</ns0:cell><ns0:cell>88.8 11.2</ns0:cell><ns0:cell cols='2'>893</ns0:cell><ns0:cell>97.0 3.0</ns0:cell><ns0:cell>0.001*</ns0:cell></ns0:row><ns0:row><ns0:cell>nsp14</ns0:cell><ns0:cell>18998C>T A6245V</ns0:cell><ns0:cell>Absent Present</ns0:cell><ns0:cell>26454 305</ns0:cell><ns0:cell>98.9 1.1</ns0:cell><ns0:cell>2836 5</ns0:cell><ns0:cell>99.8 0.2</ns0:cell><ns0:cell cols='2'>310</ns0:cell><ns0:cell>99.0 1.0</ns0:cell><ns0:cell>0.001*</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>18736T>C F6158L</ns0:cell><ns0:cell>Absent Present</ns0:cell><ns0:cell cols='3'>26751 100.0 2613 8 -228</ns0:cell><ns0:cell>92.0 8.0</ns0:cell><ns0:cell cols='2'>236</ns0:cell><ns0:cell>99.2 0.8</ns0:cell><ns0:cell>0.001*</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>19524C>T L6420L</ns0:cell><ns0:cell>Absent Present</ns0:cell><ns0:cell>26530 229</ns0:cell><ns0:cell>99.1 0.9</ns0:cell><ns0:cell>2825 16</ns0:cell><ns0:cell>99.4 0.6</ns0:cell><ns0:cell cols='2'>245</ns0:cell><ns0:cell>99.2 0.8</ns0:cell><ns0:cell>0.102</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>Total</ns0:cell><ns0:cell /><ns0:cell cols='4'>26759 100.0 2840 100.0</ns0:cell><ns0:cell /><ns0:cell cols='2'>100.0</ns0:cell></ns0:row></ns0:table></ns0:figure>
<ns0:figure type='table' xml:id='tab_1'><ns0:head>Table 3 .</ns0:head><ns0:label>3</ns0:label><ns0:figDesc>Logistic regression model of MoE and location on single variables. Each location was represented as itself (1) and others (0).</ns0:figDesc><ns0:table><ns0:row><ns0:cell>Locations</ns0:cell><ns0:cell>p</ns0:cell><ns0:cell>OR</ns0:cell><ns0:cell>95% C.I.</ns0:cell></ns0:row><ns0:row><ns0:cell>Asia</ns0:cell><ns0:cell cols='3'><0.001* 1.697 1.513 to 1.903</ns0:cell></ns0:row><ns0:row><ns0:cell>Africa</ns0:cell><ns0:cell>0.937</ns0:cell><ns0:cell cols='2'>1.015 0.703 to 1.465</ns0:cell></ns0:row><ns0:row><ns0:cell>South America</ns0:cell><ns0:cell>0.032*</ns0:cell><ns0:cell cols='2'>0.650 0.439 to 0.963</ns0:cell></ns0:row><ns0:row><ns0:cell>Europe</ns0:cell><ns0:cell cols='3'><0.001* 1.184 1.095 to 1.281</ns0:cell></ns0:row><ns0:row><ns0:cell cols='4'>North America <0.001* 0.589 0.533 to 0.650</ns0:cell></ns0:row><ns0:row><ns0:cell>Oceania</ns0:cell><ns0:cell>0.330</ns0:cell><ns0:cell cols='2'>1.085 0.921 to 1.278</ns0:cell></ns0:row><ns0:row><ns0:cell cols='4'>OR, Odds-Ratio; C.I.: confidence interval, *p-value<0.05 was statistically significant.</ns0:cell></ns0:row></ns0:table><ns0:note>PeerJ reviewing PDF | (2020:08:51800:2:0:NEW 21 Sep 2020)</ns0:note></ns0:figure>
<ns0:figure type='table' xml:id='tab_2'><ns0:head>Table 4 (on next page)</ns0:head><ns0:label>4</ns0:label><ns0:figDesc>Logistic regression model of MoE on single variables and a final model.</ns0:figDesc><ns0:table><ns0:row><ns0:cell>(Final Model A) Logistic regression model of ten mutations on final model. (Final Model B)</ns0:cell></ns0:row><ns0:row><ns0:cell>Logistic regression model of four mutations and location on final model. OR: Odds-Ratio; C.I.:</ns0:cell></ns0:row><ns0:row><ns0:cell>confidence interval; Multiple logistic regression final model was executed on all these</ns0:cell></ns0:row><ns0:row><ns0:cell>statistically significant variables, included together in the model, and selected with backward</ns0:cell></ns0:row><ns0:row><ns0:cell>stepwise method; *p-value<0.05 was statistically significant.</ns0:cell></ns0:row></ns0:table><ns0:note>PeerJ reviewing PDF | (2020:08:51800:2:0:NEW 21 Sep 2020)</ns0:note></ns0:figure>
<ns0:figure type='table' xml:id='tab_3'><ns0:head>Table 4 .</ns0:head><ns0:label>4</ns0:label><ns0:figDesc>Logistic regression model of MoE on single variables and a final model. (Final Model A) Logistic regression model of ten 2 mutations on final model. (Final Model B) Logistic regression model of four mutations and location on final model.</ns0:figDesc><ns0:table><ns0:row><ns0:cell cols='4'>1 Single Variables</ns0:cell><ns0:cell /><ns0:cell cols='2'>Final Model A</ns0:cell><ns0:cell /><ns0:cell>Final Model B</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>Mutations</ns0:cell><ns0:cell>p</ns0:cell><ns0:cell>OR</ns0:cell><ns0:cell>95% C.I.</ns0:cell><ns0:cell>p</ns0:cell><ns0:cell>OR</ns0:cell><ns0:cell>95% C.I.</ns0:cell><ns0:cell>p</ns0:cell><ns0:cell>OR</ns0:cell><ns0:cell>95% C.I.</ns0:cell></ns0:row><ns0:row><ns0:cell>Nsp7.11916</ns0:cell><ns0:cell><0.001*</ns0:cell><ns0:cell>0.172</ns0:cell><ns0:cell>0.085 to 0.346</ns0:cell><ns0:cell><0.001*</ns0:cell><ns0:cell>0.180</ns0:cell><ns0:cell>0.089 to 0.363</ns0:cell><ns0:cell>0.001*</ns0:cell><ns0:cell>0.314</ns0:cell><ns0:cell>0.154 to 0.641</ns0:cell></ns0:row><ns0:row><ns0:cell>Nsp7.12076</ns0:cell><ns0:cell>0.403</ns0:cell><ns0:cell>1.571</ns0:cell><ns0:cell>0.545 to 4.530</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell></ns0:row><ns0:row><ns0:cell>Nsp7.11919</ns0:cell><ns0:cell>0.433</ns0:cell><ns0:cell>0.448</ns0:cell><ns0:cell>0.060 to 3.334</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell></ns0:row><ns0:row><ns0:cell>Nsp7.12073</ns0:cell><ns0:cell><0.001*</ns0:cell><ns0:cell>7.341</ns0:cell><ns0:cell>2.732 to 19.728</ns0:cell><ns0:cell><0.001*</ns0:cell><ns0:cell>8.108</ns0:cell><ns0:cell>3.009 to 21.847</ns0:cell><ns0:cell><0.001*</ns0:cell><ns0:cell>9.164</ns0:cell><ns0:cell>3.311 to 25.361</ns0:cell></ns0:row><ns0:row><ns0:cell>Nsp7.11962</ns0:cell><ns0:cell>0.756</ns0:cell><ns0:cell>0.724</ns0:cell><ns0:cell>0.095 to 5.540</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell></ns0:row><ns0:row><ns0:cell>Nsp8.12478</ns0:cell><ns0:cell cols='2'><0.001* 442.707</ns0:cell><ns0:cell>108.996 to 1798.139</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell></ns0:row><ns0:row><ns0:cell>Nsp8.12550</ns0:cell><ns0:cell>0.997</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell></ns0:row><ns0:row><ns0:cell>Nsp8.12415</ns0:cell><ns0:cell>0.998</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell></ns0:row><ns0:row><ns0:cell>Nsp8.12557</ns0:cell><ns0:cell>0.998</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell></ns0:row><ns0:row><ns0:cell>Nsp8.12400</ns0:cell><ns0:cell>0.388</ns0:cell><ns0:cell>0.377</ns0:cell><ns0:cell>0.051 to 2.780</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell></ns0:row><ns0:row><ns0:cell>Nsp12.14408</ns0:cell><ns0:cell><0.001*</ns0:cell><ns0:cell>1.186</ns0:cell><ns0:cell>1.085 to 1.297</ns0:cell><ns0:cell><0.001*</ns0:cell><ns0:cell>1.310</ns0:cell><ns0:cell>1.144 to 1.500</ns0:cell><ns0:cell><0.001*</ns0:cell><ns0:cell>1.662</ns0:cell><ns0:cell>1.435 to 1.926</ns0:cell></ns0:row><ns0:row><ns0:cell>Nsp12.14805</ns0:cell><ns0:cell><0.001*</ns0:cell><ns0:cell>0.523</ns0:cell><ns0:cell>0.439 to 0.625</ns0:cell><ns0:cell>0.007*</ns0:cell><ns0:cell>0.746</ns0:cell><ns0:cell>0.603 to 0.923</ns0:cell><ns0:cell>0.073</ns0:cell><ns0:cell>0.817</ns0:cell><ns0:cell>0.655 to 1.019</ns0:cell></ns0:row><ns0:row><ns0:cell>Nsp12.13730</ns0:cell><ns0:cell><0.001*</ns0:cell><ns0:cell>0.375</ns0:cell><ns0:cell>0.242 to 0.581</ns0:cell><ns0:cell>0.002*</ns0:cell><ns0:cell>0.497</ns0:cell><ns0:cell>0.317 to 0.778</ns0:cell><ns0:cell><0.001*</ns0:cell><ns0:cell>0.393</ns0:cell><ns0:cell>0.250 to 0.619</ns0:cell></ns0:row><ns0:row><ns0:cell>Nsp12.13536</ns0:cell><ns0:cell>0.026*</ns0:cell><ns0:cell>0.582</ns0:cell><ns0:cell>0.361 to 0.938</ns0:cell><ns0:cell>0.044*</ns0:cell><ns0:cell>0.611</ns0:cell><ns0:cell>0.379 to 0.987</ns0:cell><ns0:cell>0.009*</ns0:cell><ns0:cell>0.528</ns0:cell><ns0:cell>0.327 to 0.855</ns0:cell></ns0:row><ns0:row><ns0:cell>Nsp12.13862</ns0:cell><ns0:cell>0.002*</ns0:cell><ns0:cell>0.335</ns0:cell><ns0:cell>0.165 to 0.678</ns0:cell><ns0:cell>0.004*</ns0:cell><ns0:cell>0.355</ns0:cell><ns0:cell>0.175 to 0.720</ns0:cell><ns0:cell>0.001*</ns0:cell><ns0:cell>0.293</ns0:cell><ns0:cell>0.144 to 0.594</ns0:cell></ns0:row><ns0:row><ns0:cell>Nsp14.18060</ns0:cell><ns0:cell><0.001*</ns0:cell><ns0:cell>0.440</ns0:cell><ns0:cell>0.347 to 0.559</ns0:cell><ns0:cell>0.001*</ns0:cell><ns0:cell>0.625</ns0:cell><ns0:cell>0.479 to 0.816</ns0:cell><ns0:cell>0.001*</ns0:cell><ns0:cell>1.658</ns0:cell><ns0:cell>1.244 to 2.209</ns0:cell></ns0:row><ns0:row><ns0:cell>Nsp14.18877</ns0:cell><ns0:cell><0.001*</ns0:cell><ns0:cell>5.770</ns0:cell><ns0:cell>5.002 to 6.656</ns0:cell><ns0:cell><0.001*</ns0:cell><ns0:cell>5.543</ns0:cell><ns0:cell>4.793 to 6.409</ns0:cell><ns0:cell><0.001*</ns0:cell><ns0:cell>6.437</ns0:cell><ns0:cell>5.483 to 7.557</ns0:cell></ns0:row><ns0:row><ns0:cell>Nsp14.18998</ns0:cell><ns0:cell><0.001*</ns0:cell><ns0:cell>0.153</ns0:cell><ns0:cell>0.063 to 0.370</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell></ns0:row><ns0:row><ns0:cell>Nsp14.18736</ns0:cell><ns0:cell cols='2'><0.001* 291.773</ns0:cell><ns0:cell>144.002 to 591.182</ns0:cell><ns0:cell cols='6'><0.001* 368.884 180.195 to 755.153 <0.001* 970.884 469.324 to 2008.453</ns0:cell></ns0:row><ns0:row><ns0:cell>Nsp14.19524</ns0:cell><ns0:cell>0.104</ns0:cell><ns0:cell>0.656</ns0:cell><ns0:cell>0.395 to 1.091</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell></ns0:row><ns0:row><ns0:cell>Location</ns0:cell><ns0:cell><0.001*</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell><0.001*</ns0:cell><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>Africa</ns0:cell><ns0:cell>0.019*</ns0:cell><ns0:cell>0.634</ns0:cell><ns0:cell>0.434-0.927</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>0.017*</ns0:cell><ns0:cell>0.580</ns0:cell><ns0:cell>0.391-0.860</ns0:cell></ns0:row><ns0:row><ns0:cell>South America</ns0:cell><ns0:cell><0.001*</ns0:cell><ns0:cell>0.409</ns0:cell><ns0:cell>0.273-0.612</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell><0.001*</ns0:cell><ns0:cell>0.302</ns0:cell><ns0:cell>0.198-0.461</ns0:cell></ns0:row><ns0:row><ns0:cell>Europe</ns0:cell><ns0:cell><0.001*</ns0:cell><ns0:cell>0.671</ns0:cell><ns0:cell>0.597-0.755</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell><0.001*</ns0:cell><ns0:cell>0.681</ns0:cell><ns0:cell>0.591-0.785</ns0:cell></ns0:row><ns0:row><ns0:cell>North America</ns0:cell><ns0:cell><0.001*</ns0:cell><ns0:cell>0.415</ns0:cell><ns0:cell>0.361-0.477</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell><0.001*</ns0:cell><ns0:cell>0.228</ns0:cell><ns0:cell>0.192-0.271</ns0:cell></ns0:row><ns0:row><ns0:cell>Oceania</ns0:cell><ns0:cell><0.001*</ns0:cell><ns0:cell>0.675</ns0:cell><ns0:cell>0.557-0.817</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell><0.001*</ns0:cell><ns0:cell>0.536</ns0:cell><ns0:cell>0.428-0.670</ns0:cell></ns0:row></ns0:table><ns0:note>PeerJ reviewing PDF | (2020:08:51800:2:0:NEW 21 Sep 2020)</ns0:note></ns0:figure>
<ns0:note place='foot'>*p-value<0.05 was statistically significant.</ns0:note>
<ns0:note place='foot'>Manuscript to be reviewed</ns0:note>
</ns0:body>
" | "Editor's Decision
Thank you again for your submission to PeerJ. I think that you will be able to perform these very minor changes quickly.
We are grateful to the editor and the reviewers for their time and work, and hope our revisions will have brought our manuscript to a level of standards worthy of a publication such as PeerJ.
1) Improve / extend Figure legends
We thank the editor for this comment. We have now improved the figure legend in the revised manuscript.
2) Replace the mentioned by the reviewer reference to this one:
Dilucca M, Forcelloni S, Georgakilas AG, Giansanti A, Pavlopoulou A.
Codon Usage and Phenotypic Divergences of SARS-CoV-2 Genes.
Viruses. 2020 Apr 30;12(5):498.
We understand the reviewer’s concern. However, the study that we cited in our submission
(Dilucca M, Forcelloni S, Georgakilas AG, Giansanti A, Pavlopoulou A. 2020a. Temporal evolution and adaptation of SARS-COV 2 codon usage. bioRxiv:2020.05.29.123976. DOI: 10.1101/2020.05.29.123976.)
has not been published yet as a peer reviewed article. The article suggested above has already been cited in our manuscript.
[# PeerJ Staff Note: It is PeerJ policy that additional references suggested during the peer-review process should only be included if the authors are in agreement that they are relevant and useful #]
Reviewer: Alexandros Georgakilas
Basic reporting
The revised manuscript stands much better and the authors have addressed almost all of my concerns. The interaction with host importance , I think it is underestimated.
Experimental design
The design and description is no better and more concise.
Validity of the findings
The findings are valid.
Comments for the author
The revisions have improved the quality and clarity of the manuscript. I am still though sceptical on the importance of their findings regarding the interaction with host.
We thank the reviewer for their previous suggestions, and are glad that our improvements have been satisfactory.
In addition, the Figure legends are way too small.
We emphatically agree with the reviewer’s concern on the legend for Figure 1, and have therefore expanded it as follows:
“Figure 1. The average mutation density per day for genome, S gene, and M and E genes. (A) The mutation density vs. time for the whole SARS-CoV-2 genome. (B) The mutation density vs. time for the S gene. (C) The combined mutation density vs. time for the M and E genes. Values in y-axis represent the average number of SNVs in the corresponding day, normalized by kilobase of region of interest. SNV counts of genomes are normalized by capping at the 25- and 75-percentile values to minimize the effects of potential sequencing or assembly artifacts. Correlation scores are calculated using Spearman rank correlation.”
Iam not sure also of this reference can be accepted as a regular citing document since it is in the bioRxiv.
Dilucca M, Forcelloni S, Georgakilas AG, Giansanti A, Pavlopoulou A. 2020a. Temporal evolution and adaptation of SARS-COV 2 codon usage. bioRxiv:2020.05.29.123976. DOI: 10.1101/2020.05.29.123976.
While we understand the reviewer’s concern, we believe that it is noteworthy to cite potential results in preprints in a topic as rapidly evolving such as this, unless they have already been refuted.
Reviewer: Dimitrios Vlachakis
Basic reporting
The authors of the article 'Mutations of SARS-CoV-2 nsp14 exhibit strong association with increased genome-wide mutation load' have made a significant effort to address the comments and revisions I suggested.
Experimental design
Experimental design is scientifically sound
Validity of the findings
Validity of the findings is OK
Comments for the author
I think that the article 'Mutations of SARS-CoV-2 nsp14 exhibit strong association with increased genome-wide mutation load' has been significantly improved and therefore is ready to be published
We thank the reviewer for their kind comments, as well as their time in reviewing our work.
" | Here is a paper. Please give your review comments after reading it. |
9,831 | "<ns0:abstract xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:p>Background: This study aimed to systematically profile the alterations and sex-and age-related differences in the drug metabolizing enzymes (DMEs) in a KRAS-mutant mouse model of lung cancer (KRAS mice).</ns0:p></ns0:div>
<ns0:div><ns0:head>Methodology:</ns0:head><ns0:p>In this study, the LC-MS/MS approach and a probe substrate method were used to detect the alterations in 21 isoforms of DMEs, as well as the enzymatic activities of five isoforms, respectively. Western blotting was applied to study the protein expression of four related receptors.</ns0:p></ns0:div>
<ns0:div><ns0:head>Results:</ns0:head><ns0:p>The proteins contents of CYP2C29 and CYP3A11, were significantly downregulated in the livers of male KRAS mice at 26 weeks (3.7-and 4.4-fold, respectively, p < 0.05). SULT1A1 and SULT1D1 were upregulated by 1.8-to 7.0-fold at 20 and 26 weeks (p > 0.05 and p = 0.031, respectively). There were positive correlations between protein expression and enzyme activity for CYP2E1, UGT1A9, SULT1A1 and SULT1D1 (r 2 ≥ 0.5, p < 0.001). Western blotting analysis revealed the downregulation of AHR, FXR and PPARα protein expression in male KRAS mice at 26 weeks. For sex-related differences, CYP2E1 was malepredominant and UGT1A2 was female-predominant in the kidney. UGT1A1 and UGT1A5 expression was female-predominant, whereas UGT2B1 exhibited male-predominant expression in liver tissue. For the tissue distribution of DMEs, 21 subtypes of DMEs were all expressed in liver tissue. In the intestine, the expression levels of CYP2C29, CYP27A1, UGT1A2, 1A5, 1A6a, 1A9, 2B1, 2B5 and 2B36 were under the limitation of quantification. The subtypes of CYP7A1, 1B1, 2E1 and UGT1A1, 2A3, 2B34 were detected in kidney tissue.</ns0:p></ns0:div>
<ns0:div><ns0:head>Conclusions:</ns0:head><ns0:p>This study, for the first time, unveils the variations and sex-and age-related differences in DMEs in C57 BL/6 (WT) mice and KRAS mice.</ns0:p></ns0:div>
</ns0:abstract>
<ns0:body xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:head>INTRODUCTION</ns0:head><ns0:p>Many studies have posited that disease states, as well as sex-and age-related differences can alter the expression of drug metabolizing enzymes (DMEs), and in turn change the metabolism and detoxification of drugs by remodeling the hepatic absorption, distribution, metabolism and excretion (ADME) <ns0:ref type='bibr' target='#b12'>(Court 2010;</ns0:ref><ns0:ref type='bibr' target='#b23'>Hui Li et al. 2017;</ns0:ref><ns0:ref type='bibr'>Beatrice A et al. 2014;</ns0:ref><ns0:ref type='bibr'>S. Xu et al. 2019;</ns0:ref><ns0:ref type='bibr' target='#b72'>Wu and Lin 2018)</ns0:ref>. Therefore, we specifically investigated the changes in DME expression levels in a disease model with age-and sex-related differences.</ns0:p><ns0:p>The KRAS mutation, the most frequently mutated isoform of RAS, accounts for > 85% of RAS-driven cancers <ns0:ref type='bibr'>(Ding et al. 2008;</ns0:ref><ns0:ref type='bibr' target='#b53'>Nussinov et al. 2015)</ns0:ref>. However, to date, it is still a major challenge to develop novel drugs that effectively treat KRAS mutant lung cancer <ns0:ref type='bibr' target='#b48'>(Mccormick 2015)</ns0:ref>. Considering the widespread and incurable nature of this disease, the metabolic profile of drugs is urgently needed to determine when KRAS is mutated. The mouse genome is 99% identical to the human genome, and the organs and systemic physiology of mouse have similar patterns with humans (Robert A. <ns0:ref type='bibr'>Ribeiro et al. 2006)</ns0:ref>. Hence, mice have been widely used in current cancer research (~ 59% of the total number of animals used) <ns0:ref type='bibr' target='#b68'>(Wang et al. 2020</ns0:ref>). In our study, a KRAS-mutant mouse model of lung cancer (KRAS mice, spontaneous tumors in the lung) was used to study the changes in DMEs in the development of lung cancer. KRAS mice were first observed to have small pleural nodules at one week of age, and numerous pleural lesions started to appear in 5-weeks-old mice <ns0:ref type='bibr' target='#b35'>(Leisa Johnson et al. 2001)</ns0:ref>. Corresponding to human disease process, the cancer stage of 20-week old KRAS mice began to enter advanced stage, and the life span of KRAS mice is approximately 28 weeks (Leisa <ns0:ref type='bibr' target='#b35'>Johnson et al. 2001)</ns0:ref>.</ns0:p><ns0:p>The activation and inactivation of exogenous drugs are mainly regulated by drug metabolizing enzymes (DMEs), including cytochrome P450s (CYPs), UDP-glucuronosyltransferases (UGTs), and sulfotransferases (SULTs) (T. <ns0:ref type='bibr'>Yan et al. 2014;</ns0:ref><ns0:ref type='bibr' target='#b11'>Cong Xie et al. 2017)</ns0:ref>. The changes in DMEs can further affect the efficacy of the drug and even increase the side effects. For instance, irinotecan, which is used in the treatment of metastatic colorectal cancer, causes severe intestinal toxicity attributed to damage to UGT1A <ns0:ref type='bibr'>(Ronaldo A. Ribeiro et al. 2016)</ns0:ref>. It was also reported that CYP3A has higher expression in osteosarcoma (By H. R. <ns0:ref type='bibr' target='#b5'>Dhaini et al. 2003)</ns0:ref>. The activation/inactivation of anticancer drugs metabolized by CYP3A would lead to changes in curative effect. Sex and age are important factors influencing the expression level of DMEs <ns0:ref type='bibr' target='#b28'>(Kennedy 2008;</ns0:ref><ns0:ref type='bibr'>H. Zheng et al. 2018)</ns0:ref>. For age-related differences, a related report showed that the enzyme activity of UGT1A increases before 20 years of age and then decreases <ns0:ref type='bibr' target='#b12'>(Court 2010)</ns0:ref>.</ns0:p><ns0:p>Sex-related differences characterize the metabolism of many drugs used frequently in the clinic <ns0:ref type='bibr' target='#b70'>(Waxman and Holloway 2009)</ns0:ref>. For instance, men showed a 38% faster clearance of olanzapine than women (Kristin L. <ns0:ref type='bibr'>Bigos et al. 2008)</ns0:ref>. Therefore, a thorough understanding of the variations in DMEs is beneficial and indispensable for pharmacological evaluation and rational clinical drug use.</ns0:p><ns0:p>To effectively treat the KRAS-mutant lung cancer, researchers have used many drugs, such as gefitinib, erlotinib, cisplatin, trametinib, and pazopanib (Min K. <ns0:ref type='bibr' target='#b50'>Kim et al. 2018</ns0:ref>; Jean L. K. <ns0:ref type='bibr' target='#b25'>Pujol et al. 2006)</ns0:ref>. Changes in DMEs could alter the metabolic characteristics of these drugs, further affecting their efficacy in vivo. The drugs erlotinib and cisplatin are mainly metabolized by CYP3A4, an ortholog of mouse CYP3A11 <ns0:ref type='bibr' target='#b49'>(Melanie et al. 2015;</ns0:ref><ns0:ref type='bibr'>Hanna K. Sanoff et al. 2010)</ns0:ref>.</ns0:p><ns0:p>Alterations in the activity of CYP3A4 could potentially have a pronounced effect on drug exposure. In other words, downregulation of CYP3A4 could reduce sorafenib hepatotoxicity (T. <ns0:ref type='bibr' target='#b66'>Yan et al. 2015)</ns0:ref>. However, little is known about the alterations in DMEs after KRAS mutation, thus causing some difficulties in understanding the fate of drugs in vivo and leading to confusion about the efficacy and side effects.</ns0:p><ns0:p>MS-based quantifications are different from traditional immunogenic methods and can increase the sensitivity and high throughput of the absolute quantification of proteins. In our study, an LC-MS/MS method was employed to determine the protein expression of DMEs in WT and KRAS mice (J. <ns0:ref type='bibr'>Chen et al. 2017</ns0:ref>). In addition, the possible mechanism for the variations in DMEs was investigated. We intend to provide a valuable reference for pharmacological evaluation and rational clinical drug use in patients with KRAS mutated lung cancer.</ns0:p><ns0:p>Insitute. The animals were kept at controlled temperature of 24-26 ℃ and humidity of 50-60%, with a 12 h light-dark cycle. The permission of all animal experiments was obtained from the Institutional Animal Care and Use Committee of the International Institute for Translational Chinese Medicine (IITCM_20171105). Before the experiment, animals were fasted overnight but allowed free access to water. All procedures were performed under diethyl ether anesthesia and the efforts were made to minimize suffering. After the animal experiment was completed, the animal bodies were frozen and sent to professionals for harmless treatment.</ns0:p><ns0:p>Histopathological analysis of lung tissues. Lung tissues were acquired from KRAS mice and wild-type (WT) mice. The morphology of lung tissues was observed under stereoscopic microscope (Leica, M165C), then tissues were fixed by 4% paraformaldehyde. After paraformaldehyde fixation and paraffin embedding, mouse lung tissues were sliced and stained with hematoxylin for 30 s and 0.5% eosin for 10 s, and covered with neutral gum. The images were obtained under microscopy (NIKON Eclipseci, Japan).</ns0:p><ns0:p>Preparation of mouse S9 fractions. Mouse tissue (liver, kidney and intestine) were isolated from WT and KRAS mice. The tissues were minced and washed with ice cold saline. Ice-cold homogenization buffer (50 mM potassium phosphate, 250 mM sucrose, 1 mM EDTA, PH 7.4) with 0.28 mM phenylmethylsulfonyl fluoride (PMSF) was added to the minced tissues and homogenized until an even suspension was obtained. Then the suspension was centrifuged at 9,000 × g for 20 min at 4°C. The supernatant was collected as S9 fractions (L. <ns0:ref type='bibr' target='#b34'>Tang et al. 2012;</ns0:ref><ns0:ref type='bibr' target='#b78'>Zhu et al. 2010)</ns0:ref>. Liver tissue was handled with a single sample of each mouse while kidney and intestine tissues of each group of mice (n=5) were mixed into same sample. Protein Manuscript to be reviewed concentrations of mouse S9 fractions were detected by coomassie brilliant blue and the bovine serum albumin was selected as the standard.</ns0:p><ns0:p>LC-MS/MS analysis. Eight isoforms of CYPs (CYP1B1, CYP2C29, CYP2D22, CYP2E1, CYP3A11, CYP3A25, CYP7A1 and CYP27A1), ten isoforms of UGTs (UGT1A1, UGT1A2, UGT1A5, UGT1A6a, UGT1A9, UGT2A3, UGT2B1, UGT2B5, UGT2B34 and UGT2B36) and three isoforms of SULTs (SULT1A1, SULT1B1 and SULT1D1) were analyzed. The methods of sample preparation and quantifying DMEs amounts by UHPLC/MS-MS were consistent with our previous study and dynamic MRM chromatograms of 21 subtypes were displayed in Supplemental Figure <ns0:ref type='figure' target='#fig_11'>1</ns0:ref> (J. <ns0:ref type='bibr'>Chen et al. 2017)</ns0:ref>. Samples were analyzed by using an Agilent 6490 triple quadruple mass spectrometer coupled with 1290 Infinite UHPLC system. A Poroshell C 18 column (2.1 mm × 100 nm, 2.7 μm; Agilent Technologies) was used for separation. In this study, the protein amounts of DMEs were represented in the form of pmol protein per S9 fraction protein (pmol/mg). The quantification of protein levels was performed two independent experiments. All samples were performed in triplicate in each independent experiment and data were presented as mean ± SD.</ns0:p><ns0:p>Enzyme assays of liver S9 fractions. Enzyme activities of CYP2E1, CYP3A11, UGT1A9, SULT1A1 and SULT1D1 were measured by specific probe substrates in vitro (chlorzoxazone, testosterone, propofol, p-nitrophenol and dopamine, respectively). The enzyme activities of these isoforms in mice were determined by incubating S9 fractions with appropriate substrate concentrations. Production of metabolites was quantified to value the activities of these isoforms between WT and KRAS mice at their different age. The incubation systems of CYPs, UGTs and Manuscript to be reviewed SULTs, are in accordance with our previous articles with minor modification (C. <ns0:ref type='bibr'>Xie et al. 2017;</ns0:ref><ns0:ref type='bibr'>T. Yan et al. 2015;</ns0:ref><ns0:ref type='bibr'>H. Zheng et al. 2018)</ns0:ref>. In order to terminate the reaction, 200 μL methanol with 200 nM genistein was added. Then, the solution was vortexed and thereafter centrifuged for 30 min at 18000 g. Then the supernatant of all samples was injected to analyze by LC-MS/MS.</ns0:p><ns0:p>The enzyme activity was measured from 2 independent experiments. Each sample was performed in triplicates in each independent experiment and data was presented as mean ± SD.</ns0:p><ns0:p>Western blotting. The protein levels of aryl hydrocarbon receptor (AHR), bile acid receptor (FXR), pregnane X receptor (PXR) and peroxisome proliferator-activated receptor (PPARα)</ns0:p><ns0:p>were determined in male WT and KRAS mice at 26 weeks, and β-actin was used as an internal control. The S9 samples were mixed with 5 × loading buffer and the mixture was denatured at 100 °C for 5 min. An equal amount of protein (40 μg) was separated by SDS-PAGE at a voltage of 120 V to the correct band size and the protein was subsequently transferred from the gel to the PVDF membrane. Then, the membrane was blocked for 1 h with 5% non-fat milk (w/v) in Trisbuffered saline containing 0.1% Tween-20 (TBST). The corresponding primary antibodies, against mouse peroxisome proliferator-activated receptor (PPARα, sc-398394), pregnane X receptor (PXR, ab118336), bile acid receptor (FXR, ab28480) and aryl hydrocarbon receptor (AHR, ab2769) and β-ACTIN (from Cell Signaling Technology, CST, Boston, USA) were diluted to a recommended dilution of 1:1000 with 5% non-fat milk according to the manufacturer's instructions. After blocking, the membrane was incubated with the corresponding primary antibodies at 4 °C overnight with gentle shaking and was then washed before being incubated with the corresponding secondary antibody at a dilution of 1:2000-1:3000 for 1 h at PeerJ reviewing <ns0:ref type='table'>PDF | (2020:01:45080:2:0:CHECK 3 Sep 2020)</ns0:ref> Manuscript to be reviewed room temperature. ECL chemiluminescence was used to detect the signals and each protein band was quantified by Image J (National Institutes of Health, Hercules, CA, USA). The WB analysis was performed from 2 independent experiments, and each target protein was analyzed twice in each independent experiment. The data was presented as mean ± SD.</ns0:p><ns0:p>Data analysis. One-way ANOVA analysis, non-parametric test and independent sample t tests were conducted using SPSS 19.0 to evaluate statistical differences. Correlation analyses were performed using SPSS 19.0 and GraphPad Prism 7, according to the Pearson product-moment correlation for normal related data and Spearman's rank correlation for non-normally related data.</ns0:p><ns0:p>Partial least squares discriminant analysis (PLS-DA) was performed to visualize the changes of DMEs after KRAS mutation using SIMCA-P 14.0 tool (Umetrics, Umea, Sweden). In each case, a value of p < 0.05 denotes statistical significance for all of the statistical analyses. <ns0:ref type='table'>PDF | (2020:01:45080:2:0:CHECK 3 Sep 2020)</ns0:ref> Manuscript to be reviewed</ns0:p></ns0:div>
<ns0:div><ns0:head>PeerJ reviewing</ns0:head></ns0:div>
<ns0:div><ns0:head>RESULTS</ns0:head><ns0:p>The phenotypic characteristics of KRAS mice. Pulmonary morphology was observed using a stereoscope (Fig. <ns0:ref type='figure' target='#fig_11'>1A-B</ns0:ref>). Compared with those in the WT mice, many lung nodules were observed in the lung tissues of the KRAS mice (Fig. <ns0:ref type='figure' target='#fig_11'>1B</ns0:ref>, As the arrows point), and the lung tissues appeared dull overall. Histological and pathological features of the WT and KRAS mice were detected by H&E staining (Fig. <ns0:ref type='figure' target='#fig_11'>1C-H</ns0:ref> and Supplemental Fig. <ns0:ref type='figure'>2</ns0:ref>). The morphology of lung cells in the WT mice was normal, whereas in the KRAS mice, the lung cells were hyperproliferative (Fig. <ns0:ref type='figure' target='#fig_11'>1F</ns0:ref>-H and Supplemental Fig. <ns0:ref type='figure'>2</ns0:ref>, As the arrows point). The nuclei were deeply stained and lager.</ns0:p><ns0:p>Alterations in the protein contents of DMEs by KRAS mutation. In male mice, PLS-DA analysis was applied to evaluate the clustering between the male WT and KRAS mice based on the expression of 21 DMEs (Fig. <ns0:ref type='figure'>2A-E</ns0:ref>). We observed obvious distinctions between the male WT and KRAS mice at 5, 10, 15, 20 and 26 weeks. These results demonstrated the differences in DMEs among them. As shown in Fig. <ns0:ref type='figure'>2H</ns0:ref> in liver tissue, SULT1A1 increased by 2.4-fold at 5 weeks (p = 0.016); CYP27A1 and UGT1A1 increased by 1.9-fold and 2.3-fold at 15 weeks respectively (p = 0.001 and p > 0.05, respectively); SULT1A1 and SULT1D1 were upregulated by 3.4-fold and 1.8-fold at 20 weeks, respectively (p = 0.015 and p = 0.017, respectively); and at 26 weeks, SULT1A1 and SULT1D1 were upregulated by 2.0-fold and 1.8-fold respectively (p > 0.05 and p = 0.031, respectively), and CYP2C29, CYP3A11, CYP27A1 and UGT1A5 decreased </ns0:p></ns0:div>
<ns0:div><ns0:head>Alterations in DME activities of DMEs by KRAS mutation.</ns0:head><ns0:p>As shown in Fig. <ns0:ref type='figure'>3</ns0:ref>, the activity of CYP3A11 was significantly downregulated in the male KRAS mice with aging. Furthermore, there were significant differences between the WT and KRAS mice. UGT1A9 gradually declined in the male mice from 5 to 26 weeks. SULT1A1 and SULT1D1 displayed larger differences at 20 and 26 weeks in the male KRAS mice than in the WT mice. SULT1A1 increased by 2.2-(p > 0.05) and 3.9-fold (p = 0.031), respectively. SULT1D1 was upregulated Manuscript to be reviewed by 7.0-(p = 0.007) and 3.5-fold (p > 0.05), respectively. Supplemental Fig. <ns0:ref type='figure'>4</ns0:ref> shows the activities in the female WT and KRAS mice at different ages. The activities of CYP2E1 and SULT1D1 displayed no significant differences in the WT and KRAS mice with increasing age. CYP3A11 displayed an increasing tendency from 5 to 26 weeks. The activity of UGT1A9 showed a significant decrease at 15, 20 and 26 weeks compared to that at 5 weeks. At 15 weeks, the activity of SULT1A1 was markedly different in the KRAS mice relative to the WT mice (p = 0.008).</ns0:p></ns0:div>
<ns0:div><ns0:head>Correlation of protein content and enzyme activity of DMEs.</ns0:head><ns0:p>The enzyme activities of CYP2E1, CYP3A11, UGT1A9, SULT1A1 and SULT1D1 were compared with their protein contents. The correlation analysis assessed the protein levels quantified by LC-MS/MS and the activities detected by the specific probes. As shown in Fig. <ns0:ref type='figure'>4</ns0:ref>, there was a good correlation between enzyme activity and protein content (CYP2E1, r 2 = 0.54, p < 0.001; UGT1A9, r 2 = 0.55, p < 0.001; SULT1A1, r 2 = 0.62, p < 0.001; SULT1D1, r 2 = 0.89, p < 0.001). A poor correlation for CYP3A11 was observed in the mouse liver (r 2 < 0.50, p > 0.05). <ns0:ref type='figure'>PPARα (H-2</ns0:ref>) and PXR. To evaluate the protein levels of receptors, we dissected liver tissue from the KRAS mice. As shown in Fig. <ns0:ref type='figure'>5</ns0:ref>, the protein expression levels of AHR, FXR and PPARα (H-2) were downregulated by 40.22%, 20.90% and 26.76% in the livers of the male KRAS mice, respectively (p = 0.000, 0.035 and 0.005, respectively). Compared to the WT mice, the KRAS mice showed no significant difference in the protein amount of PXR (decreased by 13.52%, p = 0.109). In the female mice (Supplemental Fig. <ns0:ref type='figure'>5</ns0:ref>), the protein expression levels of AHR, FXR, PPARα (H-2) and PXR were Manuscript to be reviewed downregulated by 12.00%, 10.14%, 5.60% and 38.06% in the liver, respectively (p = 0.466, 0.442, 0.710 and 0.074, respectively).</ns0:p></ns0:div>
<ns0:div><ns0:head>Protein expression profiles of AHR, FXR,</ns0:head><ns0:p>Tissue distribution of DMEs. To evaluate the tissue distribution of DMEs, we present data for male mice at 10 weeks as an example. CYP2C29/CYP2D22/CYP3A11/CYP3A25/CYP27A1 (LLOQ). In the kidney, UGT1A2 was detected, but the other UGT isoforms were below the lower limit of quantification; for SULT isoforms, SULT1D1 was detected, but the others were all below the lower limit of quantification. SULT1D1, CYP1B1, CYP2E1, CYP7A1 and UGT1A2 had the highest protein expression levels. DME variations based on sex. Fig. <ns0:ref type='figure' target='#fig_13'>6</ns0:ref> shows that the the sex-related changes in DMEs have a similar trend in both the WT and KRAS mice at 10 weeks of age. Therefore, we mainly discuss the differences in protein content in WT mice. In kidney tissue, CYP2E1 was male-predominant, Manuscript to be reviewed while UGT1A2 was female-predominant. In liver tissue, the content of UGT2B1 was significantly higher in the male mice than in the female mice.</ns0:p><ns0:p>Variations in DME protein content with increased age. In the liver, CYP7A1 increased with increasing age in the male mice (Fig. <ns0:ref type='figure' target='#fig_14'>7G</ns0:ref>); UGT1A9 showed a decreasing trend with increasing age (Fig. <ns0:ref type='figure' target='#fig_14'>7L</ns0:ref>). In the intestine, the CYP isoforms showed no significant changes at different ages (Fig. <ns0:ref type='figure'>8A-E</ns0:ref>). UGT2B34 showed a decreasing trend in the male WT and KRAS mice with increasing age (Fig. <ns0:ref type='figure'>8H</ns0:ref>), and SULT1A1 showed an increasing trend with increasing age (Fig. <ns0:ref type='figure'>8I</ns0:ref>).</ns0:p><ns0:p>In the kidney, CYP1B1 displayed a decreasing trend in the male WT and KRAS mice with increasing age (Fig. <ns0:ref type='figure'>8L</ns0:ref>). Supplemental Fig. <ns0:ref type='figure' target='#fig_13'>6</ns0:ref> shows the changes of protein amount with aging in female WT and KRAS mice. In the liver, the protein amount of CYP2C29 decreased 2.9-fold at 15 weeks compared to that at 5 weeks in the WT mice. This pattern of changes did not appear in the KRAS mice. High individual differences in the protein amounts in female mice were observed. There are no significant differences at different ages.</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:01:45080:2:0:CHECK 3 Sep 2020)</ns0:p><ns0:p>Manuscript to be reviewed</ns0:p></ns0:div>
<ns0:div><ns0:head>DISCUSSION</ns0:head><ns0:p>In this study, we systematically investigated the alterations of DMEs in KRAS mice of different ages and sexes, with the aim of providing a better explanation for the clinically observed variation in the efficacy and toxicity of anticancer drugs in KRAS-mutant lung cancer patients. Currently, limited information is available concerning the changes in DMEs in patients with KRAS mutant lung cancer.</ns0:p><ns0:p>The absolute protein contents of 21 metabolic enzymes in KRAS mice were simultaneously determined by the LC-MS/MS approach. In our study, the protein expression levels of CYP2C29 and CYP3A11 were significantly downregulated in the male KRAS mice at 26 weeks of age. A previous study indicated that hepatic DMEs are reduced during infection and inflammation in humans, rats, and mice (Nozomu <ns0:ref type='bibr' target='#b52'>Moriya et al. 2012)</ns0:ref>. Similar results were also reported that a decrease in Cyp gene expression and enzymatic activity was observed in a dextran sulfate sodium (DSS)-induced mouse model of ulcerative colitis <ns0:ref type='bibr'>(Yoshiiki K. et al. 2014</ns0:ref>). CYP2C29 is the major arachidonate CYP2C epoxygenase in mice <ns0:ref type='bibr'>(Komal S. et al. 2009</ns0:ref>). The decreased expression of CYP2C29 is closely related to the occurrence and development of inflammation.</ns0:p><ns0:p>Related studies have shown that this decreased expression may be triggered by an increased production of inflammatory cytokines <ns0:ref type='bibr'>(Yoshiiki K. et al. 2014)</ns0:ref> Manuscript to be reviewed expression of CYP3A11 in the KRAS mice may cause some differences in efficacy or even side effects. With respect to the SULT family, the protein expression and activities of SULT1A1 and SULT1D1 were upregulated in the male KRAS mice at 20 and 26 weeks. A number of studies have been conducted on SULT in different cancers, but many conflicting outcomes have been reported (Y. <ns0:ref type='bibr' target='#b76'>Jiang et al. 2010)</ns0:ref>. Some authors showed a potential association between SULT1A1 polymorphisms and breast cancer, but inconsistent results also exist (Y. <ns0:ref type='bibr' target='#b76'>Jiang et al. 2010)</ns0:ref>.</ns0:p><ns0:p>Relevant studies have reported that SULT1A3 may be a diagnostic marker for osteosarcoma, and SULT1A3 protein upregulation is closely related to the occurrence and development of cancer (X. <ns0:ref type='bibr'>Chen et al. 2014;</ns0:ref><ns0:ref type='bibr'>C. Xie et al. 2017)</ns0:ref>. SULT1D1 is a pseudogene in humans, Sult1d1 encodes protein expression in mice, and its functions are similar to those of human SULT1A3 (S. <ns0:ref type='bibr' target='#b64'>Wong et al. 2010)</ns0:ref>. Our results also revealed increased protein expression of SULT1D1 in male mice after KRAS mutation. This finding is beneficial to explain the metabolic characteristics of SULT1D1-metabolized drugs in KRAS mice. The expression of SULT1A3 should be further explored in KRAS-mutant lung cancer patients.</ns0:p><ns0:p>To further explore the changes in enzymatic activity, we used specific probe substrates to detect the enzymatic status in the KRAS mice. In our present study, we found the CYP2C29/CYP3A11/SULT1A1/SULT1D1 displayed significant changes in protein expression in the liver of male WT and KRAS mice. Therefore, we are intended to research their activities in the liver tissue. We failed to find an authoritative and specific probe to study the activity of CYP2C29. Chlorzoxazone and Propofol are usually used as specific substrates to study the activities of CYP2E1 and UGT1A9. Their good correlation between protein expression and Manuscript to be reviewed activities indicated the protein quantification results are credible. Therefore, we select them for the enzyme activity test. Notably, SULT1A1 and SULT1D1 activity was upregulated at 20 and 26 weeks in the male KRAS mice (Fig. <ns0:ref type='figure'>3</ns0:ref> and Supplemental Fig. <ns0:ref type='figure'>4</ns0:ref>). This result was consistent with their protein expression levels. In this context, a high degree of correlation was observed between the enzymatic activity and protein level (Fig. <ns0:ref type='figure'>4</ns0:ref>). For the poor correlation between the enzyme activity and protein level of CYP3A11, the nonspecificity of the substrate may be a possible reason. The FDA (USA) reported that testosterone was metabolized by CYP3A4 and CYP3A5. In addition, the protein structure could affect the activity. CYP3A4 showed a significant genetic polymorphism in individuals, causing a flexible three-dimensional structure of CYP3A4 <ns0:ref type='bibr' target='#b71'>(Werk and Cascorbi 2014)</ns0:ref>. Moreover, the genetic polymorphism of CYP2D6 (ortholog of CYP2D22 in mice) could induce variations in the expression or function of CYP3A4 <ns0:ref type='bibr' target='#b71'>(Werk and Cascorbi 2014)</ns0:ref>. Generally, these findings indicate that the protein expression levels of some DMEs could be applied to forecast the enzymatic activities regarding drug metabolism. Changes in the ability of DMEs to metabolize drugs could lead to differences in the ADME properties of drugs, thereby affecting drug efficacy and toxicity in the body.</ns0:p><ns0:p>The expression of DMEs is regulated by the binding of xenobiotics to receptors, such as the aryl hydrocarbon receptor (AHR), the murine pregnane X receptor (PXR), peroxisome proliferator-activated receptor (PPARα) and bile acid receptor (FXR) (S. <ns0:ref type='bibr' target='#b59'>Anakk et al. 2003;</ns0:ref><ns0:ref type='bibr'>C. HANDSCHIN 2003;</ns0:ref><ns0:ref type='bibr' target='#b22'>HONKAKOSKI and NEGISHI 2000)</ns0:ref>. The decrease in receptor levels may contribute to the emergence of changes in DME expression (S. <ns0:ref type='bibr' target='#b59'>Anakk et al. 2003;</ns0:ref><ns0:ref type='bibr'>L. Li et al. 2009</ns0:ref>; J. E. <ns0:ref type='bibr' target='#b24'>Moscovitz et al. 2018)</ns0:ref>. Moreover, some reports have suggested that disease status PeerJ reviewing PDF | (2020:01:45080:2:0:CHECK 3 Sep 2020)</ns0:p><ns0:p>Manuscript to be reviewed (e.g., cancer and inflammation) can affect the expression and activity of DMEs via specific receptors <ns0:ref type='bibr' target='#b33'>(Lamba et al. 2016;</ns0:ref><ns0:ref type='bibr'>H. Chen et al. 2014;</ns0:ref><ns0:ref type='bibr' target='#b0'>A. Schröder et al. 2011)</ns0:ref>. Therefore, we further studied the changes in receptor expression after KRAS mutation. In our study, the protein expression levels of AHR, FXR and PPARα were downregulated in the livers of the male KRAS mice compared to the WT mice at 26 weeks. This phenomenon was not significant in the female KRAS mice. Major xenobiotic-sensing transcription factors, such as AHR and PXR, are involved in the regulation of the protein expression of DMEs. Related reports revealed that most core DMEs were positively correlated with AHR, PXR and PPARα, and their protein expression was downregulated in nearly 50% of the patients with hepatocellular carcinoma (HCC) (H. <ns0:ref type='bibr'>Chen et al. 2014;</ns0:ref><ns0:ref type='bibr'>D. G. Hu et al. 2018;</ns0:ref><ns0:ref type='bibr'>S. Zhong et al. 2016)</ns0:ref>. Activation of AHR could induce the upregulation of Cyp1a/3a/Ugt1a1 mRNA expression, which would therefore not occur in Ahrnull mice (C.D. <ns0:ref type='bibr' target='#b6'>Klaassen and A.L. Slitt 2005;</ns0:ref><ns0:ref type='bibr' target='#b51'>Nakajima et al. 2003)</ns0:ref>. PPARα and PXR were implicated in the regulation of CYP3A/4A/1A1/2B6/2C8/2C9/2C19/UGT1A1 induction (J. E. <ns0:ref type='bibr' target='#b24'>Moscovitz et al. 2018)</ns0:ref>. FXR, an important regulator of lipid and glucose homeostasis, is involved in the expression of CYP7A1 and CYP27A1 <ns0:ref type='bibr' target='#b58'>(Sánchez 2018)</ns0:ref>. Hence, in our study, we speculate that these variations in DME expression may be regulated by decreased receptors of AHR, FXR and PPARα.</ns0:p><ns0:p>For sex-difference, CYP2E1 showed significant male-specificity in kidney tissue. CYP2E1 mediates the metabolism of many low molecular weight organic compounds (such as ethanol and acetone) and some drugs (such as p-nitrophenol, caffeine, chlorzoxazone, etc.) (S. <ns0:ref type='bibr' target='#b63'>Löfgren et al. 2004;</ns0:ref><ns0:ref type='bibr' target='#b79'>Zuber et al. 2002)</ns0:ref>. Therefore, in regard to the intake of these exogenous substances, we Manuscript to be reviewed should consider the effects related to sex differences in patients. For UGT enzymes, UGT2B1 exhibited male-predominant expression in the liver tissue. Conversely, UGT1A1 and UGT1A5 expression in the liver and UGT1A2 in the kidney are female-predominant, whereas UGT2B1 exhibited male-predominant expression in liver tissue. These results are consistent with previous reports (D. B. Buckley and C. D. <ns0:ref type='bibr' target='#b13'>Klaassen 2007</ns0:ref><ns0:ref type='bibr' target='#b14'>Klaassen , 2009))</ns0:ref>. The female-predominant UGT1A1 expression accounts for the higher bilirubin-UGT activity in females. UGT1A and UGT2B are the primary families of UGT enzymes, involved in the inactivation of > 30% of drugs currently used in the clinic (C. <ns0:ref type='bibr' target='#b8'>Guillemette et al. 2014</ns0:ref>). These sex-specific expressions may be crucial in understanding the mechanisms by which many drugs display variations in metabolism and clearance.</ns0:p><ns0:p>For age-related difference, except for UGT1A9, the majority of DMEs showed no significant changes from 5 to 26 weeks of age in female and male WT and KRAS mice. UGT1A9, major UGT isoforms expressed in the liver (~6% of hepatic UGT expression), is responsible for the glucuronidation of multiple endogenous substances (e.g., thyroid hormones) and drugs (e.g., acetaminophen and propofol) (S. <ns0:ref type='bibr' target='#b65'>Cho et al. 2016)</ns0:ref>. The activity and protein expression of UGT1A9 appeared to decrease in the liver of female and male WT mice. Related studies indicated that UGT1A9 activity showed a downward trend from 6 weeks to 52 weeks in mice with a FVB background (H. <ns0:ref type='bibr' target='#b17'>Zheng et al. 2018)</ns0:ref>. Therefore, the appropriate dosage should be considered when patients of different ages are prescribed drugs metabolized by UGT1A9.</ns0:p><ns0:p>For tissue-related differences, abundant CYP enzymes were expressed in the liver, predominantly CYP2D22, CYP2C29, CYP2E1 and CYP3A11 (Fig. <ns0:ref type='figure' target='#fig_13'>6A-B</ns0:ref>). A previous studies Manuscript to be reviewed also demonstrated that the protein contents of these isoforms were high in the liver (J. <ns0:ref type='bibr'>Chen et al. 2017;</ns0:ref><ns0:ref type='bibr' target='#b7'>C. Gröer et al. 2014</ns0:ref>). In the intestine, the CYP1B1, CYP2D22, CYP2E1 and CYP3A11 protein levels were significantly higher than the levels of other proteins. Mouse phase I enzymes (CYP2D22, CYP2E1 and CYP3A11) are orthologs of the corresponding human enzymes (CYP2D6, CYP2E1 and CYP3A4), in charge of major phase I-dependent metabolism in marketed drugs <ns0:ref type='bibr' target='#b47'>(Liu 2013;</ns0:ref><ns0:ref type='bibr'>G. Ruaño et al. 2012)</ns0:ref>. Hence, optimal drug administration routes should be considered when these enzymes are involved in the inactivation or activation of drugs.</ns0:p></ns0:div>
<ns0:div><ns0:head>CONCLUSION</ns0:head><ns0:p>Taken together, our data showed significant decrease in CYP3A11 and CYP2C29, but an increase in SULT1A1 and SULT1D1 in the KRAS mice at 26 weeks. These DMEs all participate in the metabolism of drugs. Therefore, we hope that these results could provide useful guidance or a theoretical basis for further drug research and implementation. (A-E) Comparative data analysis of DMEs protein content in the liver of the male WT and KRAS mice was performed using PLS-DA plot at different ages. The solid black dots represent the KRAS mice, and the open black dots represent the WT mice (n=5). (F-H) The relative expression levels of DMEs in the liver, intestine and kidney tissue in the male mice at different ages. Protein levels in the male WT mice (n=5) were normalized to those in the male KRAS mice (n=5). The data were analyzed by independent sample t tests (for normally distributed data) and Mann-Whitney U analysis (for non-normally distributed data). The symbol '*' indicates a displayed significant difference between the male WT and KRAS mice at the same age, p < 0.05.</ns0:p></ns0:div>
<ns0:div><ns0:head>Figure 3</ns0:head><ns0:p>Changes in the enzyme activity of CYP2E1, CYP3A11, UGT1A9, SULT1A1 and SULT1D1 in liver tissues of the male KRAS and WT mice at different ages (n=5).</ns0:p><ns0:p>The solid line represents male KRAS mice, and the dashed line represents male WT mice.</ns0:p><ns0:p>Each data point is presented as the mean ± SD. For the comparison between KRAS and WT at the same age, the data were analyzed by independent sample t tests (for normally distributed data) and Mann-Whitney U analysis (for non-normally distributed data). The symbol '*' indicates a significant difference between the male WT and KRAS mice at the same age, p < 0.05. For different ages compared to 5 weeks, the data were analyzed by oneway ANOVA (for normally distributed data) and Kruskal-Wallis H analysis (for non-normally distributed data). We adjusted the significance level α to 0.0125 according to the Bonferroni correction (0.05/4=0.0125). The symbols 'A' and 'a' indicate significant differences in the male WT and KRAS mice at 10, 15, 20 and 26 weeks relative to 5 weeks, p < 0.0125.</ns0:p></ns0:div>
<ns0:div><ns0:head>Figure 4</ns0:head><ns0:p>Correlation between the protein expression and activity of CYP2E1, CYP3A11, UGT1A9, SULT1A1 and SULT1D1 in the liver tissue (n=100).</ns0:p><ns0:p>The correlation between the protein expression and activity included 5, 10, 15, 20 and 26 weeks, which were analyzed together. DME in the liver was determined using an isotope label-free LC-MS/MS method. The enzyme activities of DMEs were measured using probe substrates. All measurements were performed in triplicate and the data are presented as the mean ± SD. Pearson product correlation and Spearman's rank correlation were used to analyze the correlation. Regression line is shown for significant correlation at p < 0.05.</ns0:p></ns0:div>
<ns0:div><ns0:head>Figure 5</ns0:head><ns0:p>Protein expression levels of AHR, FXR, PPARα and PXR in the male WT (n=5) and KRAS mice (n=4) at 26 weeks.</ns0:p><ns0:p>(A) The mprint of five proteins was represented and β-ACTIN was used as an internal control.</ns0:p><ns0:p>(B) The data on protein expression levels was shown as a box chart. The data were analyzed by independent sample t tests (for normally distributed data) and Mann-Whitney U analysis (for non-normally distributed data). The symbol '*' indicates a significance difference of protein expression level in the KRAS mice relative to that in the WT mice, p < 0.05. Alterations in protein levels of 8 CYPs, 9 UGTs and 2 SULTs at different ages in the liver of male KRAS and WT mice, n=5.</ns0:p><ns0:p>The dotted and solid lines represent the WT and KRAS mice, respectively. Each data point represents the mean ± SD. The data were analyzed by one-way ANOVA (for normally distributed data) and Kruskal-Wallis H analysis (for non-normally distributed data). We adjusted the significance level α to 0.0125 according to the Bonferroni correction (0.05/4=0.0125). The symbols 'A' and 'a' indicate significant differences in the male WT and KRAS mice at 10, 15, 20 and 26 weeks relative to 5 weeks, p < 0.0125.</ns0:p></ns0:div>
<ns0:div><ns0:head>Figure 8</ns0:head><ns0:p>Alterations in protein levels of CYPs, UGTs and SULTs at different ages in the intestinal and kidney tissue of male KRAS and WT mice, n=5.</ns0:p><ns0:p>A-K shows the DMEs expression in intestinal tissue. L-O shows the DMEs expression in kidney tissue. The dotted and solid lines represent the WT and KRAS mice, respectively. Each data point represents the mean ± SD. The data were analyzed by one-way ANOVA (for normally distributed data) and Kruskal-Wallis H analysis (for non-normally distributed data). We adjusted the significance level α to 0.0125 according to the Bonferroni correction (0.05/4=0.0125). The symbols 'A' and 'a' indicate significant differences in the male WT and KRAS mice at 10, 15, 20 and 26 weeks relative to 5 weeks, p < 0.0125.</ns0:p></ns0:div><ns0:figure xml:id='fig_0'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:01:45080:2:0:CHECK 3 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_1'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:01:45080:2:0:CHECK 3 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_2'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:01:45080:2:0:CHECK 3 Sep 2020)Manuscript to be reviewed by 3.7-fold (p = 0.005), 4.4-fold (p = 0.004), 2.1-fold (p = 0.043) and 2.3-fold (p = 0.014), respectively. In intestinal tissue, SULT1B1 and SULT1D1 increased by 3.2-fold and 2.9-fold at 20 weeks, respectively (p > 0.05); and SULT1A1 and SULT1D1 increased by 2.0-fold (p > 0.05) and 1.8-fold at 26 weeks (p = 0.024), respectively. In kidney tissue, SULT1D1 was upregulated by 3.0-fold at 26 weeks (p > 0.05).Supplemental Fig.3shows some changes in DMEs with KRAS mutations in female KRAS mice. In liver tissue, UGT1A9 decreased by 2.3-fold at 5 weeks (p = 0.002); UGT2B1 decreased by 2.1-fold at 10 weeks (p = 0.004); CYP2C29, CYP2D22, CYP2E1, CYP3A11, CYP27A1, UGT1A1, UGT2A3, UGT2B5 and UGT2B1 were upregulated by 2.5-(p = 0.014), 1.7-(p = 0.018), 2.8-(p = 0.001), 2.6-(p = 0.001), 1.8-(p = 0.038), 2.5-(p = 0.008), 2.2-(p = 0.005), 2.3fold (p = 0.014), and 2.4 folds (p = 0.002), respectively, at 15 weeks; SULT1A1 increased by 2.2 folds at 20 weeks (p > 0.05); and CYP3A11 decreased by 2.1-fold at 26 weeks (p = 0.033). In intestine tissue, SULT1B1 and SULT1D1 increased by 3.4-fold (p > 0.05) and 4.7-fold (p = 0.042), respectively, at 26 weeks. In kidney tissue, SULT1D1 increased by 1.9 folds at 26 weeks (p = 0.013).</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_3'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:01:45080:2:0:CHECK 3 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_4'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:01:45080:2:0:CHECK 3 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_5'><ns0:head /><ns0:label /><ns0:figDesc>Fig. 6 displays the distribution of DMEs in liver, intestine and kidney tissue. In liver, CYP2C29 > CYP2D22 > CYP3A11 ≈ CYP2E1 ≈ CYP1B1 > CYP7A1 > CYP27A1 ≈ CYP3A25; UGT2B5 > UGT2B1 > UGT1A6a > UGT1A1 > UGT2B34 ≈ UGT2B36 > UGT2A3 > UGT1A9 ≈ UGT1A5 > UGT1A2 (lower limit of quantification, LLOQ); SULT1A1 > SULT1D1 > SULT1B1 (LLOQ). The protein contents of UGT2B5, UGT2B1, UGT1A6a, CYP2C29, CYP2D22, UGT1A1 and SULT1A1 were the highest. In the intestine, CYP1B1 > CYP2D22 > CYP3A11 > CYP3A25≈CYP7A1 > CYP2E1/CYP2C29/CYP27A1 (LLOQ); UGT2B34 > UGT1A1 > UGT2A3 > UGT1A2/ UGT1A5/ UGT1A6a/UGT1A9/UGT2B1/UGT2B5/UGT2B36 (LLOQ); SULT1B1 > SULT1A1 > SULT1D1. The protein contents of CYP1B1, UGT2B34, SULT1B1, CYP2D22, CYP3A11, SULT1A1 and CYP3A25 were the highest. In the kidney, CYP1B1 > CYP2E1 > CYP7A1 ></ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_6'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:01:45080:2:0:CHECK 3 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_7'><ns0:head /><ns0:label /><ns0:figDesc>. CYP3A11 plays a vital role in the metabolism of various clinical anticancer drugs, such as erlotinib, cisplatin, sorafenib. The drug concentration in serum would change accordingly with enzymatic expression. The declining PeerJ reviewing PDF | (2020:01:45080:2:0:CHECK 3 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_8'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:01:45080:2:0:CHECK 3 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_9'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:01:45080:2:0:CHECK 3 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_10'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:01:45080:2:0:CHECK 3 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_11'><ns0:head>Figure 1</ns0:head><ns0:label>1</ns0:label><ns0:figDesc>Figure 1</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_12'><ns0:head /><ns0:label /><ns0:figDesc>Figure 2</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_13'><ns0:head>Figure 6</ns0:head><ns0:label>6</ns0:label><ns0:figDesc>Figure 6</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_14'><ns0:head>Figure 7</ns0:head><ns0:label>7</ns0:label><ns0:figDesc>Figure 7</ns0:figDesc></ns0:figure>
<ns0:note place='foot'>PeerJ reviewing PDF | (2020:01:45080:2:0:CHECK 3 Sep 2020)Manuscript to be reviewed</ns0:note>
</ns0:body>
" | "Dear Editors and reviewers:
Thank you for your hard work and generous comments on the manuscript entitled “Changes and sex- and age-related differences in the expression of drug metabolizing enzymes in a KRAS-mutant mouse model of lung cancer” (ID: ms# 45080). The comments are all valuable and very helpful for revising and improving our paper. We have revised the manuscript according to the reviewer’ comments and the revised portions are marked in blue. We hope that the revised manuscript is sufficient for publication in Peer J.
We appreciate the editor / reviewer’s comments and hard work.
Warm regards,
Zhongqiu Liu, Ph.D.
Chair Professor of Pharmaceutics
Joint Laboratory for Translational Cancer Research of Chinese Medicine of the Ministry of Education of the People’s Republic of China, International Institute for Translational Chinese Medicine
Guangzhou University of Chinese Medicine
Guangzhou 510006
China
+8620-39358061
Email: liuzq@gzucm.edu.cn
Revised portions are marked in blue in the manuscript. The responses to the reviewers’ comments are as following:
Reviewer 1: Bindu Hegde
Basic reporting
The authors have addressed the questions satisfactorily. The language has been improved considerably, with more details added to the methods section. I was satisfied with the changes made to the discussion as well.
Experimental design
No comments
Validity of the findings
No comments
Reply: Thank you very much for your hard work.
Reviewer 2
Basic reporting
The study presented here is much clearer and polished compared to the initial submission. The authors have made the changes to convert it into a more professional manuscript. The references are improved and figures have been made clearer. My comments on the initial submission have been addressed satisfactorily.
Experimental design
No comment
Validity of the findings
No comment
Comments for the author
I wish you the best for your future endeavors
Reply: Thank you very much for your hard work.
Reviewer 3
Basic reporting
Comments for the author
The authors addressed all concerns raised by the reviewers and the manuscript has been significantly improved. However, there are some points which still need further clarification;
1. Please correct - line 59: change ‘’detoxifivcation’’ to detoxification; - line 167: change the sentence to ‘’before the experiment, animals were fasted overnight but allowed free access to water’’; - Line 168: change the sentence to ‘’all procedures were performed…’’; - Line 180: change ‘’fixated’’ to fixed; - Line 224: change the sentence to ‘’Then, the solution was vortexed and thereafter centrifuged for…’’; - Line 226: change the sentence to ‘’ the enzyme activity was measured from 2 independent experiments’’; - Line 249: add ‘’was performed from 2 independent experiments’’
Reply: We are so sorry for this incorrect writing and they have been re-written in the revised manuscript.
2. Line 169: it is not clear what do the authors mean by: ‘’at the conclusion of the experiment, the animals bodies were harmlessly treated by professionals’’
Reply: We are so sorry for unclear statement. It has been corrected as “After the animal experiment was completed, the animal bodies were frozen and sent to professionals for harmless treatment” (page 8, line 136-138).
3. Line 199, the authors state that 10 isoforms of UGTs were analysed, however they show only 9 isoforms in Fig. 2 (F, G, H), UGT1A2 is missing
Reply: Thank you so much for your hard work. 10 isoforms of UGTs were detected in our study. UGT1A2 shows a significant female-predominant characteristic in kidney tissue, and it is below the lower limit of quantification in male mice. The protein amount of UGT1A2 in kidney tissue of female mice was displayed in figure 6, supplemental figure 3 and supplemental figure 7O.
4. Line 269, the authors mention that cell membranes were broken in H&E images, however this is not clear from the image, please clarify or mark it on the image.
Reply: We are so sorry for this inappropriate description and it has been removed from the revised manuscript.
5. Please mention in the results part, for Fig. 2, what do the arrows correspond for
Reply: Thank you so much for your hard work. There are no arrows in Fig. 2. But there are some arrows in Fig. 1 and Supplemental Fig. 2. We have added some explains in results part and figure legend.
6. Authors state that the cancer stage of 20-week old KRAS mice corresponds to the advanced stage of human lung cancer. Could you please explain why did you then choose 26-week old mice mice for your experiments (western blot, histology), taking into consideration that these mice live 28 weeks and as Fig. 1 in the rebuttal letter shows, it is already difficult to distinguish lung morphology in mice at 26-weeks of age. And how does this age group (26-weeks) correlate to cancer stage in humans?
Reply: Thank you so much for your hard work and generous comments. We studied the protein amount of DMEs, activities, and lung histology in WT and KRAS mice at 5, 10, 15, 20 and 26 weeks. We found some significant changes in CYP3A11, CYP2C29, SULT1A1 and SULT1D1 at 26 weeks. Therefore, we further studied the expression of receptor proteins, which regulate the expression of DMEs at 26 weeks.
We are so sorry for this vague statement. Varying grades of tumors are present from hyperplasia/dysplasia to carcinomas similar to human non-small cell lung cancer (Leisa Johnson et al. 2001). The cancer stage of 20-week old KRAS mice began to enter advanced stage, corresponding to human disease process. Therefore, age from 20 to 28 weeks are in the advanced stage of cancer. At 20 and 26 weeks, some mice died naturally, so we chose the last age point at 26 weeks.
References: Leisa Johnson, Kim Mercer, Doron Greenbaum, Roderick T. Bronsonk, Denise Crowley, David A. Tuveson & Tyler Jacks. 2001. “Somatic Activation of the K-Ras Oncogene Causes Early Onset Lung Cancer in Mice.” Nature. 410(6832): 1111-16.
7. Figure 4, did the authors correlate the enzyme activities of 26-week old mice? Please mention the age in the figure legend
Reply: Thank you so much for your kind suggestion. The correlation between the protein expression and activity included 5, 10, 15, 20 and 26 weeks, which were analyzed together. The corresponding explanation has been added in the revised figure legend.
8. I would suggest to add to the main manuscript the explanation from the rebuttal letter, (Reviewer 3- question 10) of the criteria for selection Cyp2e1, Cyp3a11, Ugt1a9 and Sult1a1 for the enzyme activity - Fig. 3
‘’In our present study, we found the CYP2C29/CYP3A11/SULT1A1/SULT1D1 displayed significant changes in protein expression in the liver of male WT and KRAS mice. Therefore, we are intended to research their activities in the liver tissue. We failed to find an authoritative and specific probe to study the activity of Cyp2c29. Chlorzoxazone and Propofol are usually used as specific substrates to study the activities of Cyp2e1 and Ugt1a9. Their good correlation between protein expression and activities indicated the protein quantification results are credible. Therefore, we select them for the enzyme activity test, intending to show a good correlation between protein expression and activity’’.
Reply: Thank you so much for your good suggestion. The corresponding correction has been added in the revised manuscript (page 18-19, line 383-391).
" | Here is a paper. Please give your review comments after reading it. |
9,832 | "<ns0:abstract xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:p>Background: This study aimed to systematically profile the alterations and sex-and age-related differences in the drug metabolizing enzymes (DMEs) in a KRAS-mutant mouse model of lung cancer (KRAS mice). Methodology: In this study, the LC-MS/MS approach and a probe substrate method were used to detect the alterations in 21 isoforms of DMEs, as well as the enzymatic activities of five isoforms, respectively. Western blotting was applied to study the protein expression of four related receptors.</ns0:p></ns0:div>
<ns0:div><ns0:head>Results:</ns0:head><ns0:p>The proteins contents of CYP2C29 and CYP3A11, were significantly downregulated in the livers of male KRAS mice at 26 weeks (3.7-and 4.4-fold, respectively, p < 0.05). SULT1A1 and SULT1D1 were upregulated by 1.8-to 7.0-fold at 20 (p = 0.015 and 0.017, respectively) and 26 weeks (p = 0.055 and 0.031, respectively). There were positive correlations between protein expression and enzyme activity for CYP2E1, UGT1A9, SULT1A1 and SULT1D1 (r 2 ≥ 0.5, p < 0.001). Western blotting analysis revealed the downregulation of AHR, FXR and PPARα protein expression in male KRAS mice at 26 weeks. For sexrelated differences, CYP2E1 was male-predominant and UGT1A2 was female-predominant in the kidney. UGT1A1 and UGT1A5 expression was female-predominant, whereas UGT2B1 exhibited male-predominant expression in liver tissue. For the tissue distribution of DMEs, 21 subtypes of DMEs were all expressed in liver tissue. In the intestine, the expression levels of CYP2C29, CYP27A1, UGT1A2, 1A5, 1A6a, 1A9, 2B1, 2B5 and 2B36 were under the limitation of quantification. The subtypes of CYP7A1, 1B1, 2E1 and UGT1A1, 2A3, 2B34 were detected in kidney tissue. Conclusions: This study, for the first time, unveils the variations and sex-and age-related differences in DMEs in C57 BL/6 (WT) mice and KRAS mice.</ns0:p></ns0:div>
</ns0:abstract>
<ns0:body xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:head>INTRODUCTION</ns0:head><ns0:p>Many studies have posited that disease states, as well as sex-and age-related differences can alter the expression of drug metabolizing enzymes (DMEs), and in turn change the metabolism and detoxification of drugs by remodeling the hepatic absorption, distribution, metabolism and excretion (ADME) <ns0:ref type='bibr'>(Court 2010;</ns0:ref><ns0:ref type='bibr'>Hui Li et al. 2017;</ns0:ref><ns0:ref type='bibr'>Beatrice A et al. 2014;</ns0:ref><ns0:ref type='bibr'>S. Xu et al. 2019;</ns0:ref><ns0:ref type='bibr'>Wu and Lin 2018)</ns0:ref>. Therefore, we specifically investigated the changes in DME expression levels in a disease model with age-and sex-related differences.</ns0:p><ns0:p>The KRAS mutation, the most frequently mutated isoform of RAS, accounts for > 85% of RAS-driven cancers <ns0:ref type='bibr'>(Ding et al. 2008;</ns0:ref><ns0:ref type='bibr'>Nussinov et al. 2015)</ns0:ref>. However, to date, it is still a major challenge to develop novel drugs that effectively treat KRAS mutant lung cancer <ns0:ref type='bibr'>(Mccormick 2015)</ns0:ref>. Considering the widespread and incurable nature of this disease, the metabolic profile of drugs is urgently needed to determine when KRAS is mutated. The mouse genome is 99% identical to the human genome, and the organs and systemic physiology of mouse have similar patterns with humans (Robert <ns0:ref type='bibr'>A. Ribeiro et al. 2006</ns0:ref>). Hence, mice have been widely used in current cancer research (~ 59% of the total number of animals used) <ns0:ref type='bibr' target='#b38'>(Wang et al. 2020</ns0:ref>). In our study, a KRAS-mutant mouse model of lung cancer (KRAS mice, spontaneous tumors in the lung) was used to study the changes in DMEs in the development of lung cancer. KRAS mice were first observed to have small pleural nodules at one week of age, and numerous pleural lesions started to appear in 5-weeks-old mice <ns0:ref type='bibr'>(Leisa Johnson et al. 2001</ns0:ref>). Advanced tumours begin to appear in the lung of KRAS mice at 20 weeks and its life span is approximately 28 weeks <ns0:ref type='bibr'>(Leisa Johnson et al. 2001</ns0:ref>).</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:01:45080:3:0:NEW 18 Sep 2020)</ns0:p><ns0:p>Manuscript to be reviewed</ns0:p><ns0:p>The activation and inactivation of exogenous drugs are mainly regulated by drug metabolizing enzymes (DMEs), including cytochrome P450s (CYPs), UDP-glucuronosyltransferases (UGTs), and sulfotransferases (SULTs) (T. <ns0:ref type='bibr'>Yan et al. 2014;</ns0:ref><ns0:ref type='bibr'>Cong Xie et al. 2017</ns0:ref>). The changes in DMEs can further affect the efficacy of the drug and even increase the side effects. For instance, irinotecan, which is used in the treatment of metastatic colorectal cancer, causes severe intestinal toxicity attributed to damage to UGT1A (Ronaldo <ns0:ref type='bibr'>A. Ribeiro et al. 2016</ns0:ref>). It was also reported that CYP3A has higher expression in osteosarcoma (By H. <ns0:ref type='bibr'>R. Dhaini et al. 2003</ns0:ref>). The activation/ inactivation of anticancer drugs metabolized by CYP3A would lead to changes in curative effect.</ns0:p><ns0:p>Sex and age are important factors influencing the expression level of DMEs <ns0:ref type='bibr'>(Kennedy 2008;</ns0:ref><ns0:ref type='bibr'>H. Zheng et al. 2018</ns0:ref>). For age-related differences, a related report showed that the enzyme activity of UGT1A increases before 20 years of age and then decreases <ns0:ref type='bibr'>(Court 2010)</ns0:ref>. Sex-related differences characterize the metabolism of many drugs used frequently in the clinic <ns0:ref type='bibr'>(Waxman and Holloway 2009)</ns0:ref>. For instance, men showed a 38% faster clearance of olanzapine than women (Kristin L. <ns0:ref type='bibr'>Bigos et al. 2008</ns0:ref>). Therefore, a thorough understanding of the variations in DMEs is beneficial and indispensable for pharmacological evaluation and rational clinical drug use.</ns0:p><ns0:p>To effectively treat the KRAS-mutant lung cancer, researchers have used many drugs, such as gefitinib, erlotinib, cisplatin, trametinib, and pazopanib (Min K. <ns0:ref type='bibr'>Kim et al. 2018</ns0:ref>; Jean L. <ns0:ref type='bibr'>K. Pujol et al. 2006</ns0:ref>). Changes in DMEs could alter the metabolic characteristics of these drugs, further affecting their efficacy in vivo. The drugs erlotinib and cisplatin are mainly metabolized by CYP3A4, an ortholog of mouse CYP3A11 <ns0:ref type='bibr'>(Melanie et al. 2015;</ns0:ref><ns0:ref type='bibr'>Hanna K. Sanoff et al. 2010</ns0:ref>).</ns0:p></ns0:div>
<ns0:div><ns0:head>Alterations in the activity of CYP3A4 could potentially have a pronounced effect on drug</ns0:head><ns0:p>PeerJ reviewing PDF | (2020:01:45080:3:0:NEW 18 Sep 2020)</ns0:p><ns0:p>Manuscript to be reviewed exposure. In other words, downregulation of CYP3A4 could reduce sorafenib hepatotoxicity (T. <ns0:ref type='bibr'>Yan et al. 2015)</ns0:ref>. However, little is known about the alterations in DMEs after KRAS mutation, thus causing some difficulties in understanding the fate of drugs in vivo and leading to confusion about the efficacy and side effects.</ns0:p><ns0:p>MS-based quantifications are different from traditional immunogenic methods and can increase the sensitivity and high throughput of the absolute quantification of proteins. In our study, an LC-MS/MS method was employed to determine the protein expression of DMEs in WT and KRAS mice (J. <ns0:ref type='bibr'>Chen et al. 2017</ns0:ref>). In addition, the possible mechanism for the variations in DMEs was investigated. We intend to provide a valuable reference for pharmacological evaluation and rational clinical drug use in patients with KRAS mutated lung cancer. </ns0:p></ns0:div>
<ns0:div><ns0:head>Animals.</ns0:head><ns0:p>Male and female C57 BL/6 mice <ns0:ref type='bibr'>(5, 10, 15, 20, 26 weeks, n=5)</ns0:ref> were obtained from Vital River Laboratory Animal Technology <ns0:ref type='bibr'>Co. Ltd (Beijing, China)</ns0:ref>. Male and female B6.129S-Kras tm3Tyj /Nci (K-ras LA2 ) <ns0:ref type='bibr'>(5, 10, 15, 20, 26 weeks, n=5)</ns0:ref> were acquired from National Cancer Insitute. The animals were kept at controlled temperature of 24-26 ℃ and humidity of 50-60%, with a 12 h light-dark cycle. The permission of all animal experiments was obtained from the Institutional Animal Care and Use Committee of the International Institute for Translational Chinese Medicine (IITCM_20171105). Before the experiment, animals were fasted overnight but allowed free access to water. All procedures were performed under diethyl ether anesthesia and the efforts were made to minimize suffering. After the animal experiment was completed, the animal bodies were frozen and sent to professionals for harmless treatment.</ns0:p><ns0:p>Histopathological analysis of lung tissues. Lung tissues were acquired from KRAS mice and wild-type (WT) mice. The morphology of lung tissues was observed under stereoscopic microscope (Leica, M165C), then tissues were fixed by 4% paraformaldehyde. After paraformaldehyde fixation and paraffin embedding, mouse lung tissues were sliced and stained with hematoxylin for 30 s and 0.5% eosin for 10 s, and covered with neutral gum. The images were obtained under microscopy (NIKON Eclipseci, Japan).</ns0:p><ns0:p>Preparation of mouse S9 fractions. Mouse tissue (liver, kidney and intestine) were isolated from WT and KRAS mice. The tissues were minced and washed with ice cold saline. Ice-cold homogenization buffer (50 mM potassium phosphate, 250 mM sucrose, 1 mM EDTA, PH 7.4) with 0.28 mM phenylmethylsulfonyl fluoride (PMSF) was added to the minced tissues and homogenized until an even suspension was obtained. Then the suspension was centrifuged at 9,000 × g for 20 min at 4°C. The supernatant was collected as S9 fractions (L. <ns0:ref type='bibr'>Tang et al. 2012;</ns0:ref><ns0:ref type='bibr'>Zhu et al. 2010</ns0:ref>). Liver tissue was handled with a single sample of each mouse while kidney and intestine tissues of each group of mice (n=5) were mixed into same sample. Protein concentrations of mouse S9 fractions were detected by coomassie brilliant blue and the bovine serum albumin was selected as the standard. Manuscript to be reviewed CYP3A11, CYP3A25, CYP7A1 and CYP27A1), ten isoforms of UGTs (UGT1A1, UGT1A2, UGT1A5, UGT1A6a, UGT1A9, UGT2A3, UGT2B1, UGT2B5, UGT2B34 and UGT2B36) and three isoforms of SULTs (SULT1A1, SULT1B1 and SULT1D1) were analyzed. The methods of sample preparation and quantifying DMEs amounts by UHPLC/MS-MS were consistent with our previous study and dynamic MRM chromatograms of 21 subtypes were displayed in Supplemental Figure <ns0:ref type='figure'>1</ns0:ref> (J. <ns0:ref type='bibr'>Chen et al. 2017</ns0:ref>). Samples were analyzed by using an Agilent 6490 triple quadruple mass spectrometer coupled with 1290 Infinite UHPLC system. A Poroshell C 18 column (2.1 mm × 100 nm, 2.7 μm; Agil m; Agilent Technologies) was used for separation. In this study, the protein amounts of DMEs were represented in the form of pmol protein per S9 fraction protein (pmol/mg). The quantification of protein levels was performed two independent experiments.</ns0:p></ns0:div>
<ns0:div><ns0:head>LC-MS</ns0:head><ns0:p>All samples were performed in triplicate in each independent experiment and data were presented as mean ± SD.</ns0:p><ns0:p>Enzyme assays of liver S9 fractions. Enzyme activities of CYP2E1, CYP3A11, UGT1A9, SULT1A1 and SULT1D1 were measured by specific probe substrates in vitro (chlorzoxazone, testosterone, propofol, p-nitrophenol and dopamine, respectively). The enzyme activities of these isoforms in mice were determined by incubating S9 fractions with appropriate substrate concentrations. Production of metabolites was quantified to value the activities of these isoforms between WT and KRAS mice at their different age. The incubation systems of CYPs, UGTs and SULTs, are in accordance with our previous articles with minor modification (C. <ns0:ref type='bibr'>Xie et al. 2017;</ns0:ref><ns0:ref type='bibr'>T. Yan et al. 2015;</ns0:ref><ns0:ref type='bibr'>H. Zheng et al. 2018</ns0:ref>). In order to terminate the reaction, 200 μm; Agil L methanol with 200 nM genistein was added. Then, the solution was vortexed and thereafter centrifuged for 30 min at 18000 g. Then the supernatant of all samples was injected to analyze by LC-MS/MS.</ns0:p><ns0:p>The enzyme activity was measured from 2 independent experiments. Each sample was performed in triplicates in each independent experiment and data was presented as mean ± SD.</ns0:p><ns0:p>Western blotting. The protein levels of aryl hydrocarbon receptor (AHR), bile acid receptor (FXR), pregnane X receptor (PXR) and peroxisome proliferator-activated receptor (PPARα) were determined in male WT and KRAS mice at 26 weeks, and β-actin was used as an internal control.</ns0:p><ns0:p>The S9 samples were mixed with 5 × loading buffer and the mixture was denatured at 100 °C for 5 min. An equal amount of protein (40 μm; Agil g) was separated by SDS-PAGE at a voltage of 120 V to the correct band size and the protein was subsequently transferred from the gel to the PVDF membrane. Then, the membrane was blocked for 1 h with 5% non-fat milk (w/v) in Tris-buffered saline containing 0.1% Tween-20 (TBST). The corresponding primary antibodies, against mouse peroxisome proliferator-activated receptor (PPARα, sc-398394), pregnane X receptor (PXR, ab118336), bile acid receptor (FXR, ab28480) and aryl hydrocarbon receptor (AHR, ab2769) and β-ACTIN (from Cell Signaling Technology, CST, Boston, USA) were diluted to a recommended dilution of 1:1000 with 5% non-fat milk according to the manufacturer's instructions. After blocking, the membrane was incubated with the corresponding primary antibodies at 4 °C overnight with gentle shaking and was then washed before being incubated with the corresponding secondary antibody at a dilution of 1:2000-1:3000 for 1 h at room temperature.</ns0:p><ns0:p>ECL chemiluminescence was used to detect the signals and each protein band was quantified by Image J (National Institutes of Health, Hercules, CA, USA). The WB analysis was performed from 2 independent experiments, and each target protein was analyzed twice in each independent experiment. The data was presented as mean ± SD.</ns0:p></ns0:div>
<ns0:div><ns0:head>Data analysis.</ns0:head><ns0:p>One-way ANOVA analysis, non-parametric test and independent sample t tests were conducted using SPSS 19.0 to evaluate statistical differences. Correlation analyses were performed using SPSS 19.0 and GraphPad Prism 7, according to the Pearson product-moment correlation for normal related data and Spearman's rank correlation for non-normally related data.</ns0:p><ns0:p>Partial least squares discriminant analysis (PLS-DA) was performed to visualize the changes of DMEs after KRAS mutation using SIMCA-P 14.0 tool (Umetrics, Umea, Sweden). In each case, a value of p < 0.05 denotes statistical significance for all of the statistical analyses.</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:01:45080:3:0:NEW 18 Sep 2020)</ns0:p><ns0:p>Manuscript to be reviewed</ns0:p></ns0:div>
<ns0:div><ns0:head>RESULTS</ns0:head><ns0:p>The phenotypic characteristics of KRAS mice. Pulmonary morphology was observed using a stereoscope (Fig. <ns0:ref type='figure'>1A-B</ns0:ref>). Compared with those in the WT mice, many lung nodules were observed in the lung tissues of the KRAS mice (Fig. <ns0:ref type='figure'>1B</ns0:ref>, As the arrows point), and the lung tissues appeared dull overall. Histological and pathological features of the WT and KRAS mice were detected by H&E staining (Fig. <ns0:ref type='figure'>1C</ns0:ref>-H and Supplemental Fig. <ns0:ref type='figure' target='#fig_10'>2</ns0:ref>). The morphology of lung cells in the WT mice was normal, whereas in the KRAS mice, the lung cells were hyperproliferative (Fig. <ns0:ref type='figure'>1F</ns0:ref>-H and Supplemental Fig. </ns0:p></ns0:div>
<ns0:div><ns0:head>Alterations in DME activities of DMEs by KRAS mutation.</ns0:head><ns0:p>As shown in Fig. <ns0:ref type='figure'>3</ns0:ref>, the activity of CYP3A11 was significantly downregulated in the male KRAS mice with aging. Furthermore, there were significant differences between the WT and KRAS mice. UGT1A9 gradually declined in the male mice from 5 to 26 weeks. SULT1A1 and SULT1D1 displayed larger differences at 20 and 26 weeks in the male KRAS mice than in the WT mice. SULT1A1 increased by 2. Protein expression profiles of AHR, FXR, PPARα (H-2) an H-2) and P ) and PXR. To evaluate the protein levels of receptors, we dissected liver tissue from the KRAS mice. As shown in Fig. <ns0:ref type='figure'>5</ns0:ref>, the protein expression levels of AHR, FXR and PPARα (H-2) were downregulated by <ns0:ref type='bibr'>40.22%, 20.90% and 26.76%</ns0:ref> in the livers of the male KRAS mice, respectively (p = 0.000, 0.035 and 0.005, respectively). Compared to the WT mice, the KRAS mice showed no significant difference in the protein amount of PXR (decreased by 13.52%, p = 0.109). In the female mice (Supplemental Fig. <ns0:ref type='figure'>5</ns0:ref>), the protein expression levels of AHR, FXR, PPARα (H-2) and PXR were downregulated by 12.00%, 10.14%, 5.60% and 38.06% in the liver, respectively (p = 0.466, 0.442, 0.710 and 0.074, respectively).</ns0:p></ns0:div>
<ns0:div><ns0:head>Tissue distribution of DMEs.</ns0:head><ns0:p>To evaluate the tissue distribution of DMEs, we present data for male mice at 10 weeks as an example. CYP2C29/CYP2D22/CYP3A11/CYP3A25/CYP27A1 (LLOQ). In the kidney, UGT1A2 was detected, but the other UGT isoforms were below the lower limit of quantification; for SULT isoforms, SULT1D1 was detected, but the others were all below the lower limit of quantification.</ns0:p><ns0:p>SULT1D1, CYP1B1, CYP2E1, CYP7A1 and UGT1A2 had the highest protein expression levels.</ns0:p><ns0:p>DME variations based on sex. <ns0:ref type='bibr'>Fig. 6</ns0:ref> shows that the the sex-related changes in DMEs have a similar trend in both the WT and KRAS mice at 10 weeks of age. Therefore, we mainly discuss the differences in protein content in WT mice. In kidney tissue, CYP2E1 was male-predominant, while UGT1A2 was female-predominant. In liver tissue, the content of UGT2B1 was significantly higher in the male mice than in the female mice.</ns0:p><ns0:p>Variations in DME protein content with increased age. In the liver, CYP7A1 increased with increasing age in the male mice (Fig. <ns0:ref type='figure' target='#fig_12'>7G</ns0:ref>); UGT1A9 showed a decreasing trend with increasing age (Fig. <ns0:ref type='figure' target='#fig_12'>7L</ns0:ref>). In the intestine, the CYP isoforms showed no significant changes at different ages (Fig. <ns0:ref type='figure' target='#fig_13'>8A-E</ns0:ref>). UGT2B34 showed a decreasing trend in the male WT and KRAS mice with Manuscript to be reviewed increasing age (Fig. <ns0:ref type='figure' target='#fig_13'>8H</ns0:ref>), and SULT1A1 showed an increasing trend with increasing age (Fig. <ns0:ref type='figure' target='#fig_13'>8I</ns0:ref>).</ns0:p><ns0:p>In the kidney, CYP1B1 displayed a decreasing trend in the male WT and KRAS mice with increasing age (Fig. <ns0:ref type='figure' target='#fig_13'>8L</ns0:ref>). Supplemental Fig. <ns0:ref type='figure' target='#fig_11'>6</ns0:ref> shows the changes of protein amount with aging in female WT and KRAS mice. In the liver, the protein amount of CYP2C29 decreased 2.9-fold at 15 weeks compared to that at 5 weeks in the WT mice. This pattern of changes did not appear in the KRAS mice. High individual differences in the protein amounts in female mice were observed. There are no significant differences at different ages.</ns0:p></ns0:div>
<ns0:div><ns0:head>DISCUSSION</ns0:head><ns0:p>In this study, we systematically investigated the alterations of DMEs in KRAS mice of different ages and sexes, with the aim of providing a better explanation for the clinically observed variation in the efficacy and toxicity of anticancer drugs in KRAS-mutant lung cancer patients.</ns0:p><ns0:p>Currently, limited information is available concerning the changes in DMEs in patients with KRAS mutant lung cancer.</ns0:p><ns0:p>The absolute protein contents of 21 metabolic enzymes in KRAS mice were simultaneously determined by the LC-MS/MS approach. In our study, the protein expression levels of CYP2C29 and CYP3A11 were significantly downregulated in the male KRAS mice at 26 weeks of age. A were upregulated in the male KRAS mice at 20 and 26 weeks. A number of studies have been conducted on SULT in different cancers, but many conflicting outcomes have been reported (Y. <ns0:ref type='bibr'>Jiang et al. 2010)</ns0:ref>. Some authors showed a potential association between SULT1A1 polymorphisms and breast cancer, but inconsistent results also exist (Y. <ns0:ref type='bibr'>Jiang et al. 2010</ns0:ref>).</ns0:p><ns0:p>Relevant studies have reported that SULT1A3 may be a diagnostic marker for osteosarcoma, and SULT1A3 protein upregulation is closely related to the occurrence and development of cancer (X. <ns0:ref type='bibr'>Chen et al. 2014;</ns0:ref><ns0:ref type='bibr'>C. Xie et al. 2017</ns0:ref>). SULT1D1 is a pseudogene in humans, Sult1d1 encodes protein expression in mice, and its functions are similar to those of human SULT1A3 (S. <ns0:ref type='bibr'>Wong et al. 2010</ns0:ref>). Our results also revealed increased protein expression of SULT1D1 in male mice after KRAS mutation. This finding is beneficial to explain the metabolic characteristics of SULT1D1metabolized drugs in KRAS mice. The expression of SULT1A3 should be further explored in KRAS-mutant lung cancer patients.</ns0:p><ns0:p>To further explore the changes in enzymatic activity, we used specific probe substrates to detect the enzymatic status in the KRAS mice. In our present study, we found the CYP2C29/CYP3A11/SULT1A1/SULT1D1 displayed significant changes in protein expression in the liver of male WT and KRAS mice. Therefore, we are intended to research their activities in the liver tissue. We failed to find an authoritative and specific probe to study the activity of CYP2C29. Chlorzoxazone and Propofol are usually used as specific substrates to study the activities of CYP2E1 and UGT1A9. Their good correlation between protein expression and activities indicated the protein quantification results are credible. Therefore, we select them for the enzyme activity test. Notably, SULT1A1 and SULT1D1 activity was upregulated at 20 and 26 weeks in the male KRAS mice (Fig. <ns0:ref type='figure'>3</ns0:ref> and Supplemental Fig. <ns0:ref type='figure'>4</ns0:ref>). This result was consistent with</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:01:45080:3:0:NEW 18 Sep 2020)</ns0:p><ns0:p>Manuscript to be reviewed their protein expression levels. In this context, a high degree of correlation was observed between the enzymatic activity and protein level (Fig. <ns0:ref type='figure'>4</ns0:ref>). For the poor correlation between the enzyme activity and protein level of CYP3A11, the nonspecificity of the substrate may be a possible reason. The FDA (USA) reported that testosterone was metabolized by CYP3A4 and CYP3A5. In addition, the protein structure could affect the activity. CYP3A4 showed a significant genetic polymorphism in individuals, causing a flexible three-dimensional structure of CYP3A4 <ns0:ref type='bibr'>(Werk and Cascorbi 2014)</ns0:ref>. Moreover, the genetic polymorphism of CYP2D6 (ortholog of CYP2D22 in mice) could induce variations in the expression or function of CYP3A4 <ns0:ref type='bibr'>(Werk and Cascorbi 2014)</ns0:ref>. Generally, these findings indicate that the protein expression levels of some DMEs could be applied to forecast the enzymatic activities regarding drug metabolism. Changes in the ability of DMEs to metabolize drugs could lead to differences in the ADME properties of drugs, thereby affecting drug efficacy and toxicity in the body.</ns0:p><ns0:p>The expression of DMEs is regulated by the binding of xenobiotics to receptors, such as the aryl hydrocarbon receptor (AHR), the murine pregnane X receptor (PXR), peroxisome proliferator-activated receptor (PPARα) and bile acid receptor (FXR) (S. <ns0:ref type='bibr'>Anakk et al. 2003;</ns0:ref><ns0:ref type='bibr'>C. HANDSCHIN 2003;</ns0:ref><ns0:ref type='bibr'>HONKAKOSKI and NEGISHI 2000)</ns0:ref>. The decrease in receptor levels may contribute to the emergence of changes in DME expression (S. <ns0:ref type='bibr'>Anakk et al. 2003;</ns0:ref><ns0:ref type='bibr'>L. Li et al. 2009</ns0:ref>; J. E. <ns0:ref type='bibr'>Moscovitz et al. 2018</ns0:ref>). Moreover, some reports have suggested that disease status (e.g., cancer and inflammation) can affect the expression and activity of DMEs via specific receptors <ns0:ref type='bibr'>(Lamba et al. 2016;</ns0:ref><ns0:ref type='bibr'>H. Chen et al. 2014;</ns0:ref><ns0:ref type='bibr'>A. Schröder et al. 2011</ns0:ref>). Therefore, we further studied the changes in receptor expression after KRAS mutation. In our study, the protein expression levels of AHR, FXR and PPARα were downregulated in the livers of the male KRAS <ns0:ref type='bibr'>Moscovitz et al. 2018</ns0:ref>). FXR, an important regulator of lipid and glucose homeostasis, is involved in the expression of CYP7A1 and CYP27A1 (Sánchez 2018). Hence, in our study, we speculate that these variations in DME expression may be regulated by decreased receptors of AHR, FXR and PPARα.</ns0:p><ns0:p>For sex-difference, CYP2E1 showed significant male-specificity in kidney tissue. CYP2E1</ns0:p><ns0:p>mediates the metabolism of many low molecular weight organic compounds (such as ethanol and acetone) and some drugs (such as p-nitrophenol, caffeine, chlorzoxazone, etc.) (S. <ns0:ref type='bibr'>Löfgren et al. 2004;</ns0:ref><ns0:ref type='bibr'>Zuber et al. 2002)</ns0:ref>. Therefore, in regard to the intake of these exogenous substances, we should consider the effects related to sex differences in patients. For UGT enzymes, UGT2B1 exhibited male-predominant expression in the liver tissue. Conversely, UGT1A1 and UGT1A5 expression in the liver and UGT1A2 in the kidney are female-predominant, whereas UGT2B1 exhibited male-predominant expression in liver tissue. These results are consistent with previous reports (D. <ns0:ref type='bibr'>B. Buckley and</ns0:ref><ns0:ref type='bibr'>C. D. Klaassen 2007, 2009)</ns0:ref>. The female-predominant UGT1A1</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:01:45080:3:0:NEW 18 Sep 2020)</ns0:p><ns0:p>Manuscript to be reviewed expression accounts for the higher bilirubin-UGT activity in females. UGT1A and UGT2B are the primary families of UGT enzymes, involved in the inactivation of > 30% of drugs currently used in the clinic (C. <ns0:ref type='bibr'>Guillemette et al. 2014</ns0:ref>). These sex-specific expressions may be crucial in understanding the mechanisms by which many drugs display variations in metabolism and clearance.</ns0:p><ns0:p>For age-related difference, except for UGT1A9, the majority of DMEs showed no significant changes from 5 to 26 weeks of age in female and male WT and KRAS mice. UGT1A9, major UGT isoforms expressed in the liver (~6% of hepatic UGT expression), is responsible for the glucuronidation of multiple endogenous substances (e.g., thyroid hormones) and drugs (e.g., acetaminophen and propofol) (S. <ns0:ref type='bibr'>Cho et al. 2016</ns0:ref>). The activity and protein expression of UGT1A9 appeared to decrease in the liver of female and male WT mice. Related studies indicated that UGT1A9 activity showed a downward trend from 6 weeks to 52 weeks in mice with a FVB background (H. <ns0:ref type='bibr'>Zheng et al. 2018</ns0:ref>). Therefore, the appropriate dosage should be considered when patients of different ages are prescribed drugs metabolized by UGT1A9.</ns0:p><ns0:p>For tissue-related differences, abundant CYP enzymes were expressed in the liver, predominantly CYP2D22, CYP2C29, CYP2E1 and CYP3A11 (Fig. <ns0:ref type='figure' target='#fig_11'>6A-B</ns0:ref>). A previous studies also demonstrated that the protein contents of these isoforms were high in the liver (J. <ns0:ref type='bibr'>Chen et al. 2017;</ns0:ref><ns0:ref type='bibr'>C. Gröer et al. 2014</ns0:ref>). In the intestine, the CYP1B1, CYP2D22, CYP2E1 and CYP3A11 protein levels were significantly higher than the levels of other proteins. Mouse phase I enzymes (CYP2D22, CYP2E1 and CYP3A11) are orthologs of the corresponding human enzymes (CYP2D6, CYP2E1 and CYP3A4), in charge of major phase I-dependent metabolism in marketed drugs <ns0:ref type='bibr'>(Liu 2013;</ns0:ref><ns0:ref type='bibr'>G. Ruaño et al. 2012</ns0:ref>). Hence, optimal drug administration routes Manuscript to be reviewed should be considered when these enzymes are involved in the inactivation or activation of drugs.</ns0:p></ns0:div>
<ns0:div><ns0:head>CONCLUSION</ns0:head><ns0:p>Taken together, our data showed significant decrease in CYP3A11 and CYP2C29, but an increase in SULT1A1 and SULT1D1 in the KRAS mice at 26 weeks. These DMEs all participate in the metabolism of drugs. Therefore, we hope that these results could provide useful guidance or a theoretical basis for further drug research and implementation.</ns0:p><ns0:note type='other'>Figure 1</ns0:note><ns0:p>Morphology and H&E staining of lung tissues from the WT and KRAS mice. Manuscript to be reviewed </ns0:p><ns0:note type='other'>Figure 3</ns0:note><ns0:p>Changes in the enzyme activity of CYP2E1, CYP3A11, UGT1A9, SULT1A1 and SULT1D1 in liver tissues of the male KRAS and WT mice at different ages (n=5).</ns0:p><ns0:p>The solid line represents male KRAS mice, and the dashed line represents male WT mice.</ns0:p><ns0:p>Each data point is presented as the mean ± SD. For the comparison between KRAS and WT at the same age, the data were analyzed by independent sample t tests (for normally distributed data) and Mann-Whitney U analysis (for non-normally distributed data). The symbol '*' indicates a significant difference between the male WT and KRAS mice at the same age, p < 0.05. For different ages compared to 5 weeks, the data were analyzed by oneway ANOVA (for normally distributed data) and Kruskal-Wallis H analysis (for non-normally distributed data). We adjusted the significance level α to 0.0125 according to the Bonferroni correction (0.05/4=0.0125). The symbols 'A' and 'a' indicate significant differences in the male WT and KRAS mice at 10, 15, 20 and 26 weeks relative to 5 weeks, p < 0.0125.</ns0:p></ns0:div>
<ns0:div><ns0:head>Figure 4</ns0:head><ns0:p>Correlation between the protein expression and activity of CYP2E1, CYP3A11, UGT1A9, SULT1A1 and SULT1D1 in the liver tissue (n=100).</ns0:p><ns0:p>The correlation between the protein expression and activity included 5, 10, 15, 20 and 26 weeks, which were analyzed together. DME in the liver was determined using an isotope label-free LC-MS/MS method. The enzyme activities of DMEs were measured using probe substrates. All measurements were performed in triplicate and the data are presented as the mean ± SD. Pearson product correlation and Spearman's rank correlation were used to analyze the correlation. Regression line is shown for significant correlation at p < 0.05.</ns0:p></ns0:div>
<ns0:div><ns0:head>Figure 5</ns0:head><ns0:p>Protein expression levels of AHR, FXR, PPARα and PXR in the male WT (n=5) and KRAS mice (n=4) at 26 weeks.</ns0:p><ns0:p>(A) The mprint of five proteins was represented and β-ACTIN was used as an internal control.</ns0:p><ns0:p>(B) The data on protein expression levels was shown as a box chart. The data were analyzed by independent sample t tests (for normally distributed data) and Mann-Whitney U analysis (for non-normally distributed data). The symbol '*' indicates a significance difference of protein expression level in the KRAS mice relative to that in the WT mice, p < 0.05. We adjusted the significance level α to 0.0125 according to the Bonferroni correction (0.05/4=0.0125). The symbols 'A' and 'a' indicate significant differences in the male WT and KRAS mice at 10, 15, 20 and 26 weeks relative to 5 weeks, p < 0.0125.</ns0:p></ns0:div><ns0:figure xml:id='fig_0'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:01:45080:3:0:NEW 18 Sep 2020) Manuscript to be reviewed MATERIALS AND METHODS Chemicals and Reagents. Ammonium bicarbonate (AB), dithiothreitol (DTT), iodoacetamide (IAA), trifluoroacetic acid (TFA) and phenylmethanesulfonyl fluoride (PMSF) were bought from Sigma-Aldrich, USA. Sequencing grade modified trypsin was obtained from Promega (Madison, WI). All peptides and internal standard (purity > 95%) were got from Your R&D Partner. HPLCgrade methanol, formic acid and acetonitrile were acquired from Merck (Darmstadt, Germany). NADPH solution A and NADPH solution B were got from BD Bioscience, USA. Alamethicin, Tetracycline, Glucosyl monophosphate, Uridine diphosphate glucuronic acid (UDPGA), 3 ，phosphoadenosine-5 ， -phosphosulfate (PAPS), MgCl2, Chlorzoxazone, Testosterone, 6βhydroxytestosterone, Propofol, Dopamine, 6-hydroxy chlorzoxazone, Propofol-glucuronide and 4-Nitrophenyl sulfate metabolite were acquired from Sigma-Aldrich, USA. Dopamine 3-Osulfate and dopamine 4-O-sulfate were got from TRC, Canada. P-Nitrophenol was bought from Aladdin, China. Genistein and ammonium acetate were got from Chengdu Mansite Biotechnology Co., Ltd. and Dalian Meilun Biotechnology Co., Ltd., respectively. Coomassie brilliant blue, providing for protein measurement, was bought from Bio-Rad (Hercules, California, USA).</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_1'><ns0:head /><ns0:label /><ns0:figDesc>/MS analysis. Eight isoforms of CYPs (CYP1B1, CYP2C29, CYP2D22, CYP2E1, PeerJ reviewing PDF | (2020:01:45080:3:0:NEW 18 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_2'><ns0:head /><ns0:label /><ns0:figDesc>2, As the arrows point). The nuclei were deeply stained and lager.Alterations in the protein contents of DMEs by KRAS mutation. In male mice, PLS-DA analysis was applied to evaluate the clustering between the male WT and KRAS mice based on the expression of 21 DMEs (Fig.2A-E). We observed obvious distinctions between the male WT and KRAS mice at 5, 10, 15, 20 and 26 weeks. These results demonstrated the differences in DMEs among them. As shown in Fig.2Hin liver tissue, SULT1A1 increased by 2.4-fold at 5 weeks (p = 0.016); CYP27A1 and UGT1A1 increased by 1.9-fold and 2.3-fold at 15 weeks respectively (p = 0.001 and p > 0.05, respectively); SULT1A1 and SULT1D1 were upregulated by 3.4-fold and 1.8-fold at 20 weeks, respectively (p = 0.015 and p = 0.017, respectively); and at 26 weeks, SULT1A1 and SULT1D1 were upregulated by 2.0-fold and 1.8-fold respectively (p > 0.05 and p = 0.031, respectively), and CYP2C29, CYP3A11, CYP27A1 and UGT1A5 decreased by 3.7-fold (p = 0.005), 4.4-fold (p = 0.004), 2.1-fold (p = 0.043) and 2.3-fold (p = 0.014), respectively. In intestinal tissue, SULT1B1 and SULT1D1 increased by 3.2-fold and 2.9-fold at PeerJ reviewing PDF | (2020:01:45080:3:0:NEW 18 Sep 2020) Manuscript to be reviewed 20 weeks, respectively (p > 0.05); and SULT1A1 and SULT1D1 increased by 2.0-fold (p > 0.05) and 1.8-fold at 26 weeks (p = 0.024), respectively. In kidney tissue, SULT1D1 was upregulated by 3.0-fold at 26 weeks (p > 0.05). Supplemental Fig. 3 shows some changes in DMEs with KRAS mutations in female KRAS mice. In liver tissue, UGT1A9 decreased by 2.3-fold at 5 weeks (p = 0.002); UGT2B1 decreased by 2.1-fold at 10 weeks (p = 0.004); CYP2C29, CYP2D22, CYP2E1, CYP3A11, CYP27A1, UGT1A1, UGT2A3, UGT2B5 and UGT2B1 were upregulated by 2.5-(p = 0.014), 1.7-(p = 0.018), 2.8-(p = 0.001), 2.6-(p = 0.001), 1.8-(p = 0.014), 2.5-(p = 0.008), 2.2-(p = 0.005), 2.3fold (p = 0.014), and 2.4 folds (p = 0.002), respectively, at 15 weeks; SULT1A1 increased by 2.2 folds at 20 weeks (p > 0.05); and CYP3A11 decreased by 2.1-fold at 26 weeks (p = 0.033). In intestine tissue, SULT1B1 and SULT1D1 increased by 3.4-fold (p > 0.05) and 4.7-fold (p = 0.042), respectively, at 26 weeks. In kidney tissue, SULT1D1 increased by 1.9 folds at 26 weeks (p = 0.013).</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_3'><ns0:head /><ns0:label /><ns0:figDesc>and 3.9-fold (p = 0.008), respectively. SULT1D1 was upregulated by 7.0-(p = 0.007) and 3.5-fold (p > 0.05), respectively. Supplemental Fig.4shows the activities in the female WT and KRAS mice at different ages. The activities of CYP2E1 and SULT1D1 displayed no significant differences in the WT and KRAS mice with increasing age. CYP3A11 displayed an increasing PeerJ reviewingPDF | (2020:01:45080:3:0:NEW 18 Sep 2020)Manuscript to be reviewed tendency from 5 to 26 weeks. The activity of UGT1A9 showed a significant decrease at 15, 20 and 26 weeks compared to that at 5 weeks. At 15 weeks, the activity of SULT1A1 was markedly different in the KRAS mice relative to the WT mice (p = 0.022).Correlation of protein content and enzyme activity of DMEs. The enzyme activities of CYP2E1, CYP3A11, UGT1A9, SULT1A1 and SULT1D1 were compared with their protein contents. The correlation analysis assessed the protein levels quantified by LC-MS/MS and the activities detected by the specific probes. As shown in Fig.4, there was a good correlation between enzyme activity and protein content (CYP2E1, r 2 = 0.54, p < 0.001; UGT1A9, r 2 = 0.55, p < 0.001; SULT1A1, r 2 = 0.62, p < 0.001; SULT1D1, r 2 = 0.89, p < 0.001). A poor correlation for CYP3A11 was observed in the mouse liver (r 2 < 0.50, p > 0.05).</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_4'><ns0:head /><ns0:label /><ns0:figDesc>Fig. 6 displays the distribution of DMEs in liver, intestine and kidney tissue. In liver, CYP2C29 > CYP2D22 > CYP3A11 ≈ CYP2E1 ≈ CYP1B1 > PeerJ reviewing PDF | (2020:01:45080:3:0:NEW 18 Sep 2020)Manuscript to be reviewedCYP7A1 > CYP27A1 ≈ CYP3A25; UGT2B5 > UGT2B1 > UGT1A6a > UGT1A1 > UGT2B34≈ UGT2B36 > UGT2A3 > UGT1A9 ≈ UGT1A5 > UGT1A2 (lower limit of quantification, LLOQ); SULT1A1 > SULT1D1 > SULT1B1 (LLOQ). The protein contents of UGT2B5, UGT2B1, UGT1A6a, CYP2C29, CYP2D22, UGT1A1 and SULT1A1 were the highest. The protein contents of CYP1B1, UGT2B34, SULT1B1, CYP2D22, CYP3A11, SULT1A1 and CYP3A25 were the highest. In the kidney, CYP1B1 > CYP2E1 > CYP7A1 ></ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_5'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:01:45080:3:0:NEW 18 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_6'><ns0:head /><ns0:label /><ns0:figDesc>previous study indicated that hepatic DMEs are reduced during infection and inflammation in humans, rats, and mice(Nozomu Moriya et al. 2012). Similar results were also reported that a decrease in Cyp gene expression and enzymatic activity was observed in a dextran sulfate sodium (DSS)-induced mouse model of ulcerative colitis(Yoshiiki K. et al. 2014). CYP2C29 is the major arachidonate CYP2C epoxygenase in mice(Komal S. et al. 2009). The decreased expression of CYP2C29 is closely related to the occurrence and development of inflammation. Related studies have shown that this decreased expression may be triggered by an increased production of inflammatory cytokines(Yoshiiki K. et al. 2014). CYP3A11 plays a vital role in the metabolism of various clinical anticancer drugs, such as erlotinib, cisplatin, sorafenib. The drug concentration in serum would change accordingly with enzymatic expression. The declining expression of CYP3A11 in the KRAS mice may cause some differences in efficacy or even side effects. With respect to the SULT family, the protein expression and activities of SULT1A1 and SULT1D1</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_7'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:01:45080:3:0:NEW 18 Sep 2020)Manuscript to be reviewed mice compared to the WT mice at 26 weeks. This phenomenon was not significant in the female KRAS mice. Major xenobiotic-sensing transcription factors, such as AHR and PXR, are involved in the regulation of the protein expression of DMEs. Related reports revealed that most core DMEs were positively correlated with AHR, PXR and PPARα, and their protein expression was downregulated in nearly 50% of the patients with hepatocellular carcinoma (HCC) (H.Chen et al. 2014; D. G. Hu et al. 2018; S. Zhong et al. 2016). Activation of AHR could induce the upregulation of Cyp1a/3a/Ugt1a1 mRNA expression, which would therefore not occur in Ahrnull mice (C.D.Klaassen and A.L. Slitt 2005; Nakajima et al. 2003). PPARα and PXR were implicated in the regulation of CYP3A/4A/1A1/2B6/2C8/2C9/2C19/UGT1A1 induction (J. E.</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_8'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:01:45080:3:0:NEW 18 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_9'><ns0:head>(</ns0:head><ns0:label /><ns0:figDesc>A-B) Pulmonary morphology in the WT and KRAS mice under a stereoscopic mirror. The arrows pointed out the lung nodules in KRAS mice. (C-H) Histological and pathological features of the WT and KRAS mice were detected by H&E staining at 26 weeks. The arrows partly pointed out the hyperproliferative lung cells in KRAS mice. PeerJ reviewing PDF | (2020:01:45080:3:0:NEW 18 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_10'><ns0:head>Figure 2 Alterations</ns0:head><ns0:label>2</ns0:label><ns0:figDesc>Figure 2</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_11'><ns0:head>Figure 6</ns0:head><ns0:label>6</ns0:label><ns0:figDesc>Figure 6</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_12'><ns0:head>Figure 7</ns0:head><ns0:label>7</ns0:label><ns0:figDesc>Figure 7</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_13'><ns0:head>Figure 8</ns0:head><ns0:label>8</ns0:label><ns0:figDesc>Figure 8</ns0:figDesc></ns0:figure>
<ns0:note place='foot'>PeerJ reviewing PDF | (2020:01:45080:3:0:NEW 18 Sep 2020)Manuscript to be reviewed</ns0:note>
</ns0:body>
" | "Dear Editor:
Thank you for your hard work and generous comments on the manuscript entitled “Changes and sex- and age-related differences in the expression of drug metabolizing enzymes in a KRAS-mutant mouse model of lung cancer” (ID: ms# 45080). The comments are all valuable and very helpful for revising and improving our paper. We have revised the manuscript according to your and the reviewer’ comments and the revised portions are marked in blue. We hope that the revised manuscript is sufficient for publication in Peer J.
We appreciate the editor / reviewer’s comments and hard work.
Warm regards,
Zhongqiu Liu, Ph.D.
Chair Professor of Pharmaceutics
Joint Laboratory for Translational Cancer Research of Chinese Medicine of the Ministry of Education of the People’s Republic of China, International Institute for Translational Chinese Medicine
Guangzhou University of Chinese Medicine
Guangzhou 510006
China
+8620-39358061
Email: liuzq@gzucm.edu.cn
Revised portions are marked in blue in the manuscript. The responses to the reviewers’ comments are as following:
Dear Editor’s
I am afraid that I seem to have found some discrepancies in different versions of Fig. 3: for example, in the first version, no significant differences were found in UGT1A9 at 20w, but upon applying the Bonferroni correction significant differences were found between KRAS and WT, as well as between WT 20w and WT 5w. This does not make sense, and I think some labeling mistakes were introduced. To clarify matters, please double-check every comparison and add the values of standard deviation and corresponding test statistics to the enzyme raw data in the Supporting information files.
Reply: Thank you very much for your hard work. We had checked every comparison and specific values were displayed in supporting information files. We did found a few labeling mistakes after the Bonferroni correction and the corresponding correction was done.
In Fig. 3C, we did found the significant differences between male WT and KRAS in 20 weeks, as well as between male KRAS 20 weeks and male KRAS 5 weeks (p =0.031 and p = 0.000, respectively). The Bonferroni correction was applied when different ages were compared with 5 weeks (padjusted = 0.0125). And the statistical analysis of criteria alpha was 0.05 between WT and KRAS mice.
In the first version, we made some mistakes when doing statistical analysis. We not only did not consider the multiple-comparison correction, but also put all the data together to consider issues such as normality and homogeneity of variance. Therefore, some different differences appeared in two versions. The detailed statistical approaches had been described in corresponding “Figure legends” and supporting information files. Meanwhile, the detailed statistical results and operation process were displayed in this letter and supporting information.
Reviewer 3
I would only ask the authors to rewrite/rephrase the sentence (line 70-71) using correct English.
Reply: Thank you very much for your good suggestion. We have rephrased the sentence as “Advanced tumours begin to appear in the lung of KRAS mice at 20 weeks and its life span is approximately 28 weeks.” in the revised manuscript (page 5, line 71-72).
Note: normally distributed data (n); non-normally distributed data (no); Test for homogeneity of variance: Uniform variance (UF); Uneven variance (UV).
" | Here is a paper. Please give your review comments after reading it. |
9,833 | "<ns0:abstract xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:p>Kidney renal clear cell carcinoma (KIRC) is the leading cause of kidney cancer-related deaths. Currently, there are no studies in tumor immunology investigating the use of signatures as a predictor of overall survival in KIRC patients. Our study attempts to establish an immune-related gene risk signature to predict clinical outcomes in KIRC. A total of 528 patients from The Cancer Genome Atlas (TCGA) database were included in our analysis and randomly divided into training (n = 315) and testing sets (n = 213). We collected 1,534 immune-related genes from the Immunology Database and Analysis Portal as candidates to construct our signature. LASSO-COX was used to find gene models with the highest predictive ability. We used survival and Cox analysis to test the model's independent prognostic ability. Univariate analysis identified 650 immune-related genes with prognostic abilities. After 1,000 iterations, we choose 14 of the most frequent and stable immune-related genes as our signature. We found that the signature was associated with M stage, T stage, and pathological staging. More importantly, the signature can independently predict clinical prognosis in KIRC patients. Gene Set Enrichment Analysis (GSEA) showed an association between our signature and critical metabolism pathways.</ns0:p><ns0:p>Our research established a model based upon 14 immune-related genes that predicted the prognosis of KIRC patients based on tumor immune microenvironments.</ns0:p></ns0:div>
</ns0:abstract>
<ns0:body xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:head>Introduction</ns0:head><ns0:p>Kidney cancer is one of the most common urological tumors worldwide, with approximately 403,262 new cases and 175,098 deaths associated with this form of cancer in 2018 <ns0:ref type='bibr' target='#b1'>(Bray et al. 2018)</ns0:ref>. Kidney renal clear cell carcinoma (KIRC) is the leading cause of kidney cancer mortality and the most common type, accounting for 85% of kidney cancers <ns0:ref type='bibr' target='#b15'>(Siegel et al. 2017;</ns0:ref><ns0:ref type='bibr' target='#b18'>Tseng 2016</ns0:ref>). Considering the aging population of KIRC patients and the rising expense of treatment, KIRC is gradually becoming the focus of geriatric cancer <ns0:ref type='bibr' target='#b17'>(Tan et al. 2015)</ns0:ref>. Despite the rapid development of cancer treatments, mortality rates in KIRC remain stagnant. With the progression of next generation sequencing and data mining techniques, it is urgent that we explore prognostic biomarkers for KIRC, using molecular characteristics and tumor immune environments to guide patient therapy.</ns0:p><ns0:p>Over the past decade our understanding of immune components, including the impact of tumor microenvironments on patient survival and therapy response <ns0:ref type='bibr' target='#b5'>(Grivennikov et al. 2010</ns0:ref>), has increased. Some studies have found that tumorinfiltrating immune cells are able to serve as either tumor suppressors or promoters in microenvironments. For example, CD8 + T cells have been associated with improved survival in cancer patients <ns0:ref type='bibr' target='#b3'>(Gajewski et al. 2013)</ns0:ref>, while tumor associated macrophages and regulatory T cells demonstrate the ability to promote tumor development <ns0:ref type='bibr' target='#b8'>(Nishikawa & Sakaguchi 2014;</ns0:ref><ns0:ref type='bibr' target='#b9'>Noy & Pollard 2014)</ns0:ref>. Considering the complexity and significance of tumor immune microenvironments, it is imperative that we investigate immune-related biomarkers for KIRC patients. Recent studies have provided insight into the KIRC immune signature <ns0:ref type='bibr' target='#b4'>(Geissler et al. 2015;</ns0:ref><ns0:ref type='bibr' target='#b13'>Şenbabaoğlu et al. 2016)</ns0:ref>. <ns0:ref type='bibr'>Şenbabaoğlu et al. (2016)</ns0:ref> found mRNA signatures with the potential to be immunotherapeutic biomarkers in KIRC. However, their investigations do not include immune-related genes for analysis nor do they establish a systematic immune-related gene-risk signature for KIRC patients. <ns0:ref type='bibr'>Khadirnaikar et al. (2019)</ns0:ref> utilized immune associated lncRNA (long noncoding RNA) to construct prognostic subtypes in KIRC patients. Our immune clusters were more robust and independent. Additionally, they concentrated on lncRNA, not the overall immune-related genes. <ns0:ref type='bibr'>Smith et al. (2018)</ns0:ref> constructed endogenous retroviral signatures for KIRC patients, but they did not investigate the prognostic ability of the signature in different subtypes of patients. Therefore, it is essential that we explore a systematic prognostic signature based on tumor immune environments in KIRC.</ns0:p><ns0:p>In our study, we used RNA-seq data from The Cancer Genome Atlas (TCGA) to find immune-related genes with prognostic ability and to establish an immune-related risk signature for KIRC. To assess the clinical potential of the signature, we investigated the association between the signature, clinical parameters, and patient survival. Gene set enrichment analysis (GSEA) was performed to explore the molecular characteristics of the signature.</ns0:p></ns0:div>
<ns0:div><ns0:head>Materials & Methods</ns0:head></ns0:div>
<ns0:div><ns0:head>Patient cohort</ns0:head><ns0:p>The TCGA database was used to collect the clinical and RNA-seq data of 528 KIRC patients. We randomly divided the dataset into training (n = 315) and test sets (n = 213). RNA-seq data was obtained to analyze the transcriptome profiling of RNA expression and were measured using fragments per kilobase of exon per million fragments mapped (FPKM). We performed a log2-based transformation to normalize RNA expression profiles. To ensure detection and total cohort <ns0:ref type='bibr' target='#b7'>(Harrell et al. 1982)</ns0:ref>. The independent prognostic ability of the immune-related risk signature was assessed using survival analysis as well as Cox analysis (R package survival, v2.42). We performed multivariable Cox regression to analyze the relationship between immune-related risk signatures and clinicopathological factors.</ns0:p></ns0:div>
<ns0:div><ns0:head>Gene set enrichment analysis</ns0:head><ns0:p>To identify the biological functions and pathways between high-and low-risk groups, we conducted GSEA to investigate potential biological mechanisms in the Molecular Signatures Database (MSigDB; <ns0:ref type='bibr' target='#b16'>(Subramanian et al. 2005)</ns0:ref>. GSEA was performed using GSEApy, a python wrapper for gene enrichment (https://pypi.org/project/gseapy/).</ns0:p><ns0:p>We selected C2 and C5, including pathway databases and GO terms, from the MSigDB. The gesa sub-command of GSEApy was used in GSEA with default parameters. Enriched gene sets with a false discovery rate (FDR) of less than 0.25 and a P-value of less than 0.05 were considered statistically significant.</ns0:p></ns0:div>
<ns0:div><ns0:head>Statistical analysis</ns0:head><ns0:p>Boxplots were created with the R package, ggplot2 (v3.0.0). The R package ComplexHeatmap (v1.18.1) was used to create heatmaps <ns0:ref type='bibr' target='#b6'>(Gu et al. 2016)</ns0:ref>. We counted C-index with R packages 'survcomp' <ns0:ref type='bibr' target='#b7'>(Harrell et al. 1982;</ns0:ref><ns0:ref type='bibr' target='#b12'>Schröder et al. 2011)</ns0:ref>. The student's t-test was performed for statistical comparison. We chose R to conduct statistical analysis (https://www.r-project.org/). P-values lower than 0.05 were considered statistically significant. The main code of analysis was pushed to github (https://github.com/huchua/KIRC).</ns0:p></ns0:div>
<ns0:div><ns0:head>Results</ns0:head></ns0:div>
<ns0:div><ns0:head>Establishment and validation of the immune-related gene signature in KIRC</ns0:head><ns0:p>Figure <ns0:ref type='figure' target='#fig_1'>1</ns0:ref> illustrates the workflow we used to develop an immune-related gene-risk signature. The immune-related gene signature was constructed within the KIRC training data set, while we applied the testing set to validate the signature.</ns0:p><ns0:p>After 1,000 iterations, seven unique gene models were selected (Figure <ns0:ref type='figure'>2A</ns0:ref>, Table <ns0:ref type='table'>S1</ns0:ref>). The model selected as the Manuscript to be reviewed immune-related gene-risk signature consisted of 14 genes and ranked the highest in frequency 282 times. Parameter selection in the LASSO-cox model was log lambda = -2.641 and alpha = 1. The univariate and multivariate Cox analysis of the 14 immune-related genes are shown in Table <ns0:ref type='table'>1 and Table 2</ns0:ref>. The principle component analysis of the 14 immune-related gene signature displayed a different distribution pattern between low-and high-risk groups when comparing the training, testing, and total cohort (Figure <ns0:ref type='figure'>2B-D</ns0:ref>). This indicated that the low-and high-risk groups had different immune phenotypes. In the training, testing, and total data set, the c-index was 0.7862, 0.6534, and 0.7367, respectively (P < 0.001; Figure <ns0:ref type='figure'>2E</ns0:ref>). In a time-dependent receiver operating characteristic curve (ROC) created for three datasets, area under the curve (AUC) values at 1, 3, and 5 years were 0.679, 0.63, 0.627; 0.65, 0.596, 0.568; and 0.618, 0.589, 0.59, respectively (Figure <ns0:ref type='figure'>S2</ns0:ref>). The 14 immune-related genes are AR, BID, BMP8A, CCL7, CCR10, FGF17, GDF1, IL20RB, IL4, KLRC2, LHB, SEMA3A, SEMA3G, and TXLNA. K-M (Kaplan-Meier) survival curves and gene expression of the 14 immune-related genes are shown in Figure <ns0:ref type='figure'>3</ns0:ref> and Figure <ns0:ref type='figure' target='#fig_7'>S3</ns0:ref>.</ns0:p></ns0:div>
<ns0:div><ns0:head>Correlation between the signature of immune-related genes and clinical parameters</ns0:head><ns0:p>Among the 14 immune-related genes, four genes (TXLNA, SEMA3G, AR, and BID) had a high expression, and 10 genes (IL20RB, CCR10, BMP8A, SEMA3A, CCL7, GDF1, KLRC2, LHB, FGF17, and IL4) had a low expression (Figure <ns0:ref type='figure' target='#fig_9'>4A</ns0:ref>, B; Figure <ns0:ref type='figure' target='#fig_2'>5</ns0:ref>). The relationship between the signature and clinical factors demonstrated that patients with advanced pathological staging, M stage, and T stage had a higher risk score than those with early stage disease (Figure <ns0:ref type='figure' target='#fig_9'>4A</ns0:ref>, B; Figure <ns0:ref type='figure' target='#fig_2'>5</ns0:ref>; Figure <ns0:ref type='figure' target='#fig_10'>6A, B, D</ns0:ref>). However, we did not find a correlation between the signature and N stage (Figure <ns0:ref type='figure' target='#fig_9'>4A</ns0:ref>, B; Figure <ns0:ref type='figure' target='#fig_2'>5</ns0:ref>; Figure <ns0:ref type='figure' target='#fig_10'>6C</ns0:ref>).</ns0:p></ns0:div>
<ns0:div><ns0:head>Influence of the immune-related gene signature on patient prognosis</ns0:head><ns0:p>We then assessed whether this signature influenced KIRC patient prognosis. Survival analysis showed that patients with a high-risk score were associated with poor survival outcomes in the training, testing, and total group sets (Figure <ns0:ref type='figure' target='#fig_9'>4C</ns0:ref>; Figure <ns0:ref type='figure' target='#fig_9'>4E</ns0:ref>; Figure <ns0:ref type='figure' target='#fig_11'>7B</ns0:ref>). We found that the signature also predicted survival outcomes in subgroups of KIRC patients, including stage I-II (Figure <ns0:ref type='figure' target='#fig_11'>7C</ns0:ref>), stage III-IV (Figure <ns0:ref type='figure' target='#fig_11'>7D</ns0:ref>), M0 stage (Figure <ns0:ref type='figure' target='#fig_11'>7E</ns0:ref>), M1 stage (Figure <ns0:ref type='figure' target='#fig_11'>7F</ns0:ref>), N0 stage (Figure <ns0:ref type='figure' target='#fig_11'>7G</ns0:ref>), N1 (Figure <ns0:ref type='figure' target='#fig_11'>7H</ns0:ref>), T1 (Figure <ns0:ref type='figure' target='#fig_11'>7I</ns0:ref>), T2 (Figure <ns0:ref type='figure' target='#fig_11'>7J</ns0:ref>), T3 (Figure <ns0:ref type='figure' target='#fig_11'>7K</ns0:ref>), and T4 (Figure <ns0:ref type='figure' target='#fig_11'>7L</ns0:ref>). Multivariate Cox analysis revealed that the risk signature was able to independently predict overall survival in KIRC patients (Figure <ns0:ref type='figure' target='#fig_9'>4D</ns0:ref>, F; Figure <ns0:ref type='figure' target='#fig_11'>7A</ns0:ref>; Table <ns0:ref type='table'>3</ns0:ref>).</ns0:p></ns0:div>
<ns0:div><ns0:head>Engagement of the immune-related gene risk signature in biological pathways and functions</ns0:head><ns0:p>GSEA was used to investigate the signature's biological pathways and functions. There were 177 KEGG pathways and 4,528 GO terms used in our investigation. Our analysis found that the signature was able to engage in a total of 19 enriched KEGG pathways (FDR < 0.25; Figure <ns0:ref type='figure' target='#fig_4'>8A</ns0:ref>). The low-risk signature was significantly correlated with 10 pathways, including the citrate cycle (TCA cycle) pathway, fatty acid metabolism pathway, propanoate metabolism pathway, butanoate metabolism pathway, peroxisome pathway, lysine degradation pathway, valine leucine and isoleucine degradation pathway, proximal tubule bicarbonate reclamation pathway, vasopressin regulated water reabsorption pathway, and the pyruvate metabolism pathway (Figure <ns0:ref type='figure' target='#fig_4'>8B</ns0:ref>). Similarly, eight GO annotations were enriched in the low-risk group (FDR < 0.25; Figure <ns0:ref type='figure' target='#fig_5'>9A, B</ns0:ref>).</ns0:p></ns0:div>
<ns0:div><ns0:head>Discussion</ns0:head><ns0:p>Our study used the TCGA database and immune-related genes to establish a KIRC signature consisting of 14 immunerelated genes. We found that patients in the high-risk group showed a positive association with M stage, T stage, and advanced pathological staging. Additionally, the signature exhibited strong prognostic abilities and independently predicted KIRC patient prognosis. Functional analysis highlighted the significance of our signature based on its involvement in many important pathways.</ns0:p><ns0:p>In our investigation, we used a total of 14 immune-related genes to construct our signature. In the signature, CCL7 increased the peripheral blood mononuclear cell recruitment in renal cell cancer through the inhibition of let-7d <ns0:ref type='bibr' target='#b11'>(Riihimaki et al. 2014</ns0:ref><ns0:ref type='bibr'>). Wyler et al. (2014)</ns0:ref> demonstrated the ability of CCL7 to recruit monocytes through CCR2, promoting renal cell cancer metastasis to the brain. This indicates that CCL7 is a major factor in the development of KIRC in tumor immune microenvironments and may be a potential immunotherapy target. Fibroblast growth factor receptor 17 (FGF17) is another immune-related gene in our signature. FGF17 demonstrates a variety of functions in cancer development. <ns0:ref type='bibr'>Gauglhofer et al. (2011)</ns0:ref> found that FGF17 was involved in the paracrine and autocrine signaling of hepatocellular carcinoma and promoted the neoangiogenesis of hepatocellular carcinoma. <ns0:ref type='bibr'>Heer et al. (2004)</ns0:ref> showed that FGF17 was overexpressed in prostate cancer and participated in prostate carcinogenesis. Our study showed that FGF17 plays an important role in KIRC based on tumor immunology. However, the underlying mechanism of FGF17 in KIRC immune microenvironments requires further investigation. In our signature, IL-4 is another well-studied immune-related biomarker for KIRC. IL-4, released by immune cells, controls the expression of B7-H1, thus altering T cell responses in KIRC <ns0:ref type='bibr' target='#b10'>(Quandt et al. 2014</ns0:ref>). The investigation conducted by <ns0:ref type='bibr'>Chang et al. (2015)</ns0:ref> demonstrated that IL-4 expression can predict KIRC patient recurrence and survival outcomes. Collectively, these investigations reinforce the significance of the 14 immune-related gene-risk signature in KIRC.</ns0:p><ns0:p>Our signature is associated with the survival outcome of KIRC patients and clinical parameters, including pathological staging, M stage, and T stage. No correlation was found between the signature and N-stage, possibly due to a lack of N-stage information for many patients. According to the immune-related gene-risk signature, our study found that clinical cohorts in KIRC have different immune-related risk factors. This signature also reflects differences in tumor immune microenvironments and predicts survival outcomes in KIRC patients; thus, demonstrating the clinical significance of our signature and its possible use as a survival predictor in KIRC.</ns0:p><ns0:p>Our study expands on the signatures association with several important pathways, especially the metabolism pathway. This may reflect the mutual interaction between tumor metabolism and tumor immunology in KIRC. Pearce et al.</ns0:p><ns0:p>(2013) showed that metabolic reprogramming can influence the fate and function of T cells in tumors. Our study further indicates the importance of metabolism pathways in KIRC immune microenvironments.</ns0:p></ns0:div>
<ns0:div><ns0:head>Conclusions</ns0:head><ns0:p>This investigation utilized RNA-seq data from the TCGA database to construct a 14 immune-related gene-risk signature with the ability to independently predict survival outcomes in KIRC; thus, providing novel clinical applications and possible immune targets for KIRC. </ns0:p></ns0:div>
<ns0:div><ns0:head>Figure legends</ns0:head></ns0:div><ns0:figure xml:id='fig_0'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2019:11:43459:1:1:NEW 22 Feb 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_1'><ns0:head>Figure 1 :</ns0:head><ns0:label>1</ns0:label><ns0:figDesc>Figure 1: Our workflow constructing the model for risk-score signatures</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_2'><ns0:head>Figure 5 :</ns0:head><ns0:label>5</ns0:label><ns0:figDesc>Figure 5: Analysis of the 14 immune-related genes predictive ability in total cohort. As demonstrated by the heatmap,</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_3'><ns0:head>Figure 6 :</ns0:head><ns0:label>6</ns0:label><ns0:figDesc>Figure 6: The K-M survival curve of (A) pathological staging, (B) M stage, (C) N stage, and (D) T stage (left). The</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_4'><ns0:head>Figure 8 :</ns0:head><ns0:label>8</ns0:label><ns0:figDesc>Figure 8: Gene set enrichment analysis for high-and low-risk groups, using 177 KEGG pathways (A) A total of 19</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_5'><ns0:head>Figure 9 :</ns0:head><ns0:label>9</ns0:label><ns0:figDesc>Figure 9: Gene set enrichment analysis for high-and low-risk groups, using 4,528 GO terms. (A) Eight significant</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_6'><ns0:head>Figure S1 :</ns0:head><ns0:label>S1</ns0:label><ns0:figDesc>Figure S1: Parameter selection in the LASSO-cox model. (A) LASSO coefficient values of the 14 immune related</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_7'><ns0:head>Figure S3 :</ns0:head><ns0:label>S3</ns0:label><ns0:figDesc>FigureS3: Expression of 14 immune-related genes in kidney renal clear cell carcinoma and normal tissues.</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_8'><ns0:head>Figure 1</ns0:head><ns0:label>1</ns0:label><ns0:figDesc>Figure 1</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_9'><ns0:head>Figure 4</ns0:head><ns0:label>4</ns0:label><ns0:figDesc>Figure 4</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_10'><ns0:head>Figure 6</ns0:head><ns0:label>6</ns0:label><ns0:figDesc>Figure 6</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_11'><ns0:head>Figure 7</ns0:head><ns0:label>7</ns0:label><ns0:figDesc>Figure 7</ns0:figDesc></ns0:figure>
<ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='21,42.52,70.87,525.00,383.25' type='bitmap' /></ns0:figure>
<ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='28,42.52,229.87,525.00,364.50' type='bitmap' /></ns0:figure>
<ns0:figure type='table' xml:id='tab_0'><ns0:head>Table1 :</ns0:head><ns0:label>Table1</ns0:label><ns0:figDesc>Univariate Cox analysis of 14 immune relate genes in all cohort</ns0:figDesc><ns0:table><ns0:row><ns0:cell>AR</ns0:cell><ns0:cell cols='2'>Variable 0.638 0.556-0.731</ns0:cell><ns0:cell>HR <0.001</ns0:cell><ns0:cell>95%CI IL20RB</ns0:cell><ns0:cell>pvalue 1.268 1.102-1.46</ns0:cell><ns0:cell>0.001</ns0:cell></ns0:row><ns0:row><ns0:cell>age</ns0:cell><ns0:cell cols='2'>KLRC2 1.352 1.143-1.599</ns0:cell><ns0:cell>20.657 <0.001</ns0:cell><ns0:cell>9.116-46.81 age</ns0:cell><ns0:cell><0.001 1.414 1.195-1.672</ns0:cell><ns0:cell><0.001</ns0:cell></ns0:row><ns0:row><ns0:cell>stage</ns0:cell><ns0:cell cols='2'>CCL7 3.725 2.708-5.124</ns0:cell><ns0:cell>2.757 <0.001</ns0:cell><ns0:cell>2.090-3.637 stage</ns0:cell><ns0:cell><0.001 3.275 2.351-4.562</ns0:cell><ns0:cell><0.001</ns0:cell></ns0:row><ns0:row><ns0:cell>gender</ns0:cell><ns0:cell cols='2'>0.943 0.687-1.294</ns0:cell><ns0:cell>0.718</ns0:cell><ns0:cell>gender</ns0:cell><ns0:cell>1.025 0.744-1.412</ns0:cell><ns0:cell>0.879</ns0:cell></ns0:row><ns0:row><ns0:cell>BID</ns0:cell><ns0:cell cols='2'>SEMA3A 1.411 1.211-1.645</ns0:cell><ns0:cell>1.748 <0.001</ns0:cell><ns0:cell>1.437-2.126 IL4</ns0:cell><ns0:cell><0.001 1.458 1.294-1.643</ns0:cell><ns0:cell><0.001</ns0:cell></ns0:row><ns0:row><ns0:cell>age</ns0:cell><ns0:cell cols='2'>SEMA3G 1.407 1.188-1.666</ns0:cell><ns0:cell>0.617 <0.001</ns0:cell><ns0:cell>0.534-0.714 age</ns0:cell><ns0:cell><0.001 1.39 1.171-1.651</ns0:cell><ns0:cell><0.001</ns0:cell></ns0:row><ns0:row><ns0:cell>stage</ns0:cell><ns0:cell cols='2'>CCR10 3.173 2.284-4.41</ns0:cell><ns0:cell>1.759 <0.001</ns0:cell><ns0:cell>1.381-2.241 stage</ns0:cell><ns0:cell><0.001 3.426 2.484-4.725</ns0:cell><ns0:cell><0.001</ns0:cell></ns0:row><ns0:row><ns0:cell>gender</ns0:cell><ns0:cell cols='2'>BMP8A 1.039 0.755-1.43</ns0:cell><ns0:cell>1.813 0.814</ns0:cell><ns0:cell>1.298-2.532 gender</ns0:cell><ns0:cell><0.001 0.999 0.726-1.375</ns0:cell><ns0:cell>0.997</ns0:cell></ns0:row><ns0:row><ns0:cell>BMP8A</ns0:cell><ns0:cell cols='2'>FGF17 1.213 1.057-1.391</ns0:cell><ns0:cell>7.335 0.006</ns0:cell><ns0:cell cols='2'>3.202-16.803 KLRC2 1.402 1.233-1.594 <0.001</ns0:cell><ns0:cell><0.001</ns0:cell></ns0:row><ns0:row><ns0:cell>age</ns0:cell><ns0:cell cols='2'>1.454 1.229-1.721</ns0:cell><ns0:cell><0.001</ns0:cell><ns0:cell>age</ns0:cell><ns0:cell>1.383 1.169-1.637</ns0:cell><ns0:cell><0.001</ns0:cell></ns0:row><ns0:row><ns0:cell>stage</ns0:cell><ns0:cell cols='2'>GDF1 3.607 2.619-4.969</ns0:cell><ns0:cell>1.483 <0.001</ns0:cell><ns0:cell>1.062-2.070 stage</ns0:cell><ns0:cell>0.021 3.429 2.483-4.737</ns0:cell><ns0:cell><0.001</ns0:cell></ns0:row><ns0:row><ns0:cell>gender</ns0:cell><ns0:cell cols='2'>IL4 1.052 0.765-1.447</ns0:cell><ns0:cell>109.106 0.754</ns0:cell><ns0:cell cols='2'>34.328-346.781 gender 0.958 0.694-1.322 <0.001</ns0:cell><ns0:cell>0.792</ns0:cell></ns0:row><ns0:row><ns0:cell>CCL7</ns0:cell><ns0:cell cols='2'>LHB 1.329 1.202-1.471</ns0:cell><ns0:cell>5.053 <0.001</ns0:cell><ns0:cell>3.240-7.881 LHB</ns0:cell><ns0:cell><0.001 1.285 1.168-1.414</ns0:cell><ns0:cell><0.001</ns0:cell></ns0:row><ns0:row><ns0:cell>age</ns0:cell><ns0:cell cols='2'>TXLNA 1.455 1.226-1.728</ns0:cell><ns0:cell>2.329 <0.001</ns0:cell><ns0:cell>1.500-3.618 age</ns0:cell><ns0:cell><0.001 1.393 1.178-1.647</ns0:cell><ns0:cell><0.001</ns0:cell></ns0:row><ns0:row><ns0:cell>stage</ns0:cell><ns0:cell cols='2'>AR 3.485 2.529-4.802</ns0:cell><ns0:cell>0.548 <0.001</ns0:cell><ns0:cell>0.463-0.649 stage</ns0:cell><ns0:cell><0.001 3.519 2.55-4.854</ns0:cell><ns0:cell><0.001</ns0:cell></ns0:row><ns0:row><ns0:cell>gender</ns0:cell><ns0:cell>1</ns0:cell><ns0:cell>IL20RB 0.726-1.377</ns0:cell><ns0:cell>1.225 1</ns0:cell><ns0:cell>1.145-1.310 gender</ns0:cell><ns0:cell><0.001 1.17 0.844-1.622</ns0:cell><ns0:cell>0.347</ns0:cell></ns0:row><ns0:row><ns0:cell>CCR10</ns0:cell><ns0:cell cols='2'>1.235 1.111-1.372</ns0:cell><ns0:cell><0.001</ns0:cell><ns0:cell>SEMA3A</ns0:cell><ns0:cell>1.214 1.107-1.33</ns0:cell><ns0:cell><0.001</ns0:cell></ns0:row><ns0:row><ns0:cell>age</ns0:cell><ns0:cell cols='2'>BID 1.382 1.166-1.639</ns0:cell><ns0:cell>3.505 <0.001</ns0:cell><ns0:cell>2.439-5.035 age</ns0:cell><ns0:cell><0.001 1.445 1.218-1.715</ns0:cell><ns0:cell><0.001</ns0:cell></ns0:row><ns0:row><ns0:cell>stage</ns0:cell><ns0:cell cols='2'>3.844 2.793-5.292</ns0:cell><ns0:cell><0.001</ns0:cell><ns0:cell>stage</ns0:cell><ns0:cell>3.549 2.572-4.898</ns0:cell><ns0:cell><0.001</ns0:cell></ns0:row><ns0:row><ns0:cell>gender</ns0:cell><ns0:cell cols='2'>1.034 0.751-1.424</ns0:cell><ns0:cell>0.836</ns0:cell><ns0:cell>gender</ns0:cell><ns0:cell>1.108 0.802-1.529</ns0:cell><ns0:cell>0.535</ns0:cell></ns0:row><ns0:row><ns0:cell>FGF17</ns0:cell><ns0:cell cols='2'>1.235 1.112-1.372</ns0:cell><ns0:cell><0.001</ns0:cell><ns0:cell cols='2'>SEMA3G 0.652 0.549-0.775</ns0:cell><ns0:cell><0.001</ns0:cell></ns0:row><ns0:row><ns0:cell>age</ns0:cell><ns0:cell cols='2'>1.399 1.181-1.657</ns0:cell><ns0:cell><0.001</ns0:cell><ns0:cell>age</ns0:cell><ns0:cell>1.379 1.164-1.633</ns0:cell><ns0:cell><0.001</ns0:cell></ns0:row><ns0:row><ns0:cell>stage</ns0:cell><ns0:cell cols='2'>3.722 2.705-5.121</ns0:cell><ns0:cell><0.001</ns0:cell><ns0:cell>stage</ns0:cell><ns0:cell>3.29 2.384-4.539</ns0:cell><ns0:cell><0.001</ns0:cell></ns0:row><ns0:row><ns0:cell>gender</ns0:cell><ns0:cell cols='2'>1.109 0.806-1.525</ns0:cell><ns0:cell>0.527</ns0:cell><ns0:cell>gender</ns0:cell><ns0:cell>0.973 0.707-1.339</ns0:cell><ns0:cell>0.867</ns0:cell></ns0:row><ns0:row><ns0:cell>GDF1</ns0:cell><ns0:cell cols='2'>1.086 0.975-1.211</ns0:cell><ns0:cell>0.135</ns0:cell><ns0:cell>TXLNA</ns0:cell><ns0:cell>1.285 1.098-1.505</ns0:cell><ns0:cell>0.002</ns0:cell></ns0:row><ns0:row><ns0:cell>age</ns0:cell><ns0:cell cols='2'>1.411 1.191-1.672</ns0:cell><ns0:cell><0.001</ns0:cell><ns0:cell>age</ns0:cell><ns0:cell>1.433 1.208-1.7</ns0:cell><ns0:cell><0.001</ns0:cell></ns0:row><ns0:row><ns0:cell>stage</ns0:cell><ns0:cell>3.76</ns0:cell><ns0:cell>2.735-5.17</ns0:cell><ns0:cell><0.001</ns0:cell><ns0:cell>stage</ns0:cell><ns0:cell>3.696 2.687-5.085</ns0:cell><ns0:cell><0.001</ns0:cell></ns0:row><ns0:row><ns0:cell>gender</ns0:cell><ns0:cell cols='2'>1.103 0.796-1.529</ns0:cell><ns0:cell>0.555</ns0:cell><ns0:cell>gender</ns0:cell><ns0:cell>1.212 0.87-1.688</ns0:cell><ns0:cell>0.255</ns0:cell></ns0:row></ns0:table></ns0:figure>
<ns0:note place='foot'>Manuscript to be reviewed</ns0:note>
<ns0:note place='foot' n='1'>Table 2. Multivariate Cox analysis of 14 immune relate genes in all cohort</ns0:note>
</ns0:body>
" | "Resubmission: A 14-immune-related-gene signature predicts clinical outcome of kidney renal clear cell carcinoma
Dear Editors,
Thank you very much for the efforts and suggestions from the editors and the reviewers. We have revised our manuscript entitled “A 14-immune-related-gene signature predicts clinical outcome of kidney renal clear cell carcinoma” according to the reviewers’ recommendation and would like to resubmit it. Based on the comments, we have carefully made major revision on the revised version. Detailed revisions are shown as follows. We have uploaded the manuscript with computer-generated tracked changes to the revision response files section. Besides, we have uploaded the a “clean” copy where the changes are not marked.
1. Reviewer 1 mentions lack of novelty. PeerJ does not require novelty, but it does require that you do more to place the article in context with prior studies that have been performed. 'The article should include sufficient introduction and background to demonstrate how the work fits into the broader field of knowledge. Relevant prior literature should be appropriately referenced.' The Introduction does touch on this briefly, but a more careful and thorough evaluation of prior work should be provided. What specifically does your article provide that others did not? Or is it a replication study?
Reply: We sincerely appreciate your constructive suggestion on our manuscript. According to your suggestion, we have added another two studies in which the authors constructed immune related signatures for renal clear cell carcinoma patients [Khadirnaikar S, Kumar P, Pandi SN, Malik R, Dhanasekaran SM, Shukla SK. Immune associated LncRNAs identify novel prognostic subtypes of renal clear cell carcinoma; Smith CC, Beckermann KE, Bortone DS, et al. Endogenous retroviral signatures predict immunotherapy response in clear cell renal cell carcinoma]. Compared with their studies, our signature is more robust with a much more independent prognostic ability. Besides, our 14-immune-related-gene signature also has prognostic ability in different subgroups of patients, which were not mentioned conducted in other studies. The detailed comparison is mentioned in the Introduction part of the manuscript. In addition, manuscript with computer-generated tracked changes were uploaded.
2. Reviewer 1 mentions that the authors did not provide any of their own data. This is okay. However, as all the reviewers mentioned, when proposing a gene signature for potential clinical relevance, there is a need to validate the signature on independent datasets. Reviewers 2 and 3 mentioned that you might be able to find these in Gene Expression Omnibus. If you do this, make sure to keep those datasets completely independent from the TCGA data.
Reply: We sincerely appreciate your constructive suggestions on our manuscript. We have tried our best to find the KIRC datasets with expression and survival information in public database including GEO. GSE40912 was only one dataset which contained expression and survival information. But its platform was IQUSP_Human_intronic_4k _v2.0 which contained only 3206 genes. We only found one of the 14 genes in the dataset. Therefore, we can’t validate our results in other gene expression datasets. We have begun to construct our kidney cancer biological sample bank and we will validate the results in the future study. Besides, manuscript with computer-generated tracked changes were uploaded.
3. Two reviewers mentioned that the English language usage needs work. I agree with them. PeerJ's criteria state, 'The article must be written in English and must use clear, unambiguous, technically correct text. The article must conform to professional standards of courtesy and expression.' I know this is difficult because English is not your first language, but it must be improved in many places throughout the manuscrip.
Reply: We sincerely appreciate your constructive comment on our manuscript. In order to improve our English language, we have applied for English Copyediting Services of PeerJ. We hope to improve our English language through this program.
4. The authors should analyze expression of the 14 immune genes in tumor and normal tissue
Reply: Thanks for your suggestion. As shown in Figure S3, we added the result of 14 gene expression.
5. The authors should present a time-dependent ROC curve .(1 year,3year and 5 year)
Reply: Thanks for your suggestion. As shown in Figure S2, we added the results of time-dependent ROC curve.
6. The Methods section needs more detail. For example, it says, 'univariate analysis was performed' but it is hard to know exactly what that means. Later it indicates 'lambda = lambda.min' but it doesn't indicate what lambda.min equals. Few details are provided on specifically how the GSEA analysis was performed.
Reply: Thanks for your suggestions. We apologize for our unclear description and we have modified the description. More detailed descriptions have been added. In addition, manuscript with computer-generated tracked changes were uploaded.
7. I'm not sure that I understand exactly how model selection was performed. For example, this language is unclear to me: 'A total of 650 immune-related genes went through the cox proportional hazards regression with 10-fold cross112 validation so as to establish an immune-related gene risk signature for KIRC. After 1,000 iterations, a total of seven 113 models consisting of different number of genes were arranged.'
Reply: Thanks for your suggestion. As you suggested, in the method section of establishment of immune-related genes signature we have modified the description of the gene screening and model building process. In addition, manuscript with computer-generated tracked changes were uploaded.
8. The 'Establishment and validation of the signature of immune-related gene in KIRC' section is confusing because it comes after other details have been provided about methods. These sections should be integrated together.
Reply: Thanks for your suggestion. As you suggested, we have modified and appropriately merged the two section.
9. It's commendable that the authors provided a GitHub repository with analysis code and specifically mention versions of R packages used. However, to make it easier for others to verify your code, please do the following:
- Provide instructions on installing packages or else use something like MyBinder.org.
- Provide data files together with the code (or else a specific URL for each data file).
Reply: Thanks for your suggestion. In the readme section of the github repository, we have provided related R package installation and data connection.
10. Cite Harrell's c-index
Reply: Thanks for your suggestion. We have cited the Harrell's c-index in the method section.
11. In many of the figures, the text is very small, making it difficult to read. In many figures, axis labels use underscores (should be replaced with space characters) or acronyms that are undefined.
Reply: Thanks for your suggestion. We are very sorry for this problem. We think that the fonts in Figures 8 and 9 are indeed too small, so we have revised the font and replaced the underscore with a space.
12. Did you use clinical covariates in the Cox proportional hazards models? If not, why?
Reply: Thanks for your suggestion. In the process of model selection, we did not use clinical information because we wanted to establish a gene signature. However, we used clinical information in the subsequent uni- and multi-factor cox analysis, and we verified the independent predictive power of our gene signature.
We hope that the revision and our responses can address your concerns and that you can consider this article suitable for acceptance. Please do not hesitate to contact us if any additional explanation or revision is required. Thank you again for your assistance.
Best regards
Chuan Hu
" | Here is a paper. Please give your review comments after reading it. |
9,834 | "<ns0:abstract xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:p>Kidney renal clear cell carcinoma (KIRC) is the leading cause of kidney cancer-related deaths. Currently, there are no studies in tumor immunology investigating the use of signatures as a predictor of overall survival in KIRC patients. Our study attempts to establish an immune-related gene risk signature to predict clinical outcomes in KIRC. A total of 528 patients from The Cancer Genome Atlas (TCGA) database were included in our analysis and randomly divided into training (n = 315) and testing sets (n = 213). We collected 1,534 immune-related genes from the Immunology Database and Analysis Portal as candidates to construct our signature. LASSO-COX was used to find gene models with the highest predictive ability. We used survival and Cox analysis to test the model's independent prognostic ability. Univariate analysis identified 650 immune-related genes with prognostic abilities. After 1,000 iterations, we choose 14 of the most frequent and stable immune-related genes as our signature. We found that the signature was associated with M stage, T stage, and pathological staging. More importantly, the signature can independently predict clinical prognosis in KIRC patients. Gene Set Enrichment Analysis (GSEA) showed an association between our signature and critical metabolism pathways.</ns0:p><ns0:p>Our research established a model based upon 14 immune-related genes that predicted the prognosis of KIRC patients based on tumor immune microenvironments.</ns0:p></ns0:div>
</ns0:abstract>
<ns0:body xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:head>Introduction</ns0:head><ns0:p>Kidney cancer is one of the most common urological tumors worldwide, with approximately 403,262 new cases and 175,098 deaths associated with this form of cancer in 2018 <ns0:ref type='bibr' target='#b2'>(Bray et al. 2018)</ns0:ref>. Kidney renal clear cell carcinoma (KIRC) is the leading cause of kidney cancer mortality and the most common type, accounting for 85% of kidney cancers <ns0:ref type='bibr'>(Siegel et al. 2017</ns0:ref>; Tseng 2016). Considering the aging population of KIRC patients and the rising expense of treatment, KIRC is gradually becoming the focus of geriatric cancer <ns0:ref type='bibr'>(Tan et al. 2015)</ns0:ref>. Despite the rapid development of cancer treatments, mortality rates in KIRC remain stagnant. With the progression of next generation sequencing and data mining techniques, it is urgent that we explore prognostic biomarkers for KIRC, using molecular characteristics and tumor immune environments to guide patient therapy.</ns0:p><ns0:p>Over the past decade our understanding of immune components, including the impact of tumor microenvironments on patient survival and therapy response <ns0:ref type='bibr' target='#b3'>(Chen & Mellman 2017;</ns0:ref><ns0:ref type='bibr' target='#b8'>Grivennikov et al. 2010</ns0:ref>), has increased. Some studies have found that tumor-infiltrating immune cells are able to serve as either tumor suppressors or promoters in microenvironments. For example, CD8 + T cells have been associated with improved survival in cancer patients <ns0:ref type='bibr' target='#b5'>(Gajewski et al. 2013;</ns0:ref><ns0:ref type='bibr' target='#b7'>Governa et al. 2017)</ns0:ref>, while tumor associated macrophages and regulatory T cells demonstrate the ability to promote tumor development <ns0:ref type='bibr' target='#b12'>(Nishikawa & Sakaguchi 2014;</ns0:ref><ns0:ref type='bibr' target='#b13'>Noy & Pollard 2014)</ns0:ref>. Considering the complexity and significance of tumor immune microenvironments, it is imperative that we investigate immune-related biomarkers for KIRC patients. Recent studies have provided insight into the KIRC immune signature <ns0:ref type='bibr' target='#b6'>(Geissler et al. 2015;</ns0:ref><ns0:ref type='bibr'>Şenbabaoğlu et al. 2016)</ns0:ref>. <ns0:ref type='bibr'>Şenbabaoğlu et al. (2016)</ns0:ref> found mRNA signatures with the potential to be immunotherapeutic biomarkers in KIRC. However, their investigations do not include immune-related genes for analysis nor do they establish a systematic immune-related gene-risk signature for KIRC patients. <ns0:ref type='bibr'>Khadirnaikar et al. (2019)</ns0:ref> utilized immune associated lncRNA (long non-coding RNA) to construct prognostic subtypes in KIRC patients. Our immune clusters were more robust and independent. Additionally, they concentrated on lncRNA, not the overall immune-related genes. <ns0:ref type='bibr'>Smith et al. (2018)</ns0:ref> constructed endogenous retroviral signatures for KIRC patients, but they did not investigate the prognostic ability of the signature in different subtypes of patients. Therefore, it is essential that we explore a systematic prognostic signature based on tumor immune environments in KIRC.</ns0:p><ns0:p>In our study, we used RNA-seq data from The Cancer Genome Atlas (TCGA) to find immune-related genes with prognostic ability and to establish an immune-related risk signature for KIRC. To assess the clinical potential of the signature, we investigated the association between the signature, clinical parameters, and patient survival. Gene set enrichment analysis (GSEA) was performed to explore the molecular characteristics of the signature.</ns0:p></ns0:div>
<ns0:div><ns0:head>Materials & Methods</ns0:head></ns0:div>
<ns0:div><ns0:head>Patient cohort</ns0:head><ns0:p>The TCGA database was used to collect the clinical and RNA-seq data of 528 KIRC patients. We randomly divided the dataset into training (n = 315) and test sets (n = 213). RNA-seq data was obtained to analyze the transcriptome profiling of RNA expression and were measured using fragments per kilobase of exon per million fragments mapped (FPKM). We performed a log2-based transformation to normalize RNA expression profiles. To ensure detection and total cohort <ns0:ref type='bibr' target='#b10'>(Harrell et al. 1982)</ns0:ref>. The independent prognostic ability of the immune-related risk signature was assessed using survival analysis as well as Cox analysis (R package survival, v2.42). We performed multivariable Cox regression to analyze the relationship between immune-related risk signatures and clinicopathological factors.</ns0:p></ns0:div>
<ns0:div><ns0:head>Gene set enrichment analysis</ns0:head><ns0:p>To identify the biological functions and pathways between high-and low-risk groups, we conducted GSEA to investigate potential biological mechanisms in the Molecular Signatures <ns0:ref type='bibr'>Database (MSigDB;</ns0:ref><ns0:ref type='bibr'>(Subramanian et al. 2005)</ns0:ref>. GSEA was performed using GSEApy, a python wrapper for gene enrichment (https://pypi.org/project/gseapy/).</ns0:p><ns0:p>We selected C2 and C5, including pathway databases and GO terms, from the MSigDB. The gesa sub-command of GSEApy was used in GSEA with default parameters. Enriched gene sets with a false discovery rate (FDR) of less than 0.25 and a P-value of less than 0.05 were considered statistically significant.</ns0:p></ns0:div>
<ns0:div><ns0:head>Statistical analysis</ns0:head><ns0:p>Boxplots were created with the R package, ggplot2 (v3.0.0). The R package ComplexHeatmap (v1.18.1) was used to create heatmaps <ns0:ref type='bibr' target='#b9'>(Gu et al. 2016)</ns0:ref>. We counted C-index with R packages 'survcomp' <ns0:ref type='bibr' target='#b10'>(Harrell et al. 1982;</ns0:ref><ns0:ref type='bibr' target='#b17'>Schröder et al. 2011)</ns0:ref>. The student's t-test was performed for statistical comparison. We chose R to conduct statistical analysis (https://www.r-project.org/). P-values lower than 0.05 were considered statistically significant. The main code of analysis was pushed to github (https://github.com/huchua/KIRC).</ns0:p></ns0:div>
<ns0:div><ns0:head>Results</ns0:head></ns0:div>
<ns0:div><ns0:head>Establishment and validation of the immune-related gene signature in KIRC</ns0:head><ns0:p>Figure <ns0:ref type='figure'>1</ns0:ref> illustrates the workflow we used to develop an immune-related gene-risk signature. The immune-related gene signature was constructed within the KIRC training data set, while we applied the testing set to validate the signature.</ns0:p><ns0:p>After 1,000 iterations, seven unique gene models were selected (Figure <ns0:ref type='figure'>2A</ns0:ref>, Table <ns0:ref type='table'>S1</ns0:ref>). The model selected as the Manuscript to be reviewed immune-related gene-risk signature consisted of 14 genes and ranked the highest in frequency 282 times. Parameter selection in the LASSO-cox model was log lambda = -2.641 and alpha = 1. The univariate and multivariate Cox analysis of the 14 immune-related genes are shown in Table <ns0:ref type='table'>1 and Table 2</ns0:ref>. The principle component analysis of the 14 immune-related gene signature displayed a different distribution pattern between low-and high-risk groups when comparing the training, testing, and total cohort (Figure <ns0:ref type='figure'>2B-D</ns0:ref>). This indicated that the low-and high-risk groups had different immune phenotypes. In the training, testing, and total data set, the c-index was 0.7862, 0.6534, and 0.7367, respectively (P < 0.001; Figure <ns0:ref type='figure'>2E</ns0:ref>). In a time-dependent receiver operating characteristic curve (ROC) created for three datasets, area under the curve (AUC) values at 1, 3, and 5 years were 0.679, 0.63, 0.627; 0.65, 0.596, 0.568; and 0.618, 0.589, 0.59, respectively (Figure <ns0:ref type='figure'>S2</ns0:ref>). The 14 immune-related genes are AR, BID, BMP8A, CCL7, CCR10, FGF17, GDF1, IL20RB, IL4, KLRC2, LHB, SEMA3A, SEMA3G, and TXLNA. K-M (Kaplan-Meier) survival curves and gene expression of the 14 immune-related genes are shown in Figure <ns0:ref type='figure'>3</ns0:ref> and Figure <ns0:ref type='figure'>S3</ns0:ref>. Because of the differences between the log-rank test and the univariate cox analysis, the results of the univariate cox analysis of other genes except GDF1 are very consistent with the results of the K-M (Kaplan-Meier) survival curves.</ns0:p></ns0:div>
<ns0:div><ns0:head>Correlation between the signature of immune-related genes and clinical parameters</ns0:head><ns0:p>Among the 14 immune-related genes, four genes (TXLNA, SEMA3G, AR, and BID) had a high expression, and 10 genes (IL20RB, CCR10, BMP8A, SEMA3A, CCL7, GDF1, KLRC2, LHB, FGF17, and IL4) had a low expression (Figure <ns0:ref type='figure'>4A, B</ns0:ref>; Figure <ns0:ref type='figure' target='#fig_5'>5</ns0:ref>). The relationship between the signature and clinical factors demonstrated that patients with advanced pathological staging, M stage, and T stage had a higher risk score than those with early stage disease (Figure <ns0:ref type='figure'>4A</ns0:ref>, B; Figure <ns0:ref type='figure' target='#fig_5'>5</ns0:ref>; Figure <ns0:ref type='figure' target='#fig_11'>6A, B, D</ns0:ref>). However, we did not find a correlation between the signature and N stage (Figure <ns0:ref type='figure'>4A</ns0:ref>, B; Figure <ns0:ref type='figure' target='#fig_5'>5</ns0:ref>; Figure <ns0:ref type='figure' target='#fig_11'>6C</ns0:ref>).</ns0:p></ns0:div>
<ns0:div><ns0:head>Influence of the immune-related gene signature on patient prognosis</ns0:head><ns0:p>PeerJ reviewing <ns0:ref type='table'>PDF | (2019:11:43459:2:0:NEW 24 Mar 2020)</ns0:ref> Manuscript to be reviewed</ns0:p><ns0:p>We then assessed whether this signature influenced KIRC patient prognosis. Survival analysis showed that patients with a high-risk score were associated with poor survival outcomes in the training, testing, and total group sets (Figure <ns0:ref type='figure'>4C</ns0:ref>; Figure <ns0:ref type='figure'>4E</ns0:ref>; Figure <ns0:ref type='figure'>7B</ns0:ref>). We found that the signature also predicted survival outcomes in subgroups of KIRC patients, including stage I-II (Figure <ns0:ref type='figure'>7C</ns0:ref>), stage III-IV (Figure <ns0:ref type='figure'>7D</ns0:ref>), M0 stage (Figure <ns0:ref type='figure'>7E</ns0:ref>), M1 stage (Figure <ns0:ref type='figure'>7F</ns0:ref>), N0 stage (Figure <ns0:ref type='figure'>7G</ns0:ref>), N1 (Figure <ns0:ref type='figure'>7H</ns0:ref>), T1 (Figure <ns0:ref type='figure'>7I</ns0:ref>), T2 (Figure <ns0:ref type='figure'>7J</ns0:ref>), T3 (Figure <ns0:ref type='figure'>7K</ns0:ref>), and T4 (Figure <ns0:ref type='figure'>7L</ns0:ref>). Multivariate Cox analysis revealed that the risk signature was able to independently predict overall survival in KIRC patients (Figure <ns0:ref type='figure'>4D</ns0:ref>, F; Figure <ns0:ref type='figure'>7A</ns0:ref>; Table <ns0:ref type='table'>3</ns0:ref>).</ns0:p></ns0:div>
<ns0:div><ns0:head>Engagement of the immune-related gene risk signature in biological pathways and functions</ns0:head><ns0:p>GSEA was used to investigate the signature's biological pathways and functions. There were 177 KEGG pathways and 4,528 GO terms used in our investigation. Our analysis found that the signature was able to engage in a total of 19 enriched KEGG pathways (FDR < 0.25; Table <ns0:ref type='table'>4</ns0:ref>). The low-risk signature was significantly correlated with 10 pathways, including the citrate cycle (TCA cycle) pathway, fatty acid metabolism pathway, propanoate metabolism pathway, butanoate metabolism pathway, peroxisome pathway, lysine degradation pathway, valine leucine and isoleucine degradation pathway, proximal tubule bicarbonate reclamation pathway, vasopressin regulated water reabsorption pathway, and the pyruvate metabolism pathway (Table <ns0:ref type='table'>4</ns0:ref>). Similarly, eight GO annotations were enriched in the low-risk group (FDR < 0.25; Table <ns0:ref type='table'>4</ns0:ref>).</ns0:p></ns0:div>
<ns0:div><ns0:head>Discussion</ns0:head><ns0:p>Our study used the TCGA database and immune-related genes to establish a KIRC signature consisting of 14 immunerelated genes. We found that patients in the high-risk group showed a positive association with M stage, T stage, and advanced pathological staging. Additionally, the signature exhibited strong prognostic abilities and independently predicted KIRC patient prognosis. Functional analysis highlighted the significance of our signature based on its Manuscript to be reviewed involvement in many important pathways.</ns0:p><ns0:p>In our investigation, we used a total of 14 immune-related genes to construct our signature. In the signature, CCL7 increased the peripheral blood mononuclear cell recruitment in renal cell cancer through the inhibition of let-7d <ns0:ref type='bibr' target='#b16'>(Riihimaki et al. 2014)</ns0:ref>. <ns0:ref type='bibr'>Wyler et al. (2014)</ns0:ref> demonstrated the ability of CCL7 to recruit monocytes through CCR2, promoting renal cell cancer metastasis to the brain. This indicates that CCL7 is a major factor in the development of KIRC in tumor immune microenvironments and may be a potential immunotherapy target. Fibroblast growth factor receptor 17 (FGF17) is another immune-related gene in our signature. FGF17 demonstrates a variety of functions in cancer development. <ns0:ref type='bibr'>Gauglhofer et al. (2011)</ns0:ref> found that FGF17 was involved in the paracrine and autocrine signaling of hepatocellular carcinoma and promoted the neoangiogenesis of hepatocellular carcinoma. <ns0:ref type='bibr'>Heer et al. (2004)</ns0:ref> showed that FGF17 was overexpressed in prostate cancer and participated in prostate carcinogenesis. Our study showed that FGF17 plays an important role in KIRC based on tumor immunology. However, the underlying mechanism of FGF17 in KIRC immune microenvironments requires further investigation. In our signature, IL-4 is another well-studied immune-related biomarker for KIRC. IL-4, released by immune cells, controls the expression of B7-H1, thus altering T cell responses in KIRC <ns0:ref type='bibr' target='#b15'>(Quandt et al. 2014)</ns0:ref> Manuscript to be reviewed significance of our signature and its possible use as a survival predictor in KIRC.</ns0:p><ns0:p>Despite of the relatively low C-index of our testing set (0.6534), the C-index of the testing set was higher than those of similar signatures. In the study of Bailiang Li et al., they developed an immune signature in non-small cell lung cancer with the C-index of 0.64, which is slightly lower than the C-index of our testing set <ns0:ref type='bibr' target='#b11'>(Li et al. 2017)</ns0:ref>. Besides, our signature achieved a higher C-index of testing set compared with the immune signature in ovarian cancer (0.625) <ns0:ref type='bibr'>(Shen et al. 2019)</ns0:ref> .More importantly, compared with the C-index of clinical staging systems in renal cancer (0.62), the C-index of our testing set showed a higher accuracy <ns0:ref type='bibr' target='#b14'>(Qu et al. 2018)</ns0:ref>. Therefore, by comparing the C-index of our testing set with other clinical and molecular signatures, we prove our immune signature to be a promising tool in predicting KIRC patients' survival outcome.</ns0:p><ns0:p>Our study expands on the signatures association with several important pathways, especially the metabolism pathway.</ns0:p><ns0:p>This may reflect the mutual interaction between tumor metabolism and tumor immunology in KIRC. Pearce et al.</ns0:p><ns0:p>(2013) showed that metabolic reprogramming can influence the fate and function of T cells in tumors. Our study further indicates the importance of metabolism pathways in KIRC immune microenvironments.</ns0:p></ns0:div>
<ns0:div><ns0:head>Conclusions</ns0:head><ns0:p>This investigation utilized RNA-seq data from the TCGA database to construct a 14 immune-related gene-risk signature with the ability to independently predict survival outcomes in KIRC; thus, providing novel clinical applications and possible immune targets for KIRC. Manuscript to be reviewed Manuscript to be reviewed </ns0:p><ns0:note type='other'>Figure legends</ns0:note><ns0:note type='other'>Figure 7</ns0:note><ns0:p>The 14 immune-related genes signature could serve as an independent prognostic factor for OS </ns0:p></ns0:div><ns0:figure xml:id='fig_0'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2019:11:43459:2:0:NEW 24 Mar 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_1'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2019:11:43459:2:0:NEW 24 Mar 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_2'><ns0:head /><ns0:label /><ns0:figDesc>. The investigation conducted by Chang et al. (2015) demonstrated that IL-4 expression can predict KIRC patient recurrence and survival outcomes. Collectively, these investigations reinforce the significance of the 14 immune-related gene-risk signature in KIRC. Our signature is associated with the survival outcome of KIRC patients and clinical parameters, including pathological staging, M stage, and T stage. No correlation was found between the signature and N-stage, possibly due to a lack of N-stage information for many patients. According to the immune-related gene-risk signature, our study found that clinical cohorts in KIRC have different immune-related risk factors. This signature also reflects differences in tumor immune microenvironments and predicts survival outcomes in KIRC patients; thus, demonstrating the clinical PeerJ reviewing PDF | (2019:11:43459:2:0:NEW 24 Mar 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_3'><ns0:head /><ns0:label /><ns0:figDesc>Luna A. 2016. Tumor immune microenvironment characterization in clear cell renal cell carcinoma identifies prognostic and immunotherapeutically relevant messenger RNA signatures. Genome biology 17:231. Shang J, Song Q, Yang Z, Li D, Chen W, Luo L, Wang Y, Yang J, and Li S. 2017. Identification of lung adenocarcinoma specific dysregulated genes with diagnostic and prognostic value across 27 TCGA cancer types. Oncotarget 8:87292. Shen S, Wang G, Zhang R, Zhao Y, Yu H, Wei Y, and Chen F. 2019. Development and validation of an immune</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_4'><ns0:head>Figure 1 :Figure 2 :Figure 4 :</ns0:head><ns0:label>124</ns0:label><ns0:figDesc>Figure 1: Our workflow constructing the model for risk-score signatures</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_5'><ns0:head>Figure 5 :</ns0:head><ns0:label>5</ns0:label><ns0:figDesc>Figure 5: Analysis of the 14 immune-related genes predictive ability in total cohort. As demonstrated by the heatmap,</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_6'><ns0:head>Figure 6 :</ns0:head><ns0:label>6</ns0:label><ns0:figDesc>Figure 6: The K-M survival curve of (A) pathological staging, (B) M stage, (C) N stage, and (D) T stage (left). The</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_7'><ns0:head>(</ns0:head><ns0:label /><ns0:figDesc>H), T1 stage (I), T2 stage (J), T3 stage (K), and T4 stage (L).</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_8'><ns0:head>Figure S1 :</ns0:head><ns0:label>S1</ns0:label><ns0:figDesc>Figure S1: Parameter selection in the LASSO-cox model. (A) LASSO coefficient values of the 14 immune related</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_9'><ns0:head>Figure S3 :Figure 1</ns0:head><ns0:label>S31</ns0:label><ns0:figDesc>FigureS3: Expression of 14 immune-related genes in kidney renal clear cell carcinoma and normal tissues.</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_10'><ns0:head>Figure 5</ns0:head><ns0:label>5</ns0:label><ns0:figDesc>Figure 5</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_11'><ns0:head>Figure 6</ns0:head><ns0:label>6</ns0:label><ns0:figDesc>Figure 6</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_12'><ns0:head>(</ns0:head><ns0:label /><ns0:figDesc>A) The multivariate Cox analysis of total cohorts demonstrated that the 14 immune-related genes signature could serve as an independent prognostic factor for OS. The K-M analysis of the risk signature in the total cohort (B) and in subgroups of patients with stage I-II (C), stage III-IV (D), M0 stage (E), M1 stage (F), N0 stage (G), N1 stage (H), T1 stage (I), T2 stage (J), T3 stage (K), and T4 stage (L)</ns0:figDesc></ns0:figure>
<ns0:figure type='table' xml:id='tab_0'><ns0:head /><ns0:label /><ns0:figDesc>Proceedings of the National Academy of Sciences 102:15545-15550. Tan H-J, Filson CP, and Litwin MS. 2015. Contemporary, age-based trends in the incidence and management of patients with early-stage kidney cancer. Urologic Oncology: Seminars and Original Investigations: Elsevier.</ns0:figDesc><ns0:table><ns0:row><ns0:cell>gene-set</ns0:cell><ns0:cell>based</ns0:cell><ns0:cell>Prognostic</ns0:cell><ns0:cell>signature</ns0:cell><ns0:cell>in</ns0:cell><ns0:cell>ovarian</ns0:cell><ns0:cell>cancer.</ns0:cell><ns0:cell>EBioMedicine</ns0:cell><ns0:cell>40:318-326.</ns0:cell></ns0:row><ns0:row><ns0:cell cols='3'>10.1016/j.ebiom.2018.12.054</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell cols='8'>Siegel R, Miller K, and Jemal A. 2017. Cancer statistics, 2018 CA: a cancer. J Clin 68: 7-30.</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell cols='9'>Subramanian A, Tamayo P, Mootha VK, Mukherjee S, Ebert BL, Gillette MA, Paulovich A, Pomeroy SL, Golub TR,</ns0:cell></ns0:row><ns0:row><ns0:cell cols='9'>and Lander ES. 2005. Gene set enrichment analysis: a knowledge-based approach for interpreting genome-</ns0:cell></ns0:row><ns0:row><ns0:cell cols='3'>wide expression profiles. p 21. e19-21. e26.</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row></ns0:table><ns0:note>Tseng C-H. 2016. Use of metformin and risk of kidney cancer in patients with type 2 diabetes. European Journal of Cancer 52:19-25.</ns0:note></ns0:figure>
<ns0:figure type='table' xml:id='tab_1'><ns0:head>Table1 :</ns0:head><ns0:label>Table1</ns0:label><ns0:figDesc>Univariate Cox analysis of 14 immune relate genes in all cohort</ns0:figDesc><ns0:table><ns0:row><ns0:cell>AR</ns0:cell><ns0:cell cols='2'>Variable 0.638 0.556-0.731</ns0:cell><ns0:cell>HR <0.001</ns0:cell><ns0:cell>95%CI IL20RB</ns0:cell><ns0:cell>pvalue 1.268 1.102-1.46</ns0:cell><ns0:cell>0.001</ns0:cell></ns0:row><ns0:row><ns0:cell>age</ns0:cell><ns0:cell cols='2'>KLRC2 1.352 1.143-1.599</ns0:cell><ns0:cell>20.657 <0.001</ns0:cell><ns0:cell>9.116-46.81 age</ns0:cell><ns0:cell><0.001 1.414 1.195-1.672</ns0:cell><ns0:cell><0.001</ns0:cell></ns0:row><ns0:row><ns0:cell>stage</ns0:cell><ns0:cell cols='2'>CCL7 3.725 2.708-5.124</ns0:cell><ns0:cell>2.757 <0.001</ns0:cell><ns0:cell>2.090-3.637 stage</ns0:cell><ns0:cell><0.001 3.275 2.351-4.562</ns0:cell><ns0:cell><0.001</ns0:cell></ns0:row><ns0:row><ns0:cell>gender</ns0:cell><ns0:cell cols='2'>0.943 0.687-1.294</ns0:cell><ns0:cell>0.718</ns0:cell><ns0:cell>gender</ns0:cell><ns0:cell>1.025 0.744-1.412</ns0:cell><ns0:cell>0.879</ns0:cell></ns0:row><ns0:row><ns0:cell>BID</ns0:cell><ns0:cell cols='2'>SEMA3A 1.411 1.211-1.645</ns0:cell><ns0:cell>1.748 <0.001</ns0:cell><ns0:cell>1.437-2.126 IL4</ns0:cell><ns0:cell><0.001 1.458 1.294-1.643</ns0:cell><ns0:cell><0.001</ns0:cell></ns0:row><ns0:row><ns0:cell>age</ns0:cell><ns0:cell cols='2'>SEMA3G 1.407 1.188-1.666</ns0:cell><ns0:cell>0.617 <0.001</ns0:cell><ns0:cell>0.534-0.714 age</ns0:cell><ns0:cell><0.001 1.39 1.171-1.651</ns0:cell><ns0:cell><0.001</ns0:cell></ns0:row><ns0:row><ns0:cell>stage</ns0:cell><ns0:cell cols='2'>CCR10 3.173 2.284-4.41</ns0:cell><ns0:cell>1.759 <0.001</ns0:cell><ns0:cell>1.381-2.241 stage</ns0:cell><ns0:cell><0.001 3.426 2.484-4.725</ns0:cell><ns0:cell><0.001</ns0:cell></ns0:row><ns0:row><ns0:cell>gender</ns0:cell><ns0:cell cols='2'>BMP8A 1.039 0.755-1.43</ns0:cell><ns0:cell>1.813 0.814</ns0:cell><ns0:cell>1.298-2.532 gender</ns0:cell><ns0:cell><0.001 0.999 0.726-1.375</ns0:cell><ns0:cell>0.997</ns0:cell></ns0:row><ns0:row><ns0:cell>BMP8A</ns0:cell><ns0:cell cols='2'>FGF17 1.213 1.057-1.391</ns0:cell><ns0:cell>7.335 0.006</ns0:cell><ns0:cell cols='2'>3.202-16.803 KLRC2 1.402 1.233-1.594 <0.001</ns0:cell><ns0:cell><0.001</ns0:cell></ns0:row><ns0:row><ns0:cell>age</ns0:cell><ns0:cell cols='2'>1.454 1.229-1.721</ns0:cell><ns0:cell><0.001</ns0:cell><ns0:cell>age</ns0:cell><ns0:cell>1.383 1.169-1.637</ns0:cell><ns0:cell><0.001</ns0:cell></ns0:row><ns0:row><ns0:cell>stage</ns0:cell><ns0:cell cols='2'>GDF1 3.607 2.619-4.969</ns0:cell><ns0:cell>1.483 <0.001</ns0:cell><ns0:cell>1.062-2.070 stage</ns0:cell><ns0:cell>0.021 3.429 2.483-4.737</ns0:cell><ns0:cell><0.001</ns0:cell></ns0:row><ns0:row><ns0:cell>gender</ns0:cell><ns0:cell cols='2'>IL4 1.052 0.765-1.447</ns0:cell><ns0:cell>109.106 0.754</ns0:cell><ns0:cell cols='2'>34.328-346.781 gender 0.958 0.694-1.322 <0.001</ns0:cell><ns0:cell>0.792</ns0:cell></ns0:row><ns0:row><ns0:cell>CCL7</ns0:cell><ns0:cell cols='2'>LHB 1.329 1.202-1.471</ns0:cell><ns0:cell>5.053 <0.001</ns0:cell><ns0:cell>3.240-7.881 LHB</ns0:cell><ns0:cell><0.001 1.285 1.168-1.414</ns0:cell><ns0:cell><0.001</ns0:cell></ns0:row><ns0:row><ns0:cell>age</ns0:cell><ns0:cell cols='2'>TXLNA 1.455 1.226-1.728</ns0:cell><ns0:cell>2.329 <0.001</ns0:cell><ns0:cell>1.500-3.618 age</ns0:cell><ns0:cell><0.001 1.393 1.178-1.647</ns0:cell><ns0:cell><0.001</ns0:cell></ns0:row><ns0:row><ns0:cell>stage</ns0:cell><ns0:cell cols='2'>AR 3.485 2.529-4.802</ns0:cell><ns0:cell>0.548 <0.001</ns0:cell><ns0:cell>0.463-0.649 stage</ns0:cell><ns0:cell><0.001 3.519 2.55-4.854</ns0:cell><ns0:cell><0.001</ns0:cell></ns0:row><ns0:row><ns0:cell>gender</ns0:cell><ns0:cell>1</ns0:cell><ns0:cell>IL20RB 0.726-1.377</ns0:cell><ns0:cell>1.225 1</ns0:cell><ns0:cell>1.145-1.310 gender</ns0:cell><ns0:cell><0.001 1.17 0.844-1.622</ns0:cell><ns0:cell>0.347</ns0:cell></ns0:row><ns0:row><ns0:cell>CCR10</ns0:cell><ns0:cell cols='2'>1.235 1.111-1.372</ns0:cell><ns0:cell><0.001</ns0:cell><ns0:cell>SEMA3A</ns0:cell><ns0:cell>1.214 1.107-1.33</ns0:cell><ns0:cell><0.001</ns0:cell></ns0:row><ns0:row><ns0:cell>age</ns0:cell><ns0:cell cols='2'>BID 1.382 1.166-1.639</ns0:cell><ns0:cell>3.505 <0.001</ns0:cell><ns0:cell>2.439-5.035 age</ns0:cell><ns0:cell><0.001 1.445 1.218-1.715</ns0:cell><ns0:cell><0.001</ns0:cell></ns0:row><ns0:row><ns0:cell>stage</ns0:cell><ns0:cell cols='2'>3.844 2.793-5.292</ns0:cell><ns0:cell><0.001</ns0:cell><ns0:cell>stage</ns0:cell><ns0:cell>3.549 2.572-4.898</ns0:cell><ns0:cell><0.001</ns0:cell></ns0:row><ns0:row><ns0:cell>gender</ns0:cell><ns0:cell cols='2'>1.034 0.751-1.424</ns0:cell><ns0:cell>0.836</ns0:cell><ns0:cell>gender</ns0:cell><ns0:cell>1.108 0.802-1.529</ns0:cell><ns0:cell>0.535</ns0:cell></ns0:row><ns0:row><ns0:cell>FGF17</ns0:cell><ns0:cell cols='2'>1.235 1.112-1.372</ns0:cell><ns0:cell><0.001</ns0:cell><ns0:cell cols='2'>SEMA3G 0.652 0.549-0.775</ns0:cell><ns0:cell><0.001</ns0:cell></ns0:row><ns0:row><ns0:cell>age</ns0:cell><ns0:cell cols='2'>1.399 1.181-1.657</ns0:cell><ns0:cell><0.001</ns0:cell><ns0:cell>age</ns0:cell><ns0:cell>1.379 1.164-1.633</ns0:cell><ns0:cell><0.001</ns0:cell></ns0:row><ns0:row><ns0:cell>stage</ns0:cell><ns0:cell cols='2'>3.722 2.705-5.121</ns0:cell><ns0:cell><0.001</ns0:cell><ns0:cell>stage</ns0:cell><ns0:cell>3.29 2.384-4.539</ns0:cell><ns0:cell><0.001</ns0:cell></ns0:row><ns0:row><ns0:cell>gender</ns0:cell><ns0:cell cols='2'>1.109 0.806-1.525</ns0:cell><ns0:cell>0.527</ns0:cell><ns0:cell>gender</ns0:cell><ns0:cell>0.973 0.707-1.339</ns0:cell><ns0:cell>0.867</ns0:cell></ns0:row><ns0:row><ns0:cell>GDF1</ns0:cell><ns0:cell cols='2'>1.086 0.975-1.211</ns0:cell><ns0:cell>0.135</ns0:cell><ns0:cell>TXLNA</ns0:cell><ns0:cell>1.285 1.098-1.505</ns0:cell><ns0:cell>0.002</ns0:cell></ns0:row><ns0:row><ns0:cell>age</ns0:cell><ns0:cell cols='2'>1.411 1.191-1.672</ns0:cell><ns0:cell><0.001</ns0:cell><ns0:cell>age</ns0:cell><ns0:cell>1.433 1.208-1.7</ns0:cell><ns0:cell><0.001</ns0:cell></ns0:row><ns0:row><ns0:cell>stage</ns0:cell><ns0:cell>3.76</ns0:cell><ns0:cell>2.735-5.17</ns0:cell><ns0:cell><0.001</ns0:cell><ns0:cell>stage</ns0:cell><ns0:cell>3.696 2.687-5.085</ns0:cell><ns0:cell><0.001</ns0:cell></ns0:row><ns0:row><ns0:cell>gender</ns0:cell><ns0:cell cols='2'>1.103 0.796-1.529</ns0:cell><ns0:cell>0.555</ns0:cell><ns0:cell>gender</ns0:cell><ns0:cell>1.212 0.87-1.688</ns0:cell><ns0:cell>0.255</ns0:cell></ns0:row></ns0:table></ns0:figure>
<ns0:note place='foot'>Manuscript to be reviewed</ns0:note>
<ns0:note place='foot'>PeerJ reviewing PDF | (2019:11:43459:2:0:NEW 24 Mar 2020)Manuscript to be reviewed</ns0:note>
<ns0:note place='foot'>PeerJ reviewing PDF | (2019:11:43459:2:0:NEW 24 Mar 2020)Manuscript to be reviewed1 Table 2. Multivariate Cox analysis of 14 immune relate genes in all cohort</ns0:note>
</ns0:body>
" | "Resubmission: A 14-immune-related-gene signature predicts clinical outcome of kidney renal clear cell carcinoma
Dear Editors,
Thank you very much for the efforts and suggestions from the editors and the reviewers. We have revised our manuscript entitled “A 14-immune-related-gene signature predicts clinical outcome of kidney renal clear cell carcinoma” according to the reviewers’ recommendation and would like to resubmit it. Based on the comments, we have carefully made major revision on the revised version. Detailed revisions are shown as follows. We have uploaded the manuscript with computer-generated tracked changes to the revision response files section. Besides, we have uploaded the a “clean” copy where the changes are not marked.
1. I apologize for taking awhile. Two of the original reviewers were unavailable to re-review the article. So I sent it out for review to a new reviewer. As you can see below, this reviewer had some concerns. Please address this reviewer's concerns about the C-index values. At a minimum, add some discussion to the paper about the potential clinical relevance (or lack thereof) of a C-index in this range.
Reply: Thank you very much for your comment. As shown in the Discussion part, we compared the C-index of our testing set with other clinical (clinical staging systems) and molecular signatures. We found that despite of the relatively low C-index of our testing set, the C-index of our testing set was still higher than those of other clinical (0.62) and molecular signatures. Therefore, our immune signature can still serve as a more promising tool in predicting KIRC patients’ survival outcome than other clinical and molecular signatures.
2. The reviewer expressed concern about the survival curves crossing in Figure 3 for the GDF1 gene. Please address that briefly. But more importantly, these p-values should be adjusted for multiple tests.
Reply: Thank you very much for your comment. we performed univariate Cox analysis to screen out immune-related genes with prognostic properties. To identify the best gene model for predicting KIRC patient prognosis, genes with a P-value lower than 0.05 were evaluated using the Cox proportional hazards model with a lasso penalty (log lambda = -2.641, alpha = 1, iteration = 1,000. The p-value of Figure 3 was from log-rank test. Because of the differences between the log-rank test and the univariate Cox analysis, the results of the univariate Cox analysis of other genes except GDF1 are very consistent with the results of the K-M (Kaplan-Meier) survival curves. We have explained this in the manuscript.
3. Regarding the GSEA analysis, an FDR threshold of 0.25 is fairly high. A threshold in the range of 0.05 to 0.20 is more customary. Please modify this or provide a short justification for using 0.25.
Reply: Thank you very much for your suggestions. The publisher of the GSEA method has used FDR <=0.25 as the threshold (Subramanian A, Tamayo P, Mootha VK, et al. Gene set enrichment analysis: a knowledge-based approach for interpreting genome-wide expression profiles. Proc Natl Acad Sci U S A. 2005;102(43):15545-50). So we chose 0.25 as the threshold.
4. Per the reviewer's comment, please review the references and update them to more recently published articles in the field.
Reply: Thank you very much for your suggestion. According to your comment, we have updated some references published within 3 years.
5. Some of the text is still much too small in Figures 3, 4, 6, 7, 8, and 9.
Reply: Thank you very much for your suggestion. We have modified the font size of Figures 3, 4, 6, 7.
6. I'm not sure that it makes sense to include the figures from the GSEA output. Consider providing a table with the GSEA results rather than showing these graphics in Figures 8 and 9.
Reply: Thank you very much for your suggestion. As shown in Table 4, we have changed the GSEA results to tables.
7. Thank you for providing slightly more detail in the GitHub repository about which packages to install. However, instead of explaining this verbally, give the reader the exact code they need to install the packages. Also, your R code should download the data files. The idea is that someone could re-run your analysis without having to modify your code.
Reply: Thank you very much for your suggestion. We have written the code for installing the R package into the R file. Because files uploaded in github cannot exceed 25M, we cannot upload our data, but we have provided a data download link to users.
We hope that the revision and our responses can address your concerns and that you can consider this article suitable for acceptance. Please do not hesitate to contact us if any additional explanation or revision is required. Thank you again for your assistance.
Best regards
Chuan Hu
" | Here is a paper. Please give your review comments after reading it. |
9,835 | "<ns0:abstract xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:p>Kidney renal clear cell carcinoma (KIRC) is the leading cause of kidney cancer-related deaths. Currently, there are no studies in tumor immunology investigating the use of signatures as a predictor of overall survival in KIRC patients. Our study attempts to establish an immune-related gene risk signature to predict clinical outcomes in KIRC. A total of 528 patients from The Cancer Genome Atlas (TCGA) database were included in our analysis and randomly divided into training (n = 315) and testing sets (n = 213). We collected 1,534 immune-related genes from the Immunology Database and Analysis Portal as candidates to construct our signature. LASSO-COX was used to find gene models with the highest predictive ability. We used survival and Cox analysis to test the model's independent prognostic ability. Univariate analysis identified 650 immune-related genes with prognostic abilities. After 1,000 iterations, we choose 14 of the most frequent and stable immune-related genes as our signature. We found that the signature was associated with M stage, T stage, and pathological staging. More importantly, the signature can independently predict clinical prognosis in KIRC patients. Gene Set Enrichment Analysis (GSEA) showed an association between our signature and critical metabolism pathways.</ns0:p><ns0:p>Our research established a model based upon 14 immune-related genes that predicted the prognosis of KIRC patients based on tumor immune microenvironments.</ns0:p></ns0:div>
</ns0:abstract>
<ns0:body xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:head>Introduction</ns0:head><ns0:p>Kidney cancer is one of the most common urological tumors worldwide, with approximately 403,262 new cases and 175,098 deaths associated with this form of cancer in 2018 <ns0:ref type='bibr' target='#b1'>(Bray et al. 2018)</ns0:ref>. Kidney renal clear cell carcinoma (KIRC) is the leading cause of kidney cancer mortality and the most common type, accounting for 85% of kidney cancers <ns0:ref type='bibr'>(Siegel et al. 2017</ns0:ref>; Tseng 2016). Considering the aging population of KIRC patients and the rising expense of treatment, KIRC is gradually becoming the focus of geriatric cancer <ns0:ref type='bibr'>(Tan et al. 2015)</ns0:ref>. Despite the rapid development of cancer treatments, mortality rates in KIRC remain stagnant. With the progression of next generation sequencing and data mining techniques, it is urgent that we explore prognostic biomarkers for KIRC, using molecular characteristics and tumor immune environments to guide patient therapy.</ns0:p><ns0:p>Over the past decade our understanding of immune components, including the impact of tumor microenvironments on patient survival and therapy response <ns0:ref type='bibr' target='#b2'>(Chen & Mellman 2017;</ns0:ref><ns0:ref type='bibr' target='#b7'>Grivennikov et al. 2010</ns0:ref>), has increased. Some studies have found that tumor-infiltrating immune cells are able to serve as either tumor suppressors or promoters in microenvironments. For example, CD8 + T cells have been associated with improved survival in cancer patients <ns0:ref type='bibr' target='#b4'>(Gajewski et al. 2013;</ns0:ref><ns0:ref type='bibr' target='#b6'>Governa et al. 2017)</ns0:ref>, while tumor associated macrophages and regulatory T cells demonstrate the ability to promote tumor development <ns0:ref type='bibr' target='#b11'>(Nishikawa & Sakaguchi 2014;</ns0:ref><ns0:ref type='bibr' target='#b12'>Noy & Pollard 2014)</ns0:ref>. Considering the complexity and significance of tumor immune microenvironments, it is imperative that we investigate immune-related biomarkers for KIRC patients. Recent studies have provided insight into the KIRC immune signature <ns0:ref type='bibr' target='#b5'>(Geissler et al. 2015;</ns0:ref><ns0:ref type='bibr' target='#b17'>Şenbabaoğlu et al. 2016)</ns0:ref>. <ns0:ref type='bibr'>Şenbabaoğlu et al. (2016)</ns0:ref> found mRNA signatures with the potential to be immunotherapeutic biomarkers in KIRC. However, their investigations do not include immune-related genes for analysis nor do they establish a systematic immune-related gene-risk signature for KIRC patients. Khadirnaikar et al.</ns0:p><ns0:p>(2019) utilized immune associated lncRNA (long non-coding RNA) to construct prognostic subtypes in KIRC patients. Our immune clusters were more robust and independent. Additionally, they concentrated on lncRNA, not the overall immune-related genes. <ns0:ref type='bibr'>Smith et al. (2018)</ns0:ref> constructed endogenous retroviral signatures for KIRC patients, but they did not investigate the prognostic ability of the signature in different subtypes of patients. Therefore, it is essential that we explore a systematic prognostic signature based on tumor immune environments in KIRC.</ns0:p><ns0:p>In our study, we used RNA-seq data from The Cancer Genome Atlas (TCGA) to find immune-related genes with prognostic ability and to establish an immune-related risk signature for KIRC. To assess the clinical potential of the signature, we investigated the association between the signature, clinical parameters, and patient survival. Gene set enrichment analysis (GSEA) was performed to explore the molecular characteristics of the signature.</ns0:p></ns0:div>
<ns0:div><ns0:head>Materials & Methods</ns0:head></ns0:div>
<ns0:div><ns0:head>Patient cohort</ns0:head><ns0:p>The TCGA database was used to collect the clinical and RNA-seq data of 528 KIRC patients. We randomly divided the dataset into training (n = 315) and test sets (n = 213). RNA-seq data was obtained to analyze the transcriptome profiling of RNA expression and were measured using fragments per kilobase of exon per million fragments mapped (FPKM). We performed a log2-based transformation to normalize RNA expression profiles. To ensure detection and total cohort <ns0:ref type='bibr' target='#b9'>(Harrell et al. 1982)</ns0:ref>. The independent prognostic ability of the immune-related risk signature was assessed using survival analysis as well as Cox analysis (R package survival, v2.42). We performed multivariable Cox regression to analyze the relationship between immune-related risk signatures and clinicopathological factors.</ns0:p></ns0:div>
<ns0:div><ns0:head>Gene set enrichment analysis</ns0:head><ns0:p>To identify the biological functions and pathways between high-and low-risk groups, we conducted GSEA to investigate potential biological mechanisms in the Molecular Signatures Database (MSigDB; <ns0:ref type='bibr'>(Subramanian et al. 2005)</ns0:ref>. GSEA was performed using GSEApy, a python wrapper for gene enrichment (https://pypi.org/project/gseapy/).</ns0:p><ns0:p>We selected C2 and C5, including pathway databases and GO terms, from the MSigDB. The gesa sub-command of GSEApy was used in GSEA with default parameters. Enriched gene sets with a false discovery rate (FDR) of less than 0.25 and a P-value of less than 0.05 were considered statistically significant.</ns0:p></ns0:div>
<ns0:div><ns0:head>Statistical analysis</ns0:head><ns0:p>Boxplots were created with the R package, ggplot2 (v3.0.0). The R package ComplexHeatmap (v1.18.1) was used to create heatmaps <ns0:ref type='bibr' target='#b8'>(Gu et al. 2016)</ns0:ref>. We counted C-index with R packages 'survcomp' <ns0:ref type='bibr' target='#b9'>(Harrell et al. 1982;</ns0:ref><ns0:ref type='bibr' target='#b16'>Schröder et al. 2011)</ns0:ref>. The student's t-test was performed for statistical comparison. We chose R to conduct statistical analysis (https://www.r-project.org/). P-values lower than 0.05 were considered statistically significant. The main code of analysis was pushed to github (https://github.com/huchua/KIRC).</ns0:p></ns0:div>
<ns0:div><ns0:head>Results</ns0:head></ns0:div>
<ns0:div><ns0:head>Establishment and validation of the immune-related gene signature in KIRC</ns0:head><ns0:p>Figure <ns0:ref type='figure'>1</ns0:ref> illustrates the workflow we used to develop an immune-related gene-risk signature. The immune-related gene signature was constructed within the KIRC training data set, while we applied the testing set to validate the signature.</ns0:p><ns0:p>After 1,000 iterations, seven unique gene models were selected (Figure <ns0:ref type='figure'>2A</ns0:ref>, Table <ns0:ref type='table'>S1</ns0:ref>). The model selected as the Manuscript to be reviewed immune-related gene-risk signature consisted of 14 genes and ranked the highest in frequency 282 times. Parameter selection in the LASSO-cox model was log lambda = -2.641 and alpha = 1. The univariate and multivariate Cox analysis of the 14 immune-related genes are shown in Table <ns0:ref type='table'>1 and Table 2</ns0:ref>. The principle component analysis of the 14 immune-related gene signature displayed a different distribution pattern between low-and high-risk groups when comparing the training, testing, and total cohort (Figure <ns0:ref type='figure'>2B-D</ns0:ref>). This indicated that the low-and high-risk groups had different immune phenotypes. In the training, testing, and total data set, the c-index was 0.7862, 0.6534, and 0.7367, respectively (P < 0.001; Figure <ns0:ref type='figure'>2E</ns0:ref>). In a time-dependent receiver operating characteristic curve (ROC) created for three datasets, area under the curve (AUC) values at 1, 3, and 5 years were 0.679, 0.63, 0.627; 0.65, 0.596, 0.568; and 0.618, 0.589, 0.59, respectively (Figure <ns0:ref type='figure'>S2</ns0:ref>). The 14 immune-related genes are AR, BID, BMP8A, CCL7, CCR10, FGF17, GDF1, IL20RB, IL4, KLRC2, LHB, SEMA3A, SEMA3G, and TXLNA. K-M (Kaplan-Meier) survival curves and gene expression of the 14 immune-related genes are shown in Figure <ns0:ref type='figure'>3</ns0:ref> and Figure <ns0:ref type='figure'>S3</ns0:ref>. Because of the differences between the log-rank test and the univariate cox analysis, the results of the univariate cox analysis of other genes except GDF1 are very consistent with the results of the K-M (Kaplan-Meier) survival curves.</ns0:p></ns0:div>
<ns0:div><ns0:head>Correlation between the signature of immune-related genes and clinical parameters</ns0:head><ns0:p>Among the 14 immune-related genes, four genes (TXLNA, SEMA3G, AR, and BID) had a high expression, and 10 genes (IL20RB, CCR10, BMP8A, SEMA3A, CCL7, GDF1, KLRC2, LHB, FGF17, and IL4) had a low expression (Figure <ns0:ref type='figure'>4A, B</ns0:ref>; Figure <ns0:ref type='figure' target='#fig_4'>5</ns0:ref>). The relationship between the signature and clinical factors demonstrated that patients with advanced pathological staging, M stage, and T stage had a higher risk score than those with early stage disease (Figure <ns0:ref type='figure'>4A</ns0:ref>, B; Figure <ns0:ref type='figure' target='#fig_4'>5</ns0:ref>; Figure <ns0:ref type='figure' target='#fig_9'>6A, B, D</ns0:ref>). However, we did not find a correlation between the signature and N stage (Figure <ns0:ref type='figure'>4A</ns0:ref>, B; Figure <ns0:ref type='figure' target='#fig_4'>5</ns0:ref>; Figure <ns0:ref type='figure' target='#fig_9'>6C</ns0:ref>).</ns0:p></ns0:div>
<ns0:div><ns0:head>Influence of the immune-related gene signature on patient prognosis</ns0:head><ns0:p>PeerJ reviewing <ns0:ref type='table'>PDF | (2019:11:43459:3:0:NEW 30 Mar 2020)</ns0:ref> Manuscript to be reviewed</ns0:p><ns0:p>We then assessed whether this signature influenced KIRC patient prognosis. Survival analysis showed that patients with a high-risk score were associated with poor survival outcomes in the training, testing, and total group sets (Figure <ns0:ref type='figure'>4C</ns0:ref>; Figure <ns0:ref type='figure'>4E</ns0:ref>; Figure <ns0:ref type='figure'>7B</ns0:ref>). We found that the signature also predicted survival outcomes in subgroups of KIRC patients, including stage I-II (Figure <ns0:ref type='figure'>7C</ns0:ref>), stage III-IV (Figure <ns0:ref type='figure'>7D</ns0:ref>), M0 stage (Figure <ns0:ref type='figure'>7E</ns0:ref>), M1 stage (Figure <ns0:ref type='figure'>7F</ns0:ref>), N0 stage (Figure <ns0:ref type='figure'>7G</ns0:ref>), N1 (Figure <ns0:ref type='figure'>7H</ns0:ref>), T1 (Figure <ns0:ref type='figure'>7I</ns0:ref>), T2 (Figure <ns0:ref type='figure'>7J</ns0:ref>), T3 (Figure <ns0:ref type='figure'>7K</ns0:ref>), and T4 (Figure <ns0:ref type='figure'>7L</ns0:ref>). Multivariate Cox analysis revealed that the risk signature was able to independently predict overall survival in KIRC patients (Figure <ns0:ref type='figure'>4D</ns0:ref>, F; Figure <ns0:ref type='figure'>7A</ns0:ref>; Table <ns0:ref type='table'>3</ns0:ref>).</ns0:p></ns0:div>
<ns0:div><ns0:head>Engagement of the immune-related gene risk signature in biological pathways and functions</ns0:head><ns0:p>GSEA was used to investigate the signature's biological pathways and functions. There were 177 KEGG pathways and 4,528 GO terms used in our investigation. Our analysis found that the signature was able to engage in a total of 19 enriched KEGG pathways (FDR < 0.25; Table <ns0:ref type='table'>4</ns0:ref>). The low-risk signature was significantly correlated with 10 pathways, including the citrate cycle (TCA cycle) pathway, fatty acid metabolism pathway, propanoate metabolism pathway, butanoate metabolism pathway, peroxisome pathway, lysine degradation pathway, valine leucine and isoleucine degradation pathway, proximal tubule bicarbonate reclamation pathway, vasopressin regulated water reabsorption pathway, and the pyruvate metabolism pathway (Table <ns0:ref type='table'>4</ns0:ref>). Similarly, eight GO annotations were enriched in the low-risk group (FDR < 0.25; Table <ns0:ref type='table'>4</ns0:ref>).</ns0:p></ns0:div>
<ns0:div><ns0:head>Discussion</ns0:head><ns0:p>Our study used the TCGA database and immune-related genes to establish a KIRC signature consisting of 14 immunerelated genes. We found that patients in the high-risk group showed a positive association with M stage, T stage, and advanced pathological staging. Additionally, the signature exhibited strong prognostic abilities and independently predicted KIRC patient prognosis. Functional analysis highlighted the significance of our signature based on its Manuscript to be reviewed involvement in many important pathways.</ns0:p><ns0:p>In our investigation, we used a total of 14 immune-related genes to construct our signature. In the signature, CCL7 increased the peripheral blood mononuclear cell recruitment in renal cell cancer through the inhibition of let-7d <ns0:ref type='bibr' target='#b15'>(Riihimaki et al. 2014)</ns0:ref>. <ns0:ref type='bibr'>Wyler et al. (2014)</ns0:ref> demonstrated the ability of CCL7 to recruit monocytes through CCR2, promoting renal cell cancer metastasis to the brain. This indicates that CCL7 is a major factor in the development of KIRC in tumor immune microenvironments and may be a potential immunotherapy target. Fibroblast growth factor receptor 17 (FGF17) is another immune-related gene in our signature. FGF17 demonstrates a variety of functions in cancer development. <ns0:ref type='bibr'>Gauglhofer et al. (2011)</ns0:ref> found that FGF17 was involved in the paracrine and autocrine signaling of hepatocellular carcinoma and promoted the neoangiogenesis of hepatocellular carcinoma. <ns0:ref type='bibr'>Heer et al. (2004)</ns0:ref> showed that FGF17 was overexpressed in prostate cancer and participated in prostate carcinogenesis. Our study showed that FGF17 plays an important role in KIRC based on tumor immunology. However, the underlying mechanism of FGF17 in KIRC immune microenvironments requires further investigation. In our signature, IL-4 is another well-studied immune-related biomarker for KIRC. IL-4, released by immune cells, controls the expression of B7-H1, thus altering T cell responses in KIRC <ns0:ref type='bibr' target='#b14'>(Quandt et al. 2014)</ns0:ref> Manuscript to be reviewed significance of our signature and its possible use as a survival predictor in KIRC.</ns0:p><ns0:p>Despite of the relatively low C-index of our testing set (0.6534), the C-index of the testing set falls within a similar range as those of other related studies. For instance, in the study of Bailiang Li et al., they developed an immune signature in non-small cell lung cancer with the C-index of 0.64 <ns0:ref type='bibr' target='#b10'>(Li et al. 2017)</ns0:ref>. Besides, our signature achieved a similar C-index of testing set with the immune signature in ovarian cancer (0.625) <ns0:ref type='bibr'>(Shen et al. 2019</ns0:ref>). More importantly, the C-index of our testing set showed a similar accuracy with the C-index of clinical staging systems in renal cancer (0.62) <ns0:ref type='bibr' target='#b13'>(Qu et al. 2018</ns0:ref>).</ns0:p><ns0:p>Our study expands on the signatures association with several important pathways, especially the metabolism pathway.</ns0:p><ns0:p>This may reflect the mutual interaction between tumor metabolism and tumor immunology in KIRC. Pearce et al.</ns0:p><ns0:p>(2013) showed that metabolic reprogramming can influence the fate and function of T cells in tumors. Our study further indicates the importance of metabolism pathways in KIRC immune microenvironments.</ns0:p></ns0:div>
<ns0:div><ns0:head>Conclusions</ns0:head><ns0:p>This investigation utilized RNA-seq data from the TCGA database to construct a 14 immune-related gene-risk signature with the ability to independently predict survival outcomes in KIRC; thus, providing novel clinical applications and possible immune targets for KIRC. Manuscript to be reviewed </ns0:p><ns0:note type='other'>Figure legends</ns0:note><ns0:note type='other'>Figure 7</ns0:note><ns0:p>The 14 immune-related genes signature could serve as an independent prognostic factor for OS </ns0:p></ns0:div><ns0:figure xml:id='fig_0'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2019:11:43459:3:0:NEW 30 Mar 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_1'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2019:11:43459:3:0:NEW 30 Mar 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_2'><ns0:head /><ns0:label /><ns0:figDesc>. The investigation conducted by Chang et al. (2015) demonstrated that IL-4 expression can predict KIRC patient recurrence and survival outcomes. Collectively, these investigations reinforce the significance of the 14 immune-related gene-risk signature in KIRC. Our signature is associated with the survival outcome of KIRC patients and clinical parameters, including pathological staging, M stage, and T stage. No correlation was found between the signature and N-stage, possibly due to a lack of N-stage information for many patients. According to the immune-related gene-risk signature, our study found that clinical cohorts in KIRC have different immune-related risk factors. This signature also reflects differences in tumor immune microenvironments and predicts survival outcomes in KIRC patients; thus, demonstrating the clinical PeerJ reviewing PDF | (2019:11:43459:3:0:NEW 30 Mar 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_3'><ns0:head>Figure 1 :Figure 2 :Figure 4 :</ns0:head><ns0:label>124</ns0:label><ns0:figDesc>Figure 1: Our workflow constructing the model for risk-score signatures</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_4'><ns0:head>Figure 5 :</ns0:head><ns0:label>5</ns0:label><ns0:figDesc>Figure 5: Analysis of the 14 immune-related genes predictive ability in total cohort. As demonstrated by the heatmap,</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_5'><ns0:head>Figure 6 :</ns0:head><ns0:label>6</ns0:label><ns0:figDesc>Figure 6: The K-M survival curve of (A) pathological staging, (B) M stage, (C) N stage, and (D) T stage (left). The</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_6'><ns0:head>Figure S1 :</ns0:head><ns0:label>S1</ns0:label><ns0:figDesc>Figure S1: Parameter selection in the LASSO-cox model. (A) LASSO coefficient values of the 14 immune related</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_7'><ns0:head>Figure S3 :Figure 1</ns0:head><ns0:label>S31</ns0:label><ns0:figDesc>FigureS3: Expression of 14 immune-related genes in kidney renal clear cell carcinoma and normal tissues.</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_8'><ns0:head>Figure 5</ns0:head><ns0:label>5</ns0:label><ns0:figDesc>Figure 5</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_9'><ns0:head>Figure 6</ns0:head><ns0:label>6</ns0:label><ns0:figDesc>Figure 6</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_10'><ns0:head>(</ns0:head><ns0:label /><ns0:figDesc>A) The multivariate Cox analysis of total cohorts demonstrated that the 14 immune-related genes signature could serve as an independent prognostic factor for OS. The K-M analysis of the risk signature in the total cohort (B) and in subgroups of patients with stage I-II (C), stage III-IV (D), M0 stage (E), M1 stage (F), N0 stage (G), N1 stage (H), T1 stage (I), T2 stage (J), T3 stage (K), and T4 stage (L)</ns0:figDesc></ns0:figure>
<ns0:figure type='table' xml:id='tab_0'><ns0:head /><ns0:label /><ns0:figDesc>Oncotarget 8:87292. Shen S, Wang G, Zhang R, Zhao Y, Yu H, Wei Y, and Chen F. 2019. Development and validation of an immune</ns0:figDesc><ns0:table><ns0:row><ns0:cell>gene-set</ns0:cell><ns0:cell>based</ns0:cell><ns0:cell>Prognostic</ns0:cell><ns0:cell>signature</ns0:cell><ns0:cell>in</ns0:cell><ns0:cell>ovarian</ns0:cell><ns0:cell>cancer.</ns0:cell><ns0:cell>EBioMedicine</ns0:cell><ns0:cell>40:318-326.</ns0:cell></ns0:row><ns0:row><ns0:cell cols='3'>10.1016/j.ebiom.2018.12.054</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell cols='8'>Siegel R, Miller K, and Jemal A. 2017. Cancer statistics, 2018 CA: a cancer. J Clin 68: 7-30.</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell cols='9'>Subramanian A, Tamayo P, Mootha VK, Mukherjee S, Ebert BL, Gillette MA, Paulovich A, Pomeroy SL, Golub TR,</ns0:cell></ns0:row><ns0:row><ns0:cell cols='9'>and Lander ES. 2005. Gene set enrichment analysis: a knowledge-based approach for interpreting genome-</ns0:cell></ns0:row><ns0:row><ns0:cell cols='8'>wide expression profiles. Proceedings of the National Academy of Sciences 102:15545-15550.</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell cols='9'>Tan H-J, Filson CP, and Litwin MS. 2015. Contemporary, age-based trends in the incidence and management of</ns0:cell></ns0:row><ns0:row><ns0:cell cols='9'>patients with early-stage kidney cancer. Urologic Oncology: Seminars and Original Investigations: Elsevier.</ns0:cell></ns0:row><ns0:row><ns0:cell cols='2'>p 21. e19-21. e26.</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell cols='9'>Tseng C-H. 2016. Use of metformin and risk of kidney cancer in patients with type 2 diabetes. European Journal of</ns0:cell></ns0:row><ns0:row><ns0:cell cols='2'>Cancer 52:19-25.</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row></ns0:table><ns0:note>types.</ns0:note></ns0:figure>
<ns0:figure type='table' xml:id='tab_1'><ns0:head>Table1 :</ns0:head><ns0:label>Table1</ns0:label><ns0:figDesc>Univariate Cox analysis of 14 immune relate genes in all cohort</ns0:figDesc><ns0:table><ns0:row><ns0:cell>AR</ns0:cell><ns0:cell cols='2'>Variable 0.638 0.556-0.731</ns0:cell><ns0:cell>HR <0.001</ns0:cell><ns0:cell>95%CI IL20RB</ns0:cell><ns0:cell>pvalue 1.268 1.102-1.46</ns0:cell><ns0:cell>0.001</ns0:cell></ns0:row><ns0:row><ns0:cell>age</ns0:cell><ns0:cell cols='2'>KLRC2 1.352 1.143-1.599</ns0:cell><ns0:cell>20.657 <0.001</ns0:cell><ns0:cell>9.116-46.81 age</ns0:cell><ns0:cell><0.001 1.414 1.195-1.672</ns0:cell><ns0:cell><0.001</ns0:cell></ns0:row><ns0:row><ns0:cell>stage</ns0:cell><ns0:cell cols='2'>CCL7 3.725 2.708-5.124</ns0:cell><ns0:cell>2.757 <0.001</ns0:cell><ns0:cell>2.090-3.637 stage</ns0:cell><ns0:cell><0.001 3.275 2.351-4.562</ns0:cell><ns0:cell><0.001</ns0:cell></ns0:row><ns0:row><ns0:cell>gender</ns0:cell><ns0:cell cols='2'>0.943 0.687-1.294</ns0:cell><ns0:cell>0.718</ns0:cell><ns0:cell>gender</ns0:cell><ns0:cell>1.025 0.744-1.412</ns0:cell><ns0:cell>0.879</ns0:cell></ns0:row><ns0:row><ns0:cell>BID</ns0:cell><ns0:cell cols='2'>SEMA3A 1.411 1.211-1.645</ns0:cell><ns0:cell>1.748 <0.001</ns0:cell><ns0:cell>1.437-2.126 IL4</ns0:cell><ns0:cell><0.001 1.458 1.294-1.643</ns0:cell><ns0:cell><0.001</ns0:cell></ns0:row><ns0:row><ns0:cell>age</ns0:cell><ns0:cell cols='2'>SEMA3G 1.407 1.188-1.666</ns0:cell><ns0:cell>0.617 <0.001</ns0:cell><ns0:cell>0.534-0.714 age</ns0:cell><ns0:cell><0.001 1.39 1.171-1.651</ns0:cell><ns0:cell><0.001</ns0:cell></ns0:row><ns0:row><ns0:cell>stage</ns0:cell><ns0:cell cols='2'>CCR10 3.173 2.284-4.41</ns0:cell><ns0:cell>1.759 <0.001</ns0:cell><ns0:cell>1.381-2.241 stage</ns0:cell><ns0:cell><0.001 3.426 2.484-4.725</ns0:cell><ns0:cell><0.001</ns0:cell></ns0:row><ns0:row><ns0:cell>gender</ns0:cell><ns0:cell cols='2'>BMP8A 1.039 0.755-1.43</ns0:cell><ns0:cell>1.813 0.814</ns0:cell><ns0:cell>1.298-2.532 gender</ns0:cell><ns0:cell><0.001 0.999 0.726-1.375</ns0:cell><ns0:cell>0.997</ns0:cell></ns0:row><ns0:row><ns0:cell>BMP8A</ns0:cell><ns0:cell cols='2'>FGF17 1.213 1.057-1.391</ns0:cell><ns0:cell>7.335 0.006</ns0:cell><ns0:cell cols='2'>3.202-16.803 KLRC2 1.402 1.233-1.594 <0.001</ns0:cell><ns0:cell><0.001</ns0:cell></ns0:row><ns0:row><ns0:cell>age</ns0:cell><ns0:cell cols='2'>1.454 1.229-1.721</ns0:cell><ns0:cell><0.001</ns0:cell><ns0:cell>age</ns0:cell><ns0:cell>1.383 1.169-1.637</ns0:cell><ns0:cell><0.001</ns0:cell></ns0:row><ns0:row><ns0:cell>stage</ns0:cell><ns0:cell cols='2'>GDF1 3.607 2.619-4.969</ns0:cell><ns0:cell>1.483 <0.001</ns0:cell><ns0:cell>1.062-2.070 stage</ns0:cell><ns0:cell>0.021 3.429 2.483-4.737</ns0:cell><ns0:cell><0.001</ns0:cell></ns0:row><ns0:row><ns0:cell>gender</ns0:cell><ns0:cell cols='2'>IL4 1.052 0.765-1.447</ns0:cell><ns0:cell>109.106 0.754</ns0:cell><ns0:cell cols='2'>34.328-346.781 gender 0.958 0.694-1.322 <0.001</ns0:cell><ns0:cell>0.792</ns0:cell></ns0:row><ns0:row><ns0:cell>CCL7</ns0:cell><ns0:cell cols='2'>LHB 1.329 1.202-1.471</ns0:cell><ns0:cell>5.053 <0.001</ns0:cell><ns0:cell>3.240-7.881 LHB</ns0:cell><ns0:cell><0.001 1.285 1.168-1.414</ns0:cell><ns0:cell><0.001</ns0:cell></ns0:row><ns0:row><ns0:cell>age</ns0:cell><ns0:cell cols='2'>TXLNA 1.455 1.226-1.728</ns0:cell><ns0:cell>2.329 <0.001</ns0:cell><ns0:cell>1.500-3.618 age</ns0:cell><ns0:cell><0.001 1.393 1.178-1.647</ns0:cell><ns0:cell><0.001</ns0:cell></ns0:row><ns0:row><ns0:cell>stage</ns0:cell><ns0:cell cols='2'>AR 3.485 2.529-4.802</ns0:cell><ns0:cell>0.548 <0.001</ns0:cell><ns0:cell>0.463-0.649 stage</ns0:cell><ns0:cell><0.001 3.519 2.55-4.854</ns0:cell><ns0:cell><0.001</ns0:cell></ns0:row><ns0:row><ns0:cell>gender</ns0:cell><ns0:cell>1</ns0:cell><ns0:cell>IL20RB 0.726-1.377</ns0:cell><ns0:cell>1.225 1</ns0:cell><ns0:cell>1.145-1.310 gender</ns0:cell><ns0:cell><0.001 1.17 0.844-1.622</ns0:cell><ns0:cell>0.347</ns0:cell></ns0:row><ns0:row><ns0:cell>CCR10</ns0:cell><ns0:cell cols='2'>1.235 1.111-1.372</ns0:cell><ns0:cell><0.001</ns0:cell><ns0:cell>SEMA3A</ns0:cell><ns0:cell>1.214 1.107-1.33</ns0:cell><ns0:cell><0.001</ns0:cell></ns0:row><ns0:row><ns0:cell>age</ns0:cell><ns0:cell cols='2'>BID 1.382 1.166-1.639</ns0:cell><ns0:cell>3.505 <0.001</ns0:cell><ns0:cell>2.439-5.035 age</ns0:cell><ns0:cell><0.001 1.445 1.218-1.715</ns0:cell><ns0:cell><0.001</ns0:cell></ns0:row><ns0:row><ns0:cell>stage</ns0:cell><ns0:cell cols='2'>3.844 2.793-5.292</ns0:cell><ns0:cell><0.001</ns0:cell><ns0:cell>stage</ns0:cell><ns0:cell>3.549 2.572-4.898</ns0:cell><ns0:cell><0.001</ns0:cell></ns0:row><ns0:row><ns0:cell>gender</ns0:cell><ns0:cell cols='2'>1.034 0.751-1.424</ns0:cell><ns0:cell>0.836</ns0:cell><ns0:cell>gender</ns0:cell><ns0:cell>1.108 0.802-1.529</ns0:cell><ns0:cell>0.535</ns0:cell></ns0:row><ns0:row><ns0:cell>FGF17</ns0:cell><ns0:cell cols='2'>1.235 1.112-1.372</ns0:cell><ns0:cell><0.001</ns0:cell><ns0:cell cols='2'>SEMA3G 0.652 0.549-0.775</ns0:cell><ns0:cell><0.001</ns0:cell></ns0:row><ns0:row><ns0:cell>age</ns0:cell><ns0:cell cols='2'>1.399 1.181-1.657</ns0:cell><ns0:cell><0.001</ns0:cell><ns0:cell>age</ns0:cell><ns0:cell>1.379 1.164-1.633</ns0:cell><ns0:cell><0.001</ns0:cell></ns0:row><ns0:row><ns0:cell>stage</ns0:cell><ns0:cell cols='2'>3.722 2.705-5.121</ns0:cell><ns0:cell><0.001</ns0:cell><ns0:cell>stage</ns0:cell><ns0:cell>3.29 2.384-4.539</ns0:cell><ns0:cell><0.001</ns0:cell></ns0:row><ns0:row><ns0:cell>gender</ns0:cell><ns0:cell cols='2'>1.109 0.806-1.525</ns0:cell><ns0:cell>0.527</ns0:cell><ns0:cell>gender</ns0:cell><ns0:cell>0.973 0.707-1.339</ns0:cell><ns0:cell>0.867</ns0:cell></ns0:row><ns0:row><ns0:cell>GDF1</ns0:cell><ns0:cell cols='2'>1.086 0.975-1.211</ns0:cell><ns0:cell>0.135</ns0:cell><ns0:cell>TXLNA</ns0:cell><ns0:cell>1.285 1.098-1.505</ns0:cell><ns0:cell>0.002</ns0:cell></ns0:row><ns0:row><ns0:cell>age</ns0:cell><ns0:cell cols='2'>1.411 1.191-1.672</ns0:cell><ns0:cell><0.001</ns0:cell><ns0:cell>age</ns0:cell><ns0:cell>1.433 1.208-1.7</ns0:cell><ns0:cell><0.001</ns0:cell></ns0:row><ns0:row><ns0:cell>stage</ns0:cell><ns0:cell>3.76</ns0:cell><ns0:cell>2.735-5.17</ns0:cell><ns0:cell><0.001</ns0:cell><ns0:cell>stage</ns0:cell><ns0:cell>3.696 2.687-5.085</ns0:cell><ns0:cell><0.001</ns0:cell></ns0:row><ns0:row><ns0:cell>gender</ns0:cell><ns0:cell cols='2'>1.103 0.796-1.529</ns0:cell><ns0:cell>0.555</ns0:cell><ns0:cell>gender</ns0:cell><ns0:cell>1.212 0.87-1.688</ns0:cell><ns0:cell>0.255</ns0:cell></ns0:row></ns0:table></ns0:figure>
<ns0:note place='foot'>Manuscript to be reviewed</ns0:note>
<ns0:note place='foot'>PeerJ reviewing PDF | (2019:11:43459:3:0:NEW 30 Mar 2020)Manuscript to be reviewed</ns0:note>
<ns0:note place='foot'>PeerJ reviewing PDF | (2019:11:43459:3:0:NEW 30 Mar 2020)Manuscript to be reviewed1 Table 2. Multivariate Cox analysis of 14 immune relate genes in all cohort</ns0:note>
</ns0:body>
" | "Resubmission: A 14-immune-related-gene signature predicts clinical outcome of kidney renal clear cell carcinoma
Dear Editors,
Thank you very much for the efforts and suggestions from the editors and the reviewers. We have revised our manuscript entitled “A 14-immune-related-gene signature predicts clinical outcome of kidney renal clear cell carcinoma” according to the reviewers’ recommendation and would like to resubmit it. Based on the comments, we have carefully made major revision on the revised version. Detailed revisions are shown as follows. We have uploaded the manuscript with computer-generated tracked changes to the revision response files section. Besides, we have uploaded the a “clean” copy where the changes are not marked.
1. You have added commentary about the C-index for this model compared to other published articles. Although your C-index is slightly higher than these other studies, you cannot claim that yours is better without formally evaluating that. There are ways to do that, but I assume it is beyond what you want to accomplish here, especially because some of these articles are for other cancer types and you would need to re-analyze those datasets as well. Instead, you should soften the language in this paragraph to indicate that your C-index falls within a similar range as other related studies. Furthermore, as the reviewer suggested, a C-index of 0.65 is likely not good enough to be used in clinical practice. Thus you should remove or modify your claim that 'we prove our immune signature to be a promising tool in predicting KIRC patients’ survival outcome.' That is a subjective statement. At a minimum, you would need to evaluate this claim on multiple other datasets.
Reply: Thank you very much for your comment. According to your suggestion, we have carefully revised the commentary and statement about the C-index for this model in the Discussion part.
2. I am sorry, but I don't fully understand your response regarding Figure 3. You mention differences between the log-rank test and the univariate Cox analysis, but I am not sure how that addresses the reviewer's concern about lack of significance for GDF1. In addition, you did not address the comment about performing multiple-testing correction.
Reply: Thank you very much for your comment. GDF1 is significant in univariate cox regression, but not significant in log-rank test and multivariate cox analysis (Figure 3, Table1 and Table2). With reference to the method of Li et al (Li J, et al. Identification of a five-lncRNA signature for predicting the risk of tumor recurrence in patients with breast cancer. Int J Cancer. 2018;143(9):2150-2160), we screened prognosis-related genes based on univariate cox analysis. Is there any specific method for multiple-testing correction? We think it is multivariate cox analysis, as shown in Table 2 we have done multivariate cox analysis.
3. Thank you for justifying the use of the 0.25 FDR threshold.
Reply: Thank you very much for your comment.
4. The Tracked Changes version of the manuscript does not show any differences in the References section. Please clarify your statement that you have updated some of the references.
Reply: Thank you very much for your comment. According to your suggestion, the updated references in the References section were highlighted by using red colored text.
5. Thank you for increasing the font sizes on some of the figures. However, in Figures 3 and 7, there is still text that is far too small. I can see it with a magnifying class only. It's up to you on how to solve this, but one option would be to split these into multiple figures.
Reply: Thank you very much for your suggestion. According to your suggestion, we have continued to increase the figure3 and figure7 fonts. Now the fonts of these two figures are displayed clearly on our computer.
6. Thank you for creating Table 4.
Reply: Thank you very much for your comment.
7. You do not need to put the data files in GitHub. However, your R script uses different file names than those listed on your site. read.table() can pull the data files directly from Xena Browser. I didn't attempt to run your script yet (because the file names were different), but I will do that when you resubmit to make sure the script will work and that I can reproduce your figures.
Reply: Thank you very much for your suggestion. According to your suggestion, we have modified our code and pushed it to github.
We hope that the revision and our responses can address your concerns and that you can consider this article suitable for acceptance. Please do not hesitate to contact us if any additional explanation or revision is required. Thank you again for your assistance.
Best regards
Chuan Hu
" | Here is a paper. Please give your review comments after reading it. |
9,836 | "<ns0:abstract xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:p>Kidney renal clear cell carcinoma (KIRC) is the leading cause of kidney cancer-related deaths. Currently, there are no studies in tumor immunology investigating the use of signatures as a predictor of overall survival in KIRC patients. Our study attempts to establish an immune-related gene risk signature to predict clinical outcomes in KIRC. A total of 528 patients from The Cancer Genome Atlas (TCGA) database were included in our analysis and randomly divided into training (n = 315) and testing sets (n = 213). We collected 1,534 immune-related genes from the Immunology Database and Analysis Portal as candidates to construct our signature. LASSO-COX was used to find gene models with the highest predictive ability. We used survival and Cox analysis to test the model's independent prognostic ability. Univariate analysis identified 650 immune-related genes with prognostic abilities. After 1,000 iterations, we choose 14 of the most frequent and stable immune-related genes as our signature. We found that the signature was associated with M stage, T stage, and pathological staging. More importantly, the signature can independently predict clinical prognosis in KIRC patients. Gene Set Enrichment Analysis (GSEA) showed an association between our signature and critical metabolism pathways.</ns0:p><ns0:p>Our research established a model based upon 14 immune-related genes that predicted the prognosis of KIRC patients based on tumor immune microenvironments.</ns0:p></ns0:div>
</ns0:abstract>
<ns0:body xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:head>Introduction</ns0:head><ns0:p>Kidney cancer is one of the most common urological tumors worldwide, with approximately 403,262 new cases and 175,098 deaths associated with this form of cancer in 2018 <ns0:ref type='bibr' target='#b2'>(Bray et al. 2018)</ns0:ref>. Kidney renal clear cell carcinoma (KIRC) is the leading cause of kidney cancer mortality and the most common type, accounting for 85% of kidney cancers <ns0:ref type='bibr' target='#b24'>(Siegel et al. 2017;</ns0:ref><ns0:ref type='bibr' target='#b27'>Tseng 2016</ns0:ref>). Considering the aging population of KIRC patients and the rising expense of treatment, KIRC is gradually becoming the focus of geriatric cancer <ns0:ref type='bibr' target='#b26'>(Tan et al. 2015)</ns0:ref>. Despite the rapid development of cancer treatments, mortality rates in KIRC remain stagnant. With the progression of next generation sequencing and data mining techniques, it is urgent that we explore prognostic biomarkers for KIRC, using molecular characteristics and tumor immune environments to guide patient therapy.</ns0:p><ns0:p>Over the past decade our understanding of immune components, including the impact of tumor microenvironments on patient survival and therapy response <ns0:ref type='bibr' target='#b3'>(Chen & Mellman 2017;</ns0:ref><ns0:ref type='bibr' target='#b9'>Grivennikov et al. 2010</ns0:ref>), has increased. Some studies have found that tumor-infiltrating immune cells are able to serve as either tumor suppressors or promoters in microenvironments. For example, CD8 + T cells have been associated with improved survival in cancer patients <ns0:ref type='bibr' target='#b5'>(Gajewski et al. 2013;</ns0:ref><ns0:ref type='bibr' target='#b8'>Governa et al. 2017)</ns0:ref>, while tumor associated macrophages and regulatory T cells demonstrate the ability to promote tumor development <ns0:ref type='bibr' target='#b13'>(Nishikawa & Sakaguchi 2014;</ns0:ref><ns0:ref type='bibr' target='#b14'>Noy & Pollard 2014)</ns0:ref>. Considering the complexity and significance of tumor immune microenvironments, it is imperative that we investigate immune-related biomarkers for KIRC patients. Recent studies have provided insight into the KIRC immune signature <ns0:ref type='bibr' target='#b6'>(Geissler et al. 2015;</ns0:ref><ns0:ref type='bibr' target='#b20'>Şenbabaoğlu et al. 2016)</ns0:ref>. <ns0:ref type='bibr'>Şenbabaoğlu et al. (2016)</ns0:ref> found mRNA signatures with the potential to be immunotherapeutic biomarkers in KIRC. However, their investigations do not include immune-related genes for analysis nor do they establish a systematic immune-related gene-risk signature for KIRC patients. Khadirnaikar et al.</ns0:p><ns0:p>(2019) utilized immune associated lncRNA (long non-coding RNA) to construct prognostic subtypes in KIRC patients. Our immune clusters were more robust and independent. Additionally, they concentrated on lncRNA, not the overall immune-related genes. <ns0:ref type='bibr'>Smith et al. (2018)</ns0:ref> constructed endogenous retroviral signatures for KIRC patients, but they did not investigate the prognostic ability of the signature in different subtypes of patients. Therefore, it is essential that we explore a systematic prognostic signature based on tumor immune environments in KIRC.</ns0:p><ns0:p>In our study, we used RNA-seq data from The Cancer Genome Atlas (TCGA) to find immune-related genes with prognostic ability and to establish an immune-related risk signature for KIRC. To assess the clinical potential of the signature, we investigated the association between the signature, clinical parameters, and patient survival. Gene set enrichment analysis (GSEA) was performed to explore the molecular characteristics of the signature.</ns0:p></ns0:div>
<ns0:div><ns0:head>Materials & Methods</ns0:head></ns0:div>
<ns0:div><ns0:head>Patient cohort</ns0:head><ns0:p>The TCGA database from Xena browser was used to collect the clinical and RNA-seq data of 528 KIRC patients <ns0:ref type='bibr' target='#b7'>(Goldman et al. 2019)</ns0:ref>. We randomly divided the dataset into training (n = 315) and test sets (n = 213). RNAseq data was obtained to analyze the transcriptome profiling of RNA expression and were measured using fragments per kilobase of exon per million fragments mapped (FPKM). We performed a log2-based transformation to normalize We used Harrell's c-index to estimate the predictive ability of the immune-related risk signature in the training, testing, and total cohort <ns0:ref type='bibr' target='#b11'>(Harrell et al. 1982)</ns0:ref>. The independent prognostic ability of the immune-related risk signature was assessed using survival analysis as well as Cox analysis (R package survival, v2.42). We performed multivariable Cox regression to analyze the relationship between immune-related risk signatures and clinicopathological factors.</ns0:p></ns0:div>
<ns0:div><ns0:head>Gene set enrichment analysis</ns0:head><ns0:p>To identify the biological functions and pathways between high-and low-risk groups, we conducted GSEA to investigate potential biological mechanisms in the Molecular Signatures Database (MSigDB; <ns0:ref type='bibr' target='#b25'>(Subramanian et al. 2005)</ns0:ref>. GSEA was performed using GSEApy, a python wrapper for gene enrichment (https://pypi.org/project/gseapy/).</ns0:p><ns0:p>We selected C2 and C5, including pathway databases and GO terms, from the MSigDB. The gesa sub-command of GSEApy was used in GSEA with default parameters. Enriched gene sets with a false discovery rate (FDR) of less than 0.25 and a P-value of less than 0.05 were considered statistically significant.</ns0:p></ns0:div>
<ns0:div><ns0:head>Statistical analysis</ns0:head><ns0:p>Boxplots were created with the R package, ggplot2 (v3.0.0). The R package ComplexHeatmap (v1.18.1) was used to create heatmaps <ns0:ref type='bibr' target='#b10'>(Gu et al. 2016)</ns0:ref>. We counted C-index with R packages 'survcomp' <ns0:ref type='bibr' target='#b11'>(Harrell et al. 1982;</ns0:ref><ns0:ref type='bibr' target='#b18'>Schröder et al. 2011)</ns0:ref>. The student's t-test was performed for statistical comparison. We chose R to conduct statistical analysis (https://www.r-project.org/). P-values lower than 0.05 were considered statistically significant. The main code of analysis was pushed to github (https://github.com/huchua/KIRC).</ns0:p></ns0:div>
<ns0:div><ns0:head>Results</ns0:head></ns0:div>
<ns0:div><ns0:head>Establishment and validation of the immune-related gene signature in KIRC</ns0:head><ns0:p>Figure <ns0:ref type='figure' target='#fig_4'>1</ns0:ref> illustrates the workflow we used to develop an immune-related gene-risk signature. The immune-related gene signature was constructed within the KIRC training data set, while we applied the testing set to validate the signature. Manuscript to be reviewed After 1,000 iterations, seven unique gene models were selected (Figure <ns0:ref type='figure'>2A</ns0:ref>, Table <ns0:ref type='table'>S1</ns0:ref>). The model selected as the immune-related gene-risk signature consisted of 14 genes and ranked the highest in frequency 282 times. Parameter selection in the LASSO-cox model was log lambda = -2.641 and alpha = 1. The univariate and multivariate Cox analysis of the 14 immune-related genes are shown in Table <ns0:ref type='table'>1 and Table 2</ns0:ref>. The principle component analysis of the 14 immune-related gene signature displayed a different distribution pattern between low-and high-risk groups when comparing the training, testing, and total cohort (Figure <ns0:ref type='figure'>2B-D</ns0:ref>). This indicated that the low-and high-risk groups had different immune phenotypes. In the training, testing, and total data set, the c-index was 0.7862, 0.6534, and 0.7367, respectively (P < 0.001; Figure <ns0:ref type='figure'>2E</ns0:ref>). In a time-dependent receiver operating characteristic curve (ROC) created for three datasets, area under the curve (AUC) values at 1, 3, and 5 years were 0.679, 0.63, 0.627; 0.65, 0.596, 0.568; and 0.618, 0.589, 0.59, respectively (Figure <ns0:ref type='figure'>S2</ns0:ref>). The 14 immune-related genes are AR, BID, BMP8A, CCL7, CCR10, FGF17, GDF1, IL20RB, IL4, KLRC2, LHB, SEMA3A, SEMA3G, and TXLNA. K-M (Kaplan-Meier) survival curves and gene expression of the 14 immune-related genes are shown in Figure <ns0:ref type='figure' target='#fig_10'>3</ns0:ref> and Figure <ns0:ref type='figure' target='#fig_8'>S3</ns0:ref>. Because of the differences between the log-rank test and the univariate cox analysis, the results of the univariate cox analysis of other genes except GDF1 are very consistent with the results of the K-M (Kaplan-Meier) survival curves.</ns0:p></ns0:div>
<ns0:div><ns0:head>Correlation between the signature of immune-related genes and clinical parameters</ns0:head><ns0:p>Among the 14 immune-related genes, four genes (TXLNA, SEMA3G, AR, and BID) had a high expression, and 10 genes (IL20RB, CCR10, BMP8A, SEMA3A, CCL7, GDF1, KLRC2, LHB, FGF17, and IL4) had a low expression (Figure <ns0:ref type='figure'>4A</ns0:ref>, B; Figure <ns0:ref type='figure' target='#fig_5'>5</ns0:ref>). The relationship between the signature and clinical factors demonstrated that patients with advanced pathological staging, M stage, and T stage had a higher risk score than those with early stage disease (Figure <ns0:ref type='figure'>4A</ns0:ref>, B; Figure <ns0:ref type='figure' target='#fig_5'>5</ns0:ref>; Figure <ns0:ref type='figure' target='#fig_11'>6A, B, D</ns0:ref>). However, we did not find a correlation between the signature and N stage (Figure <ns0:ref type='figure'>4A</ns0:ref>, B; Figure <ns0:ref type='figure' target='#fig_5'>5</ns0:ref>; Figure <ns0:ref type='figure' target='#fig_11'>6C</ns0:ref>). Manuscript to be reviewed</ns0:p></ns0:div>
<ns0:div><ns0:head>Influence of the immune-related gene signature on patient prognosis</ns0:head><ns0:p>We then assessed whether this signature influenced KIRC patient prognosis. Survival analysis showed that patients with a high-risk score were associated with poor survival outcomes in the training, testing, and total group sets (Figure <ns0:ref type='figure'>4C</ns0:ref>; Figure <ns0:ref type='figure'>4E</ns0:ref>; Figure <ns0:ref type='figure' target='#fig_12'>7B</ns0:ref>). We found that the signature also predicted survival outcomes in subgroups of KIRC patients, including stage I-II (Figure <ns0:ref type='figure' target='#fig_12'>7C</ns0:ref>), stage III-IV (Figure <ns0:ref type='figure' target='#fig_12'>7D</ns0:ref>), M0 stage (Figure <ns0:ref type='figure' target='#fig_12'>7E</ns0:ref>), M1 stage (Figure <ns0:ref type='figure' target='#fig_12'>7F</ns0:ref>), N0 stage (Figure <ns0:ref type='figure' target='#fig_12'>7G</ns0:ref>), N1 (Figure <ns0:ref type='figure' target='#fig_12'>7H</ns0:ref>), T1 (Figure <ns0:ref type='figure' target='#fig_12'>7I</ns0:ref>), T2 (Figure <ns0:ref type='figure' target='#fig_12'>7J</ns0:ref>), T3 (Figure <ns0:ref type='figure' target='#fig_12'>7K</ns0:ref>), and T4 (Figure <ns0:ref type='figure' target='#fig_12'>7L</ns0:ref>). Multivariate Cox analysis revealed that the risk signature was able to independently predict overall survival in KIRC patients (Figure <ns0:ref type='figure'>4D</ns0:ref>, F; Figure <ns0:ref type='figure' target='#fig_12'>7A</ns0:ref>; Table <ns0:ref type='table'>3</ns0:ref>).</ns0:p></ns0:div>
<ns0:div><ns0:head>Engagement of the immune-related gene risk signature in biological pathways and functions</ns0:head><ns0:p>GSEA was used to investigate the signature's biological pathways and functions. There were 177 KEGG pathways and 4,528 GO terms used in our investigation. Our analysis found that the signature was able to engage in a total of 19 enriched KEGG pathways (FDR < 0.25; Table <ns0:ref type='table'>4</ns0:ref>). The low-risk signature was significantly correlated with 10 pathways, including the citrate cycle (TCA cycle) pathway, fatty acid metabolism pathway, propanoate metabolism pathway, butanoate metabolism pathway, peroxisome pathway, lysine degradation pathway, valine leucine and isoleucine degradation pathway, proximal tubule bicarbonate reclamation pathway, vasopressin regulated water reabsorption pathway, and the pyruvate metabolism pathway (Table <ns0:ref type='table'>4</ns0:ref>). Similarly, eight GO annotations were enriched in the low-risk group (FDR < 0.25; Table <ns0:ref type='table'>4</ns0:ref>).</ns0:p></ns0:div>
<ns0:div><ns0:head>Discussion</ns0:head><ns0:p>Our study used the TCGA database and immune-related genes to establish a KIRC signature consisting of 14 immunerelated genes. We found that patients in the high-risk group showed a positive association with M stage, T stage, and advanced pathological staging. Additionally, the signature exhibited strong prognostic abilities and independently Manuscript to be reviewed predicted KIRC patient prognosis. Functional analysis highlighted the significance of our signature based on its involvement in many important pathways.</ns0:p><ns0:p>In our investigation, we used a total of 14 immune-related genes to construct our signature. In the signature, CCL7 increased the peripheral blood mononuclear cell recruitment in renal cell cancer through the inhibition of let-7d <ns0:ref type='bibr' target='#b17'>(Riihimaki et al. 2014)</ns0:ref>. <ns0:ref type='bibr'>Wyler et al. (2014)</ns0:ref> demonstrated the ability of CCL7 to recruit monocytes through CCR2, promoting renal cell cancer metastasis to the brain. This indicates that CCL7 is a major factor in the development of KIRC in tumor immune microenvironments and may be a potential immunotherapy target. Fibroblast growth factor receptor 17 (FGF17) is another immune-related gene in our signature. FGF17 demonstrates a variety of functions in cancer development. <ns0:ref type='bibr'>Gauglhofer et al. (2011)</ns0:ref> found that FGF17 was involved in the paracrine and autocrine signaling of hepatocellular carcinoma and promoted the neoangiogenesis of hepatocellular carcinoma. <ns0:ref type='bibr'>Heer et al. (2004)</ns0:ref> showed that FGF17 was overexpressed in prostate cancer and participated in prostate carcinogenesis. Our study showed that FGF17 plays an important role in KIRC based on tumor immunology. However, the underlying mechanism of FGF17 in KIRC immune microenvironments requires further investigation. In our signature, IL-4 is another well-studied immune-related biomarker for KIRC. IL-4, released by immune cells, controls the expression of B7-H1, thus altering T cell responses in KIRC <ns0:ref type='bibr' target='#b16'>(Quandt et al. 2014)</ns0:ref>. The investigation conducted by <ns0:ref type='bibr'>Chang et al. (2015)</ns0:ref> demonstrated that IL-4 expression can predict KIRC patient recurrence and survival outcomes. Collectively, these investigations reinforce the significance of the 14 immune-related gene-risk signature in KIRC.</ns0:p><ns0:p>Our signature is associated with the survival outcome of KIRC patients and clinical parameters, including pathological staging, M stage, and T stage. No correlation was found between the signature and N-stage, possibly due to a lack of N-stage information for many patients. According to the immune-related gene-risk signature, our study found that clinical cohorts in KIRC have different immune-related risk factors. This signature also reflects differences in tumor Manuscript to be reviewed immune microenvironments and predicts survival outcomes in KIRC patients; thus, demonstrating the clinical significance of our signature and its possible use as a survival predictor in KIRC.</ns0:p><ns0:p>Despite of the relatively low C-index of our testing set (0.6534), the C-index of the testing set falls within a similar range as those of other related studies. For instance, in the study of Bailiang Li et al., they developed an immune signature in non-small cell lung cancer with the C-index of 0.64 <ns0:ref type='bibr' target='#b12'>(Li et al. 2017)</ns0:ref>. Besides, our signature achieved a similar C-index of testing set with the immune signature in ovarian cancer (0.625) <ns0:ref type='bibr' target='#b22'>(Shen et al. 2019</ns0:ref>). More importantly, the C-index of our testing set showed a similar accuracy with the C-index of clinical staging systems in renal cancer (0.62) <ns0:ref type='bibr' target='#b15'>(Qu et al. 2018</ns0:ref>).</ns0:p><ns0:p>Our study expands on the signatures association with several important pathways, especially the metabolism pathway.</ns0:p><ns0:p>This may reflect the mutual interaction between tumor metabolism and tumor immunology in KIRC. Pearce et al. <ns0:ref type='bibr'>(2013)</ns0:ref> showed that metabolic reprogramming can influence the fate and function of T cells in tumors. Our study further indicates the importance of metabolism pathways in KIRC immune microenvironments.</ns0:p></ns0:div>
<ns0:div><ns0:head>Conclusions</ns0:head><ns0:p>This investigation utilized RNA-seq data from the TCGA database to construct a 14 immune-related gene-risk signature with the ability to independently predict survival outcomes in KIRC; thus, providing novel clinical applications and possible immune targets for KIRC. </ns0:p></ns0:div>
<ns0:div><ns0:head>Figure legends</ns0:head></ns0:div><ns0:figure xml:id='fig_0'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2019:11:43459:4:0:NEW 10 Apr 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_1'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2019:11:43459:4:0:NEW 10 Apr 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_2'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2019:11:43459:4:0:NEW 10 Apr 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_3'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2019:11:43459:4:0:NEW 10 Apr 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_4'><ns0:head>Figure 1 :</ns0:head><ns0:label>1</ns0:label><ns0:figDesc>Figure 1: Our workflow constructing the model for risk-score signatures</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_5'><ns0:head>Figure 5 :</ns0:head><ns0:label>5</ns0:label><ns0:figDesc>Figure 5: Analysis of the 14 immune-related genes predictive ability in total cohort. As demonstrated by the heatmap,</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_6'><ns0:head>Figure 6 :</ns0:head><ns0:label>6</ns0:label><ns0:figDesc>Figure 6: The K-M survival curve of (A) pathological staging, (B) M stage, (C) N stage, and (D) T stage (left). The</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_7'><ns0:head>Figure S1 :</ns0:head><ns0:label>S1</ns0:label><ns0:figDesc>Figure S1: Parameter selection in the LASSO-cox model. (A) LASSO coefficient values of the 14 immune related</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_8'><ns0:head>Figure S3 :</ns0:head><ns0:label>S3</ns0:label><ns0:figDesc>FigureS3: Expression of 14 immune-related genes in kidney renal clear cell carcinoma and normal tissues.</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_9'><ns0:head>Figure 1</ns0:head><ns0:label>1</ns0:label><ns0:figDesc>Figure 1</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_10'><ns0:head>Figure 3 K</ns0:head><ns0:label>3</ns0:label><ns0:figDesc>Figure 3</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_11'><ns0:head>Figure 6</ns0:head><ns0:label>6</ns0:label><ns0:figDesc>Figure 6</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_12'><ns0:head>Figure 7</ns0:head><ns0:label>7</ns0:label><ns0:figDesc>Figure 7</ns0:figDesc></ns0:figure>
<ns0:figure type='table' xml:id='tab_0'><ns0:head>Table1 :</ns0:head><ns0:label>Table1</ns0:label><ns0:figDesc>Univariate Cox analysis of 14 immune relate genes in all cohort</ns0:figDesc><ns0:table><ns0:row><ns0:cell>AR</ns0:cell><ns0:cell cols='2'>Variable 0.638 0.556-0.731</ns0:cell><ns0:cell>HR <0.001</ns0:cell><ns0:cell>95%CI IL20RB</ns0:cell><ns0:cell>pvalue 1.268 1.102-1.46</ns0:cell><ns0:cell>0.001</ns0:cell></ns0:row><ns0:row><ns0:cell>age</ns0:cell><ns0:cell cols='2'>KLRC2 1.352 1.143-1.599</ns0:cell><ns0:cell>20.657 <0.001</ns0:cell><ns0:cell>9.116-46.81 age</ns0:cell><ns0:cell><0.001 1.414 1.195-1.672</ns0:cell><ns0:cell><0.001</ns0:cell></ns0:row><ns0:row><ns0:cell>stage</ns0:cell><ns0:cell cols='2'>CCL7 3.725 2.708-5.124</ns0:cell><ns0:cell>2.757 <0.001</ns0:cell><ns0:cell>2.090-3.637 stage</ns0:cell><ns0:cell><0.001 3.275 2.351-4.562</ns0:cell><ns0:cell><0.001</ns0:cell></ns0:row><ns0:row><ns0:cell>gender</ns0:cell><ns0:cell cols='2'>0.943 0.687-1.294</ns0:cell><ns0:cell>0.718</ns0:cell><ns0:cell>gender</ns0:cell><ns0:cell>1.025 0.744-1.412</ns0:cell><ns0:cell>0.879</ns0:cell></ns0:row><ns0:row><ns0:cell>BID</ns0:cell><ns0:cell cols='2'>SEMA3A 1.411 1.211-1.645</ns0:cell><ns0:cell>1.748 <0.001</ns0:cell><ns0:cell>1.437-2.126 IL4</ns0:cell><ns0:cell><0.001 1.458 1.294-1.643</ns0:cell><ns0:cell><0.001</ns0:cell></ns0:row><ns0:row><ns0:cell>age</ns0:cell><ns0:cell cols='2'>SEMA3G 1.407 1.188-1.666</ns0:cell><ns0:cell>0.617 <0.001</ns0:cell><ns0:cell>0.534-0.714 age</ns0:cell><ns0:cell><0.001 1.39 1.171-1.651</ns0:cell><ns0:cell><0.001</ns0:cell></ns0:row><ns0:row><ns0:cell>stage</ns0:cell><ns0:cell cols='2'>CCR10 3.173 2.284-4.41</ns0:cell><ns0:cell>1.759 <0.001</ns0:cell><ns0:cell>1.381-2.241 stage</ns0:cell><ns0:cell><0.001 3.426 2.484-4.725</ns0:cell><ns0:cell><0.001</ns0:cell></ns0:row><ns0:row><ns0:cell>gender</ns0:cell><ns0:cell cols='2'>BMP8A 1.039 0.755-1.43</ns0:cell><ns0:cell>1.813 0.814</ns0:cell><ns0:cell>1.298-2.532 gender</ns0:cell><ns0:cell><0.001 0.999 0.726-1.375</ns0:cell><ns0:cell>0.997</ns0:cell></ns0:row><ns0:row><ns0:cell>BMP8A</ns0:cell><ns0:cell cols='2'>FGF17 1.213 1.057-1.391</ns0:cell><ns0:cell>7.335 0.006</ns0:cell><ns0:cell cols='2'>3.202-16.803 KLRC2 1.402 1.233-1.594 <0.001</ns0:cell><ns0:cell><0.001</ns0:cell></ns0:row><ns0:row><ns0:cell>age</ns0:cell><ns0:cell cols='2'>1.454 1.229-1.721</ns0:cell><ns0:cell><0.001</ns0:cell><ns0:cell>age</ns0:cell><ns0:cell>1.383 1.169-1.637</ns0:cell><ns0:cell><0.001</ns0:cell></ns0:row><ns0:row><ns0:cell>stage</ns0:cell><ns0:cell cols='2'>GDF1 3.607 2.619-4.969</ns0:cell><ns0:cell>1.483 <0.001</ns0:cell><ns0:cell>1.062-2.070 stage</ns0:cell><ns0:cell>0.021 3.429 2.483-4.737</ns0:cell><ns0:cell><0.001</ns0:cell></ns0:row><ns0:row><ns0:cell>gender</ns0:cell><ns0:cell cols='2'>IL4 1.052 0.765-1.447</ns0:cell><ns0:cell>109.106 0.754</ns0:cell><ns0:cell cols='2'>34.328-346.781 gender 0.958 0.694-1.322 <0.001</ns0:cell><ns0:cell>0.792</ns0:cell></ns0:row><ns0:row><ns0:cell>CCL7</ns0:cell><ns0:cell cols='2'>LHB 1.329 1.202-1.471</ns0:cell><ns0:cell>5.053 <0.001</ns0:cell><ns0:cell>3.240-7.881 LHB</ns0:cell><ns0:cell><0.001 1.285 1.168-1.414</ns0:cell><ns0:cell><0.001</ns0:cell></ns0:row><ns0:row><ns0:cell>age</ns0:cell><ns0:cell cols='2'>TXLNA 1.455 1.226-1.728</ns0:cell><ns0:cell>2.329 <0.001</ns0:cell><ns0:cell>1.500-3.618 age</ns0:cell><ns0:cell><0.001 1.393 1.178-1.647</ns0:cell><ns0:cell><0.001</ns0:cell></ns0:row><ns0:row><ns0:cell>stage</ns0:cell><ns0:cell cols='2'>AR 3.485 2.529-4.802</ns0:cell><ns0:cell>0.548 <0.001</ns0:cell><ns0:cell>0.463-0.649 stage</ns0:cell><ns0:cell><0.001 3.519 2.55-4.854</ns0:cell><ns0:cell><0.001</ns0:cell></ns0:row><ns0:row><ns0:cell>gender</ns0:cell><ns0:cell>1</ns0:cell><ns0:cell>IL20RB 0.726-1.377</ns0:cell><ns0:cell>1.225 1</ns0:cell><ns0:cell>1.145-1.310 gender</ns0:cell><ns0:cell><0.001 1.17 0.844-1.622</ns0:cell><ns0:cell>0.347</ns0:cell></ns0:row><ns0:row><ns0:cell>CCR10</ns0:cell><ns0:cell cols='2'>1.235 1.111-1.372</ns0:cell><ns0:cell><0.001</ns0:cell><ns0:cell>SEMA3A</ns0:cell><ns0:cell>1.214 1.107-1.33</ns0:cell><ns0:cell><0.001</ns0:cell></ns0:row><ns0:row><ns0:cell>age</ns0:cell><ns0:cell cols='2'>BID 1.382 1.166-1.639</ns0:cell><ns0:cell>3.505 <0.001</ns0:cell><ns0:cell>2.439-5.035 age</ns0:cell><ns0:cell><0.001 1.445 1.218-1.715</ns0:cell><ns0:cell><0.001</ns0:cell></ns0:row><ns0:row><ns0:cell>stage</ns0:cell><ns0:cell cols='2'>3.844 2.793-5.292</ns0:cell><ns0:cell><0.001</ns0:cell><ns0:cell>stage</ns0:cell><ns0:cell>3.549 2.572-4.898</ns0:cell><ns0:cell><0.001</ns0:cell></ns0:row><ns0:row><ns0:cell>gender</ns0:cell><ns0:cell cols='2'>1.034 0.751-1.424</ns0:cell><ns0:cell>0.836</ns0:cell><ns0:cell>gender</ns0:cell><ns0:cell>1.108 0.802-1.529</ns0:cell><ns0:cell>0.535</ns0:cell></ns0:row><ns0:row><ns0:cell>FGF17</ns0:cell><ns0:cell cols='2'>1.235 1.112-1.372</ns0:cell><ns0:cell><0.001</ns0:cell><ns0:cell cols='2'>SEMA3G 0.652 0.549-0.775</ns0:cell><ns0:cell><0.001</ns0:cell></ns0:row><ns0:row><ns0:cell>age</ns0:cell><ns0:cell cols='2'>1.399 1.181-1.657</ns0:cell><ns0:cell><0.001</ns0:cell><ns0:cell>age</ns0:cell><ns0:cell>1.379 1.164-1.633</ns0:cell><ns0:cell><0.001</ns0:cell></ns0:row><ns0:row><ns0:cell>stage</ns0:cell><ns0:cell cols='2'>3.722 2.705-5.121</ns0:cell><ns0:cell><0.001</ns0:cell><ns0:cell>stage</ns0:cell><ns0:cell>3.29 2.384-4.539</ns0:cell><ns0:cell><0.001</ns0:cell></ns0:row><ns0:row><ns0:cell>gender</ns0:cell><ns0:cell cols='2'>1.109 0.806-1.525</ns0:cell><ns0:cell>0.527</ns0:cell><ns0:cell>gender</ns0:cell><ns0:cell>0.973 0.707-1.339</ns0:cell><ns0:cell>0.867</ns0:cell></ns0:row><ns0:row><ns0:cell>GDF1</ns0:cell><ns0:cell cols='2'>1.086 0.975-1.211</ns0:cell><ns0:cell>0.135</ns0:cell><ns0:cell>TXLNA</ns0:cell><ns0:cell>1.285 1.098-1.505</ns0:cell><ns0:cell>0.002</ns0:cell></ns0:row><ns0:row><ns0:cell>age</ns0:cell><ns0:cell cols='2'>1.411 1.191-1.672</ns0:cell><ns0:cell><0.001</ns0:cell><ns0:cell>age</ns0:cell><ns0:cell>1.433 1.208-1.7</ns0:cell><ns0:cell><0.001</ns0:cell></ns0:row><ns0:row><ns0:cell>stage</ns0:cell><ns0:cell>3.76</ns0:cell><ns0:cell>2.735-5.17</ns0:cell><ns0:cell><0.001</ns0:cell><ns0:cell>stage</ns0:cell><ns0:cell>3.696 2.687-5.085</ns0:cell><ns0:cell><0.001</ns0:cell></ns0:row><ns0:row><ns0:cell>gender</ns0:cell><ns0:cell cols='2'>1.103 0.796-1.529</ns0:cell><ns0:cell>0.555</ns0:cell><ns0:cell>gender</ns0:cell><ns0:cell>1.212 0.87-1.688</ns0:cell><ns0:cell>0.255</ns0:cell></ns0:row></ns0:table></ns0:figure>
<ns0:note place='foot'>Manuscript to be reviewed</ns0:note>
<ns0:note place='foot'>PeerJ reviewing PDF | (2019:11:43459:4:0:NEW 10 Apr 2020)Manuscript to be reviewed</ns0:note>
<ns0:note place='foot'>PeerJ reviewing PDF | (2019:11:43459:4:0:NEW 10 Apr 2020)Manuscript to be reviewed1 Table 2. Multivariate Cox analysis of 14 immune relate genes in all cohort</ns0:note>
</ns0:body>
" | "Resubmission: A 14-immune-related-gene signature predicts clinical outcome of kidney renal clear cell carcinoma
Dear Editors,
Thank you very much for the efforts and suggestions from the editors and the reviewers. We have revised our manuscript entitled “A 14-immune-related-gene signature predicts clinical outcome of kidney renal clear cell carcinoma” according to the reviewers’ recommendation and would like to resubmit it. Based on the comments, we have carefully made major revision on the revised version. Detailed revisions are shown as follows. We have uploaded the manuscript with computer-generated tracked changes to the revision response files section. Besides, we have uploaded the a “clean” copy where the changes are not marked.
1. The Methods section states, 'The TCGA database was used to collect the clinical and RNA-seq data of 528 KIRC patients.' Please be more specific. You should also cite the Xena browser (there is a paper for it).
Reply: Thank you very much for your comment. According to your suggestion, we have cited the Xena browser.
2. Cox regression accounts for covariates, but it doesn't account for multiple tests. The following articles provides some insights on methods of correcting for multiple tests: https://www.nature.com/articles/nbt1209-1135.
Reply: Thank you very much for your comment. We are very grateful for the references you provided. From the literature, we know the multiple test, but we still do not know how to apply the multiple test to the univariate cox analysis. We read a lot of articles about prognosis signatures, and they didn’t use the multiple test method (JAMA Oncol. 2017 Nov; 3(11): 1529–1537; Mol Cancer. 2017; 16: 98; EBioMedicine. 2018 May;31:182-189). We would like to consult if there is a better way to solve this problem.
3. The R script uses the following packages, but it appears that some/all of them have not been cited in the manuscript. Please either cite them or remove them from the R script if you did not use them.
circlize, survminer, survivalROC, copynumber, clusterProfiler, maftools, scatterplot3d, survRM2
Reply: Thank you very much for your comment. According to your suggestion, we have deleted unused R packages.
4. I attempted to run the R script but received an error message toward the beginning (see below). I believe the error is because read.table doesn't work with gzipped files. The readr package does have a function to do this. Or there are functions in R to download and gunzip a file. Please make sure the R script runs from beginning to end (and that it repeats your analysis successfully) before resubmitting.
Reply: Thank you very much for your comment. We have used the readr package to read data according to your requirements, but may not have successfully read in the data due to network reasons. We can use the local data to run the entire process smoothly.
We hope that the revision and our responses can address your concerns and that you can consider this article suitable for acceptance. Please do not hesitate to contact us if any additional explanation or revision is required. Thank you again for your assistance.
Best regards
Chuan Hu
" | Here is a paper. Please give your review comments after reading it. |
9,838 | "<ns0:abstract xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:p>Kidney renal clear cell carcinoma (KIRC) is the leading cause of kidney cancer-related deaths. Currently, there are no studies in tumor immunology investigating the use of signatures as a predictor of overall survival in KIRC patients. Our study attempts to establish an immune-related gene risk signature to predict clinical outcomes in KIRC. A total of 528 patients from The Cancer Genome Atlas (TCGA) database were included in our analysis and randomly divided into training (n = 315) and testing sets (n = 213). We collected 1,534 immune-related genes from the Immunology Database and Analysis Portal as candidates to construct our signature. LASSO-COX was used to find gene models with the highest predictive ability. We used survival and Cox analysis to test the model's independent prognostic ability. Univariate analysis identified 650 immune-related genes with prognostic abilities. After 1,000 iterations, we choose 14 of the most frequent and stable immune-related genes as our signature. We found that the signature was associated with M stage, T stage, and pathological staging. More importantly, the signature can independently predict clinical prognosis in KIRC patients. Gene Set Enrichment Analysis (GSEA) showed an association between our signature and critical metabolism pathways.</ns0:p><ns0:p>Our research established a model based upon 14 immune-related genes that predicted the prognosis of KIRC patients based on tumor immune microenvironments.</ns0:p></ns0:div>
</ns0:abstract>
<ns0:body xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:head>Introduction</ns0:head><ns0:p>Kidney cancer is one of the most common urological tumors worldwide, with approximately 403,262 new cases and 175,098 deaths associated with this form of cancer in 2018 <ns0:ref type='bibr' target='#b1'>(Bray et al. 2018)</ns0:ref>. Kidney renal clear cell carcinoma (KIRC) is the leading cause of kidney cancer mortality and the most common type, accounting for 85% of kidney cancers <ns0:ref type='bibr' target='#b22'>(Siegel et al. 2017;</ns0:ref><ns0:ref type='bibr' target='#b25'>Tseng 2016</ns0:ref>). Considering the aging population of KIRC patients and the rising expense of treatment, KIRC is gradually becoming the focus of geriatric cancer <ns0:ref type='bibr' target='#b24'>(Tan et al. 2015)</ns0:ref>. Despite the rapid development of cancer treatments, mortality rates in KIRC remain stagnant. With the progression of next generation sequencing and data mining techniques, it is urgent that we explore prognostic biomarkers for KIRC, using molecular characteristics and tumor immune environments to guide patient therapy.</ns0:p><ns0:p>Over the past decade our understanding of immune components, including the impact of tumor microenvironments on patient survival and therapy response <ns0:ref type='bibr' target='#b2'>(Chen & Mellman 2017;</ns0:ref><ns0:ref type='bibr' target='#b8'>Grivennikov et al. 2010</ns0:ref>), has increased. Some studies have found that tumor-infiltrating immune cells are able to serve as either tumor suppressors or promoters in microenvironments. For example, CD8 + T cells have been associated with improved survival in cancer patients <ns0:ref type='bibr' target='#b4'>(Gajewski et al. 2013;</ns0:ref><ns0:ref type='bibr' target='#b7'>Governa et al. 2017)</ns0:ref>, while tumor associated macrophages and regulatory T cells demonstrate the ability to promote tumor development <ns0:ref type='bibr' target='#b12'>(Nishikawa & Sakaguchi 2014;</ns0:ref><ns0:ref type='bibr' target='#b13'>Noy & Pollard 2014)</ns0:ref>. Considering the complexity and significance of tumor immune microenvironments, it is imperative that we investigate immune-related biomarkers for KIRC patients. Recent studies have provided insight into the KIRC immune signature <ns0:ref type='bibr' target='#b5'>(Geissler et al. 2015;</ns0:ref><ns0:ref type='bibr' target='#b18'>Şenbabaoğlu et al. 2016)</ns0:ref>. <ns0:ref type='bibr'>Şenbabaoğlu et al. (2016)</ns0:ref> found mRNA signatures with the potential to be immunotherapeutic biomarkers in KIRC. However, their investigations do not include immune-related genes for analysis nor do they establish a systematic immune-related gene-risk signature for KIRC patients. Khadirnaikar et al.</ns0:p><ns0:p>(2019) utilized immune associated lncRNA (long non-coding RNA) to construct prognostic subtypes in KIRC patients. Our immune clusters were more robust and independent. Additionally, they concentrated on lncRNA, not the overall immune-related genes. <ns0:ref type='bibr'>Smith et al. (2018)</ns0:ref> constructed endogenous retroviral signatures for KIRC patients, but they did not investigate the prognostic ability of the signature in different subtypes of patients. Therefore, it is essential that we explore a systematic prognostic signature based on tumor immune environments in KIRC.</ns0:p><ns0:p>In our study, we used RNA-seq data from The Cancer Genome Atlas (TCGA) to find immune-related genes with prognostic ability and to establish an immune-related risk signature for KIRC. To assess the clinical potential of the signature, we investigated the association between the signature, clinical parameters, and patient survival. Gene set enrichment analysis (GSEA) was performed to explore the molecular characteristics of the signature.</ns0:p></ns0:div>
<ns0:div><ns0:head>Materials & Methods</ns0:head></ns0:div>
<ns0:div><ns0:head>Patient cohort</ns0:head><ns0:p>The TCGA database from Xena browser was used to collect the clinical and RNA-seq data of 528 KIRC patients <ns0:ref type='bibr' target='#b6'>(Goldman et al. 2019)</ns0:ref>. We randomly divided the dataset into training (n = 315) and test sets (n = 213). RNAseq data was obtained to analyze the transcriptome profiling of RNA expression and were measured using fragments per kilobase of exon per million fragments mapped (FPKM). We performed a log2-based transformation to normalize We used Harrell's c-index to estimate the predictive ability of the immune-related risk signature in the training, testing, and total cohort <ns0:ref type='bibr' target='#b10'>(Harrell et al. 1982)</ns0:ref>. The independent prognostic ability of the immune-related risk signature was assessed using survival analysis as well as Cox analysis (R package survival, v2.42). We performed multivariable Cox regression to analyze the relationship between immune-related risk signatures and clinicopathological factors.</ns0:p></ns0:div>
<ns0:div><ns0:head>Gene set enrichment analysis</ns0:head><ns0:p>To identify the biological functions and pathways between high-and low-risk groups, we conducted GSEA to investigate potential biological mechanisms in the Molecular Signatures Database (MSigDB; <ns0:ref type='bibr' target='#b23'>(Subramanian et al. 2005)</ns0:ref>. GSEA was performed using GSEApy, a python wrapper for gene enrichment (https://pypi.org/project/gseapy/).</ns0:p><ns0:p>We selected C2 and C5, including pathway databases and GO terms, from the MSigDB. The gesa sub-command of GSEApy was used in GSEA with default parameters. Enriched gene sets with a false discovery rate (FDR) of less than 0.25 and a P-value of less than 0.05 were considered statistically significant.</ns0:p></ns0:div>
<ns0:div><ns0:head>Statistical analysis</ns0:head><ns0:p>Boxplots were created with the R package, ggplot2 (v3.0.0). The R package ComplexHeatmap (v1.18.1) was used to create heatmaps <ns0:ref type='bibr' target='#b9'>(Gu et al. 2016)</ns0:ref>. We counted C-index with R packages 'survcomp' <ns0:ref type='bibr' target='#b10'>(Harrell et al. 1982;</ns0:ref><ns0:ref type='bibr' target='#b17'>Schröder et al. 2011)</ns0:ref>. The student's t-test was performed for statistical comparison. We chose R to conduct statistical analysis (https://www.r-project.org/). P-values lower than 0.05 were considered statistically significant. The main code of analysis was pushed to github (https://github.com/huchua/KIRC).</ns0:p></ns0:div>
<ns0:div><ns0:head>Results</ns0:head></ns0:div>
<ns0:div><ns0:head>Establishment and validation of the immune-related gene signature in KIRC</ns0:head><ns0:p>Figure <ns0:ref type='figure'>1</ns0:ref> illustrates the workflow we used to develop an immune-related gene-risk signature. The immune-related gene signature was constructed within the KIRC training data set, while we applied the testing set to validate the signature. <ns0:ref type='table' target='#tab_2'>PDF | (2019:11:43459:6:0:CHECK 21 Sep 2020)</ns0:ref> Manuscript to be reviewed After 1,000 iterations, seven unique gene models were selected (Figure <ns0:ref type='figure' target='#fig_7'>2A</ns0:ref>, Table <ns0:ref type='table'>S1</ns0:ref>). The model selected as the immune-related gene-risk signature consisted of 14 genes and ranked the highest in frequency 282 times. Parameter selection in the LASSO-cox model was log lambda = -2.641 and alpha = 1. The univariate and multivariate Cox analysis of the 14 immune-related genes are shown in Table <ns0:ref type='table' target='#tab_2'>1 and Table 2</ns0:ref>. The principle component analysis of the 14 immune-related gene signature displayed a different distribution pattern between low-and high-risk groups when comparing the training and testing (Figure <ns0:ref type='figure' target='#fig_7'>2B-C</ns0:ref>). This indicated that the low-and high-risk groups had different immune phenotypes. In the training and testing, the c-index was 0.7862 and 0.6534, respectively (P < 0.001; Figure <ns0:ref type='figure' target='#fig_7'>2D</ns0:ref>). In a time-dependent receiver operating characteristic curve (ROC) created for training and testing datasets, area under the curve (AUC) values at 1, 3, and 5 years were 0.679, 0.63, 0.627 and 0.65, 0.596, 0.568, respectively (Figure <ns0:ref type='figure' target='#fig_7'>S2</ns0:ref>). The 14 immune-related genes are AR, BID, BMP8A, CCL7, CCR10, FGF17, GDF1, IL20RB, IL4, KLRC2, LHB, SEMA3A, SEMA3G, and TXLNA. K-M (Kaplan-Meier) survival curves and gene expression of the 14 immune-related genes are shown in Figure <ns0:ref type='figure' target='#fig_8'>3</ns0:ref> and Figure <ns0:ref type='figure' target='#fig_8'>S3</ns0:ref>. Because of the differences between the log-rank test and the univariate cox analysis, the results of the univariate cox analysis of other genes except GDF1 are very consistent with the results of the K-M (Kaplan-Meier) survival curves.</ns0:p></ns0:div>
<ns0:div><ns0:head>PeerJ reviewing</ns0:head></ns0:div>
<ns0:div><ns0:head>Correlation between the signature of immune-related genes and clinical parameters</ns0:head><ns0:p>Among the 14 immune-related genes, four genes (TXLNA, SEMA3G, AR, and BID) had a high expression, and 10 genes (IL20RB, CCR10, BMP8A, SEMA3A, CCL7, GDF1, KLRC2, LHB, FGF17, and IL4) had a low expression (Figure <ns0:ref type='figure'>4A, B</ns0:ref>). The relationship between the signature and clinical factors demonstrated that patients with advanced pathological staging, M stage, and T stage had a higher risk score than those with early stage disease (Figure <ns0:ref type='figure'>4A, B</ns0:ref>; Figure <ns0:ref type='figure' target='#fig_5'>5A, B, D</ns0:ref>). However, we did not find a correlation between the signature and N stage (Figure <ns0:ref type='figure'>4A</ns0:ref>, B; Figure <ns0:ref type='figure' target='#fig_5'>5C</ns0:ref>). Manuscript to be reviewed</ns0:p></ns0:div>
<ns0:div><ns0:head>Influence of the immune-related gene signature on patient prognosis</ns0:head><ns0:p>We then assessed whether this signature influenced KIRC patient prognosis. Survival analysis showed that patients with a high-risk score were associated with poor survival outcomes in the training and testing (Figure <ns0:ref type='figure'>4C</ns0:ref>; Figure <ns0:ref type='figure'>4E</ns0:ref>).</ns0:p><ns0:p>We found that the signature also predicted survival outcomes in subgroups of KIRC patients, including stage I-II (Figure <ns0:ref type='figure'>6A</ns0:ref>), stage III-IV (Figure <ns0:ref type='figure'>6B</ns0:ref>), M0 stage (Figure <ns0:ref type='figure'>6C</ns0:ref>), M1 stage (Figure <ns0:ref type='figure'>6D</ns0:ref>), N0 stage (Figure <ns0:ref type='figure'>6E</ns0:ref>), N1 (Figure <ns0:ref type='figure'>6F</ns0:ref>), T1 (Figure <ns0:ref type='figure'>6G</ns0:ref>), T2 (Figure <ns0:ref type='figure'>6H</ns0:ref>), T3 (Figure <ns0:ref type='figure'>6I</ns0:ref>), and T4 (Figure <ns0:ref type='figure'>6J</ns0:ref>). Multivariate Cox analysis revealed that the risk signature was able to independently predict overall survival in KIRC patients (Figure <ns0:ref type='figure'>4D</ns0:ref>, F, Table <ns0:ref type='table'>3</ns0:ref>).</ns0:p></ns0:div>
<ns0:div><ns0:head>Engagement of the immune-related gene risk signature in biological pathways and functions</ns0:head><ns0:p>GSEA was used to investigate the signature's biological pathways and functions. There were 177 KEGG pathways and 4,528 GO terms used in our investigation. Our analysis found that the signature was able to engage in a total of 19 enriched KEGG pathways (FDR < 0.25; Table <ns0:ref type='table'>4</ns0:ref>). The low-risk signature was significantly correlated with 10 pathways, including the citrate cycle (TCA cycle) pathway, fatty acid metabolism pathway, propanoate metabolism pathway, butanoate metabolism pathway, peroxisome pathway, lysine degradation pathway, valine leucine and isoleucine degradation pathway, proximal tubule bicarbonate reclamation pathway, vasopressin regulated water reabsorption pathway, and the pyruvate metabolism pathway (Table <ns0:ref type='table'>4</ns0:ref>). Similarly, eight GO annotations were enriched in the low-risk group (FDR < 0.25; Table <ns0:ref type='table'>4</ns0:ref>).</ns0:p></ns0:div>
<ns0:div><ns0:head>Discussion</ns0:head><ns0:p>Our study used the TCGA database and immune-related genes to establish a KIRC signature consisting of 14 immunerelated genes. We found that patients in the high-risk group showed a positive association with M stage, T stage, and advanced pathological staging. Additionally, the signature exhibited strong prognostic abilities and independently predicted KIRC patient prognosis. Functional analysis highlighted the significance of our signature based on its Manuscript to be reviewed involvement in many important pathways.</ns0:p><ns0:p>In our investigation, we used a total of 14 immune-related genes to construct our signature. In the signature, CCL7 increased the peripheral blood mononuclear cell recruitment in renal cell cancer through the inhibition of let-7d <ns0:ref type='bibr' target='#b16'>(Riihimaki et al. 2014)</ns0:ref>. <ns0:ref type='bibr'>Wyler et al. (2014)</ns0:ref> demonstrated the ability of CCL7 to recruit monocytes through CCR2, promoting renal cell cancer metastasis to the brain. This indicates that CCL7 is a major factor in the development of KIRC in tumor immune microenvironments and may be a potential immunotherapy target. Fibroblast growth factor receptor 17 (FGF17) is another immune-related gene in our signature. FGF17 demonstrates a variety of functions in cancer development. <ns0:ref type='bibr'>Gauglhofer et al. (2011)</ns0:ref> found that FGF17 was involved in the paracrine and autocrine signaling of hepatocellular carcinoma and promoted the neoangiogenesis of hepatocellular carcinoma. <ns0:ref type='bibr'>Heer et al. (2004)</ns0:ref> showed that FGF17 was overexpressed in prostate cancer and participated in prostate carcinogenesis. Our study showed that FGF17 plays an important role in KIRC based on tumor immunology. However, the underlying mechanism of FGF17 in KIRC immune microenvironments requires further investigation. In our signature, IL-4 is another well-studied immune-related biomarker for KIRC. IL-4, released by immune cells, controls the expression of B7-H1, thus altering T cell responses in KIRC <ns0:ref type='bibr' target='#b15'>(Quandt et al. 2014)</ns0:ref> Manuscript to be reviewed significance of our signature and its possible use as a survival predictor in KIRC.</ns0:p><ns0:p>Despite of the relatively low C-index of our testing set (0.6534), the C-index of the testing set falls within a similar range as those of other related studies. For instance, in the study of Bailiang Li et al., they developed an immune signature in non-small cell lung cancer with the C-index of 0.64 <ns0:ref type='bibr' target='#b11'>(Li et al. 2017)</ns0:ref>. Besides, our signature achieved a similar C-index of testing set with the immune signature in ovarian cancer (0.625) <ns0:ref type='bibr' target='#b20'>(Shen et al. 2019)</ns0:ref>. More importantly, the C-index of our testing set showed a similar accuracy with the C-index of clinical staging systems in renal cancer (0.62) <ns0:ref type='bibr' target='#b14'>(Qu et al. 2018</ns0:ref>).</ns0:p><ns0:p>Our study expands on the signatures association with several important pathways, especially the metabolism pathway.</ns0:p><ns0:p>This may reflect the mutual interaction between tumor metabolism and tumor immunology in KIRC. <ns0:ref type='bibr'>Pearce et al. (2013)</ns0:ref> showed that metabolic reprogramming can influence the fate and function of T cells in tumors. Our study further indicates the importance of metabolism pathways in KIRC immune microenvironments.</ns0:p><ns0:p>We acknowledge the limitation of our evaluation scheme, including randomly dividing the full data set into training and testing data set, which result in the inherent bias for the specific study. Another limitation of our research is the lack of independent validation data sets in our evaluation scheme. Therefore, our study should be further validated through a prospective cohort data to further illustrate the robustness of the model.</ns0:p></ns0:div>
<ns0:div><ns0:head>Conclusions</ns0:head><ns0:p>This investigation utilized RNA-seq data from the TCGA database to construct a 14 immune-related gene-risk signature with the ability to independently predict survival outcomes in KIRC; thus, providing novel clinical applications and possible immune targets for KIRC. Manuscript to be reviewed Manuscript to be reviewed 283 Figure <ns0:ref type='figure' target='#fig_8'>S3</ns0:ref>: Expression of 14 immune-related genes in kidney renal clear cell carcinoma and normal tissues.</ns0:p><ns0:note type='other'>Figure legends</ns0:note></ns0:div>
<ns0:div><ns0:head>Figure 1</ns0:head><ns0:p>The workflow of model construction.</ns0:p><ns0:p>Our workflow constructing the model for risk-score signatures. Manuscript to be reviewed Manuscript to be reviewed </ns0:p></ns0:div><ns0:figure xml:id='fig_0'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2019:11:43459:6:0:CHECK 21 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_1'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2019:11:43459:6:0:CHECK 21 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_2'><ns0:head /><ns0:label /><ns0:figDesc>. The investigation conducted by Chang et al. (2015) demonstrated that IL-4 expression can predict KIRC patient recurrence and survival outcomes. Collectively, these investigations reinforce the significance of the 14 immune-related gene-risk signature in KIRC. Our signature is associated with the survival outcome of KIRC patients and clinical parameters, including pathological staging, M stage, and T stage. No correlation was found between the signature and N-stage, possibly due to a lack of N-stage information for many patients. According to the immune-related gene-risk signature, our study found that clinical cohorts in KIRC have different immune-related risk factors. This signature also reflects differences in tumor immune microenvironments and predicts survival outcomes in KIRC patients; thus, demonstrating the clinical PeerJ reviewing PDF | (2019:11:43459:6:0:CHECK 21 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_3'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2019:11:43459:6:0:CHECK 21 Sep 2020)</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_4'><ns0:head>Figure 1 :Figure 4 :</ns0:head><ns0:label>14</ns0:label><ns0:figDesc>Figure 1: Our workflow constructing the model for risk-score signatures</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_5'><ns0:head>Figure 5 :</ns0:head><ns0:label>5</ns0:label><ns0:figDesc>Figure 5: The K-M survival curve of (A) pathological staging, (B) M stage, (C) N stage, and (D) T stage (left). The</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_6'><ns0:head>Figure S1 :</ns0:head><ns0:label>S1</ns0:label><ns0:figDesc>Figure S1: Parameter selection in the LASSO-cox model. (A) LASSO coefficient values of the 14 immune related</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_7'><ns0:head>Figure 2</ns0:head><ns0:label>2</ns0:label><ns0:figDesc>Figure 2</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_8'><ns0:head>Figure 3 K</ns0:head><ns0:label>3</ns0:label><ns0:figDesc>Figure 3</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_9'><ns0:head>(</ns0:head><ns0:label /><ns0:figDesc>A, B) As demonstrated by the heatmap, the expression distribution of the 14 immunerelated genes differed from each other in training and testing cohorts. Each column represents the same patient and corresponds to the point below showing risk score distribution, survival status, and time in KIRC patients. Each point represented one patient sorted by the rank of the risk score. Patients with high-risk, low-risk, deceased, and alive were marked by red, blue, black, and grey, respectively. The advanced-stage patients showed high-risk scores in training and testing cohorts without stratification. (C, E) Survival analysis showed that patients with high-risk scores correlated with poor survival outcomes.(D, F) The multivariate Cox analysis in training and testing cohorts. The 14 immune-related gene signature was able to serve as an independent prognostic factor for OS.</ns0:figDesc></ns0:figure>
<ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='22,42.52,280.87,525.00,316.50' type='bitmap' /></ns0:figure>
<ns0:figure type='table' xml:id='tab_0'><ns0:head>Table1 :</ns0:head><ns0:label>Table1</ns0:label><ns0:figDesc>Univariate Cox analysis of 14 immune relate genes in all cohort</ns0:figDesc><ns0:table><ns0:row><ns0:cell>Variable</ns0:cell><ns0:cell>HR</ns0:cell><ns0:cell>95%CI</ns0:cell><ns0:cell>pvalue</ns0:cell></ns0:row><ns0:row><ns0:cell>KLRC2</ns0:cell><ns0:cell>20.657</ns0:cell><ns0:cell>9.116-46.81</ns0:cell><ns0:cell><0.001</ns0:cell></ns0:row><ns0:row><ns0:cell>CCL7</ns0:cell><ns0:cell>2.757</ns0:cell><ns0:cell>2.090-3.637</ns0:cell><ns0:cell><0.001</ns0:cell></ns0:row><ns0:row><ns0:cell>SEMA3A</ns0:cell><ns0:cell>1.748</ns0:cell><ns0:cell>1.437-2.126</ns0:cell><ns0:cell><0.001</ns0:cell></ns0:row><ns0:row><ns0:cell>SEMA3G</ns0:cell><ns0:cell>0.617</ns0:cell><ns0:cell>0.534-0.714</ns0:cell><ns0:cell><0.001</ns0:cell></ns0:row><ns0:row><ns0:cell>CCR10</ns0:cell><ns0:cell>1.759</ns0:cell><ns0:cell>1.381-2.241</ns0:cell><ns0:cell><0.001</ns0:cell></ns0:row><ns0:row><ns0:cell>BMP8A</ns0:cell><ns0:cell>1.813</ns0:cell><ns0:cell>1.298-2.532</ns0:cell><ns0:cell><0.001</ns0:cell></ns0:row><ns0:row><ns0:cell>FGF17</ns0:cell><ns0:cell>7.335</ns0:cell><ns0:cell>3.202-16.803</ns0:cell><ns0:cell><0.001</ns0:cell></ns0:row><ns0:row><ns0:cell>GDF1</ns0:cell><ns0:cell>1.483</ns0:cell><ns0:cell>1.062-2.070</ns0:cell><ns0:cell>0.021</ns0:cell></ns0:row><ns0:row><ns0:cell>IL4</ns0:cell><ns0:cell>109.106</ns0:cell><ns0:cell>34.328-346.781</ns0:cell><ns0:cell><0.001</ns0:cell></ns0:row><ns0:row><ns0:cell>LHB</ns0:cell><ns0:cell>5.053</ns0:cell><ns0:cell>3.240-7.881</ns0:cell><ns0:cell><0.001</ns0:cell></ns0:row><ns0:row><ns0:cell>TXLNA</ns0:cell><ns0:cell>2.329</ns0:cell><ns0:cell>1.500-3.618</ns0:cell><ns0:cell><0.001</ns0:cell></ns0:row><ns0:row><ns0:cell>AR</ns0:cell><ns0:cell>0.548</ns0:cell><ns0:cell>0.463-0.649</ns0:cell><ns0:cell><0.001</ns0:cell></ns0:row><ns0:row><ns0:cell>IL20RB</ns0:cell><ns0:cell>1.225</ns0:cell><ns0:cell>1.145-1.310</ns0:cell><ns0:cell><0.001</ns0:cell></ns0:row><ns0:row><ns0:cell>BID</ns0:cell><ns0:cell>3.505</ns0:cell><ns0:cell>2.439-5.035</ns0:cell><ns0:cell><0.001</ns0:cell></ns0:row></ns0:table><ns0:note>PeerJ reviewing PDF | (2019:11:43459:6:0:CHECK 21 Sep 2020)</ns0:note></ns0:figure>
<ns0:figure type='table' xml:id='tab_1'><ns0:head>Table 2 (on next page)</ns0:head><ns0:label>2</ns0:label><ns0:figDesc>Multivariate Cox analysis of 14 immune relate genes in all cohort.</ns0:figDesc><ns0:table /><ns0:note>PeerJ reviewing PDF | (2019:11:43459:6:0:CHECK 21 Sep 2020)</ns0:note></ns0:figure>
<ns0:figure type='table' xml:id='tab_2'><ns0:head>Table 2 . Multivariate Cox analysis of 14 immune relate genes in all cohort 2</ns0:head><ns0:label>2</ns0:label><ns0:figDesc /><ns0:table><ns0:row><ns0:cell>1 AR</ns0:cell><ns0:cell cols='2'>0.638 0.556-0.731</ns0:cell><ns0:cell><0.001</ns0:cell><ns0:cell>IL20RB</ns0:cell><ns0:cell>1.268 1.102-1.46</ns0:cell><ns0:cell>0.001</ns0:cell></ns0:row><ns0:row><ns0:cell>age</ns0:cell><ns0:cell cols='2'>1.352 1.143-1.599</ns0:cell><ns0:cell><0.001</ns0:cell><ns0:cell>age</ns0:cell><ns0:cell>1.414 1.195-1.672</ns0:cell><ns0:cell><0.001</ns0:cell></ns0:row><ns0:row><ns0:cell>stage</ns0:cell><ns0:cell cols='2'>3.725 2.708-5.124</ns0:cell><ns0:cell><0.001</ns0:cell><ns0:cell>stage</ns0:cell><ns0:cell>3.275 2.351-4.562</ns0:cell><ns0:cell><0.001</ns0:cell></ns0:row><ns0:row><ns0:cell>gender</ns0:cell><ns0:cell cols='2'>0.943 0.687-1.294</ns0:cell><ns0:cell>0.718</ns0:cell><ns0:cell>gender</ns0:cell><ns0:cell>1.025 0.744-1.412</ns0:cell><ns0:cell>0.879</ns0:cell></ns0:row><ns0:row><ns0:cell>BID</ns0:cell><ns0:cell cols='2'>1.411 1.211-1.645</ns0:cell><ns0:cell><0.001</ns0:cell><ns0:cell>IL4</ns0:cell><ns0:cell>1.458 1.294-1.643</ns0:cell><ns0:cell><0.001</ns0:cell></ns0:row><ns0:row><ns0:cell>age</ns0:cell><ns0:cell cols='2'>1.407 1.188-1.666</ns0:cell><ns0:cell><0.001</ns0:cell><ns0:cell>age</ns0:cell><ns0:cell>1.39 1.171-1.651</ns0:cell><ns0:cell><0.001</ns0:cell></ns0:row><ns0:row><ns0:cell>stage</ns0:cell><ns0:cell cols='2'>3.173 2.284-4.41</ns0:cell><ns0:cell><0.001</ns0:cell><ns0:cell>stage</ns0:cell><ns0:cell>3.426 2.484-4.725</ns0:cell><ns0:cell><0.001</ns0:cell></ns0:row><ns0:row><ns0:cell>gender</ns0:cell><ns0:cell cols='2'>1.039 0.755-1.43</ns0:cell><ns0:cell>0.814</ns0:cell><ns0:cell>gender</ns0:cell><ns0:cell>0.999 0.726-1.375</ns0:cell><ns0:cell>0.997</ns0:cell></ns0:row><ns0:row><ns0:cell>BMP8A</ns0:cell><ns0:cell cols='2'>1.213 1.057-1.391</ns0:cell><ns0:cell>0.006</ns0:cell><ns0:cell>KLRC2</ns0:cell><ns0:cell>1.402 1.233-1.594</ns0:cell><ns0:cell><0.001</ns0:cell></ns0:row><ns0:row><ns0:cell>age</ns0:cell><ns0:cell cols='2'>1.454 1.229-1.721</ns0:cell><ns0:cell><0.001</ns0:cell><ns0:cell>age</ns0:cell><ns0:cell>1.383 1.169-1.637</ns0:cell><ns0:cell><0.001</ns0:cell></ns0:row><ns0:row><ns0:cell>stage</ns0:cell><ns0:cell cols='2'>3.607 2.619-4.969</ns0:cell><ns0:cell><0.001</ns0:cell><ns0:cell>stage</ns0:cell><ns0:cell>3.429 2.483-4.737</ns0:cell><ns0:cell><0.001</ns0:cell></ns0:row><ns0:row><ns0:cell>gender</ns0:cell><ns0:cell cols='2'>1.052 0.765-1.447</ns0:cell><ns0:cell>0.754</ns0:cell><ns0:cell>gender</ns0:cell><ns0:cell>0.958 0.694-1.322</ns0:cell><ns0:cell>0.792</ns0:cell></ns0:row><ns0:row><ns0:cell>CCL7</ns0:cell><ns0:cell cols='2'>1.329 1.202-1.471</ns0:cell><ns0:cell><0.001</ns0:cell><ns0:cell>LHB</ns0:cell><ns0:cell>1.285 1.168-1.414</ns0:cell><ns0:cell><0.001</ns0:cell></ns0:row><ns0:row><ns0:cell>age</ns0:cell><ns0:cell cols='2'>1.455 1.226-1.728</ns0:cell><ns0:cell><0.001</ns0:cell><ns0:cell>age</ns0:cell><ns0:cell>1.393 1.178-1.647</ns0:cell><ns0:cell><0.001</ns0:cell></ns0:row><ns0:row><ns0:cell>stage</ns0:cell><ns0:cell cols='2'>3.485 2.529-4.802</ns0:cell><ns0:cell><0.001</ns0:cell><ns0:cell>stage</ns0:cell><ns0:cell>3.519 2.55-4.854</ns0:cell><ns0:cell><0.001</ns0:cell></ns0:row><ns0:row><ns0:cell>gender</ns0:cell><ns0:cell>1</ns0:cell><ns0:cell>0.726-1.377</ns0:cell><ns0:cell>1</ns0:cell><ns0:cell>gender</ns0:cell><ns0:cell>1.17 0.844-1.622</ns0:cell><ns0:cell>0.347</ns0:cell></ns0:row><ns0:row><ns0:cell>CCR10</ns0:cell><ns0:cell cols='2'>1.235 1.111-1.372</ns0:cell><ns0:cell><0.001</ns0:cell><ns0:cell>SEMA3A</ns0:cell><ns0:cell>1.214 1.107-1.33</ns0:cell><ns0:cell><0.001</ns0:cell></ns0:row><ns0:row><ns0:cell>age</ns0:cell><ns0:cell cols='2'>1.382 1.166-1.639</ns0:cell><ns0:cell><0.001</ns0:cell><ns0:cell>age</ns0:cell><ns0:cell>1.445 1.218-1.715</ns0:cell><ns0:cell><0.001</ns0:cell></ns0:row><ns0:row><ns0:cell>stage</ns0:cell><ns0:cell cols='2'>3.844 2.793-5.292</ns0:cell><ns0:cell><0.001</ns0:cell><ns0:cell>stage</ns0:cell><ns0:cell>3.549 2.572-4.898</ns0:cell><ns0:cell><0.001</ns0:cell></ns0:row><ns0:row><ns0:cell>gender</ns0:cell><ns0:cell cols='2'>1.034 0.751-1.424</ns0:cell><ns0:cell>0.836</ns0:cell><ns0:cell>gender</ns0:cell><ns0:cell>1.108 0.802-1.529</ns0:cell><ns0:cell>0.535</ns0:cell></ns0:row><ns0:row><ns0:cell>FGF17</ns0:cell><ns0:cell cols='2'>1.235 1.112-1.372</ns0:cell><ns0:cell><0.001</ns0:cell><ns0:cell cols='2'>SEMA3G 0.652 0.549-0.775</ns0:cell><ns0:cell><0.001</ns0:cell></ns0:row><ns0:row><ns0:cell>age</ns0:cell><ns0:cell cols='2'>1.399 1.181-1.657</ns0:cell><ns0:cell><0.001</ns0:cell><ns0:cell>age</ns0:cell><ns0:cell>1.379 1.164-1.633</ns0:cell><ns0:cell><0.001</ns0:cell></ns0:row><ns0:row><ns0:cell>stage</ns0:cell><ns0:cell cols='2'>3.722 2.705-5.121</ns0:cell><ns0:cell><0.001</ns0:cell><ns0:cell>stage</ns0:cell><ns0:cell>3.29 2.384-4.539</ns0:cell><ns0:cell><0.001</ns0:cell></ns0:row><ns0:row><ns0:cell>gender</ns0:cell><ns0:cell cols='2'>1.109 0.806-1.525</ns0:cell><ns0:cell>0.527</ns0:cell><ns0:cell>gender</ns0:cell><ns0:cell>0.973 0.707-1.339</ns0:cell><ns0:cell>0.867</ns0:cell></ns0:row><ns0:row><ns0:cell>GDF1</ns0:cell><ns0:cell cols='2'>1.086 0.975-1.211</ns0:cell><ns0:cell>0.135</ns0:cell><ns0:cell>TXLNA</ns0:cell><ns0:cell>1.285 1.098-1.505</ns0:cell><ns0:cell>0.002</ns0:cell></ns0:row><ns0:row><ns0:cell>age</ns0:cell><ns0:cell cols='2'>1.411 1.191-1.672</ns0:cell><ns0:cell><0.001</ns0:cell><ns0:cell>age</ns0:cell><ns0:cell>1.433 1.208-1.7</ns0:cell><ns0:cell><0.001</ns0:cell></ns0:row><ns0:row><ns0:cell>stage</ns0:cell><ns0:cell>3.76</ns0:cell><ns0:cell>2.735-5.17</ns0:cell><ns0:cell><0.001</ns0:cell><ns0:cell>stage</ns0:cell><ns0:cell>3.696 2.687-5.085</ns0:cell><ns0:cell><0.001</ns0:cell></ns0:row><ns0:row><ns0:cell>gender</ns0:cell><ns0:cell cols='2'>1.103 0.796-1.529</ns0:cell><ns0:cell>0.555</ns0:cell><ns0:cell>gender</ns0:cell><ns0:cell>1.212 0.87-1.688</ns0:cell><ns0:cell>0.255</ns0:cell></ns0:row></ns0:table></ns0:figure>
</ns0:body>
" | "Article ID: 43459
PeerJ
Yong Zou et al.
A 14-immune-related-gene signature predicts clinical outcome of kidney renal clear cell carcinoma
Dear Dr. Jun Chen,
Thank you very much for giving us an opportunity to revise our manuscript. The constructive suggestions from you have greatly helped us to improve our manuscript. Our point-by-point response to your points are detailed in the following pages. Our computer-generated tracked changes and “clean” copy manuscript have been uploaded.
We hope that you will find our revised manuscript acceptable for publication in PeerJ.
Sincerely yours,
Chuan Hu
Point-by-point response
1. To avoid overfitting, please evaluate your trained model (14 immune-related gene signature) on the test data set only. Please remove all the evaluation results (figures, tables, etc.) based on the training or full (training+test) data set.
Reply: Thank you very much for the great suggestion. According to your suggestion, we removed the evaluation results based on the training data set or the full data set to avoid the possibility of overfitting in the process of evaluating the robustness of our model. As a result, we finally deleted Figure 2D-E, Figure 5, Figure 6A-B, Figure S2C and Table 3 related results and the updated figures and tables have been uploaded to the system again.
2. In the discussion, please acknowledge the limitation of your evaluation scheme (i.e., randomly dividing the full data set into training and testing data set). Such evaluation is subject to the inherent bias for the specific study (e.g., confounding, batch effects). Ideally, an independent data set should be used for evaluation.
Reply: Thank you very much for your suggestions with which we completely agree. According to your suggestion, we added explanations about the limitations of our evaluation scheme in the discussion session. Detailed descriptions are as follows: ‘We acknowledge the limitation of our evaluation scheme, including randomly dividing the full data set into training and testing data set, which result in the inherent bias for the specific study. Another limitation of our research is the lack of independent validation data sets in our evaluation scheme. Therefore, our study should be further validated through a prospective cohort data to further illustrate the robustness of the model’
" | Here is a paper. Please give your review comments after reading it. |
9,839 | "<ns0:abstract xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:p>The breakdown of plant biomass in rumen depends on interactions between bacteria, archaea, fungi, and protozoa. However, the majority of studies of the microbiome of ruminants, including the few studies of the rumen of camels, only studied one of these microbial groups. In this study, we applied total rRNA sequencing to identify active microbial communities in twenty-two solid and liquid rumen samples from eleven camels reared under three feeding systems. These camels were separated in three groups, G1 (n=3), G2 (n=6) and G3 (n=2) and fed Egyptian clover hay and wheat straw and concentrates feed mixture, fresh Egyptian clover, and wheat straw, respectively. Bacteria dominated, followed protozoa, archaea, and fungi, libraries of reads generated from all camel rumen samples. Firmicutes, Thermoplasmatales, Diplodinium, and Neocallimastix dominated bacterial, archaeal, protozoal and fungal communities, respectively in all samples. Feeding systems influenced the microbial diversity and relative abundance of microbial groups; libraries generated from camels fed fresh clover showed the highest alpha diversity. Principal co-ordinate analysis and linear discriminate analysis showed clusters associated with feeding system and that the relative abundance of microbes varied between liquid and solid fractions. In addition, the analysis showed positive and negative correlations between the microbial groups. This study is the first to assess all the active microbial profiles in the rumen of camels under different feeding systems to expand our knowledge regarding microbial communities and their symbiotic and competitive interactions for maintaining the normal functions of the rumen.</ns0:p></ns0:div>
</ns0:abstract>
<ns0:body xmlns:ns0='http://www.tei-c.org/ns/1.0'>
<ns0:div><ns0:head>Introduction</ns0:head><ns0:p>Camel (Camelus dromedaries) provides food security in arid and semi-arid countries with the increase of global warming due to its ability to produce milk and meat in hot climate <ns0:ref type='bibr' target='#b83'>(Samsudin et al., 2011;</ns0:ref><ns0:ref type='bibr' target='#b21'>Faye, 2013)</ns0:ref>. Camel also provide textiles (fiber and hair) and it is also commonly used for daily human activities such as transportation, agriculture, tourism, race and riding <ns0:ref type='bibr' target='#b76'>(Rabee et al., 2019)</ns0:ref>. This unique animal is well adapted to arid conditions in the hot deserts by its unique feeding behavior and the functional structure of its digestive tract <ns0:ref type='bibr' target='#b44'>(Kay et al., 1989)</ns0:ref>. The retention time of feed particles in the camel forestomach is longer than cows and sheep, which prolongs the exposure of plant biomasses to the symbiotic microorganisms and helps in the efficient digestion <ns0:ref type='bibr'>(Lechner-Dolland and Engelhardt, 1989)</ns0:ref>. Camel production lies under three systems based on feeding type, camels in traditional extensive system depend on low quality feeds; while, camels in semi-intensive system depend on highquality forage and camels in intensive system depend on high-quality forage and concentrates supplements <ns0:ref type='bibr' target='#b21'>(Faye, 2013)</ns0:ref>. Many factors affecting the microbial communities in the rumen, including age, animal breed; however, feeding system, including diet composition and feeding plan, is the main determiner of the diversity of rumen microbial communities <ns0:ref type='bibr' target='#b31'>(Henderson et al., 2015)</ns0:ref>. The chemical composition of the diet is the major shaper of fermentation in the rumen. For instance, cellulolytic and hemicellulytic diets favor the fibrolytic microbes; while, starch and sugars are the major fermentation components of concentrate-based diets; thus, favoring the amylolytic microbes <ns0:ref type='bibr' target='#b12'>(Carberry et al., 2012)</ns0:ref>. Also, the microbial composition and diversity varied between liquid and solid rumen fractions, which might indicate different roles in rumen fermentation; for instance, plant-adherent microbiota might have a role in fiber degradation <ns0:ref type='bibr' target='#b78'>(Ren et al., 2020)</ns0:ref>. Digestion in the camel depends on microbial fermentation in the rumen <ns0:ref type='bibr' target='#b83'>(Samsudin et al., 2011)</ns0:ref>. The efficiency of microbial fermentations in the rumen depends on interactions between a wide variety of microbial groups, including bacteria, archaea, fungi and protozoa <ns0:ref type='bibr' target='#b97'>(Yanagita et al., 2000;</ns0:ref><ns0:ref type='bibr' target='#b43'>Kamra, 2005)</ns0:ref>. Camels can utilize lignocelulolytic shrubs that other domestic ruminants avoid <ns0:ref type='bibr' target='#b83'>(Samsudin et al., 2011)</ns0:ref>. Consequently, camel rumen microbes must have the capacity to degrade such poor-quality feeds <ns0:ref type='bibr' target='#b25'>(Gharechahi et al., 2015)</ns0:ref>. However, the microbial community in the rumen of dromedary camel received less attention than other domesticated ruminants. The investigation of rumen microbial community has many implications, including the possibility of improving animal productivity and the reduction of greenhouse gas emission <ns0:ref type='bibr' target='#b31'>(Henderson et al., 2015)</ns0:ref>. The development of the next-generation sequencing technologies offer the possibility to use various metagenomic and metatranscriptomic techniques for the rapid identification of rumen microbiomes and overcome the intrinsic constraints of traditional culture-based methods <ns0:ref type='bibr' target='#b83'>(Samsudin et al., 2011;</ns0:ref><ns0:ref type='bibr' target='#b36'>Ishaq and Wright, 2014)</ns0:ref>. Most of PCR-based assessments of microbial groups in the rumen have relied on amplicon sequencing, which target a specific variable region on 16S rRNA gene <ns0:ref type='bibr' target='#b56'>(Li et al. 2016)</ns0:ref>. This approach needs a wide range of primers to study different microbial communities <ns0:ref type='bibr' target='#b46'>(Kittelmann et al., 2013)</ns0:ref>. Therefore, the output could be biased due to the primer selection and amplification cycling conditions <ns0:ref type='bibr' target='#b27'>(Guo et al., 2015;</ns0:ref><ns0:ref type='bibr' target='#b56'>Li et al., 2016;</ns0:ref><ns0:ref type='bibr' target='#b20'>Elekwachi et al., 2017)</ns0:ref>. Total RNA sequencing (RNA-Seq) offers the advantage of specifically targeting active microbes and avoids biases associated with primer selection and chimera generation in PCR <ns0:ref type='bibr' target='#b23'>(Gaidos et al., 2011;</ns0:ref><ns0:ref type='bibr' target='#b27'>Guo et al., 2015;</ns0:ref><ns0:ref type='bibr' target='#b56'>Li et al., 2016)</ns0:ref>. In addition, RNA-Seq approach is capable of identifying novel microbes as it is not reliant on primers for known microbes <ns0:ref type='bibr' target='#b56'>(Li et al., 2016)</ns0:ref>. High-throughput metatranscriptome sequencing provides a comprehensive understanding of the biological systems by characterization of different groups of organisms in the same environment based on the sequencing of coding and noncoding RNA <ns0:ref type='bibr' target='#b20'>(Elekwachi et al., 2017)</ns0:ref>. Total RNA-Seq was applied to investigate microbial communities in many different systems including, for example, the microbial community in human gut <ns0:ref type='bibr' target='#b75'>(Qin et al., 2012)</ns0:ref>, and cow rumen <ns0:ref type='bibr' target='#b56'>(Li et al., 2016;</ns0:ref><ns0:ref type='bibr' target='#b20'>Elekwachi et al., 2017 )</ns0:ref>.</ns0:p><ns0:p>All the microbiome studies on the camel rumen characterized one or two microbial groups using classical or molecular approaches. For example, the protozoal community in camel rumen was studied heavily by conventional microscopic methods <ns0:ref type='bibr' target='#b24'>(Ghali et al., 2005;</ns0:ref><ns0:ref type='bibr' target='#b3'>Baraka, 2012)</ns0:ref>. Only three molecular-based studies are available on the bacterial community <ns0:ref type='bibr' target='#b83'>(Samsudin et al., 2011;</ns0:ref><ns0:ref type='bibr' target='#b6'>Bhatt et al., 2013;</ns0:ref><ns0:ref type='bibr' target='#b25'>Gharechahi et al., 2015)</ns0:ref>. Furthermore, only one study classified the rumen archaea <ns0:ref type='bibr' target='#b25'>(Gharechahi et al., 2015)</ns0:ref>. Regarding the anaerobic fungi, a new fungal genus, Oontomyces was isolated from the rumen of Indian camel <ns0:ref type='bibr' target='#b14'>(Dagar et al., 2015)</ns0:ref>, and only one study investigated the whole fungal community in the gut of the camel <ns0:ref type='bibr' target='#b76'>(Rabee et al., 2019)</ns0:ref>.</ns0:p><ns0:p>Moreover, no study provided a comprehensive analysis of potential active rumen microbiotas in the camel. In the present study, total rRNA sequencing was applied to 1) get insight into the composition of active microbiota in the rumen of camels reared under different feeding systems; 2) describe the distribution of microbial groups among the solid and liquid rumen fractions; 3) investigate the correlations between all the microbial groups.</ns0:p></ns0:div>
<ns0:div><ns0:head>Materials and Methods</ns0:head></ns0:div>
<ns0:div><ns0:head>Rumen samples</ns0:head><ns0:p>Rumen samples were collected from eleven adult dromedary camels under three different feeding systems. Camels in group G1 (n=3) were housed in the Maryout Research Station, Alexendria, Egypt and were fed on Egyptian clover hay (Trifolium alexandrinum), wheat straw and concentrates feed mixture. Camels in group G2 (n=6) were fed on fresh Egyptian clover (100 % high-quality forage diet) then slaughtered in the Kom Hammada slaughterhouse, Elbehera, Egypt. Animals of group G3 (n=2) were fed on wheat straw (100 % low-quality forage diet) then were slaughtered in Pasateen slaughterhouse, Cairo, Egypt. All the animals kept on the diet for at least one month before the sampling time. The proximate analysis of feeds illustrated in supplementary table <ns0:ref type='table'>S1</ns0:ref>. Details regarding the camel rumen samples in this study presented in Supplementary table <ns0:ref type='table'>S2</ns0:ref>. The rumen contents were strained immediately by two layers cheesecloth to separate the liquid and solid to form twenty-two samples, then were frozen using liquid nitrogen and stored at -80 o C before further processing <ns0:ref type='bibr' target='#b20'>(Elekwachi et al., 2017)</ns0:ref>. The project was approved and all samples were collected according to the Institutional Animal Care and Use Committee, Faculty of Veterinary Medicine, University of Sadat City, Egypt (Approval number: VUSC00003).</ns0:p></ns0:div>
<ns0:div><ns0:head>RNA isolation, quality and quantity estimation and sequencing</ns0:head><ns0:p>The frozen rumen samples were ground using liquid nitrogen. About 0.5 gram of frozen fine powder was used for total RNA isolation using Trizol-Reagent protocol (Invitrogen, Carlsbad, CA), followed by RNA clean up using MEGA clear Kit (Invitrogen). Total RNA quality and quantity were estimated using an Agilent 2100 bioanalyzer (Agilent Technologies, USA) and RNA 6000 Nano kit (Agilent Technologies, USA). One hundred Nanogram of total RNA was reversetranscribed into first strand cDNA and sequenced using Illumina rRNA MiSeq preparation kit (Illumina, USA) by Illumina MiSeq platform.</ns0:p></ns0:div>
<ns0:div><ns0:head>Bioinformatic data analysis</ns0:head><ns0:p>The generated RNA sequence reads were analysed using pipeline developed by <ns0:ref type='bibr' target='#b20'>Elekwachi et al. (2017)</ns0:ref>. Briefly, the sequence quality was checked using the FastQC program v. 0.11.4 <ns0:ref type='bibr' target='#b0'>(Andrews, 2010)</ns0:ref>, then Trimmomatic program v. 0.35 <ns0:ref type='bibr' target='#b8'>(Bolger et al., 2014)</ns0:ref> was used to trim adaptors, barcodes, ambiguous and low quality reads. PEAR program v. 0.9.6 <ns0:ref type='bibr' target='#b99'>(Zhang et al., 2014)</ns0:ref> was used to merge read 1 and read 2 using default options. Then after, the hidden Markov models rRNA-HMM tool of the rapid analysis of multiple metagenomes with a clustering and annotation pipeline (RAMMCAP) <ns0:ref type='bibr' target='#b55'>(Li, 2009)</ns0:ref> was used to sort the reads into archaea and bacteria (16S, 23S), and eukaryote (18S, 23S) rRNA sequences. Merged sequence files were then sub-sampled as needed using <ns0:ref type='bibr'>MEME program v. 4.10.2 (Bailey et al., 2009)</ns0:ref>. For each sample, 70,000 reads were run through the pipeline. For subsequent analysis steps, 20 000, 10 000, and 2000 sequences were used for bacteria, eukaryote and archaea, respectively. Taxonomy binning for eukaryote and archaeal SSU rRNA sequences was performed using BLASTN. The sub-sampled query sequences were searched against the SILVA SSURef-111 database using an e-value of 1e -5 . Bacterial SSU sequences were binned into operational taxonomic units (OTUs) using the 'classify. seqs' command of Mothur v. 1.33.1 program <ns0:ref type='bibr' target='#b84'>(Schloss et al., 2009)</ns0:ref>. The SSURef -108 gene and the SSURef-108b taxonomy databases were used. Principal co-ordinate analysis (PCoA) using Bray Curtis dissimilarity and alpha diversity indices (Chao1, Shannon and Inverse Simpson) were evaluated by Mothur <ns0:ref type='bibr' target='#b84'>(Schloss et al., 2009)</ns0:ref> based on sub-sampling of 70,000 reads per sample according the protocol 'Community Structure Analysis Based on OTU Clustering' outlined in <ns0:ref type='bibr' target='#b20'>Elekwachi et al. (2017)</ns0:ref>.</ns0:p></ns0:div>
<ns0:div><ns0:head>Statistical analyses</ns0:head><ns0:p>Data of relative abundance of bacterial phyla, protozoal genera, fungal genera and archaea genera and order Thermoplasmatales were tested for normality and homogeneity using Shapiro-Wilk test and variables that were deemed non-normal were then arcsine transformed. Linear Discriminate Analysis (LDA) and Bray Curtis Permutational Multivariate Analysis of Variance (PERMANOVA) tests depended on the relative abundance of bacterial phyla. All the protozoal, fungal and archaeal genera and the order Thermoplasmatales were used to show the differences in community structure and to compare the clustering of samples. Pearson correlation analysis was used to identify correlation within and between microbial communities and the correlation scores were visualized as a heatmap. The statistical analyses were performed using the SPSS v. 20.0 software package (SPSS, 1999) and PAST <ns0:ref type='bibr' target='#b30'>(Hammer et al., 2001)</ns0:ref>. All the sequences were deposited to the sequence read archive (SRA) under the accession number: SRP107370.</ns0:p></ns0:div>
<ns0:div><ns0:head>Results</ns0:head></ns0:div>
<ns0:div><ns0:head>The composition and diversity of active microbial community</ns0:head><ns0:p>Total rRNA sequencing in twenty-two solid and liquid rumen samples from eleven camels resulted in a total of 3958591 reads with average of 359871.9 ± 85365.7 (mean ± standard error (SE)) reads per animal in the solid fraction (SF) and 3386392 reads with an average of 307853.8 ± 60989.6 reads per animal in the liquid fraction (LF). The sequence reads of bacteria dominated the active microbial community, followed by protozoa, archaea and fungi (Table <ns0:ref type='table'>1</ns0:ref>). The relative abundance of protozoa was higher in LF-G1 (liquid fraction of G1), while the relative abundance of bacteria was higher in SF-G1 (solid fraction of G1). The highest population of archaea was observed in G2 camels. Additionally, G3 camels showed the highest relative abundance of fungi (Table <ns0:ref type='table'>1</ns0:ref>; Supplementary Figure <ns0:ref type='figure' target='#fig_0'>1</ns0:ref>). Number of OTUs and Alpha-diversity indices, Chao1, Shannon and Inverse Simpson, were higher in the rumen of LF-G2 samples (Table <ns0:ref type='table'>1</ns0:ref>).</ns0:p></ns0:div>
<ns0:div><ns0:head>Bacterial community</ns0:head><ns0:p>The composition of bacterial community varied little between treatments and consisted of 12 phyla. The five most predominant phyla were Firmicutes, Bacteroidetes, Proteobacteria, Spirochaetes and Fibrobacteres, respectively (Table <ns0:ref type='table'>2</ns0:ref>). Phylum Firmicutes dominated the bacterial community in all groups and was higher in G2 followed by G1 and G3 camels, respectively, and was also higher in SF compared to LF (Table <ns0:ref type='table'>2</ns0:ref>). On the family level, the Firmicutes phylum was dominated by Lachnospiraceae and Ruminococcuceae. In addition, six genera dominated this phylum, including Butyrivibrio, RFN8-YE57, Ruminococcus, vadinHA42, Acetitomaculum and Blautia (Fig. <ns0:ref type='figure' target='#fig_0'>1a</ns0:ref> and supplementary Table <ns0:ref type='table'>S3</ns0:ref>). The second largest phylum, Bacteroidetes, showed the highest relative abundance in G3 followed by G1 and G2 camels and was higher in LF than SF (Fig. <ns0:ref type='figure' target='#fig_0'>1a</ns0:ref> and supplementary Table <ns0:ref type='table'>S3</ns0:ref>). On the family and genera levels, Bacteroidetes was dominated by three families (Prevotellaceae, BS11_ gut_ group, Rikenellaceae) and two genera (Prevotella, RC9_gut_group) besides uncultured Bacteroidetes. Proteobacteria, phylum showed a higher relative abundance in LF-G1 samples and was dominated by Succinivibrionaceae family and Desulfovibrio genus (Table <ns0:ref type='table'>2</ns0:ref>, Fig. <ns0:ref type='figure' target='#fig_0'>1a</ns0:ref>, supplementary Table <ns0:ref type='table'>S3</ns0:ref>). The Spirochaetes phylum was higher in the SF-G3 and it was classified into two families including Spirochaetaceae and PL-11B10 and was dominated by Treponema genus. The Fibrobacteres phylum was higher in SF-G3 (Table <ns0:ref type='table'>2</ns0:ref>, Fig. <ns0:ref type='figure' target='#fig_0'>1a</ns0:ref>, supplementary Table <ns0:ref type='table'>S3</ns0:ref>). The other phyla, including Actinobacteria,thatwas higher in SF-G2 samples, Tenricutes phylum was higher in the LF-G1 samples andLentisphaeraephylum,was about 3-fold higher in the LF as relative to SF and accounted for a large population in the camels of G3 (Table <ns0:ref type='table'>2</ns0:ref>). Additionally, several minor bacterial phyla were also observed in the rumen of camels such as Verrucomicrobia, Elusimicrobia, Cyanobacteria and Chloroflexi (Table <ns0:ref type='table'>2</ns0:ref>).</ns0:p><ns0:p>All Bacterial genera were observed in all groups except seven genera, including uncultured Marinilabiaceae (Bacteroidetes), Quinella (Firmicutes) and Streptococcus (Firmicutes) that were observed only in G2 and G3 camels. Ruminobacter (Proteobacteria) was observed only in G1 and G2 camels. On the other hand, Arcobacter and Succinivibrio within phylum Proteobacteria were observed only in G1 camels and Betaproteobacteria (Proteobacteria) was observed only in G3 camels. Moreover, many unclassified bacteria were observed across samples and accounted for 38.53% of total bacterial reads. Most of these unclassified bacterial reads were observed in phylum Firmicutes and Bacteroidetes.</ns0:p></ns0:div>
<ns0:div><ns0:head>Archaeal community</ns0:head><ns0:p>All archaeal reads were assigned to the phylum Euryacheota. The order level classification revealed three orders, including Thermoplasmatales, Methanobacteriale and Methanomicrobial. Thermoplasmatales dominated the archaeal community and showed the highest population in LF-G3 camels, this order was not classified out of order level (Table <ns0:ref type='table'>3</ns0:ref>, Fig. <ns0:ref type='figure' target='#fig_0'>1b</ns0:ref>). All the Methanobacteriale reads were belonged to family Methanobacteriacea that classified into three genera; Methanobrevibacter, Methanophera and Methanobacterium. Methanobrevibacter is the second largest contributor in archaeal population and was higher in SF-G1 camels. Methanosphaera exhibited higher relative abundance in SF-G2 camels. Methanobacterium was absent in G3 camels; however, a small proportion of this genus was found in the camels of G1 and G2. Methanomicrobium genus, which belongs to order Methanomicrobiales and family Methanomicrobiaceae was the least contributor in archaeal population and was more prevalent in LF-G3 camels (Table <ns0:ref type='table'>3</ns0:ref>, Fig. <ns0:ref type='figure' target='#fig_0'>1b</ns0:ref>).</ns0:p></ns0:div>
<ns0:div><ns0:head>Protozoal community</ns0:head><ns0:p>The protozoal population in camels of the current study was grouped in two cultured families, Ophryoscolecidae and Isotrichidae (Table <ns0:ref type='table'>4</ns0:ref>). The Ophryoscolecidae family consisted of seven genera, Diplodinium, Ophryoscolex, Entodinium, Polyplastron, Eudiplodinium, Epidinium and Trichostomatia. In addition, Isotrichidae consisted of two genera, Dasytricha and Isotricha. The variation among the camels in protozoal population was clearly observed and seemed to be higher than other microbial communities; however, the protozoal community composition was similar among the camels (Table <ns0:ref type='table'>4</ns0:ref>, Fig. <ns0:ref type='figure' target='#fig_0'>1c</ns0:ref>). The most dominant protozoal genera were Diplodinium, Ophryoscolex and Entodinium. Camels in G1 had the highest population of Entodinium and Epidinium. Camels in G2 had the greatest population of Eudiplodinium, Ophryoscolex, Isotricha and Dasytricha. The camels in G3 had the greatest population of Diplodinium, Polyplastron and Trichostomatia. On the sample fraction level, the solid fraction had a higher representation of Ophryoscolex, Polyplastron, Eudiplodinium, Epidinium and Diplodinium while the liquid fraction had a higher representation of Entodinium, Isotricha and Dasytricha (Table <ns0:ref type='table'>4</ns0:ref>, Fig. <ns0:ref type='figure' target='#fig_0'>1c</ns0:ref>).</ns0:p></ns0:div>
<ns0:div><ns0:head>Anaerobic rumen fungal community</ns0:head><ns0:p>The characterization of rumen fungi revealed four fungal genera; three of which were anaerobic fungi related to phylum Neocallimastigomycota and family Neocallimasticeceae including Neocallimastix, which dominated the fungal community in the current study, followed by Piromyces and Cyllamyces (Table <ns0:ref type='table'>5</ns0:ref>, Fig. <ns0:ref type='figure' target='#fig_0'>1d</ns0:ref>). These anaerobic fungal genera represented > 99.5 % of the fungal population. In addition, genus Spizellomyces, which is related to phylum Chytridiomycota and family Spizellomycetaceae, was noted in a very small proportion (<0.5 %) (Table <ns0:ref type='table'>5</ns0:ref>). Neocallimastix was more abundant in the SF-G1 samples while Piromyces and Cyllamyces were more abundant in LF-G2 and SF-G3 respectively (Table <ns0:ref type='table'>5</ns0:ref>, Fig. <ns0:ref type='figure' target='#fig_0'>1d</ns0:ref>).</ns0:p></ns0:div>
<ns0:div><ns0:head>Effect of feeding system on the composition of microbial communities</ns0:head><ns0:p>Multivariate analysis separated libraries by feeding system distinctly (Figs. <ns0:ref type='figure' target='#fig_4'>2 and 3</ns0:ref>). Also, bacteria, dominated by phylum Firmicutes were the main driver of differences between animals (Fig. <ns0:ref type='figure' target='#fig_4'>3</ns0:ref>). Furthermore, Entodinium, Thermoplasmatales, Neocallimastix were the main drivers of differences in protozoal, archaeal and fungal communities, respectively. PERMANOVA analysis revealed that the difference between camel groups was significant (P < 0.01) in all microbial groups (Supplementary Table <ns0:ref type='table'>S4</ns0:ref>). Pairwise comparison between camel groups based on Bonferroni-corrected p-value demonstrated that the difference was significant (P < 0.05) between camels of G2 and G3 in bacterial and archaeal communities (Supplementary Table <ns0:ref type='table'>S4</ns0:ref>). Moreover, the difference was significant between the three groups in the protozoal community (P < 0.05) whereas, in the fungal community, the difference was significant only between camels in group G1 and G2 (Supplementary Table <ns0:ref type='table'>S4</ns0:ref>).</ns0:p></ns0:div>
<ns0:div><ns0:head>Pearson correlation between microbes in the rumen of dromedary camel</ns0:head><ns0:p>Pearson correlation analysis (Fig. <ns0:ref type='figure' target='#fig_6'>4A, 4B</ns0:ref>), revealed many significant positive and negative correlations (P < 0.05). For example, in active bacteria, Bacteroidetes correlated positively with Cyllamycesand negativelywith Butyrivibrio, Methanosphaera and Trichostomatia. Prevotellaceae correlated positively with Neocallimastix and Entodinium and negatively with Ruminococcaceae, Methanosphaera and Diplodinium. Fibrobacteres correlated positively with Cyllamyces, Methanomicrobium, Thermoplasmatales and Diplodinium and negatively with Methanosphaera, Epidinium, Ruminococcaceae and Butyrivibrio. Firmicutes correlated positively with Methanosphaera and negatively with Piromyces, Thermoplasmatales and Methanomicrobium.</ns0:p><ns0:p>In active archaea, Thermoplasmatales correlated positively with Diplodinium and negatively with Methanobrevibacter and Methanosphaera. In active protozoa, there was a negative correlation between Polyplastron, Entodinium, Ophryoscolex and Epidinium. In active fungi, a negative correlation was observed between Cyllamyces, Neocallimastix and Piromyces and between Piromyces and Entodinium.</ns0:p></ns0:div>
<ns0:div><ns0:head>Discussion</ns0:head><ns0:p>Rumen microbes can ferment a wide variety of feed components, including cellulose, xylan, amylose and protein <ns0:ref type='bibr' target='#b31'>(Henderson et al., 2015)</ns0:ref> and produce volatile fatty acids that provide the animal with approximately 70% of daily energy requirements <ns0:ref type='bibr' target='#b4'>(Bergman, 1990)</ns0:ref>. Furthermore, the rumen fermentation generates methane, which contributes to global warming and represents 2-12% loss of feed energy for the animal <ns0:ref type='bibr' target='#b42'>(Johnson and Ward, 1996;</ns0:ref><ns0:ref type='bibr' target='#b12'>Carberry et al., 2012;</ns0:ref><ns0:ref type='bibr' target='#b39'>Jami et al., 2014)</ns0:ref>. Therefore, investigation of these microbial communities is the key to understand their roles and maximize ruminal fermentation and fiber digestion <ns0:ref type='bibr' target='#b54'>(Lee et al., 2012)</ns0:ref>. The structure of microbial community in the camel rumen was similar in the composition; however, feeding system had an impact on the microbial diversity and the relative abundance of microbial groups. PCoA, LDA and PERMANOVA analyses confirmed the finding of this study and was in agreement with the results of other ruminant studies <ns0:ref type='bibr' target='#b31'>(Henderson et al., 2015)</ns0:ref>. Camels in the present study were fed on different forages; Egyptian clover and wheat straw (Supplementary Table <ns0:ref type='table'>S1</ns0:ref>). The Egyptian clover is considered the most balanced fodder in Egypt as it has high nutritive value regarding crude protein, crude fiber, mineral content and soluble carbohydrate compared to wheat straw and concentrates mixture <ns0:ref type='bibr' target='#b12'>(Carberry et al., 2012;</ns0:ref><ns0:ref type='bibr' target='#b1'>Bakheit, 2013;</ns0:ref><ns0:ref type='bibr' target='#b87'>Shrivastava et al., 2014)</ns0:ref>, which might supported the high microbial diversity in G2 camels compared to other groups (Table <ns0:ref type='table'>1</ns0:ref>). This was consistent with previous studies on cows <ns0:ref type='bibr' target='#b72'>(Pitta et al., 2010;</ns0:ref><ns0:ref type='bibr' target='#b86'>Shanks et al., 2011;</ns0:ref><ns0:ref type='bibr' target='#b49'>Kumar et al., 2015)</ns0:ref>. Highly degradable carbohydrates support the bacterial and protozoal growth <ns0:ref type='bibr' target='#b18'>(Dijkstra and Tamminga, 1995;</ns0:ref><ns0:ref type='bibr' target='#b49'>Kumar et al., 2015)</ns0:ref>, which could demonstrate their higher population in G1 camels. Additionally, the higher bacterial population slows the fungi growth <ns0:ref type='bibr' target='#b90'>(Stewart et al., 1992;</ns0:ref><ns0:ref type='bibr' target='#b66'>Orpin and Joblin, 1997)</ns0:ref>, which was illustrated by the low fungal population in G1 camels.</ns0:p></ns0:div>
<ns0:div><ns0:head>Bacterial community</ns0:head><ns0:p>Firmicutes phylum was found to be more abundant than Bacteroidetes and both phyla comprised > 75% of all bacterial reads (Table <ns0:ref type='table'>2</ns0:ref>), which is in agreement with the results of previous studies on different animals including camels <ns0:ref type='bibr' target='#b83'>(Samsudin et al., 2011)</ns0:ref>, Surti Buffalo <ns0:ref type='bibr' target='#b67'>(Pandya et al., 2010)</ns0:ref> and Muskoxen <ns0:ref type='bibr' target='#b82'>(Salgado-Flores et al., 2016)</ns0:ref>. The majority of Firmicutes' members have a potential role in fiber digestion, which might illustrate their higher population in G2 camels that were fed on high-quality forage and also in solid fraction. This speculation was supported by the high proportion of Ruminococcaceae and Lachnospiraceae families that found to be active in fiber digestion in the rumen <ns0:ref type='bibr' target='#b70'>(Pitta et al., 2014a;</ns0:ref><ns0:ref type='bibr' target='#b62'>Nathani et al., 2015)</ns0:ref>. Bothe Blautia and Acetitomaculum genera are known to have a key role as reductive acetogens (Le <ns0:ref type='bibr' target='#b50'>Van et al., 1998;</ns0:ref><ns0:ref type='bibr' target='#b98'>Yang et al., 2016)</ns0:ref> and found to be varied according to the feeding system among the camel groups in this study. This finding could indicate that the reductive acetogenesis pathway could be maximized by diet to minimize methane production (Le <ns0:ref type='bibr' target='#b50'>Van et al., 1998)</ns0:ref>. Bacteroidetes were higher in poor quality forage (G3), which was similar to results found in cattle <ns0:ref type='bibr' target='#b69'>(Pitta et al., 2014b)</ns0:ref>, and this phylum was dominated by family Prevotellaceae that was in agreement with the study of <ns0:ref type='bibr' target='#b25'>Gharechahi et al. (2015)</ns0:ref> on camels. The members of Bacteroidetes possess diverse enzymes that can target cellulose, pectin and soluble polysaccharides released in the liquid phase <ns0:ref type='bibr' target='#b59'>(Mackenzie et al., 2015)</ns0:ref>. Additionally, Prevotella genus is involved in propionate production that is used for energy by the host <ns0:ref type='bibr' target='#b62'>(Nathani et al., 2015)</ns0:ref>. Taken together, we speculate that Bacteroidetes play a key role to improve the digestion and better utilization of poor-quality feeds and contribute to adaptation of camels to arid conditions. Further studies on structure and function of these enzymes will be necessary to determine their molecular roles.</ns0:p><ns0:p>The RC9_gut_group found in this study belongs to uncultured genera and was found also in the gut of Rhinoceros hindgut <ns0:ref type='bibr' target='#b7'>(Bian et al., 2013)</ns0:ref>. Unclassified Bacteroidetes are specialized in lignocellulose degradation <ns0:ref type='bibr' target='#b59'>(Mackenzie et al., 2015)</ns0:ref>, which could support their high proportion in G3 camels. The Fibrobacteres was higher (3.1%) in this study compared to the other findings on camels <ns0:ref type='bibr' target='#b25'>(Gharechahi et al., 2015)</ns0:ref>. Interestingly, Fibrobacteres has been shown in previous studies to be the principal cellulolytic bacteria in the rumen <ns0:ref type='bibr' target='#b77'>(Ransom-Jones et al., 2012;</ns0:ref><ns0:ref type='bibr' target='#b62'>Nathani et al., 2015)</ns0:ref> which might illustrate its higher relative abundance in solid fraction and in the rumen of camels fed on wheat straw (G3) (Table <ns0:ref type='table'>2</ns0:ref>) that is rich in lignocellulose. We also identified that the members of Proteobacteria were lower in G2 and G3 camels that were fed on diet rich in fiber contents. These findings highlighted this phylum's function as a protein-degrading bacteria as it was reported by <ns0:ref type='bibr' target='#b57'>Liu et al. (2017)</ns0:ref>. The abundance of Treponema was higher in the solid fraction and in G3 camels (Figure <ns0:ref type='figure' target='#fig_0'>1a</ns0:ref>). Treponema is the dominant genus in Spirochaetes phylum and it is fiber-associated bacteria, which could indicate to its cellulytic and xylanolytic activities <ns0:ref type='bibr' target='#b35'>(Ishaq and Wright, 2012)</ns0:ref>.</ns0:p><ns0:p>The dominant bacterial genera in this study were Butyriovibrio, RFN8-YE57, Ruminococcus, Prevotella, Fibrobacter, Treponema and VadinHA. These genera were higher in the SF except RFN8-YE57 compared to the LF; this finding was consistent with a previous study on camels <ns0:ref type='bibr' target='#b25'>(Gharechahi et al., 2015)</ns0:ref> and which confirm the speculation that the solid attached microbial population seemed to play a major role in ruminal fiber digestion <ns0:ref type='bibr' target='#b40'>(Jewell et al., 2015;</ns0:ref><ns0:ref type='bibr' target='#b64'>Noel et al., 2017)</ns0:ref>. Further work is needed to examine this community which could lead to the assigning of the fibrolytic bacteria and as a consequence, could ultimately help increase our understanding and improving the fiber degradation in the rumen. Most of Elusimicrobia in this study were uncultured; some members of this phylum were isolated from the termite's gut that degrades cellulose <ns0:ref type='bibr' target='#b32'>(Herlemann et al., 2009)</ns0:ref>. Therefore, we speculate that this phylum has a role in fiber digestion and that might illustrate their high proportion in G3 camels. Actinobacteria observed also in the rumen of moose and some members of this phylum have acetogenic activities <ns0:ref type='bibr' target='#b37'>(Ishaq et al., 2015)</ns0:ref>. Some members of Victivallis within Lentisphaerae phylum were involved in cellobiose degrading activity <ns0:ref type='bibr' target='#b100'>(Zoetendal et al., 2003)</ns0:ref>.</ns0:p><ns0:p>Unclassified bacteria in our study (38.53% of total bacterial reads) were less than the percentage found in the study on Muskoxen (53.7-59.3%) <ns0:ref type='bibr' target='#b82'>(Salgado-Flores et al., 2016)</ns0:ref>. The presence of unclassified bacteria in the gut was commonly observed <ns0:ref type='bibr' target='#b26'>(Gruninger et al., 2016)</ns0:ref> and could be a result of the presence of new bacteria that has the ability to ferment plant biomass <ns0:ref type='bibr' target='#b82'>(Salgado-Flores et al., 2016)</ns0:ref> or related to the sequencing approach used where short reads were generated from RNA-seq <ns0:ref type='bibr' target='#b56'>(Li et al., 2016)</ns0:ref>. These unclassified bacteria need more studies to enable their isolation and identification.</ns0:p></ns0:div>
<ns0:div><ns0:head>Archaeal community</ns0:head><ns0:p>The archaeal population has important roles in the rumen and in methane emission mitigation strategies as they convert the H 2 and CO 2 produced in the rumen to methane <ns0:ref type='bibr' target='#b33'>(Hook et al., 2010)</ns0:ref>. Additionally, acetate produced in fiber breakdown is used to provide a methyl group for methanogenesis; therefore, methanogens population could be shifted by alteration of diet composition or feed additives and plant compounds <ns0:ref type='bibr' target='#b33'>(Hook et al., 2010;</ns0:ref><ns0:ref type='bibr' target='#b91'>Tapio et al., 2017 )</ns0:ref>, which could demonstrate the variation in the relative abundance of archaea between camel groups. Camels of the second group (G2) that fed fresh clover, showed the highest archaeal population (Table <ns0:ref type='table'>2</ns0:ref>) and the archaeal community was dominated by Thermoplasmatales, a methylotrophic methanogens order (Table <ns0:ref type='table'>3</ns0:ref>) which was consistent with the results on cattle <ns0:ref type='bibr' target='#b11'>(Carberry et al., 2014)</ns0:ref> and camels <ns0:ref type='bibr' target='#b25'>(Gharechahi et al., 2015)</ns0:ref>. Thermoplasmatales produces methane from methyl amine and its population was decreased by the addition of rapeseed oil to animal diet, making it a high potential target in future strategies to mitigate methane emissions <ns0:ref type='bibr'>(Poulsen et al., 2013)</ns0:ref>. The Methanobrevibacter, Methanosphaera, Methanomicrobium and Methanobacterium (Table <ns0:ref type='table'>4</ns0:ref>) are the other dominant archaea that were also observed in this study and in accordance with the results found in beef cattle <ns0:ref type='bibr' target='#b11'>(Carberry et al., 2014)</ns0:ref>. Methanobrevibacter dominated the methanogens in other ruminant <ns0:ref type='bibr' target='#b31'>(Henderson et al., 2015)</ns0:ref> and was associated with high methane emissions <ns0:ref type='bibr' target='#b91'>(Tapio et al., 2017)</ns0:ref>. Moreover, Methanomicrobium had its highest proportion with the feeding system of poor quality forage diet (G3), which was similar to results found in buffalo <ns0:ref type='bibr' target='#b22'>(Franzolin and Wright, 2016)</ns0:ref>, and In vitro <ns0:ref type='bibr' target='#b93'>(Wang et al., 2018)</ns0:ref>. In rumen, Methanomicrobium has been shown to be responsible for the conversion of H2 and/or formate into CH4 <ns0:ref type='bibr' target='#b51'>(Leahy et al., 2013)</ns0:ref>. The abundance of Thermoplasmatales was also negatively correlated with Methanobrevibacter which is consistent with previous results <ns0:ref type='bibr' target='#b15'>(Danielsson et al., 2017;</ns0:ref><ns0:ref type='bibr' target='#b60'>McGovern et al., 2017)</ns0:ref>.</ns0:p></ns0:div>
<ns0:div><ns0:head>Protozoal community</ns0:head><ns0:p>The majority of protozoal reads were related to Diplodinium, Ophryoscolex, Entodinium, Polyplastron, Eudiplodinium and Epidinium (Table <ns0:ref type='table'>4</ns0:ref>). Similar findings were observed in other study on different ruminants <ns0:ref type='bibr' target='#b3'>(Baraka, 2012)</ns0:ref>. The relative abundance of protozoal was influenced by feeding system, which was in the same line with results on cattle <ns0:ref type='bibr' target='#b34'>(Hristov et al., 2001;</ns0:ref><ns0:ref type='bibr' target='#b94'>Weimer, 2015)</ns0:ref>. The Diplodinium dominated the protozoal community and was prevalent in the G3 camels, which highlighted the cellulolytic activity of this genus <ns0:ref type='bibr' target='#b13'>(Coleman et al., 1976)</ns0:ref>. Also, some species of genus Diplodinium were discovered in the rumen of Egyptian camel and is considered to be peculiar in camel such as Diplodinium cameli, <ns0:ref type='bibr' target='#b48'>(Kubesy and Dehority, 2002)</ns0:ref>. In addition, Entodinium was higher in G1 camels that were fed on concentrates feed mixture that increase the protozoa. Also, previous studies showed that this genus was dominant in rumen of camels <ns0:ref type='bibr' target='#b85'>(Selim et al., 1999;</ns0:ref><ns0:ref type='bibr' target='#b24'>Ghali et al., 2005)</ns0:ref> and cattle <ns0:ref type='bibr' target='#b12'>(Carberry et al., 2012)</ns0:ref>. Moreover, the study of <ns0:ref type='bibr' target='#b47'>Kittelmann and Janssen (2011)</ns0:ref> showed that the Polyplastron was the dominant genus in cattle. On the function level, all the genus Eudiplidinum, Epidinum and Diplodinum have cellulolytic activity <ns0:ref type='bibr' target='#b13'>(Coleman et al., 1976)</ns0:ref> whereas, Polyplastrone and Epidinium have a xylanolytic activity <ns0:ref type='bibr' target='#b17'>(Devillard, 1999;</ns0:ref><ns0:ref type='bibr' target='#b5'>Béra-Maillet et al., 2005)</ns0:ref>.</ns0:p></ns0:div>
<ns0:div><ns0:head>Anaerobic rumen fungal community</ns0:head><ns0:p>The highest fungal population was observed in the solid fraction and rumen of G3 camels (Table <ns0:ref type='table'>1</ns0:ref>). These findings were in agreement with the results of different studies stated that the fibre-based diets stimulated the fungal growth <ns0:ref type='bibr' target='#b65'>(Orpin, 1977;</ns0:ref><ns0:ref type='bibr' target='#b79'>Roger et al., 1993;</ns0:ref><ns0:ref type='bibr' target='#b43'>Kamra et al., 2005;</ns0:ref><ns0:ref type='bibr' target='#b28'>Haitjema et al., 2014)</ns0:ref>. This speculation could explain the low fungal population in G1 camels in our study. Moreover, the longer retention time and neutral pH in camel's forestomach <ns0:ref type='bibr' target='#b80'>(Russell and Wilson, 1996)</ns0:ref> make it more suitable for the survival of rumen fungi. The genus Neocallimastix dominated the fungal community and found to be higher in the G1 camels which was similar to other results on sheep and camels <ns0:ref type='bibr' target='#b46'>(Kittelmann et al., 2013</ns0:ref><ns0:ref type='bibr' target='#b76'>, Rabee et al., 2019)</ns0:ref>. This genus produces enzymes capable of hydrolyzing cellulose, xylan and starch <ns0:ref type='bibr' target='#b68'>(Pearce and Bauchop, 1985)</ns0:ref>. In the other side, Cyllamyces that was observed in small population, has the ability to degrade poor-quality feeds <ns0:ref type='bibr' target='#b89'>(Sridhar et al., 2014)</ns0:ref>, which might explain its high population in solid fraction and G3 camels. Piromyces was the second dominant genus in the camel rumen of this study and has been shown to produce cellulolytic and xylanolytic enzymes <ns0:ref type='bibr' target='#b92'>(Teunissen et al., 1992)</ns0:ref>. Therefore, the fungi were more prevalent in ruminants of G2 camels, which fed high-quality forage with high fiber contents than in G2 and G3 camels. The genus Spizellomyces is closely related to Chytridiomctes <ns0:ref type='bibr' target='#b9'>(Bowman et al., 1992)</ns0:ref>, and common in grassland and crop soil <ns0:ref type='bibr' target='#b58'>(Lozupone and</ns0:ref><ns0:ref type='bibr'>Klein, 2002, Kittelmann et al., 2012)</ns0:ref>. Thus, the presence of this fungus in the camel rumen in the current study could be explained by a contamination of the forages by soil.</ns0:p></ns0:div>
<ns0:div><ns0:head>Correlation between rumen microbes</ns0:head><ns0:p>The interactions between rumen microbes are the main driver of feed degradation and methane formation in the rumen, which influence the animal production and the environment <ns0:ref type='bibr' target='#b96'>(Williams et al., 1994;</ns0:ref><ns0:ref type='bibr' target='#b54'>Lee et al., 2012;</ns0:ref><ns0:ref type='bibr' target='#b31'>Henderson et al., 2015)</ns0:ref>. Positive and negative correlations were observed within and between microbial communities in this study (Fig. <ns0:ref type='figure' target='#fig_6'>4</ns0:ref>). Methanogens colonize the protozoa and this relationship was believed to enhance methane formation <ns0:ref type='bibr' target='#b63'>(Newbold et al., 1995)</ns0:ref>, which highlighted some positive correlations between protozoa and archaea. Additionally, the fibrolytic bacteria produce the important substrates mainly hydrogen and methyl groups that methanogens use for growth, <ns0:ref type='bibr' target='#b41'>(Johnson and Johnson, 1995)</ns0:ref>, which demonstrated the positive correlations found between Fibrobacteres and some methanogens. Also, positive correlation between the methylotrophic Methanosphaera and Lachnospiraceae that has been implicated in pectin degradation and provides methanol as a substrate for the methylotrophs <ns0:ref type='bibr' target='#b16'>(Dehority, 1969)</ns0:ref>. On the other hand, Prevotella is a hydrogen utilizer and produces propionate which has a negative impact on methanogenesis in the rumen <ns0:ref type='bibr' target='#b70'>(Pitta et al., 2014a;</ns0:ref><ns0:ref type='bibr' target='#b57'>Liu et al., 2017)</ns0:ref>, which also illustrated the negative correlation obtained in this study between Prevotellaceae and archaea.</ns0:p><ns0:p>Since the rumen anaerobic fungi produce abundant H 2 through the fermentation of carbohydrate; they can interact positively with H 2 utilizers such as archaea, Prevotellaceae, Blautia and Acetitomaculum <ns0:ref type='bibr' target='#b66'>(Orpin and Joblin, 1997;</ns0:ref><ns0:ref type='bibr' target='#b50'>Le Van et al., 1998;</ns0:ref><ns0:ref type='bibr' target='#b98'>Yang et al., 2016;</ns0:ref><ns0:ref type='bibr' target='#b57'>Liu et al., 2017)</ns0:ref>. Additionally, the anaerobic fungi penetrate plant tissue, which provides an increased surface area for bacterial colonization <ns0:ref type='bibr' target='#b66'>(Orpin and Joblin, 1997)</ns0:ref>, which could explain the positive correlation between fungi and both Butyrivibrio and Fibrobacteres in this study. However, fungi are known to be negatively impacted by the presence of some bacteria and protozoa as the fungal zoospores are likely to be a prey for protozoa <ns0:ref type='bibr' target='#b61'>(Morgavi et al., 1994)</ns0:ref>, which demonstrated the negative correlation between both Neocallimastix and Piromyces with Diplodinium and Entodinium. Furthermore, Ruminococcus produces compounds that inhibit the growth of rumen fungi <ns0:ref type='bibr' target='#b90'>(Stewart et al., 1992)</ns0:ref>, which support the negative correlation between Neocallimastix and Ruminococcaceae. Polyplastron predates upon other protozoa like Epidinium, Eudiplodinium, Diplodinium, and Ostracodinium <ns0:ref type='bibr' target='#b19'>(Eadie, 1967)</ns0:ref>, which might explained the negative correlation between Polyplastron and other Protozoa.</ns0:p></ns0:div>
<ns0:div><ns0:head>Conclusions</ns0:head><ns0:p>This study applied total rRNA sequencing to get insight into the active microbial groups in the rumen of dromedary camels. However, using the DNA-amplicon sequencing with RNA sequencing is recommended in the future studies to compare the composition of active microbial groups (from RNA sequencing) with the composition of the whole microbial community.</ns0:p><ns0:p>As a major conclusion of our study, the microbial community in camel rumen was diverse and similar in composition between the camels. However, the feeding system impacted the relative abundance of active microbial communities where the fresh Egyptian clover provided the highest microbial diversity. The majority of camel rumen microbes (bacteria, fungi, and protozoa) were fibrolytic or have a possible role in fiber digestion, which might illustrate the ability of camel to live in desert harsh conditions under poor feeds. Moreover, the structure of microbial community in rumen of camel found to be similar to other ruminant studies with a shown difference in the relative abundances. The present results should open new perspectives for further cultivation and isolation studies on the unclassified microorganisms found in the rumen of camels to classify them and assign their functions.</ns0:p></ns0:div>
<ns0:div><ns0:head>Figure 1</ns0:head><ns0:p>The relative abundance of microbial groups </ns0:p></ns0:div><ns0:figure xml:id='fig_0'><ns0:head>Figure 1 :</ns0:head><ns0:label>1</ns0:label><ns0:figDesc>Figure 1: Comparison of relative abundance of genera of the microbiota in dromedary camel. bacterial (a), archaeal (b), protozoal (c) and fungi (d) in ruminal solid (SF) and liquid (LF) fractions of camels under different feeding systems.</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_1'><ns0:head /><ns0:label /><ns0:figDesc>Figure 2</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_2'><ns0:head>Figure 2 :</ns0:head><ns0:label>2</ns0:label><ns0:figDesc>Figure 2: Principal Co-ordinated analysis derived from OTUs from twenty-two ruminal liquid (LF) and solid (SF) samples distributed on three camel groups. G1 camels (red circles), G2 (white circle and G3(blue circles).</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_3'><ns0:head /><ns0:label /><ns0:figDesc>Figure 3</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_4'><ns0:head>Figure 3 :</ns0:head><ns0:label>3</ns0:label><ns0:figDesc>Figure 3: Linear Discriminant analysis of microbial communities in the samples based on the relative abundance of genera of active bacteria (a), archaea (b), protozoa (c) and fungi i (d) in ruminal solid (SF), and liquid (LF) fractions of camels under three feeding systems, G1 (black dots), G2 (blue squars) and G3 (coral triangles).</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_5'><ns0:head /><ns0:label /><ns0:figDesc>Figure 4</ns0:figDesc></ns0:figure>
<ns0:figure xml:id='fig_6'><ns0:head>Figure 4 :</ns0:head><ns0:label>4</ns0:label><ns0:figDesc>Figure 4: Heatmap based on Pearson correlation coefficients between and within the relative abundance of bacteria, archaea, protozoa and fungi in solid (a) and liquid (b) rumen fractions of dromedary camel. The black boxed ellipses refer to the significant correlations at P < 0.05.</ns0:figDesc></ns0:figure>
<ns0:note place='foot'>PeerJ reviewing PDF | (2019:12:44048:1:0:NEW 10 Mar 2020)</ns0:note>
</ns0:body>
" | "
Desert Research Center 10th March 2020
1Mathaf El Matariya St.B.O.P.11753
Matariya- Cairo,Egypt
Phone: (+202)26332846 - 26374800
FAX: (+202) 26357858
Email: alaa.bakr.stu@gebri.usc.edu.eg
Rabee_a_m@yahoo.com
Dear Editors,
We thank the Reviewers for their constructive comments and corrections. We responded to all comments and have extensively modified many sections in the paper to correct grammatical and style errors. I also enclosed unclean paper including all comments colored by three colors, yellow for Editor’s comments, green for the first Reviewer’s comments, turquoise for the second Reviewer’s comments, and pink for the third Reviewer’s comments. Below are the responses to all the comments.
We appreciate the opportunity to submit our manuscript to Peer J.
Yours sincerely,
Dr. Alaa Rabee
Researcher at Desert Research Center, Egypt
On behalf of all authors
Comments
Manuscript title: “Comparative analysis of the metabolically active microbial communities in the rumen of dromedary camels under different feeding systems using total rRNA sequencing ”
In the unclean or marked manuscript you could notice that colored comments using three colors, yellow for Editor’s comments, green for the first Reviewer’s comments, turquoise for the second Reviewer’s comments, and pink for the third Reviewer’s comments.
Editor Reviewer 1 Reviewer 2 Reviewer 3
Editor’s comments
Line 38. Here and throughout, write succinctly. For example, replace “The plant biomass breakdown in the rumen depends on the complex microbiota that consists of bacteria, archaea, fungi, and protozoa and their interactions.” With “Breakdown of plant biomass in rumen depends on interactions between bacteria, archaea, fungi, and protozoa.”
>> Modified, thank you.
Line 40. Again, write succinctly. Replace “… majority of rumen microbiome studies characterized separate microbial groups to understand the microbial fermentation in the gut of the herbivorous animals including camels. This study…” with “the majority of studies of the microbiome of rumininants, including the few studies of the rumen of camels, only studied one of these microbial groups. In this study, we”
>> Modified, thank you.
Line 42. Here and throughout, write directly and clearly. Replace “get a collective insight into the potential” with “identify”
>> replaced.
Line 45. Delete “on”
>> Deleted.
Line 46. Avoid the passive voice. Replace “.. respectively (spelling). The active bacterial community was the most dominant in the rumen of camel followed by protozoal, archaeal and fungal communities, respectively.” With “Bacteria dominated, followed protozoa, archaea and fungi, libraries of reads generated from the rumen of camels.” See also line 48 (were the most).
>> Modified, thank you.
Line 47. Avoid fillers like “Our results showed” It should be obvious to the reader when you are presenting your results. See also lines 198, 220 (“in the present study”).
>> removed.
Line 50. Be specific. Replace “groups, where…indices.” with “…groups; libraries generated from camels fed fresh clover showed the highest alpha diversity.”
>> Replaced.
Line 52-53. Replace “.. showed that the samples of camel groups clustered separately. Variations in the relative abundance of microbial communities across the sample fractions were observed.” With “… showed clusters associated with feeding system and that the relative abundance of microbes varied between liquid and solid fractions of.”
>> Replaced, thank you.
Line 62. Write succinctly. Replace “has high value to support the” with “provides”
>> Replaced.
Line 64. Repetitive. Replace “In addition to the high….milk. Camels can provide….race riding..” with “Camels also provide textiles…”
>> Replaced.
Line 72. Avoid superfluous determiner “The” in “The Digestion..” Also, delete “like other ruminants”
>> Modified, thank you.
Line 74. Delete “in different forms of”
>> Deleted.
Line 75. Delete “that cooperate…host animal”
>> Deleted, thank you.
Line 77. Delete superfluous “the” and revise to write actively. Replace “shrubs that are mostly avoided by other domestic ruminant” with “shrubs that other domestic ruminants avoid”
>> Modified.
Line 83-86. Revise this section.
>> Improved.
Line 87. Delete “RNA/DNA”
>> Removed.
Line 89. Delete “in the same environment”
>> Removed.
Line 92-99. Discuss published work in the present tense. Replace “…outperformed amplicon
Sequencing… chimera structures” with “offers the advantage of specifically targeting active microbes and avoids biases associated with primer selection and chimer generation in PCR”
>> Modified, thank you.
Line 100. Join this discussion of the benefits of RNAseq with the previous paragraph and replace “it doesn’t just based” with “it doesn’t depend”
>> Thank you, improved.
Line 145. Here and throughout. Avoid superfluous “The.” Actually delete “The analysis…feeding systems.”
>>Deleted.
Line 170. Shapiro and Wilk are surnames.
>> Modified, thank you.
Line 173. Missing a verb.
>> Modified.
Line 181-184. Delete this section, which repeats Introduction and Methods.
>>Deleted.
Line 197-199. Wordy. Replace “… community was similar… and the five most predominant…” with “…community varied little between treatments and consisted of 12 phyla. The five most predominant…”
>>Replaced.
Line 204. Again, write actively. Replace “this phylum was found to be dominated by six genera” with “six genera dominated this phylum”
>> Modified.
Line 227. Revise “were mainly observed in”
>> Modified, thank you.
Line 242. As above, delete phrases like “The results showed that” and “of the current study” We know results show results. Also, delete “observed in this study (line 247)” and “In this study (line 257).”
>> Modified, thank you.
Line 267-269. Wordy. Replace “Multivariate analysis …depending on the feeding system.” With “Multivariate analysis separated libraries by feeding system distinctly…”
>> Replaced.
Line 274-277. Do not repeat Methods. Replace “We performed…metric. The results revealed…” with “PERMANOVA revealed…”
>> Modified.
Line 285. As above, revise to “Pearson correlation analysis revealed many…”
>> Modified.
Line 303-305. Revise to “Rumen microbes can ferment a wide variety.. and produce volatile fatty acids…”
>> Improved.
Line 311. Again with “In this study” and “in the present study”
>> Removed, thank you.
Line 314. Write actively. Revise to “PCoA…confirmed this finding…”
>> Modified.
Line 317. Missing a verb here.
>>Modified.
Line 321. Delete “the results obtained from”
>> Removed.
Line 326. Replace “have been found to affect negatively the” with “slows”
>> Replaced.
Line 328. Sample fractions don’t impact the community. Sampling may impact your results.
>> This part is confusing, so that we removed it and we added a new section in the introduction about the importance of identification the microbial communities in liquid and solid rumen fractions.
Line 330. Have a native English speaker read this out loud.
>> This part is confusing, so that we removed it and we added a new section in the introduction about the importance of identification the microbial communities in liquid and solid rumen fractions.
Line 509. Carefully proofread references. (Physiol. Rev.). Italicize genus and species (line 552). Delete “42”
>> Modified, we revised the reference section, thank you.
line 591. Only cap proper nouns “A proposed taxonomy (line 610).”
>> Modified.
Reviewer 1
Line 43: liquid rumen……. Camels are classified as pseudo-ruminants, please check whether to use the term rumen
>>Thank you, The classification of camel as a true or pseudo-ruminant is controversial and there is no agreement between the scientists around this point and it needs more studies. I collected most of views around this point it is introduced below.
Camel rumen
There is no agreement between the scientists regarding to the number of chambers in camel stomach. Yagil, 1982, stated that, although camels ruminate, they are not true ruminants, as they lack the four well-defined stomachs of the ruminants; the rumen, reticulum, omasum and abomasum. The anatomy of all members of the Camelidae is considered to be similar, but most of the available data on the anatomy of the alimentary canal have been obtained mainly from the llama. In spite of the fact that the camel ruminates, its ingested feeds are subject to microbial digestion and the final metabolic products are similar to those in true ruminants, it is classified as pseudo ruminant, but this classification is mainly due to the significant differences in the structure and function of the digestive system of the camelids (Tylo-pods) and the true ruminants (Bhattacharya, 1986). Camels are classified as pseudo-ruminants because its foregut is differentiated into three compartments (rumen, reticulum and the gastric secreting abomasums instead of four compartments like true ruminants (Engelhardt et al., 2007). Hegazi, 1950, described the camel as having the same four compartments as other ruminants, but with the external constrictions between the omasum and abomasum being less well defined in the camel. Wardeh, 2004, indicated that, the rumen (first compartment) of the camel is characterized by its unique exterior glandular sacs which secrete a mucus-like substance that differs in composition from rumen liquid. The third compartment is absent from the honeycomb-like structure which does not distinctively separate from the fourth compartment. The motility pattern of the compound stomach of the camel differs from that of the true ruminants. The biochemical pattern of microbial fermentation is generally similar between ruminants and camelids although concentrations of short chain fatty acids in the forestomach contents are always higher in camels.
References
Bhattacharya, A.N. (1986). Structural peculiarities in the digestive system of camel. Al-Jouf Range and Animal Development Centre. Saudi Arabia.
Engelhardt, W.; Dycker, C. and Lechner-Doll, M. (2007). Absorption of short-chain fatty acids, sodium and the foregut of camels. J. Comp. Physiol. B. Biochem. Syst. Environ. Physiol. 177:631–640.
Hegazi, A.H. (1950). The stomach of the camel. Brit. Vet. J. 106: 209–213.
Wardeh, M. F. (2004). The nutrient requirements of the Dromedary camel. J. Camel Science. 1: 37-45.
Yagil, R. (1982). Camels and camel milk . Food and Agriculture Organization of United Nations. Rome, FAO.
Line 44: G1 (n=3), G2 (n=6) and G3 (n=2)…… What is the rationale of using different number of test camels for each group?
>>Thank you, we collected equal number of samples. However, the isolated RNA in some samples was degraded so that we discarded those samples, which generated unequal animal number in the groups.
Line 46-47: The active bacterial community was the most dominant……… Please clarify if this observation is true for all three feeding regimes.
>>Clarified.
Line 47-49: Different taxonomic levels have been used to describe the dominance of different communities. For example, Firmicutes are a phylum, Thermoplasmatales a order, while Diplodinium and Neocallimastix are genera. It would be better to have the same level of classification. Also, please mention if this observation is true for all three feeding regimes.
>> Thank you, we discovered many bacterial phyla (12), including a lot of genera. In the same time the archaeal reads are related to one phylum (Euryacheota), fungal reads also are related to one phylum (Neocallimastigomycota), and protozoal reads are related to one phylum (Ciliophora), including few genera. Also, the order Thermoplasmatales, the dominant archaea, was not classified out of order level in our study. Therefore, it is not possible to use one level of classification in the analysis.
Line 49: Feeding system to… Feeding systems
>>corrected.
Line 51-52 : Principal co-ordinate analysis and linear discriminate analysis showed ………….., Please indicate the relevance of this observation.
>>Modified and clarified, thank you.
Line 53: Does sample fractions indicate the solid and liquid samples?
>>Yes it does, clarified
Line 54: to This study is
>>Modified.
Line 56: to enhance our….rephrase this part.
>>Modified.
Line 57: normal function of rumen to …..normal functions
>>corrected.
Line 62: Camel (Camelus dromedaries)…. Italicise
>> modified
Line 70: Please rephrase
>>Rephrased
Line 72: rumen or rumen like chamber, pseudo-rumen?
>>We clarified this point previously at Line 43.
Line 83: remove (Recently).
>>Removed
Line 84: use various molecular methods ………, Rephrase.
>>Rephrased.
Line 87: Specify, the term 'molecular' is vague.
>>Modified.
Line 98: remove (region which is biased due to primer properties)
>>Modified.
Line 100: it doesn’t just based………?
>>modified
Line 114: groups in the rumen of camels were not investigated yet…….Rephrase the sentence.
>>Rephrased.
Line 116: No introduction is given related to the rationale of rearing camels on different feeding systems or how diet is expected to affect the microbial communities?
Line 117: Introduction is missing on potential difference between solid and liquid fractions
>> Thank you, we have added a paragraph to the introduction about the role of feeding in shaping the microbial community and the benefit of identification of rumen microbiota in different rumen fractions.
Line 124: Trifolium alexandrinum….Italicise
>> Corrected.
Line 125: 100 % high-quality forage…., Which forage? In abstract, only fresh clover hay is mentioned. Also, no details available in the supplementary tables
>>We made modification in the text and in the supplementary table S2 to make it clearer, thank you.
Line 126: What about the camels fed on clover hay? Were they slaughtered too?
>> No, they were not slaughtered.
Line 127: 100 % low-quality forage diet….. Which one?
>>We made modification in the text here and in the supplementary table S2 to make it clearer, thank you.
Line 128: The proximate analysis of feeds were illustrated …Remove were.
>>Done.
Line 129: Details regarding the camel rumen samples in this study were presented….. Remove were
>>done.
Line 130: Use same terms to describe the forage (clover hay or barseem hay or fresh barseem of fresh clover hay) in both supplementary tables.
>>Yes, we modified it, thank you
Line 132: stored at -80oC for further processing……Remove for and add reference
>> Done.
Line 169: Please explain why different taxonomic levels are mentioned
>> Thank you, we discovered many bacterial phyla (12), including a lot of genera. In the same time the archaeal reads are related to on phylum (Euryacheota), fungal reads also are related to one phylum (Neocallimastigomycota), and protozoal reads are related to one phylum (Ciliophora), including few genera. Also, the order Thermoplasmatales, the dominant archaea, was not classified out of order level in our study. Therefore, it is not possible to use one level of classification in the analysis.
Line 181: we characterized the potential active bacteria….functionally active.
Line 183: What about different fractions?
>> Thank you, we removed this paragraph based on the request of the Editor as this paragraph is repeated in material and method section.
Line 203: Lachnospiraceae…. Italicize here, and every other place in the MS for names of families.
>> Thank you, but the family name should not be italicized according to Peer J rules.
Line 205: Butyrivibrio, RFN8-YE57, Ruminococcus, vadinHA42…..Don't italicize the strain numbers. Make relevant changes throughout MS .. and Add space.
>>Thank you, we modified RFN8-YE57. Regarding the vadinHA42, we noticed there is no space in the name you could check this paper: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3937448/
Line 210: RC9_gut_group……Why italicized?
>>Corrected.
Line 212: Desulfovibrio ……Italicize
>>Corrected.
Line 213: PL-11B10…..Is it a family name?
>> Yes, it is uncultured family.
Line 237: Methanobacterium was absent in the rumen contents of G3 camels; however, traces of it was….Rephrase.
>> Modified.
Line 240: What about the results of solid and liquid fractions?
>>Thank you, LF-G3 in the sentence refers to liquid fraction of G3.
Line 256: 'Analysis of active' is repetitively used for all microbial groups. It can be removed and described as subsections.
>> We modified the titles of subsections.
Line 257: Please rephrase this sentence. It makes the reader wonder if there are multiple orders and families in the phylum Neocallimastigomycota.
>> Thank you for this important note, Improved.
Line 264: What about the results of solid and liquid fractions?
>>Thank you, LF and SF in the sentence refers to liquid and solid fraction respectively. These symbols were explained at the beginning of result section.
Line 271: change was to were….What about different fractions?
>> Thank you, modified, here we talk about the general trend in the experiment as PERMANOVA is a multivariate analysis, we used it to measure the dissimilarity between the groups based on all given data.
Line 271: Please rephrase the sentence.
>> Thank you, Modified.
Line 302: Discussion
1) if any microbial community is specific to the camel that help camel survive in such extreme conditions, as was observed in case of an anaerobic fungal genus in camel.
2. The microbial communities usually observed in the camel rumen using metagenomic approaches, and how are they different from functionally active microbial communities observed in this study.
>>Thank you so much, our discussion came in this context, you could observe that we highlighted the function of microbial communities in the camel and improved our introduction, discussion and conclusion to focus on this idea. For example, in the bacteria discussion we showed that the dominant bacteria were, Ruminococcaceae and Lachnospiraceae families, RC9_gut_group, Unclassified Bacteroidetes, Fibrobacteres, Treponema, and Elusimicrobia. All these bacteria have lignocellulolytic activities. In protozoa discussion, we improved it, you could notice that the dominant genus is Diplodinium, which has cellulolytic activities and linked to camel rumen. Also, more than 99.5% of fungal community in our study have cellulolytic activities. All these facts highlight the ability of camel to survive in harsh condition, we made this idea clear in the discussion and in conclusion. Furthermore, we were interested to compare our results to the results of other studies.
Line 304: Remove final
>> Thank you, all the sentence was modified based on the editor’s comment.
Line 305: Are volatile
>> Thank you, all the sentence was modified based on the editor’s comment.
Line 311: In this study, …. was similar…to what?
>> Modified.
Line 315: In materials and methods, you have also mentioned about other diets.
>> Thank you, improved.
Line 320: Why discuss only G2 here, what about other groups?
>> Thank you, improved. Here, we would like to highlight the high diversity in G2 camels compared to other groups.
Line 322: Bacteria are dominant in the rumen of most animals, not relevant.
>>We removed this sentence as we used different identification technique.
Line 326: What about diet? How does diet affect the fungal growth?
>> We discussed that in fungi section.
Line 328: How?
>> We removed this paragraph as it is confusing and need more studies.
Line 345: Bacteroidetes was high…..change was to were.
>> Changed.
Line 351: What about the role of Firmicutes?
>> We discussed their role before Bacteriodetes and focused on two families, Ruminococcaceae and Lachnospiraceae.
Line 355: Are you talking about RC9? If so, clarify in the earlier sentence as well
>> Thank you, here, we talk about all unclassified Bacteroidetes including RC9.
Line 360: Please explain here and elsewhere the possible reasons why cellulolytic bacteria dominate the solid fraction. Is it because of the fact that the microbes present in the solid fraction attach themselves to the substrate.
>> yes, this opinion is clear in the next paragraph, thank you.
Line 365: Please recheck this statement. Treponema spp. are ususally saccharolytic or pectinolytic but non cellulolytic. Some of them are known to degrade xylan, but most exist as partners with cellulolytic bacteria like Fibrobacter.
>> Thank you, we made slight modification in the sentence, and below some papers about the potential lignocellulolytic activities of Treponema.
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC6276622/pdf/main.pdf
https://www.pnas.org/content/pnas/115/51/E11996.full.pdf
https://www.nature.com/articles/s41396-018-0255-1.pdf
Line 368-369: RFN8-YE57 and VadinHA…… Dont italicize here and elsewhere
>> Done.
Line 400: Methyl Amine ….. to methyl amine.
>> Modified
Line 405: Methanogens to methanogens.
>> Modified
Line 408: Refer to latest papers describing variations in methangoenic community with respect to diet….Italicez
>> Modified and a new reference was added, thank you.
Line 411: Methanobrevibacter …….. Italicize
>> Done.
Line 431 : What does it mean?
>> Thank you, it means that the high-fiber diets such as G3 diet stimulate the fungal growth, and this idea is clear through the discussion.
Line 435: All anaerobic fungal cultures produce these enzymes, though the activities might vary. Please refer to latest papers describing variations in activities of different genera.
>> Most of studies on the cellulolytic and xylanolytic activities of anaerobic fungi were conducted before 2019, and we have mentioned the recent studies such as (Rabee et al., 2019).
Line 437: Again, no such thing that Cyllamyces only degrade poor quality feeds. Its true for most genera of anaerobic fungi
>>Thank you, we mentioned this piece of information as all in vitro studies that have been conducted on this fungal genera used poor-quality substrates (fibrous).
Line 440: Not exclusive to only Piromyces.
>> Thank you, we improved this sentence.
Line 456: change Fibrobacters to Fibrobacteres
>> Done.
Line 466: rephrase, others do not penetrate plant tissues.
>>Done, thank you.
Reviewer 2
line 137: change 'grinded' to 'ground'
>> modified, thank you
line 154: should the eukaryotic RNA be '(18S, 28S)'?
>>The characterization of rumen protozoa based widely on 18S rDNA as a taxonomic marker. While the characterization of rumen fungi based on 18S rDNA and ITS1. The 18S rDNA is not variable enough to phylogenetic differentiation and there is a variability of ITS1 size between the strains. Therefore, there is a move towards the use of 28S rDNA or the Large Subunit(LSU) as a barcoding locus. However, using LSU needs more isolation and sequencing work to deposit more sequences into the ribosomal databases especially with the emergence of Third-generation Sequencing that sequence longer DNA/RNA segments. Also, more studies are needed to find suitable primers for 28S rDNA to reach the optimum amplicon size for Next-generation Sequencing (NGS) and to find primer combinations able to detect all the strains, which open the door to depend solely on 28S rDNA as a genetic marker. I would recommend combining the use of ITS1, 18S rDNA, and 28S rDNA in the future to check the reliability of modern and old data sets. In our study, we used 18S rRNA shotgun sequencing which enables the measurement of transcripts, allowing us to identify known and novel strains instead of trying many primers.
You can some more details regarding my response in:
Edwards, J. E., Forster, R. J., Callaghan, T. M., Dollhofer, V., Dagar, S. S., Cheng, Y., Chang, J., Kittelmann, S., Fliegerova, K., Puniya, A. K., Henske, J. K., Gilmore, S. P., O'Malley, M. A., Griffith, G. W., & Smidt, H. (2017). PCR and Omics Based Techniques to Study the Diversity, Ecology and Biology of Anaerobic Fungi: Insights, Challenges and Opportunities. Frontiers in microbiology, 8, 1657. https://doi.org/10.3389/fmicb.2017.01657
Elekwachi, C. O., Wang, Z., Wu, X., Rabee, A., & Forster, R. J. (2017). Total rRNA-Seq Analysis Gives Insight into Bacterial, Fungal, Protozoal and Archaeal Communities in the Rumen Using an Optimized RNA Isolation Method. Frontiers in microbiology, 8, 1814. https://doi.org/10.3389/fmicb.2017.01814
Kounosu, A., Murase, K., Yoshida, A. et al. Improved 18S and 28S rDNA primer sets for NGS-based parasite detection. Sci Rep 9, 15789 (2019). https://doi.org/10.1038/s41598-019-52422-z
Reviewer 3
L48: Use italics for genus/species names (here and later e.g. L62)
>> Modified, thank you.
L60-70- it is worth reporting in the introduction that one new species of anaerobic fungus was recently discovered from camel and appeared to be specific to camels. Protozoa can show distinct patterns of host specificity, so would be useful to mention if any are distinctive to camles. Useful to link this to the pseudoruminant point below.
>> Thank you, we have supported the introduction with some information about Oontomyces fungi, that was isolated from camel rumen, and we highlighted the protozoal genus Diplodinium, which include some species such as Diplodinium cameli, which was discovered in the rumen of Egyptian camel .
L72-80: should mention that camels are not ruminant but rather pseudoruminants and discuss potential implications of this difference (ie example of convergent evolution)
>> Thank you, The classification of camel as a true or pseudo-ruminant is controversial and there is no agreement between the scientists around this point and it needs more studies. I collected most of views around this point and enclosed below.
>> also we add some sentence regarding the retention time of feed particles in the rumen of camel compared to other ruminants.
Camel rumen
There is no agreement between the scientists regarding to the number of chambers in camel stomach. Yagil, 1982, stated that, although camels ruminate, they are not true ruminants, as they lack the four well-defined stomachs of the ruminants; the rumen, reticulum, omasum and abomasum. The anatomy of all members of the Camelidae is considered to be similar, but most of the available data on the anatomy of the alimentary canal have been obtained mainly from the llama. In spite of the fact that the camel ruminates, its ingested feeds are subject to microbial digestion and the final metabolic products are similar to those in true ruminants, it is classified as pseudo ruminant, but this classification is mainly due to the significant differences in the structure and function of the digestive system of the camelids (Tylo-pods) and the true ruminants (Bhattacharya, 1986). Camels are classified as pseudo-ruminants because its foregut is differentiated into three compartments (rumen, reticulum and the gastric secreting abomasums instead of four compartments like true ruminants (Engelhardt et al., 2007). Hegazi, 1950, described the camel as having the same four compartments as other ruminants, but with the external constrictions between the omasum and abomasum being less well defined in the camel. Wardeh, 2004, indicated that, the rumen (first compartment) of the camel is characterized by its unique exterior glandular sacs which secrete a mucus-like substance that differs in composition from rumen liquid. The third compartment is absent from the honeycomb-like structure which does not distinctively separate from the fourth compartment. The motility pattern of the compound stomach of the camel differs from that of the true ruminants. The biochemical pattern of microbial fermentation is generally similar between ruminants and camelids although concentrations of short chain fatty acids in the forestomach contents are always higher in camels.
References
Bhattacharya, A.N. (1986). Structural peculiarities in the digestive system of camel. Al-Jouf Range and Animal Development Centre. Saudi Arabia.
Engelhardt, W.; Dycker, C. and Lechner-Doll, M. (2007). Absorption of short-chain fatty acids, sodium and the foregut of camels. J. Comp. Physiol. B. Biochem. Syst. Environ. Physiol. 177:631–640.
Hegazi, A.H. (1950). The stomach of the camel. Brit. Vet. J. 106: 209–213.
Wardeh, M. F. (2004). The nutrient requirements of the Dromedary camel. J. Camel Science. 1: 37-45.
Yagil, R. (1982). Camels and camel milk . Food and Agriculture Organization of United Nations. Rome, FAO.
L87: Correct to “Most of molecular-based assessments of microbial groups in the rumen have either to relied on RNA/DNA-amplicon sequencing”. Here and elsewhere are examples of minor grammatical errors. One final proofreading from native English speaker would help expunge these.
>Modified, thank you.
L100: as it is not reliant on primers for known microbes
>> Modified, thank you.
L122: need more details of the feed. For example was G2 all fresh forage; authors state forage diet OR clover. Seems that the camels in this group did not receive a uniform diet. Can they state for each camel what the feed was? and for how long prior to slaughter the camels receive each diet
what was interval between last feeding and slaughter?
>> Thank you, we improved the section of animals’ diets in the whole paper and in supplementary table S2.
L137: were ground using liquid nitrogen.
>> Modified, thank you.
L186: A useful initial figure (clearer than Table 1 would be to show relative abundance of transcripts of the main microbial groups. Fungi are lower than other groups but unexpectedly not more abundant in the solid fraction. Is anything known about relative rRNA abundance per unit active biomass in bacteria vs archaea/eukaryotes. Or do these findings mean that bacteria comprise ca 90% of total microbial biomass in this habitat?
>>Thank you so much:
> Regarding the table, we converted the relative abundance of microbial groups to paragraph figure and enclosed it as a supplementary.
> Regarding the fungi we noted that their relative abundance was high in high-fiber diet (G2 and G3), that was expected; also it was higher in solid fraction as overall mean. However, it was higher in liquid fraction in G1 and G3, which was not expected. We repeated the analysis two times and we got the same results. The liquid fraction might has more active released zoospores than the solid fraction, especially we used RNA sequencing that targets active cells.
> Unfortunately, we did not conduct any qPCR to evaluate the copy number in this study, but it will be useful in future studies to use it with RNA sequencing.
L186-195: Was any rRNA originating from the feed components detected, for example chloroplast rRNA. One treatment contained fresh forage so this would be expected.
> Thank you, our bioinformatics pipeline did not show something like that.
L234: Methanobrevibacter
>> Modified, thank you.
L239: Methanomicrobiales /Methanomicrobiaceae? (need to check the organism names, families etc carefully!)
>> Corrected, thank you.
L241-264: Identification of eukaryotes was done with 18S but for fungi at least this locus shows poor resolution. Use of LSU would provide more accurate taxon identification, and probably for protozoa too.
>>The characterization of rumen protozoa based widely on 18S rDNA as a taxonomic marker. While the characterization of rumen fungi based on 18S rDNA and ITS1. The 18S rDNA is not variable enough to phylogenetic differentiation and there is a variability of ITS1 size between the strains. Therefore, there is a move towards the use of 28S rDNA or the Large Subunit(LSU) as a barcoding locus. However, using LSU needs more isolation and sequencing work to deposit more sequences into the ribosomal databases especially with the emergence of Third-generation Sequencing that sequence longer DNA/RNA segments from single strains. Also, more studies are needed to find suitable primers to reach the optimum amplicon size for Next-generation Sequencing (NGS) and to find primer combinations able to detect all the strains, which will open the door to depend solely on 28S rDNA as a genetic marker. I would recommend combining the use of ITS1, 18S rDNA, and 28S rDNA in the future to check the reliability of modern and old data sets. In our study, we used 18S rRNA shotgun sequencing which enables the measurement of transcripts, allowing us to identify known and novel strains instead of trying many primers.
You can some more details regarding my response in:
Edwards, J. E., Forster, R. J., Callaghan, T. M., Dollhofer, V., Dagar, S. S., Cheng, Y., Chang, J., Kittelmann, S., Fliegerova, K., Puniya, A. K., Henske, J. K., Gilmore, S. P., O'Malley, M. A., Griffith, G. W., & Smidt, H. (2017). PCR and Omics Based Techniques to Study the Diversity, Ecology and Biology of Anaerobic Fungi: Insights, Challenges and Opportunities. Frontiers in microbiology, 8, 1657. https://doi.org/10.3389/fmicb.2017.01657
Elekwachi, C. O., Wang, Z., Wu, X., Rabee, A., & Forster, R. J. (2017). Total rRNA-Seq Analysis Gives Insight into Bacterial, Fungal, Protozoal and Archaeal Communities in the Rumen Using an Optimized RNA Isolation Method. Frontiers in microbiology, 8, 1814. https://doi.org/10.3389/fmicb.2017.01814
Kounosu, A., Murase, K., Yoshida, A. et al. Improved 18S and 28S rDNA primer sets for NGS-based parasite detection. Sci Rep 9, 15789 (2019). https://doi.org/10.1038/s41598-019-52422-z
L261: Odd to see Spizellomyces. Not an anaerobe but is associated with dung sometimes. Maybe was on the ingested forage.
>>Thank you for this, your opinion supports what we have mentioned in fungi discussion regarding this fungus.
Fig. 2: Some of the labels on the plot are difficult to read- can the plot be re-done to avoid this problem.
>> improved, thank you.
L45 Fibrobacter spp.
>> thank you.
Experimental design
The manuscript describes original primary research within Scope of the journal. As noted above, the methods are appropriate and innovative. Good that samples were separated into liquid/solid components to allow comparison of planktonic vs attached microbial communities. Similarly bioinformatic approaches are also appropriate. I assume that experiments were conducted in compliance with ethical /legal guidelines in Egypt.
The main thing that perplexes me with this study us the unusual choice of treatments and the uneven number of replicates (6,3,2; see L123-128). This is odd and is not explained. Having only 2 reps for one treatments greatly constrains subsequent statistical analysis.
>>Thank you, we collected equal number of samples. However, the isolated RNA in some samples was degraded so that we discarded those samples, which generated unequal animal number in the groups.
I was also unclear why the three treatments were chosen. More precise detail of these is needed as noted above. Why did the authors not just pick simpler diets, e.g. 100% straw, 100% green clover and 100% clover hay. This would have more clearly addressed their first hypothesis (“get insight into the composition of 116 active microbiota in the rumen of camels reared under different feeding systems”)
>> thank you so much, we will consider this in our future work. Additionally, the diet G1 contained the two forages that have been used in G2 and G3 besides concentrates mixture.
Validity of the findings
The study is very interesting in that it assesses abundance of active biomass in a more effective way that DNA metabarcoding. It think that the identification of the eukaryote communities would be enhanced by analysis of LSU (28S) transcripts to obtain more accurate taxon identification. It is also valuable in that the relative proportions of transcripts of members of the bacterial, archaeal, protozoan and fungal communities can been assessed together (and this aspect should be discussed in greater detail in the discussion).
>>Thank you so much for this, we discussed this point above.
Sadly, the odd experimental design and uneven/low replicate numbers compromises what can be deduced from the feeding treatments but I think the study is nevertheless valuable.
>>Thank you, we discussed this point above and we must consider this point in the future.
Comments for the Author
I think that the correct spelling of the 2nd author’s name is “Robert J Forster”
>> Thank you, the name of Prof. Forster is correct in the manuscript. However, we will correct it in the paper’s account.
" | Here is a paper. Please give your review comments after reading it. |